The Effect of Sector Level FDI on Income Inequality Within OECD Countries

(1)

The Effect of Sector Level FDI on Income Inequality Within

OECD Countries

University of Groningen

Faculty of Economics and Business

MSc Thesis International Economics and Business

Name Student: Jamila Rekhaoui Student ID number: S2569701

Student email: j.rekhaoui@student.rug.nl Date Paper: 18-06-19

(2)

Abstract

The paper analyzes the impact of sector level FDI on income inequality in 11 OECD countries in 6 time periods each of around 5 years over the years 1986-2005. The paper adopts a skill biased perspective to analyze the effect of FDI inflows into different sectors. Panel analysis is carried out to empirically test whether there exists differential effect concerning FDI inflows in different sectors on income inequality. The results show that the high-skill service sector shows a positive and the low-skill service sector a negative association with income inequality, whereas the manufacturing high and low skill sectors does not show a significant effect on income inequality. The agriculture sector demonstrates a non-linear association between sectoral FDI and income inequality.

(3)

As a result of economic globalization foreign direct investment (FDI) has increased rapidly as an important source of international capital transfer over the last few decades (Dellis, Sondermann & Vansteenkiste, 2017). Although there has been a decline in FDI inflows in recent years, many countries continue to employ incentives to attract FDI (UNCTAD, 2018). Consequently, for OECD as well as non-OECD countries “incentive based competition” for FDI has become an increasingly common phenomenon (Dellis, Sondermann & Vansteenkiste, 2017). From 1980 to 2018 the world’s FDI stock (defined as the total level of direct investment at a given point in time) increased from 1 trillion US dollars (6% of world GDP) to nearly 33 trillion US dollars (39% of world GDP) (Dellis, Sondermann & Vansteenkiste, 2017; OECD (2019)). This fast growth in this time period is largely a consequence of the reduction in trade as well as investment barriers between OECD countries. Besides this, many OECD countries have harmonized markets (e.g. the European Union) and have made conscious efforts to remove domestic impediments to FDI through reform and privatisation (OECD, 2002).

(5)

The academic debate has increasingly paid attention to the effect of FDI on income inequality. Improvements in the availability of data have generated a substantial number of works on the relationship between FDI and inequality in the past two decades (Bogliaccini & Egan, 2017). However, researchers have remained inconclusive and the evidence remains limited regarding the link between FDI and income inequality in developed countries (Chintrakarn, Herzer & Nunnenkamp, 2010). In other words, some researches point toward a positive and others toward a negative relationship. Some argue that FDI has an equalizing effect on income inequality, through for example positive spill-over effects between industries, and the skill upgrading of low-skilled workers (Chintrakarn, Herzer & Nunnenkamp, 2010). Whereas others point toward the disproportional benefits of high-skilled workers which points toward a positive effect of FDI on income inequality (Bogliaccini & Egan, 2017). Most of these researches have been devoted to the effect of FDI on inequality as aggregate. However, the impact of FDI on a host country could potentially be different depending on the sector at which it is aimed. This research avenue remains underexplored and extremely limited especially concerning developed countries, therefore leading to the core inquiry of the present paper.

(6)

This paper argues that skill biased technological change can explain why the effect of sectoral FDI inflows differ in their effect on income inequality. The service sector is a more heterogeneous sector compared to the manufacturing and primary sector, which means that there exists a greater discrepancy between high and low skilled workers. It is therefore favorable to split the service sector into relatively low and high skilled workers. Analyzing the service sector, there exists evidence for a wage skill premium. This means that when FDI flows into the high-skill service sector, it is likely that income inequality increases. This is different in the manufacturing sector, as this sector is less heterogeneous, where FDI inflows prove to be complementary to both low and high skill workers. It could therefore have a negative effect on income inequality, however for the OECD countries included in this paper the low-skill manufacturing sector is small and declining. It is therefore expected that inflows in the low-skill manufacturing industry have no significant effect on income inequality. Furthermore, as the manufacturing sector does not have a large polarized wage structure it is likely that there is no skill wage premium, indicating no significant effect of FDI on income inequality into the high-skill manufacturing sector. Concerning the primary sector, high-skill biased technological change could explain increases in FDI in this sector, when new technologies are adopted. However, as the already relatively small employment in the primary sector, also faces a decline, it is expected that the primary industry does not contribute to income inequality.

This research contributes to the academic debate by exploring the impact of the inflow of FDI in different industries on income inequality in OECD countries. It aims to establish whether the effect of FDI inflows on income inequality is sector dependent, i.e. whether the inflow of FDI into some industries relatively contributes more (or less) to income inequality. This is done by panel analysis for sectorial inequality data, based on the Gini coefficient (based on household earnings), which is available for 11 OECD countries, in 6 time periods between 1980 and 2005. This measures the effect of FDI inflows on income inequality in various sectors. The central question of this paper is therefore as follows: “Is the relationship between FDI and income inequality in OECD countries sector dependent?”

(7)

an association with income inequality, while the primary industry demonstrates a non-linear relationship. However, FDI inflows at the sub-sector level within the low-skill manufacturing sector shows both positive and negative assosications with income inequality. Likewise, in the primary sector, agriculture demonstrates to have a positive effect on income inequality.

The first section reviews the most important literature in the field of the subject and elaborates upon the mechanisms which explain the link between sectoral FDI and income inequality. The second section discusses the empirical methods and model. The third section presents the analysis of the outcomes of the empirical section. The final section concludes the research by addressing the effect of FDI on income inequality, and whether this relationship is sector dependent.

II. Theoretical Framework and Literature Review

First of all, it is essential to define income inequality. Income inequality considers the entire population and highlights the gap between different levels of income (Haughton, & Khandker, 2009: 101). This research focuses primarily on individual earnings which are primarily determined by income from labor, because wages from labor are the primary source of income for most individuals (Figini and Görge, 2011). More specifically, it focuses on the Gini coefficient, which is a statistical measure representing the income distribution of a country but which can also be measured at the industry level (Haughton, & Khandker, 2009: 101-106). Figure 1 shows the trends in inequality at the country level for OECD countries when all sectors are pooled. The calculations are based on ‘all individuals aged 25-54 with non-zero earnings excluding those not classified in a sector’ (Thewissen, Wang & van Vliet, 2013). It can be observed that OECD countries follow a similar path, with an increase in income inequality over the time period 1985 – 2005.

(8)

entails the taking over or merging of capital, assets and liabilities of existing enterprises (UNCTAD, 2009). As previously discussed, FDI inflows have shown an enormous increase in the service sector and seem to be the sector receiving the largest inflows of FDI within OECD countries, this is evident from figure 1. FDI inflows demonstrate to be most important in the high-skill service sector. FDI inflows can be regarded considerable for both manufacturing high-skill and service low-skill sectors. The figure also demonstrates the small inflow of FDI the primary and manufacturing low-skill sectors compared to the other sectors.

Figure 1 FDI inflows by Sector and Skill for OECD countries 1985-20051

i. Channels of the effect of FDI on wages

As wage is predominantly the primary source of income for most individuals, it has strong implications for income inequality (Figini and Görge, 2011). It is therefore useful to establish how foreign firms determine wages. Following the argument of Hijzen et al. 2013, with perfect market competition, firms are price-takers resulting in all workers being paid their marginal product. Therefore, it is expected that firms established abroad pay the same wages to workers

1_{It should be noted that this model is based upon the panel data used for the empirical analysis, which consists of} gaps and therefore does not display a totally accurate picture. It is merely to illustrate the large differences of inflows between different sectors.

0 100000 200000 300000 400000 500000 600000 700000 Primary Manufacturing

Low-Skill Manufacturing High-Skill Service Low-Skill Service High-Skill

F

D

I Infl

ow

s

(9)

that possess similar skills holding similar positions. However, individuals having a similar position in a foreign owned firm, might earn more than in a national firm. Hijzen et al. 2013, outline three perspectives on this.

Firstly, there might be search friction. This is because a multinational company often has limited access to local networks. Therefore, they might need to offer better wages in order to attract qualified workers. Secondly, foreign firms might offer better wages, because they want to prevent productivity spillovers to competitors, by reducing the worker turnover. Foreign firms are usually more productive compared to domestic firms (Hijzen et al., 2013). This productivity advantage can be due to their specific knowledge, which is inevitably (partly) obtained by the workers at this firm. Therefore, foreign firms might offer a higher wage, as it reduces the likelihood that workers will leave and pass this firm-specific knowledge onto other firms. Thirdly, higher wages might be offered to motivate workers, especially as the legal and cultural system differs from the host country. These higher wages are offered as a compensation to overcome the difficulties concerning cultural and legal differences. This way foreign firms can manage workforce differences more effectively.

(10)

Moreover, job creation as a result of increased FDI is most likely to be complementary to skilled labor whereas it is often a substitute to low skilled labor (Harrison & Hanson, 1999; Bogliaccini & Egan, 2016). This demonstrates that the inflow of capital complements skilled works and reasonably increases the relative demand for skilled workers (Bogliaccini & Egan, 2016). The next section will explore this skill bias in further detail.

ii. Skill bias and the effect of FDI and income inequality

Economic literature has dealt extensively with the effect of FDI on income inequality. Much literature has found that it increases the wage gap between high- and and low-skilled workers, increasing income inequality (Aitken, Harrison & Lipsey, 1996; Feenstra and Hanson 2001; Velde 2002; Figini and Görge 2011). Much of the academic literature can be placed within two theoretical perspectives dealing with skill biases: international trade and technological progress (Krugman 2000; Figini and Görge 2011; Feenstra and Hanson 2001).

International trade

A considerable amount of literature builds upon equilibrium trade models explaining the relationship between FDI and inequality, implying that international trade causes an increase in high-skill labor demand (Levy & Murnane, 1992; Feenstra and Hanson 1997; Feenstra and Hanson, 2001). Models concerning North-South FDI models are often based on comparative advantage and relative factor endowments. In these models, FDI potentially exacerbates inequality between low and high skill workers, in both the developed and the developing country (Chintrakarn, Herzer & Nunnenkamp, 2010). In this strand of research international trade is an important explanation for the increase in the wage gap.

(11)

consist of relatively high-skilled production. This results in an increase of skilled workers in each region. Due to outsourcing, the demand for low-skilled labor in the North decreases resulting in an increase in demand for relatively high-skilled labor in the South. This increases the relative wage for skilled workers in the North as well as in the South.

There exists much support for the notion that FDI increases inequality by increasing the wage premium for skilled labor in developing and developed countries as a result of increased trade (Herzer, Hühne & Nunnenkamp, 2012). For example, it has been established by multiple researches that in the case of Latin America, the skill premium has been confirmed. Aitken et al. (1996) analyze Mexican, Venezuelan and US manufacturing plants, using an ordinary least squares (OLS) and two stage least squares (2SLS) model. They find that even though there are very different levels of development between these countries, a robust result that increased levels of FDI, are associated with higher wages for skilled workers in each country. Gopinath and Chen (2003), use time series data, using a fixed effects model, on 11 developing and 15 developed countries. They test the general equilibrium model based on relative factor endowments, stating that FDI increases wages in host countries. They find that across countries, developing and developed, FDI reduces income inequality. However, they also find evidence that FDI inflows increase the income gap between low- and high-skilled laborers in developing countries (they do not test this for developed countries, presumably because they focus on inward FDI, which cannot explain inward FDI wage gaps within developed countries based on relative factor endowments). Likewise, similar results are obtained by Nunnenkamp, Schweickert and Wiebelt (2007) on the basis of a computable general equilibrium (CGE) model, for Bolivia demonstrates the trade-off effect between economic growth and income inequality due to FDI inflows in the medium to long-run. This trade off effect emerges as a result skill premium effect, benefitting skilled workers a developing country.

(12)

Technological Change

Kaiser (2000), defines Skill-Biased technological change (SBTC) as follows: ‘Skill-biased technological change is said to exist if new technology is complementary to skilled labor and substitutive to unskilled labor”. Te Velde (2004) elaborates upon this perspective, stating that FDI potentially increases productivity growth. This can happen in foreign firms due to technological advancements and can also have spill-over effects to domestic firms. If skilled workers benefit most from this productivity growth, FDI potentially raises SBTC.

Furthermore, technology skill biases perspectives can be explained by the endogenous growth model of Aghion and Howitt (1998). This model aims to explain different skill level effects on wages and output (Figini & Görge, 2011). Figini and Görge 2011, who build upon Aghion and Howitt (1998), develop the following production structure:

𝑌 = !𝐴_!! ! 𝓍!!𝑑𝑖

!/!

, with 0 ≤ 𝛼 ≤ 1 (1)

Figini and Görge (2011) explain the model as follows: in this production structure A is defined as the technology parameter, if A is 1 then the old technology is still in place, but when A is larger than 1 the new technology is implemented. This means that A is a shift variable, because when there is an increase in technology, A is increased by a constant. The authors theorize that there are two stages when there is an increase in technology due to FDI. In the first stage, skilled labor is used to experiment with the new technology, while the old technology is still being used, because the investment in the new technology is yet too small in order to employ the supply of skilled labor. So initially wage inequality is low. However, in stage two the firm implements the new technology, and now only uses this, therefore the wage of high-skilled labor increases, whereas the wage for low-skilled labor decreases. This is often used to develop a theoretical framework to explain how technological spillovers can account for the development of income inequality. It theorizes a non-linear relationship of FDI on wages.

(13)

However, in the second step when there are increased inflows of FDI into the economy, domestic firms will adopt the new technologies due to spill-over effects. This in turn reduces income inequality (Figini & Görge, 2011).

Non-linarites have been supported by various empirical research. Figini and Görg (1999) investigate wages between blue-collar and white-collar workers in Ireland within 17 different industries between 1979 and 1995. They find supporting evidence for the endogenous growth model of Aighon and Howitt (1998). Based on a pooled model using panel data, they find an inverted-U shape association between income inequality and the inflow of FDI. Furthermore, Taylor and Driffield (2005) conduct a similar study for industry-level panel data, using a fixed effect and GMM model, over the period 1983 to 1992 in the United Kingdom. They find a positive effect of FDI on wage inequality, but at a decreasing rate, which implies a non-linear relationship. However, Figini and Görge (2011) find mixed results for OECD and non-OECD countries. Using a fixed effect model, they find that there exists a non-linear effect of FDI on income inequality for non-OECD countries. However, they do not find this result for OECD countries. It should be noted that the aforementioned studies evaluating non-linearities focus on the manufacturing industry, and do not address the primary and tertiary sectors.

iii. Sectoral FDI, skill biases and the effect on income inequality

(14)

Thus far, there exists very limited literature discussing the effect of sectoral FDI on income inequality. Suanes (2016) explores the relationship between FDI and income inequality in different sectors; the primary sector (raw materials), manufacturing industry and services industry. She analyzed 13 developing countries over the time period 1980-2009. Based on a fixed effects, TSLS and first difference GMM model, a differential effect between the different sectors was found. In all models a positive effect between FDI inflow and income inequality in the services and manufacturing sector was discoverd, but no effect on the primary sector. Furthermore, Bogliaccini and Egan (2017) analyzed 60 middle income countries over the period 1989 to 2010. They find that FDI in services has a positive relationship with income inequality, using an error correction model with country fixed effects and year dummy variables. However, FDI in the manufacturing sector does not show such a strong relationship with income inequality. The researchers claim that skill biases explain this finding. It is therefore reasonable to assume that FDI has a different effect depending on the sector where it is aimed at.

Building on the skill bias argument, several observations can be made on the industry level. It is reasonable now to assume that inward FDI flowing into different industries varies in the way it affects income inequality. The structure of the different sectors hints at potential explanations for these different effects. The services sector is characterized by an extremely broad range of jobs and it can therefore be classified as a strongly heterogeneous sector (Bogliaccini & Egan, 2016). As discussed by Evans and Timberlake (1980), this sector ranges from waiters at restaurants and housekeepers to financial experts and doctors. The service sector cannot be categorized as labor- or capital-intensive sector (Suanes, 2016). It is therefore useful to divide the sector by skill-level to make sense of different effects of FDI in this sector. Sectors like retail trade and hotels and restaurants are more labor intensive using relatively more low-skilled labor, whereas the real estate and finance sector are more capital intensive using relatively more high-skilled labor (Suanes, 2016). If a sector presents a more heterogeneous nature, there exists a greater discrepancy between high and low skilled workers. This undoubtedly has implications for income structure within the service sector.

(15)

therefore most likely increase in income inequality. As mentioned before, the inflow of FDI into the high skilled service sector is extremely large for OECD countries. Increased inflows of FDI into these sectors are therefore likely to have a strong impact on income inequality. Concerning the inflow of FDI in the low-skill service sector, foreign firms potentially place a downward pressure on income inequality. This effect is only expected to happen when FDI is directed at those sectors in which there is relatively more low-skilled labor.

Hypothesis 1: “FDI inflows into the high-skill service sector have a positive effect on income inequality, whereas FDI inflows into the low-skill service sector have a negative effect on income inequality.

The manufacturing sector as a whole proves to be less heterogeneous in comparison to the service sector (Suanes, 2016). There exists supporting evidence that wage differentials are smaller in the manufacturing industry compared to the service industry (Bogliaccini & Egan, 2017). Pinto and Pinto (2008) claim that foreign investments in the manufacturing industry are, compared to other sectors, more complementary to low- as well as high-skill labor. This suggests that FDI in the manufacturing sector may offer similar benefits to high as well as low-skill labor. Suanes (2016) argues that this potentially improves the income distribution in host countries. However, there is limited offshoring into low-skilled manufacturing, within OECD countries (OECD, 2008), as this region as a whole presents a more high-skilled work force, and offshoring to low-skilled manufacturing is generally aimed at developing countries. Combined with the fact that there exist income structures within the declining low-skill manufacturing sector in OECD countries, it is expected that there is no significant effect of FDI inflows on low-skill manufacturing. Furthermore, as this sector does not demonstrate a large polarized wage structure, it is likely that there is no skill wage premium. This in turn means that in absence of this wage premium it is unlikely that high-skill manufacturing affects income inequality.

Hypothesis 2: “There is no significant effect of FDI inflows into the low- and high- skill manufacturing sector on income inequality”

(16)

inequality are ambiguous. It is plausible to expect that FDI inflows in the primary sector bring in new technologies leading to skill biased technological change (Bonassi, Giovanneti & Ricchiuti, 2006). However, investments in the primary industry generally involves large capital, but relatively little labor. Suanes (2016) argues that, even though agricultural jobs are often low-skill, most of the agricultural sector is concentrated in the hands of a few, since employment is extremely low. If in that case the sector benefits from new technology, or technology spill-overs, these people might benefit relatively compared to the rest of the economy. Conversely, there has been a large decline in employment in the primary sector in OECD countries as a whole. As employment in the already labor-scarce in primary sector is declining even further, it is expected that FDI inflows within this sector do not contribute to income inequality (Bonassi, Giovanneti & Ricchiuti, 2006).

Hypothesis 3: “FDI inflows in the primary sector are not associated with income inequality”

III. Data and Methods

i. Data

(17)

Thewissen & van Vliet, 2014). Furthermore, the Gini employed is based upon the first order correction, to overcome an underestimation bias for small samples (Wang, Thewissen & van Vliet, 2014).

The independent variable FDI is obtained from the OECD database. There exists two often used measures for FDI, which are FDI inward stock or FDI inward flows. In general stocks capture the long-run effect of FDI better, whereas the FDI inflows prove to fluctuate more and are better to measure short-run effects. For the purpose of this research FDI inflows are used, because of the availability of FDI data on the sectoral level. Considering the timeframe of the study, this does not pose a strong limitation on the analysis. FDI is taken as a percentage of value added in order to control for country size.

Databases used for variables at the industry level are based upon the International Standard Industrial Classification (ISIC), which is a standardized classification scheme, making the data easily comparable. In order to demonstrate whether the service sectors contribute more to income inequality than the manufacturing sector, this research divides the sectors into sub-sectors. These sectors depend on the skill required. The sectors will be categorized as follows: primary sector, service sector high-skill, service sector low-skill, manufacturing low-skill, and manufacturing high-skill. Table 1 gives an overview of the various sectors which are part of the aforementioned categories

Table 1 Sub-sectors categorized by skill level Primary sector Manufacturing low-skill Manufacturing high-skill

Service low-skill Service high-skill

Agriculture and fishing

Food, beverages and tobacco

Chemicals and chemical products

Electricity, gas and water Finance, and insurance Mining and quarring Textiles, textile products

Motor vehicles Hotel and restaurant Real estate

Other non-metallic mineral products

Machinery and equipment

Community, social and personal services

Business services

Transport equipment Postal and courier services

(18)

However, it should be noted that wood and products of wood products, and pulp, paper, paper products, printing and publishing is excluded from the analysis, as these sectors do not match the OECD database. This is because there are two Gini coefficients available for these sectors, however only one category for FDI. It is therefore unclear how much FDI is attributed to each sector.

ii. Control Variables

Control variables are included in the model, in order to control for variables which potentially influence income inequality as opposed to (sectoral) FDI. These control variables are in line with the research of Figini and Görge (2011) and Bhandari (2007). An overview of the control variables and expected signs can be found in table 2.

• Economic development. In order to control for economic development value added is included in the model, as GDP is not available at the sector level. Value added is likely to show an inverted U relationship with income inequality based upon the Kuznets curve, theorizing that in the beginning stages of economic development income inequality increases, but once it reaches a certain point, inequality declines (Bhandari, 2007). Value added squared is added to include a non-linear effect of value added on income inequality. This variable is obtained from the OECD STAN database ISIC Rev.4.

(19)

value added (also obtained from STAN ISIC Rev.3) to control for country size. The older version of the OECD STAN database has been used due to data availability.2

• Unemployment rate. The unemployment rate is expected to have a positive relationship with income inequality. If there is a larger share of the population without a job, there exists a larger supply of labor. The number of people looking for a job is larger than the amount of jobs available. This in turn affects wages, as employers are not incentivized to pay higher wages as labor is easily substitutable. However, this most likely happens in the low-skill sector, where laborers are relatively easy to replace. Therefore, it is most likely that when the unemployment rate is high, low skilled labor wages will be affected. High skilled labor is less likely to be affected, as these workers are not easily replaceable. It is therefore expected that a higher unemployment rate leads to increased inequality. Unemployment data is not available on the sector level, and has been obtained from the World Bank database and is measured as the unemployment rate as a percentage of the total labor force.

• Openness to trade. With respect to openness to trade openness, it is suggested by the Stolper–Samuelson theorem that when trade increases, inequality decreases in countries which are relatively low-skill labor abundant, as opposed to an increase in countries which are relatively high-skill labor (Bhandari 2007). Trade openness is defined and measured as total imports plus total exports over GDP (Figini and Görge, 2011) Therefore the sign of this control variable has a positive sign, as the countries included are high skill abundant countries. Trade openness is obtained from the World Bank and is only available on the country level.

• Education. In line with the research of Bhandari (2007) and Figini & Gorge (2011), education is measured as the secondary education enrollment rates as a percentage of the population. Mihaylova (2015) argues that an increase in the number of skilled workers places a downward pressure on the skilled wage premium, resulting in less inequality. It

(20)

can therefore be argued that human capital is positively affected by education (Mihaylova, 2015). Its therefore expected that Education has a negative sign. However, education is not available on the sector level and is therefore based upon secondary school enrollment rates, in line with previous research (Figini and Görge, 2011). Education is obtained from the World Bank database.

Table 2 independent variables

Variable Description Expected

sign

Value added Value added per sector in logs +

Value added Squared

Value added per sector in logs squared - Domestic

investment

Gross fixed capital formation (%value added) -/+ Unemployment Unemployment rate (% total labor force) + Openness to trade Total imports plus total exports over GDP + Education Secondary enrollment (%total labor force) -

FDI Inflows per sector (%value added) +

iii. Empirical model

Panel analysis is used to analyze the effects of sectoral FDI on income inequality over sectors, countries and time. The sample consists of 11 OECD countries, 6 periods each of around 5 years between 1980-2005 with gaps. This forms a unbalanced panel of around 439 observations, depending on which variables are included in the model. The countries Czech Republic, Denmark, Finland, Germany, Ireland, Sweden, UK, US, Austria, Belgium, Poland and Spain are included in the regression. These regions are included because the availability of sectoral inequality data. Furthermore, the study period is 1980-2005, also because of the availability of data.

(21)

carried out (Appendix A2). This test gives a significant result, confirming that a fixed effects model is most suitable given the data. In order to take into account the influence of different sectors on income inequality, dummy variables are created for aforementioned sub-categories. Consisting of primary, manufacturing low-skill, manufacturing high-skill, service low-skill, and service high-skill sectors. Subsequently these sectorial dummies are multiplied with the FDI variables, in order to attempt to capture the potential sectoral effect on income inequality. Thus, the sectoral categories represent the reference group, and will each be compared A seperate model will also test potential non-linearities as theorized by Aghion and Howitt (1998), done by adding the square term to FDI in one model, but also to each of the dummy variables in a separate model. Additonally, a sub-sector analysis is carried out, by taking dummies of each of the sectors displayed in table 1, to see whether the regressions based upon the high skill and low skill categorization is in line with the underlying sub-sectors.

Furthermore, robustness checks will be carried out, to control for the fact that the United States has especially extremely high FDI inflows of high skilled service activities (OECD, 2019). Therefore this sector will be dropped for the United States and a robustness check will ben carried out for the other countries.

The basic model to determine the effect of industry level FDI on industry level income inequality is as follows:

GINIijt = β0 +β1FDIijt+ β2logVAijt+β3logVA2ijt+β4GFCFijt+β5EDUCit+β6TRADEit+β7UNEMPjt + 𝛼ij + 𝛿t+

uijt

GINI = sectoral gini coefficient FDI = Foreign Direct Investments ED = Economic Development

ED2_{= Economic Development Squared}

GFCF = Gross Fixed Capital Formation EDUC = Education

TRADE = Openness to trade

For the model including dummy variables the following equation is used:

GINIijt = β0+β1FDI*primaryijt+β2FDI*MLSijt β3FDI*MHSijt β4FDI*SLSijt +β5FDI*SHSijt +

(22)

Where, in both equations, subscript i denotes industries, j denotes years and t time. the alfa indicates that fixed effects are included in the model for both year and industry. Moreover, the model controls for year fixed effects which is indicated by delta with subscript t. The Gini coefficient is available at the country and industry level, as well as FDI, value added and gross fixed capital formation. However, the control variables education, openness to trade and unemployment are not available at the industry level and therefore are restricted to the country level.

As theorized value added does not only demonstrate a linear relationship (appendix A1) but also highly skewed relationship. Subsequently, value added and its squared are included as logs in the model. Gross fixed capital formation shows a weak negative relationship. Furthermore, FDI demonstrates a weak positive relationship with the Gini coefficient. It is clear from this scatter plot that many values for FDI demonstrate low values, and are highly concentrated around 0. This is because FDI is taken as a percentage of value added, and can therefore be regarded as logical. This is because generally speaking value added into a sector demonstrates a much higher number compared to the amount of FDI flowing into that same sector. Figure 1 demonstrates the correlation between the Gini coefficient and FDI, where a weak positive relationship is observable. Unemployment, trade openness and education are all measured at the country level and all demonstrate a similar pattern showing a week negative relationship.

Figure 1 Scatterplot Gini coefficient and FDI

(23)

Table 3 shows the correlation matrix, as highly correlated variables could potentially lead to issues of multicollinearity. This table demonstrates no serious issues of correlated variables. Additionally, a vif test is used to confirm that there are no issues concerning highly correlated variables. The vif test (Table A3) demonstrates that there is extremely high multicollinearity between the log of value added and the log of value added squared, which is logical and can be ignored. In analyzing the vif test, these numbers are all below 4-5 (this is often regarded as a guideline for multicollinearity using the vif test). Thus, based on the correlation matix and the vif test, there are no multicollinearity issues present in this model.

Year dummies are included into the fixed effects model as they will pick up the variation in outcome over time which is not attributable to the explanatory variables. This is added because in the fixed effects model just the individual, so in this case industry and country fixed effects are eliminated. It is therefore useful to include year fixed effects as aggregate trends influencing the regression are removed. A test (Table A4), confirms that fixed effects are useful in this model.

(24)

Table 3 Correlation matrix

IV. Analysis

(25)

Table 4 Descriptives

Variable Obs Mean Std. Dev Min Max

Gini coefficient 674 28.901 7.574 8.2845 59.711

FDI 409 2.366 6.206 -33.990 36.262

Value added 576 186183.3 480967 131.888 3968035

Gross fixed capital

formation 520 22.556 16.139 2.194 163.572 Unemployment 627 8.401 4.120 3.894 22.675 Trade openness 684 65.932 33.466 16.859 146.698 Education 646 106.512 12.716 83.92 151.851 Primary* FDI 377 0.147 1.427 -5.547 24.825 ManuLowSkill * FDI 377 0.258 1.56 -13.755 28.571 ManuHighSkill * FDI 377 1.244 4.745 -33.991 34.624 ServLowSkill*FDI 377 0.765 2.783 -7.203 16.096 ServHighSkill*FDI 377 0.228 3.009 -9.840 36.262

The results of the fixed effects regression including cluster-robust standard errors are presented in table 5. The first column includes the control variables, the second column shows the results when FDI as an aggregate is added, and the third column shows the results of the model including the interaction effect between FDI and sectors. By analyzing the first model, including only the control variables it is evident that there are 446 observations for 9 countries. The following countries dropped out the regression: Belgium and Ireland, this is caused due to missing observations for several years for these countries. The control variables for value added and value added squared in logs demonstrate the expected signs. However, do not display significance. Domestic investment, included as gross fixed capital formation demonstrates a negative effect and therefore leads to a decline in income inequality. This effect is significant at the 1% level. Furthermore, as expected unemployment is positively associated with income inequality and is significant at the 5% level. Furthermore, trade openness is significant at the 1% level, however demonstrates the opposite sign as theorized. This implies from the regression that greater trade openness is associated with lower income inequality. Furthermore, education also shows the opposite sign, suggesting an increase in income inequality due to higher secondary school enrollment, but this effect is insignificant.

(26)

countries (the same countries are dropped from the regression as in the first model). The number of observations is considerably less compared to the first model. This can be explained by the large amount of missing values for FDI. The sign of the FDI coefficient is positive, implying a positive relationship with income inequality, but the effect is insignificant. This suggests that there is no aggregate relationship between FDI and income inequality. In this model the control variables demonstrate the same signs as in the first model, but drop in significance level. Gross fixed capital formation has become insignificant, unemployment is now only significant at the 10% level and trade openness at the 5% level.

(27)

Table 5: Fixed Effects estimation after adjusting for heteroskedasticity and autocorrelation SE

In order to test whether there exist non-linearities between FDI and income inequality, as explained in the theory, table 6 includes the squared terms of FDI. The first model analyzes whether FDI as an aggregate demonstrates a non-linear relationship with income inequality. In this model FDI, does not show support for a non-linear relationship, as the coefficients for FDI and FDI squared are both positive and insignificant. Most control variables are insignificant, only unemployment demonstrates a positive effect effect on income inequality, which is significant at the 5% level. The second model in table 6 tests the relationship between the FDI sector interaction dummies. In this model again, most control variables demonstrate an insignificant effect, only trade openness is negatively associated with income inequality, and is significant at the 5% level. More interestingly, FDI in the primary sector suggests a non-linear relationship with income inequality. Both the primary sector interaction variables are significant at the 5% level. The positive sign for the

Model (1) Model (2) Model(3)

Log Value Added 1.643 (5.277) 6.167 (6.413) 6.228 (7.538) Log Value Added Squared -0.080 (0.245) -0.280 (0.272) -0.291 (0.321) Gross Fixed Capital

Investment -0.065 (0.017)*** -0.042 (0.043) -0.139 (0.037) Unemployment 0.781 (0.327)** 0.675 (0.353)* 0.611 (0.381) Trade Openness -0.368 (0.092)*** -0.281 (0.09)** -0.253 (0.101)** Education 0.020 (0.050) 0.006 (0.052) 0.006 (0.059) FDI 0.014 (0.051) Primary* FDI -0.278 (0.170) ManuLowSkill * FDI -0.147 (0.131) ManuHighSkill * FDI 0.105 (0.068) ServLowSkill*FDI -0.123 (0.06)** ServHighSkill*FDI 0.068 (0.029)** N obs 446 339 312 N countries 9 9 9 R2 _Within Between Overall Adjusted R2 0.250 0.127 0.176 0.222 0.298 0.133 0.199 0.260 0.321 0.122 0.184 0.272 Chi2 Prob > chi2 6.97 0.0000 6.25 0.000 5.96 0.000

Year dummies Yes Yes Yes

Significance levels *≤0.10 **≤0.05 ***≤0.01

(28)

interaction between FDI and the primary industry and the negative sign for the squared interaction between FDI and the primary sector, suggests an inverted U relationship.

Table 6: Fixed Effects estimation after adjusting for heteroskedasticity and autocorrelation SE

Figure 2 shows a coefficient plot, based on the dummy interaction variables between the sub-sectors and FDI. This coefficient plot is constructed to analyze whether the effects of the primary, high- and low-skill manufacturing and services, are supported by their sub-sectors. For the primary sector, this is not the case as agriculture shows a positive effect on income inequality, which is significant. Because agriculture it is grouped with mining (which is insignificant) in the primary sector, the effect of the agricultural sector could be pulled toward an insignificant effect. For the

Model (1) Model (2)

Log Value Added 6.317 (6.738) 6.759 (6.729) Log Value Added Squared -0.2845 (0.289) -0.272 (0.299) Gross Fixed Capital Investment -0.043 (0.044) 0.023 (0.034) Unemployment 0.688 (0.035)** 0.558 (0.361) Trade Openness -0.283 (0.094) -0.225 (0.094)** Education 0.005 (0.052) 0.012 (0.059) FDI 0.010 (0.104) FDI squared 0.000 (0.004) Primary* FDI 1.435 (0.629)** Primary*FDI squared -0.079 (0.026)** ManuLowSkill * FDI -0.403 (0.293) ManuLowSkill*FDI squared 0.022 (0.019) ManuHighSkill * FDI 0.141 (0.180) ManuHighSkill*FDI squared -0.001 (0.008) ServLowSkill*FDI -0.034 (0.137) ServLowSkill*FDI squared -0.040 (0.009) ServHighSkill*FDI 0.039 (0.182) ServHighSkill*FDI squared 0.001 (0.006) N obs 334 339 N countries 9 9 R2 Within Between Overall Adjusted R2 0.295 0.132 0.1983 0.255 0.298 0.133 0.199 0.260 Chi2 Prob > chi2 5.46 0.0000 6.25 0.000

Year dummies Yes Yes

(29)

manufacturing low-skill sector (including the food, textile, non-metallic mineral product sectors) this also proves to be the case, as the textile industry shows a negative effect on income inequality, whereas the food industry shows a slight positive effect on income inequality. The other sectors of the manufacturing low-skill industry are insignificant. This might explain why overall manufacturing low-skill industry is significant, as there are alternating forces at play. Concerning the high-skill manufacturing sector, all sub-sectors demonstrate an insignificant effect on income inequality, which is in accordance with the outcome of the general model in table 5. Within the service low-skill sector all sub-sectors (electricity, construction, wholesale, postal, and social services) demonstrate a negative effect on income inequality, however only electricity, gas and water supply is significant. Concerning the high-skill service sector it is evident that both transport and telecommunication services, as well as the finance, business and real estate sector demonstrate a positive effect on income inequality. This is in agreement with both the previous results in table 6, as well as the theorized effect of the high-skill service sector on income inequality.

Figure 2 Coefficient plot for sub-industry levels after adjusting for heteroskedasticity and autocorrelation SE

i. Robustness checks

(30)

panel setting, reverse causality might be an issue. It could be the case that FDI is attracted by inequality, meaning that there are larger inflows of FDI as a result of larger inequality. However, in this research there exists different effects for different sectors. If it was the case that FDI is attracted by inequality, it would be expected that FDI does not go into one direction into only the service high-skill sector. The high-skill service sector has a fairly equal income distribution. So if reverse causality would be the case, it would be expected to flow into other sectors as well. Furthermore, robustness checks are carried out to address the issue that inflows within the United States high-skill service sector are much larger compared to the other countries. Table 7 therefore excludes the high-skill service sector for the United States. Based upon the model it is evident that the service high- and low skill- sector remain significant, representing the and the identical results compared to table 5. This implies that the larger inflows of FDI in the high-skill service sector in the US do not influence the results of the regression.

Table 7: Fixed Effects after adjusting for heteroskedasticity and autocorrelation SE Model

Log Value Added 6.240 (7.898) Log Value Added Squared -0.289 (0.340)

GFCF -0.144 (0.037) Unemployment 0.623 (0.381) Trade Openness -0.256 (0.101)** Education 0.006 (0.059) Primary* FDI -0.123 (0.061) ManuLowSkill * FDI -0.147 (0.131) ManuHighSkill * FDI 0.105 (0.068) ServLowSkill*FDI -0.122 (0.061)** ServHighSkill*FDI 0.068 (0.297)** N obs 305 N countries 9 R2 Within Between Overall Adjusted R2 0.316 0.121 0.186 0.265 Chi2 Prob > chi2 5.18 0.0000

Year dummies Yes

(31)

V. Conclusion

This research has analyzed the effect of sectoral FDI on income inequality, determining whether the effect of FDI on income inequality is sector dependent. This allows for a thorough analysis, leading to a more complete understanding of the link between FDI and inequality. Especially in light of recent shifts of increased FDI toward the service sector,. This research has used a panel data of 11 countries in the time period 1980 – 2005, with gaps. The main objective of this paper is to answer the question: “Is the relationship between FDI and income inequality in OECD countries sector dependent?” Based upon the empirical analysis, the relationship between FDI inflows and income inequality confirms to be sector the dependent The high-skill service sector demonstrates a positive relationship with income inequality, while the low-skill service sector has a negative association with income inequality. The manufacturing high and low skill sectors do not show any relationship with income inequality Thus the hypotheses, are therefore confirmed. Furthermore, the primary industry suggests a non-linear association with income inequality, supporting the idea that when new technology gets introduced, income inequality increases, but tends to fall back after the new technology is fully implemented.

Based upon the empirical findings, FDI inflows show a differential association with income inequality based on sectors. The high-skill service sector is positively associated with an increase in income inequality, whereas the low-skill service sector shows a negative relationship with income inequality. However, the manufacturing high and low skill sectors does not demonstrate an association with income inequality. FDI directed at the primary industry shows a non-linear relationship. These results prove to be robust to outliers. Though, at the sub-sector level, there exists some different result, where sub-sectors of the low-skill manufacturing sector shows both positive and negative associations with income inequality. Likewise, FDI in the primary sector, agriculture demonstrates to have a positive effect on income inequality. For both the high-skilled manufacturing, and high-skilled service sector, the results of the sub-sector analysis are in line with the sectoral analysis.

(32)

(33)

VI. References

Aitken, B., Harrison, A., & Lipsey, R. E. 1996. Wages and foreign ownership A comparative study of Mexico, Venezuela, and the United States. Journal of international Economics, 40(3-4), 345-371.

Aghion, P., Howitt, P., Brant-Collett, M., & García-Peñalosa, C. 1998. Endogenous growth theory. MIT press.

Bhandari, B. 2007. Effect of Inward Foreign Direct Investment on Income Inequality in Transition Countries. Journal of Economic Integration, 22(4): 888-928.

Blomstrom, M., & Kokko, A. (2003). Human capital and inward FDI. CERP Discussion Paper No. 3762

Bogliaccini, J. A., & Egan, P. J. 2017. Foreign direct investment and inequality in developing countries: Does sector matter?. Economics & fPolitics, 29(3), 209-236.

Bonassi, Giovannetti & Ricchiuti, 2006. The Effects of FDI on Growth and Inequality. Pro-poor macroeconomics: potential and limitations:119-144. Basingstoke: Palgrave Macmillan. Dellis, S., Sondermann, D., & Vansteenkiste, I. 2017. Determinants of FDI inflows in advanced economies: Does the quality of economic structures matter?. Working Paper Series No 2066, European Central Bank.

Evans, P., & Timberlake, M. 1980. Dependence, inequality, and the growth of the tertiary: A comparative analysis of less developed countries. American Sociological Review, 45(4), 531– 552

Feenstra, R. C., & Hanson, G. H. 1997. Foreign direct investment and relative incomes: Evidence from Mexico's maquiladoras. Journal of international economics, 42(3-4), 371-393. Feenstra, R., & Hanson, G. 2001. Global production sharing and rising inequality: A survey of trade and wages working paper No.w8372. National Bureau of Economic Research.

Figini, P., & Görg, H. 1999. Multinational companies and wage inequality in the host country: the case of Ireland. Weltwirtschaftliches Archiv, 135(4), 594-612.

Figini, P., & Görg, H. 2011. Does foreign direct investment affect wage inequality? An empirical investigation. The World Economy, 34(9), 1455-1475.

(34)

Haughton, J., & Khandker, S. R. 2009. Handbook on poverty and inequality. Washington, DC: World Bank.

Hijzen, A., Martins,S. P., Schank, T., & Upward, R. 2013. Foreign-owned firms around the world: A comparative analysis of wages and employment at the micro-level. European Economic review, 60(C), pages 170-188.

Kaiser, U. 2000. New technologies and the demand for heterogeneous labor: firm-level evidence for the German business-related services sector. Economics of Innovation and New Technology, 9(5), 465-486.

Krugman, P. R. 2000. Technology, trade and factor prices. Journal of international Economics, 50(1), 51-71.

Levy, F., & Murnane, R. J. 1992. US earnings levels and earnings inequality: A review of recent trends and proposed explanations. Journal of economic literature, 30(3), 1333-1381. Lipsey, R. E., & Sjöholm, F. 2004. Foreign direct investment, education and wages in Indonesian manufacturing. Journal of Development Economics, 73(1), 415-422

Nunnenkamp, P., R. Schweickert and M. Wiebelt 2007. Distributional effects of FDI: how the interaction of FDI and economic policy affects poor households in Bolivia. Development Policy Review 25 (4): 429-450.

OECD, 2002. “Trends and recent developments in foreign direct investment”, International Investment Perspectives, Paris.

OECD, 2008a. Social impact of foreign direct investment. OECD Policy Brief.

OECD, 2008b. Foreign Direct Investment Glossary. 2009. OECD Benchmark Definition of Foreign Direct Investment 2008, 227–243.

OECD 2019, FDI flows (indicator) Accessed 11 June 2019.

Pinto, P., & Pinto, S. 2008. The politics of investment partisanship and the sectoral allocation of foreign direct investment. Economics and Politics, 20(2), 216–254.

Suanes, M. 2016. Foreign direct investment and income inequality in Latin America: a sectoral analysis. Cepal Review.

Taylor, K., & Driffield, N. 2005. Income inequality and the role of multinationals: Evidence from UK panel data. Labour Economics, 12(2), 223-249.

(35)

Tsai, P.L. 1995. Foreign direct investment and income inequality: Further evidence. World Development, 23(3): 469–483.

UNCTAD publications on TNCs and FDI 2009. UNCTAD Training Manual on Statistics for FDI and the Operations of TNCs. April 2, 2018.

(36)

VII. Appendix

Figure A1 Scatterplots

Scatter Gini and Value Added

Scatterplot Gini and gross fixed capital formation

10 20 30 40 50 60 G in i C o e ffici e n t 0 1000000 2000000 3000000 4000000 Value Added 10 20 30 40 50 60 G in i C o e ffici e n t 0 20 40 60 80 100

(37)

Scatterplot Gini and Unemployment

Scatterplot Gini and Tradeopenness

Scatterplot Gini and Education

10 20 30 40 50 60 G in i C o e ffici e n t 5 10 15 20 25

Unemployment rate (%total labor force)

10 20 30 40 50 60 G in i C o e ffici e n t 0 50 100 150 Trade Openness 10 20 30 40 50 60 G in i C o e ffici e n t 80 100 120 140 160

(38)

Table A2. Results from Hausman test

Chi-square 20.72

p-value 0.0042

Table A3 Vif test for multicollinearity

Table A4 Test for time-fixed effects

F 4.77

Prob > F 0.000

Table A5 Modified Wald test for groupwise heteroskedasticity in fixed effect regression model

H0: sigma(i)^2 = sigma^2 for all i

Chi-square (134) 6.9e+35

Prob>chi2 0.0000

Variabele VIF 1/VIF

VAlog 91.64 0.011 VAlog Squared 90.32 0.011 Trade 1.46 0.684 Education 1.31 0.766 Unemployment 1.14 0.875 FDI 1.11 0.901

Gross Fixed Capital Formation

1.10 0.906

The Effect of Sector Level FDI on Income Inequality Within OECD Countries