Student name: Xiaole Wang Student number: S1938991
Email address: x.wang.15@student.rug.nl
Institution: Faculty of Economics and Business, University of Groningen Study program: Master in International Economics and Business
Thesis supervisor name: Dr. Dirk Akkermans (e-mail: d.h.m.akkermans@rug.nl) Methodology supervisor name: Prof. Dr. Erik Dietzenbacher (e-mail:
h.w.a.dietzenbacher@rug.nl) University of Groningen
Effects of FDI on income distribution in
host countries.
Content
Abstract 03
1. Introduction 04
1.1 Problem statement 04
1.2 Research question 05
2. Literature review and Hypotheses 07
3. Data and methodology 15
3.1 Data and measurements 15
3.2 Research methodology 20
4. Empirical results 26
5. Conclusion 30
5.1 Concluding remarks 30
5.2 Research limitations and future study implications 31
5.3 Policy implications 32
Acknowledgements 33
Abstract
For decades, income inequality has been a popular topic in the academic field. Researchers studied it from various aspects, for example, Galor and Zeira (1993) studied the relationship between income inequality and economic activities. Besides, many factors that generate the changes of income inequality have been discussed in previous studies. In this research, Foreign Direct Investment (FDI) is the focus among these different factors. Additionally, four other elements are also considered—economic growth rate, country development level, trade openness and tertiary education enrollment. Since the effects of FDI on income inequality vary among countries with different development levels, countries under discussion are divided into two groups: developed countries (DCs) and less developed countries (LDCs)1. Therefore, the emphasis of the study is comparing effects of FDI on income inequality between developed countries and less developed countries. The empirical results show that FDI measured as transnational corporate penetration increases income inequality in both developed and less developed countries, but the situation in less developed countries is more noticeable.
Key words
FDI; Income inequality; Less Developed Countries; Developed Countries
1
1. Introduction 1.1 Problem statement
Income inequality has been investigated in the academic circles persistently since middle of 20th century. As one aspect of the social conditions, income distribution significantly influences the aggregate economic activities both in the short run and in the long run (Galor & Zeira, 1993). Furthermore, as the common denominator of social stratification in the market societies (Bornschier, 2002), income is the most influential factor that reflects people‟s living standard. Therefore, the situation of unequal income distribution has attracted lots of attention. For instance, based on preceding researches, Gottschalk and Smeeding (1997) further elaborated the changes of income inequality in the U.S. They figured out that most of the countries had experienced at least a modest increase in the income inequality between 1980 and 1990, and the growth of income dispersion in the U.S. was almost the largest. Although income inequality cannot completely display the whole picture of the quality of life, it is still capable of revealing how wide the living standard gap between different groups of people is. Therefore, income inequality will be discussed in this paper.
1.2 Research question
Numerous researches were conducted to study the effects of FDI on income inequality in host countries. However, most of these studies put emphasis on the situation of a particular area or country. For example, Feenstra and Hanson (1997) found in data analysis of Mexico from 1975 to 1988 that foreign capital inflow is positively
correlated with wage inequality. Taylor and Driffield (2005) also found that FDI
significantly impacts wage inequality in UK, even after controlling for the two most common explanations of wage inequality—technology and trade. Besides, some researchers mainly explored a certain sector, such as Vijaya and Kaltani (2007). They investigated this issue through manufacturing sector, and found that FDI flows have a negative impact on overall wages in the this sector, which means that FDI can lead to unequally distributed income among different sectors. However, there are also vast numbers of papers discussing the influences across different countries. Choi (2006) studied the relationship between FDI and income inequality with cross-country data and concluded that FDI measured by inward, outward and stock as a percentage of GDP are all proved to widen income gaps. Since the effects of FDI on income inequality are still open to discussion, further cross-country and cross-sector evidence on the impacts will be provided in this study.
Among many different types of income inequality, factor income inequality and household income inequality are two most remarkable types. For example, in Shorrocks‟ (1982) paper, he discussed factor income inequality, which represents how income is disaggregated into various production factor components. However, household income inequality also receives a body of research (Kuznets, 1955; Edwards, 1997). Since this research emphatically discusses the effects on people‟s life, household income inequality in host countries will be discussed.
country groups with different development levels, developed countries and less developed countries are most commonly studied. However, very few researches are about comparing effects of FDI on income inequality between developed countries and less developed countries. Therefore, this paper will investigate the relative differences in the effects of FDI on income inequality in these two groups of countries, and the study will be carried out through a period of time—8 years. Does FDI really affect household income inequality in host countries? Will FDI make income inequality in one country severer or not? Are there any differences of these inside-country effects between developed and less developed countries?
2. Literature review and Hypotheses
Income inequality has been looked into overtime across countries. In Bornschier‟s (2002) paper, the changing conditions of income inequality in many countries were listed. For example, the inequality in U.S. had decreased because of New Deal of 1930s, while it has started increasing considerably since the end of 1960s; For India a decrease of inequality was observed from the early 1950s to 1970 and an increase thereafter. Kuznets (1955) concluded from data of the U.S., England and Germany that the relative distribution of income, as measured by annual income incidence in rather broad classes, has been moving toward equality since 1920s. Heshmati (2006) reviewed papers on income inequality and discussed several investigation approaches. In the research, he studied two kinds of income inequality—individual income inequality (estimating income distribution across individuals and characterizing its dynamics through time) and international income inequality (distribution of income between countries). Since this research is trying to find out how household income inequality changes through time periods, individual income inequality in different countries will be discussed.
Usually, GDP is used to explain how economic development level of a country can affect income inequality. However, since the purpose of this paper is to compare developed countries (DCs) with less developed countries (LDCs), Human Development Index (HDI) from UN is applied to distinguish them: high HDI represents developed countries, while low HDI holds for less developed countries. HDI is a statistical measure that implies a country‟s level of human development, which is also strongly correlated with the prosperity level of an economy. The concrete lists of DCs and LDCs used in this research are from those of IMF advanced economies and emerging & developing economies, respectively, and both of them use HDI as the dividing criteria.
and DCs have been investigated by many scholars. Bornschier (2002) declared in his work that income inequality in core countries is on average considered lower than that in developing countries, but the differences in the extent of inequality are more pronounced for developing countries. Chan (1989) analyzed Kuznets curve in his paper, and pointed out that mature industrialized economies above the threshold experience increasingly lower income inequality, while the intermediate level countries are with the highest income inequality. Modernization or developmental theories of income distribution predict that developing nations will exhibit higher levels of income inequality relative to both non-industrial and industrialized countries (Beer, 1999).
Therefore, in this study, countries under discussion will be divided into two different groups—developed countries and less developed countries, and a dummy variable—LDCs is added to the model (country is less developed, when dummy value=1; country is developed, when dummy value=0).
would not otherwise have them, the income inequality would be decreased. Since regardless of LDCs or DCs, only international and related industries can directly absorb the spillovers of FDI, and the introduction of FDI further increases employees‟ incomes in these industries, in which the income was already at a high level before FDI comes in. In this sense, FDI increases income inequality in host countries with various development levels.
However, under the circumstances of FDI, the income gaps get even larger in LDCs than they do in DCs. The result of Evans and Timberlake‟s (1980) paper suggests that less developed countries which were hosts to large amounts of foreign investment were likely to experience high levels of income inequality. Furthermore, Blonigen and Wang (2004) examined whether the experiences with FDI systematically vary between LDCs and DCs. What they concluded from the result is that differences do exist. As suggested in their paper, “the coefficient terms on these variables give the incremental difference in the variable‟s effect on FDI when the observation is connected with an LDC rather than a DC”. Therefore, in order to present these differences in the model, an interaction term of FDI and LDCs/DCs is also added in this research, as in Bolnigen and Wang (2004). More elaboration of this interacted effect will be made below.
In developed countries, FDI enlarges income inequality, since FDI causes high-skilled labors to earn more and low-skilled labors to earn less. Real wages of low-skilled labors tend to decrease, so that these labors can compete for industrial production with less developed countries, where low-skilled labors are low paid (Bornschier, 2002). In UK, the relative wages shift away from unskilled workers to their counterparts because FDI causes a high demand for the high skilled labors (Taylor & Driffield, 2005). Therefore, the difference in income between high-skilled and low-skilled labors is further remarkable.
inequality in less developed countries. The evidence in Evans and Timberlake‟s (1980) paper suggests that less developed countries which were hosts to large amounts of foreign investment were likely to experience high levels of inequality. Study from
Gissinger and Gleditsch (1999) reveals the increasing income inequality in LDCs. They quoted dependency theory from Galtung (1971) that the penetration of foreign capital into peripheral economies leads to the exploitation of local human and natural resources and a transfer of profit back to the imperial centers. As industrialization proceeds with the presence of FDI, it is widely observed in host LDCs that employees in the international sectors tend to earn much more than normal wages, while unemployment rate increases in the traditional sectors. Evans and Timberlake (1980) pointed out that foreign investment causes high levels of inequality by distorting the evolution of the labor-force structure in LDCs. Besides, Feenstra and Hanson (1997) also found in their research of Mexico that FDI led skilled workers to earn more than
before, comparing to unskilled workers, and this increased income inequality as well.
Therefore, both in developed and less developed countries, FDI increases income inequality. Yet, this trend is more apparent in less developed countries. Vijaya and Kaltani (2007) figured out in their study that increased capital mobility caused by FDI decreases the bargaining power of labor and can therefore have a negative impact on the wages of manufacturing sector. Since this sector contains many low-skilled labors whose income level is relatively lower, income inequality will increase under the condition described above. Compared with developed countries, FDI in less developed countries are more likely to flow into manufacturing sector. Therefore, income inequality increases more in less developed countries under the circumstances of FDI.
In addition to FDI and country development levels, there are many other factors affecting income inequality in host countries. In this research, economic, policy and social elements will be added to the model. As for economic factors, GDP growth rate are commonly used in many papers (Ahluwalia, 1976; Beer & Boswell, 2002). Since GDP growth rate is the most important indicator of economic health (Amadeo, 2009), it is applied in the model to represent economic growth rate, which offers an insight to how fast the overall economy grows, and GDP growth rate is supposed to decrease income inequality. If GDP is growing, business, job opportunities and personal incomewill also increase; while if it is slowing down, business will hold off investing and income level will be affected (Amadeo, 2009). Beer (1999) summarized from former theorists that continuous economic growth expands the middle class and increases employment and saving rates among the poor, which leads to decreasing income disparity. Besides, in Tsai‟s (1995) paper, he used short-term economic growth in the model to study the effects of FDI on income inequality, and he also discovered from the data analysis that there might be a negative correlation between growth rate and income inequality. To sum up, growing economic growth rate may decrease income inequality in host countries, regardless of developed countries or less developed countries.
H2: GDP growth rate negatively affects income inequality in both developed countries and less developed countries. (β3<0)
Then referring to economic policy, trade openness has increasingly affected the performances of domestic economic activities. Edwards‟ (1997) paper studied the links between openness or trade liberalization and income inequality in many countries, and he found no evidence to prove these links. However, Reuveny and Li (2003) confirmed in their research that trade reduces income inequality, and
Richardson (1995) concluded from the research that trade is mainly responsible for
growing income inequality and even declining median wages. This conflict occurs
Samuelson‟s factor abundant model, Jensen and Rosas (2007) suggested that in developed countries, a liberalization of trade should increase income inequality by increasing the returns of capital and decreasing that of labor; conversely, in developing countries, trade liberalization should decrease income inequality by
increasing the returns to labor and decreasing the returns to capital. Therefore, an
interaction term concerning the interactive effect of country development level and trade openness is added to the model, and further evidence on this relationship in LDCs and DCs will be demonstrated in this paper.
Most papers discuss this issue on the basis of Heckscher-Ohlin model of international
trade (Reuveny & Li, 2003; Spilimbergo, Londoño & Székely, 1999; Jensen & Rosas,
2007). Spilimbergo, Londoño and Székely (1999) studied the empirical links among
factor endowments, trade and personal income distribution in their works and the
central idea was that how trade openness canaffect income inequality depends on the
country‟s relative factor endowments. Besides, Reuveny and Li (2003) reviewed the prediction of Stolper and Samuelson (1941) that trade raises income inequality in the
DCs and reduces that in the LDCs. This is because DCs are relatively well endowed
with skilled labor and capital, and their imports are expected to hurt their unskilled labor, whereas their exports should benefit their capital owners and skilled labors. In contrast, LDCs are relatively endowed with unskilled labor, and the situation goes to the opposite side.
H3: the opener a country’s trade is, the higher the income inequality becomes in developed countries; in contrast, trade openness decreases income inequality in less developed countries. (β4>0, β7<0)
inequality among high-income countries are heavily influenced by the rewards for educational attainment (Sullivan & Smeeding, 1997)”. Gregorio and Lee‟s (2002) findings indicated that the education factors - higher attainment and more equal distribution of education - play an essential role in making income distribution more equal. O‟Neill (1995) showed in his paper that the relationship between education and income can help demonstrate why less developed countries perform poorly relative to developed countries. Therefore, education level is influential to income distribution.
Yet different levels of education lead to different effects on income distribution. Tilak (1989) confirmed in his investigation that education decreases income inequality, and the education level which contributes to this effect could change from primary to secondary education. In O‟Neill‟s (1995) paper, he utilized gross secondary school enrollment to measure a country‟s level of human capital, and discovered that education decreases income inequality. However, in this paper, the effect of tertiary education on income distribution is discussed. it is because that people who complete high school education and enroll in college are apparently distinguished as higher educated, and “young people without a college or a complete high school education are not being adequately prepared for work” (Becker, 1993). Besides, for the reason that graduation rate can be affected by many unespective factors, such as some students may drop out from school for personal reasons, it is better to use enrollment rate which can present one‟s willingness of achieving higher education as well as ability of learning and working. Since tertiary education is selected to present how education can affect income inequality, the tertiary education enrollment is used in this research.
returns to tertiary education raised skill price after 1993. Since wage of skilled-labor was already higher than unskilled-labor before this changing, it indicates that higher tertiary education enrollment will further increase income inequality. Therefore, it is assumed that tertiary education will increase income inequality both in LDCs and DCs.
H4: The higher enrollment rate of tertiary education is, the larger income gaps will be in both less developed countries and developed countries. (β5>0)
Therefore the economic theoretical model in this paper is constructed as follows:
Income Inequality = β0+β1FDI+β2LDC+β3GDPgrowth rate+ β4trade openness +β5
lag2_tertiary education enrollment +β6LDC*FDI+ β7LDC*trade openness
3. Data and methodology 3.1 Data and measurements
The main dataset applied in the paper is from World Development Indicators & Global Development Finance of World Bank and Inequality Project in University of Texas. The specific information about data is described below and displayed as table1.
Table 1: description of variables
Indicator Variable Measurement Source
Gini Income Inequality Gini index WDI&GDF(Poverty); Inequality Project,
University of Texas
Transnational Corporate Penetration (TCP)
FDI FDI, net/ GDP(all in current US$) WDI&GDF
(Economic Policy & Debt)
LDC Dummy
(0=DC,1=Less DC)
Human Development Index (HDI) IMF
GDPGR GDP Growth Rate GDP growth (annual %) WDI&GDF (Economic Policy & Debt)
Trade Trade openness Imports + Export of goods and services / GDP (all in constant 2000 US$)
WDI&GDF (Economic Policy & Debt);
Ter-Edu Tertiary Education School enrollment, tertiary (% gross)
WDI&GDF (Education)
Though FDI has started from early 20th century, its influence was broadly expanded around the world in 1980s. Therefore, 1990s is appropriate to study the spreading global effects of FDI, and data from 1992 to 1999 is collected on the bases of data availability. Aiming at decreasing research bias and because of data limitation, all countries around the world with sufficient data during 1992 and 1999 are chosen, namely 41 countries, including 20 developed countries and 21 less developed countries. The country list is exhibited below.
Developed Country
Australia, Austria, Cyprus, Denmark, Finland, France, Greece, Ireland, Israel, Italy, Japan, Malta, Netherlands, New Zealand, Norway, Slovenia, Spain, Sweden, United Kingdom, United States
Less Developed Country
As demonstrated before, Human Development Index from UN is applied to decide a country‟s development level. Advanced economies and emerging & developing economies from IMF are utilized to represent DCs and LDCs, respectively. Since the lists of DCs and LDCs are changing over time, the lists of IMF in 2011 are used as criteria to separate the countries and to study how FDI impacts their income inequality in 1990s.
According to the theoretical model displayed above, an econometric model is constructed:
Gini = β0+β1TCP+β2LDC+β3GDPGR +β4Trade +β5 lag2_Ter-Edu +β6LDC_TCP+ β7LDC_trade
+e
In World Bank Poverty analysis, many measurements of income inequality are listed: Gini-coefficient of inequality, Theil-index, Decile dispersion ratio and Share of income/consumption of the poorest x%. Beer and Boswell (2002) declared in their paper that the two most commonly used measures in comparative studies of income inequality are the Gini coefficient and the concentration of income received by the top 20% of the population. The Gini index from WDI measures the extent to which the distribution of income (or, in some cases, consumption expenditure) among individuals or households within an economy deviates from a perfectly equal distribution. A Gini index of zero represents perfect equality, while an index of 100 implies perfect inequality (Choi, 2006).
Though Gini coefficient is one of many measures that are commonly used to describe
Squire, 1996). However, this problem can be solved if income share held by the lowest/highest 20% is added, since these measurements concentrate on people standing at two poles, and explain how income changes among different income groups.
Though using both measurements can better our research, Gini index is finally used for two reasons. First of all, the concrete changing trend of income among groups is not the focus, and Gini index is sufficient to study how FDI influences income inequality. Secondly, data for income inequality is limited, since the income share held by the lowest/highest 20% is unavailable in most of the countries.
Then, the independent variable—FDI is measured by “Transnational corporate penetration”, which is the ratio of “Foreign direct investment, net (BoP, current US$)” to “GDP (current US$)” from WDI. FDI data are usually reported in terms of stocks and flows. FDI stock refers to the value of capital and reserves plus net indebtedness, whereas FDI flow refers to capital provided by or received from a foreign direct investor to an FDI enterprise (UNCTAD, 2006). In many researches on FDI, several indicators have been used. In Tsai‟s (1995) paper, he used the stock of FDI as a proportion of GDP to explain the significance of FDI to host countries. Vijaya and Kaltani (2007) introduced FDI variable as both a stock and a flow variable in their paper, since they can better represent capital mobility across countries.
who work in transnational corporation and in normal firms go wider (London & Robinson, 1989). Later, Beer (1999) further studied this issue by using new data in mid-1980s and confirmed this theory and revealed that “foreign economic penetration is a significant and robust predictor of the concentration of income in the upper portion of the population”, which indicates higher income inequality. Beer and Boswell (2002) pointed out studies of „capital dependency‟ or TNC „penetration‟ contend that disproportionate control over host economies by transnational corporations increases inequality by altering the development patterns of these nations.
Concerning transnational corporate penetration, all the measurements are based on Bornschier and Chase-Dunn (1985), which is the ratio of foreign-owned capital stock to total capital and labor. Boswell and Dixon (1990) recommended this measure in their paper because “it directly measures the internal influence of foreign capital and it taps the extent of dependency among increasingly industrialized semi-peripheral countries”. In Beer‟s (1999) paper, transnational corporate penetration is measured as the ratio between FDI inward stock (from the world Investment Report 1996) and market GDP (from the World Tables 1995) in 1985. Furthermore, Beer and Boswell (2002) applied the ratio between inward foreign direct investment (FDI) stock and GDP to measure transnational corporate penetration. Since FDI stock data is difficult to collect, the ratio of “Foreign direct investment, net (BoP, current US$)” to “GDP (current US$)” is applied to indicate transnational corporate penetration.
Besides, there are three control variables—GDP growth rate, trade openness and tertiary education enrollment. GDP growth is indicated by “GDP growth (annual %)”, which displayed how GDP grows annually.
import penetration ratios (MGDP) and exports shares in GDP (XGDP) to measure the openness of a country‟s trade. Besides, in Reuveny and Li‟s (2003) paper about the relationship of trade openness and income inequality, they also measured trade openness as the sum of the total import and export values as a share of a country‟s GDP. Considering problem of comparability and tremendously various inflation rates in DCs and LDCs, imports of services and goods, exports of services and goods and GDP are measured in constant 2000US$. Besides, it is noticed that GDP is not measured by the same criteria in the whole model. However, this is not a problem since the percentage data is used for the indicators.
At last, tertiary education is represented by “School enrollment, tertiary (% gross)”. O‟Neill (1995) also applied school enrollment ratios to stand for human capital which is created by education and introduced a time lag on this measurement. He pointed out that under the hypothesis that changes in education cause changes in income, one would expect variation in education levels to precede changes in output. Kijima (2006) applied tertiary education enrollment to investigate the correlation between high education and skill price. As discussed above as well as in theoretical part, “School enrollment, tertiary (% gross)” is used in this paper to reveal how higher education in one country will make changes for income distribution.
3.2 Research methodology
In this sub-section, how methodology works out in this paper will be elaborated. Since different countries in a series of time are under discussion, panel data estimation is applied. Firstly, in order to examine the goodness of data, normality tests will be done. Secondly, multicollinearity is checked to find out if variables are collinear. Besides, for the purpose to examine whether multiple regression estimation can be used in this model, autocorrelation is tested, hence, heteroskedasticity is checked.
At first, the normality of dependent variable—Gini and error term are tested. In Statistics, “normality tests are used to determine whether a data set is well-modeled by a normal distribution or not, or to compute how likely an underlying random variable is to be normally distributed” (Henry & Thode, 2002). In this analysis, a histogram is firstly drawn. The histogram of residual normality check is displayed as graph 1 below. It is presented in bell shapes and indicates that residuals may be normally distributed.
Graph 1: Histogram of residuals
autocorrelation and heteroskedasticity may exist in this research, and further tests on them will be elaborated later.
After the test of residual normality, a Q-Q plot is drawn to check dependent variable‟s normality, and as shown in Diagram 1 that dependent variable is normally distributed.
Diagram 1: Normal Q-Q plot of Gini
Following the normality tests, the collinearity of this model will be examined. It is important, since “the interpretation and conclusions based on the size of the regression coefficients, their standard errors, or the associated t-tests may be misleading because of the potentially confounding effects of collinearity” (Mason & Perreault, 1991). Thus, the correlation matrix is constructed below as table 3. It is clear from the table that variables in the model are not highly correlated. This implies that collinearity is not a problem in the model.
Table 3: Correlation matrix
Gini GDPgrowth rate TCP Trade Ter-Edu Less-DC Less-DC_TCP Less-DC_trade
Gini 1
GDPgrowth rate 0.092 1
TCP 0.222** 0.107 1
Trade -0.082 0.028 0.256** 1
Less-DC 0.456** -0.154** 0.276** 0.120* -0.524** 1
Less-DC_TCP 0.285** 0.086 0.884** 0.135* -0.217** 0.321** 1
Less-DC_trade 0.261** -0.133* 0.292** 0.547** -0.248** 0.650** 0.391** 1
**.Correlation is significant at the 0.01 level (2-tailed). *.Correlation is significant at the 0.05 level (2-tailed).
Then, autocorrelation test and heteroskedasticity check will be carried out, since the normality problem exists in the residuals of the model. Besides, for the sake of a better analysis, an autocorrelation test is necessary to check if current effects are influenced by former shocks. Normally, there are two different kinds of autocorrelation: spatial autocorrelation and serial autocorrelation. Hill, Griffiths and Lim (2008) defined in their book that when errors in different time periods are correlated, autocorrelation exists. They gave one example that a shock may take several periods to work through the system, so error in one period contains not only the effects of current shock, but also the carryovers from previous ones. Above are the explanations of serial autocorrelation. As declared by Cliff and Ord (1970), spatial autocorrelation exists when samples are correlated with each other, for example, situation in country 1 is connected with that in country 2. As for how to detect autocorrelation, Hill, Griffiths and Lim (2008) suggested three in their book: residual correlogram, Lagrange multiplier test and Durbin-Watson test. However, they are not quite suitable for panel data analysis. Wooldridge (2002) proposed a new test for panel data autocorrelation, and the command “xtserial depvar indepvars” written by David Drukker (http://www.stata.com/support/faqs/stat/panel.html) can be applied in STATA to carry out this test. Finally, as a result of this test, autocorrelation is detected, and this problem can be managed later with the command “xtregar” when doing panel data analysis.
for all observations are not the same. Besides, three ways are proposed in the book to detect heteroskedasticity: residual plots, Goldfeld-Quandt test and variance function test. In this research, variance function test, which is often referred to as Preusch-Pagan test (LM test), is utilized; since the reason for different variances in our sample is clear to be the countries‟ different development levels. After the Breusch-Pagan test (LM test), the result is significant with calculated statistic chi2(1) being 289.1. Therefore, heteroskedasticity is also detected from the model. Therefore, in order to solve this problem, the robust test is expected to be done, and command “robust” is used in STATA.
Thereafter, panel data analysis is to be done. A panel of data consists of a group of cross-sectional units (people, households, firms, states and countries) who are observed over time (Hill, Griffiths and Lim, 2008). In Beer and Boswell‟s (2002) paper, they used panel data of 65 countries and two time periods, and illustrated the usefulness of panel models in understanding the dynamics of income distribution. They also pointed out that while cross-sectional analyses may reveal long-term structural tendencies, panel analyses are essential for revealing causal forces affecting changes in income inequality. In a word, in order to inspect the dynamics of income inequality among countries and through time periods, panel data is applied, and the procedure of panel data analysis is described below.
variables are difficult to measure, and mistaken measure results to biased estimation. In contrast, the major attraction of fixed effects methods in nonexperimental research is the ability to control for all stable characteristics of the individuals in the study, thereby eliminating potentially large sources of bias (Allison, 2005). Besides, Hill, Griffiths and Lim (2008) stated in their book that the fixed effects model can be applied to panel data with any number of individual, cross-sectional observations. Since all countries with sufficient data are used in order to expand the significance of the research. Therefore, theoretically, fixed-effects model is more likely to satisfy the needs of this study. However, for the purpose of bettering the research and avoiding the mistakes, further investigations will be done later to further test which model is more suitable.
Thus, Hausman test is applied to help with model selection. This test is the standard procedure used in empirical panel data analysis in order to discriminate between the fixed effects and random effects model (O‟Brien, Patacchini, 2003). Hausman test compares the coefficient estimates of fixed effects with that of random effects: if both estimators are consistent, they are similar in large samples; if fixed effects estimator is still consistent, while random effects estimator is inconsistent which means that its additional random individual difference is correlated with any explanatory variables, then the converged value in random effects model is no longer true. Therefore, after the test, if the null hypothesis is rejected and the random error is correlated with any of the explanatory variables, then it is better to use fixed effects model (Hill, Griffiths and Lim, 2008).
4. Empirical results
First of all, before the analysis, a summary of variables is displayed below. Table 4: variable summary
Variable Obs Mean Std.Dev. Min Max
Country 410 21 11.84662 1 41 Year 410 1994.5 2.875791 1990 1999 Gini 318 39.49324 6.113485 24.18 61.76 GDPgrowth 328 2.250807 6.182171 -41.8 18.66484 TCP 319 .0192768 .046892 -.07532 .661166 Trade 324 .7822764 .4785862 .1469553 2.511332 TerEdu 368 34.76129 17.3044 .4193568 82.39955 LDC 410 .5121951 .5004619 0 1 LDC_TCP 319 .0161194 .0426665 0 .661166 LDC_trade 324 .4318249 .5420016 0 2.099145 TerEdu2yearsago 366 34.67018 17.22961 .4193568 82.39955
The results are revealed as table 4.
Table 4: empirical results
Variable (1) fix effects (robust) (2) LDCdum=0 (robust) (3) LDCdum=1 (robust) TCP 7.678 (6.057) 10.487* (5.866) 19.511* (9.29) LDC -- -- -- GDPgrowth 0.0426 (0.0582) 0.0559 (0.0661) 0.0453 (0.0633) Trade 2.979* (1.605) 3.671** (1.53) -1.187 (2.575) lag2_Ter-Edu 0.0689*** (0.0206) 0.0345** (0.0172) 0.112** (0.0438) LDC_TCP 17.097 (10.974) -- -- LDC_trade -3.7703 (2.584) -- -- Cons 36.14*** (1.285) 32.276*** (1.1404) 38.858*** (2.216) No. of obs. 271 129 142
In the table above, results of models for (1) fixed effects, (2) dummy=0 and (3) dummy=1 are represented. The whole model is significant at 1% significance level. However, when scanning the coefficients of variables, only a few of them are significant, which may be resulted from data bias in this research.
In model 1, the combining effects of the interaction terms indicate how the situation will change in different groups of countries. Generally, most of the estimated results are as expected in the hypotheses. First and foremost, Hypothesis 1 is not statically proved in our model, though the coefficients of TCP and LDC_TCP are positive as anticipated. The coefficient of TCP reveals that transnational corporate penetration (TCP) possibly increase income inequality in both DCs and LDCs. Besides, the coefficient of TCP_LDC is also above zero, which suggests that TCP can increase income inequality more in LDCs than in DCs. Therefore, there is a chance that transnational corporate penetration (TCP) can positively affect Gini index, though it is not significant in this model.
Secondly, Hypothesis 2 is not supported in model (1). As shown from the results, the coefficient is suggested to be positive though insignificant, which means that GDP growth rate may increase income inequality in both DCs and Less DCs in this study. Since few previous researches had the exact empirical data results to show economic growth rate does decrease income inequality and most of them are just theoretical analysis, our data analysis may suggest the new idea that higher GDP growth rate may increase income inequality.
Last but not least, the outcome for tertiary education enrollment is as expected in model (1). The coefficient is positive and highly significant at 5% significance level. Therefore, Hypothesis 4 is strongly supported by the data analysis. the result represents that the more people enroll in tertiary education, two years later, the worse income inequality gets and the wider income gaps becomes in both developed and less developed countries.
In model (2), when the value of the dummy variable equals to zero which stands for developed countries (DCs), most of the coefficients are significant. The coefficient of TCP is positive and significant at 10% significance level. This suggests that deeper transnational corporate penetration leads to higher income inequality in developed countries. Then, the result for GDP growth rate is also negative and insignificant as model (1). Thereafter, trade openness is proved to significantly (at 5% significance level) increase income inequality in developed countries. At last, the outcome for tertiary education enrollment indicates that the positive effect of education on increasing income inequality is significant (at 5% significance level) in developed countries.
Besides, in model (3) when Less-DC==1 (representing less developed countries), some of the results are the same with model developed countries while others are not. Firstly, deeper transnational corporate penetration is also statistically proved (coefficient is significant at 10% significance level) to increase income inequality in less developed countries. Therefore, by combining the results for TCP in model (2) and (3), hypothesis 1 is partially proved. Secondly, higher enrollment of tertiary education also widens income gaps in less developed countries. However, the coefficient of trade openness in model (3) is negative. Though it is not significant, it can also present the possibility that trade openness decreases income inequality in less developed countries.
significant, while other assumptions still require more sufficient data to support. Firstly, concluding from model (2) and (3), income gaps get wider with deeper transnational corporate penetration in both developed and less developed countries. This indicates that FDI is not a helpful way to narrow down the income gaps. Secondly, GDP growth rate may also increase income inequality both in developed and less developed countries. However, this still needs more complete data to testify. Thirdly, tertiary education enrollment significantly increases income inequality both in developed and less developed countries. For this research, the major reason leading to insignificant outcome is the data bias. Data applied in this paper are from 1992-1999 in 21 Less DCs and 20 DCs, and the selection is decided by data sufficiency. This indicates that the research outcome can be biased. Furthermore, even in these selected countries, data is still limited and incomplete, and the most deficient data are for transnational corporate penetration and Gini.
5. Conclusion
5.1 Concluding remarks
To sum up, differences do occur to income inequality in countries of different development levels, and income gaps are wider in less developed countries than in developed countries. Implied by our data analysis, the transnational corporate penetration is possible to increase income inequality in host countries, regardless of developed or less developed countries. However the evidence is lack in our data analysis. Furthermore, trade openness widens income gaps in both developed and less developed countries, but and the gaps maybe narrower in less developed countries. As expected, tertiary education enrollment is significantly correlated with income inequality both in developed and less developed countries. Higher tertiary education enrollment rate leads to higher income inequality two years after the enrollment. During this estimation, only the assumption for GDP growth rate is out of expectation. As assumed in theoretical model, GDP growth rate should be negatively correlated with income inequality; while data analysis finally disapproved the assumption. In this research, GDP growth rate is demonstrated to increase income inequality both in developed and less developed countries.
5.2 Research limitations and future study implications
Though many efforts have been put in this research, there still exist some limitations. At first, in this study, countries are selected according to data availability, so countries included are only those with sufficient data, which indicates the disadvantage of our dataset. If better data can be found in future, much more evidence will support the assumptions. Secondly, countries under discussion are in different areas and the effects of geography are out of consideration in our research. Therefore, in future studies, on the basis of better data, how geographical factors influence the effect of FDI on income inequality can be deeply analyzed and countries in different areas can be investigated separately.
5.3 Policy implications
Though there are still some limitations in our research, it can provide applicable implications for policy makers. After data analysis, we concluded from the result that FDI may have positive effects on income inequality in both Less DCs and DCs. This reminds the policy makers that FDI may have positive effects on economic development; however, for labors it may be not as significant as that for macroeconomic environment. If governments want to improve income level of labors and decrease income inequality, they should do more in addition to just introducing FDI. For example, governments can pay more attention to manufacturers who are unable to attract foreign capitals, and make some policies or subsidy which can encourage their developments.
Acknowledgements
It is a great pleasure to express my gratitude to all those who gave me the possibility to complete this thesis. First and foremost, I would like to extend my sincere gratitude to my thesis supervisor—Dr. Dirk Akkermans, for his instructive advice and useful suggestions on my thesis. He provides me not only the academic knowledge, but also the inspiration of innovation. Throughout the whole writing process, I have benefited so much from his profound knowledge and patient guidance. Then, I also want to show great thanks to my methodology supervisor—Prof. Dr. Erik Dietzenbacher. His comments and advices on methodology make my analysis more reasonable and logical, and also further benefit my whole paper. Last but not least, I would like to thank my family and friends. They have shown great supports on my thesis writing, and encouraged me whenever I need.
At last, I would like to show my great gratitude again to people who help me during thesis writing and also the whole master program year.
References
Ahluwalia, M. 1976. Income distribution and development: some stylized facts. The
American Economic Review, 66(2): 128-135.
Allison, 2005. Fixed Effects Regression Methods for Longitudinal Data using
SAS.
Amadeo, K. 2009. What is the difference between GDP and growth rate? http://useconomy.about.com/od/grossdomesticproduct/f/GDP_Growth_Rate.htm Beer, L. 1999. Income inequality and transnational corporate penetration. Journal of
World-Systems Research, 5(1): 1-25.
Beer, L. & Boswell, T. 2002. The resilience of dependency effects in explaining income inequality in the global economy: a cross-national analysis, 1975-1995.
Journal of World-Systems Research, 8(1): 30-59.
Becker, G. S. 1993. Encyclopedia essay on human capital. The Concise
Encyclopedia of Economics.
Blonigen, B. & Wang, M. 2004. Inappropriate Pooling of Wealthy and Poor
Countries in Empirical FDI Studies. Working paper from National Bureau of
Economic Research.
Bornschier, V. 2002. Changing income inequality in the second half of the 20th century: preliminary findings and propositions for explanations. Journal of
World-Systems Research, 8(1): 100-127.
Bornschier, V. & Chase-Dunn, C. 1985. Transnational Corporations and
Underdevelopment.
Borensztein, E., Gregorio, J. & Lee, J. 1998. How does foreign direct investment affect economic growth? Journal of International Economics, 45: 115-135.
Boswell, T. & Dixon, W. 1990. Dependency and rebellion: a cross-national analysis.
American Sociological Review, 55(4): 540-559.
Choi, C. 2006. Does foreign direct investment affect domestic income inequality?
Applied Economics Letters, 13: 811-814.
Cliff, A. & Ord, K. 1970. Spatial autocorrelation: a review of existing and new measures with applications. Economic Geography, 46: 269-292.
Deininger, K. & Squire, L. 1996. A new data set measuring income inequality. The
World Bank Economic Review, 10(3): 565-569.
Edwards, S. 1997. Trade policy, growth, and income distribution. The American
Economic Review, 87(2):205-210.
Evans, P. & Timberlake, M. 1980. Dependence, inequality, and the growth of the tertiary: a comparatieve analysis of less developed countries. American Sociological
Review, 45(4): 531-552.
Feenstra, R. & Hanson, G. 1997. Foreign direct investment and relative wages: evidence from Mexico‟s maquiladoras. Journal of International Economics, 42: 371-393.
Fields, G. & Yoo, G. 2000. Falling labor income inequality in Korea‟s economic growth: patterns and underlying causes. Review of Income and wealth, 46(2).
Galor, O. & Zeira, J. 1993. Income distribution and macroeconomics. Review of
Economic Studies, 60: 35-52.
Galtung, J. 1971. A structural theory of imperialism. Journal of Peace Research, 8(2): 81-117.
Gissinger, R. & Gleditsch, N. 1999. Globalization and conflict: welfare, distribution, and political unrest. Journal of World-Systems Research, 5(2): 327-365.
Gregorio, J. & Lee, J. 2002. Education and income inequality: new evidence from cross-country data. Review of Income and Wealth, 48: 395-416.
Gottschalk, P. & Smeeding, T. 1997. Cross-national comparison of earnings and income inequality. Journal of Economic Literature, 35: 633-687.
Heshmati, A. 2006. The world distribution of income and income inequality: a review of the economics. Journal of World-Systems Research, 12(1): 61-107.
Henry, C. & Thode, Jr. 2002. Testing for Normality. Published by Marcel Dekker, New York, 479.
Jensen, N. & Rosas, G. 2007. Foreign direct investment and income inequality in Mexico, 1990-2000. International Organization, 61: 467-487.
Kuznets, S. 1955. Economic growth and income inequality. The American Economic Review, 65(1).
Kijima, Y. 2006. Why did wage inequality increase? Evidence from urban India 1983-99. Journal of Development Economics, 81(1): 97-117.
London, B. & Robinson, T. 1989. The effect of international dependence on income inequality and political violence. American Sociological Review, 54(2): 305-308. Mason, C. & Perreault, W. Jr. 1991. Collinearity, power, and interpretation of multiple regression analysis. Journal of Marketing Research, 28(3): 268-280.
Nielsen, F. & Alderson, A. 1995. Income inequality, development, and dualism: results from an unbalanced cross-national panel. American Sociological Review, 60: 674-701.
O‟Neill, D. 1995. Education and income growth: implications for cross-country inequality. The Journal of Political Economy, 103(6): 1289-1301.
O‟Brien, R. & Patacchini, E. 2003. Testing the Exogeneity Assumption in Panel
Data Models with “Non Classical” Disturbances. Discussion paper in Economics
and Econometrics, University of Southampton.
Reuveny, R. & Li, Q. 2003. Economic openness, democracy, and income inequality—an empirical analysis. Comparative Political Studies, 36(5): 575-601. Richardson, J. 1995. Income inequality and trade: how to think, what to conclude.
The Journal of Economic Perspectives, 9(3): 33-55.
Shorrocks, A. 1982. Inequality decomposition by factor components. Econometrica, 50(1).
Spilimbergo, A., Londoño, J. & Székely, M. 1999. Income distribution, factor endowments, and trade openness. Journal of Development Economics, 59: 77-101. Stolper, F. & Samuelson, P. 1941. Protection and real wages. Review of Economic
Studies, 9: 58-73.
Sylwester, K. 2005. Foreign direct investment, growth and income inequality in less developed countries. International Review of Applied Economics, 19(3): 289-300. Taylor, K. & Driffield, N. 2005. Wage inequality and the role of multinational: evidence from UK panel data. Lobour Economics, 12: 223-249.
Tilak, j. 1989. Education and its relation to economic growth, poverty, and income distribution—past evidence and further analysis. World Bank Comparative Studies, research paper No.3.
Tsai, P. 1995. Foreign direct investment and income inequality: further evidence.
World Development, 23(3): 469-483.
Vijaya, B. & Kaltani, L. 2007. Foreign direct investment and wages: a bargaining power approach. Journal of World-Systems Research, 13(1): 83-95.
Velde, D. 2003. Foreign direct investment and income inequality in Latin
America (experience and policy implications). Working paper from Overseas
Development Institute.
Wooldridge, J. 2002. Econometric analysis of cross section and panel data. Massachusetts Institute of Technology, U.S.
UNCTAD (United Nations Conference on Trade and Development), 2006.
Yanikkaya, H. 2003. Trade openness and economic growth: a cross-country empirical investigation. Journal of Development Economics, 72: 57-89.
