What is the effect of trade openness on
the income inequality in developing
countries compared to OECD countries?
Idse Kuipers
12 ec
10422358
Macro Economics
The effect of trade openness on the income distribution.
Abstract
Income inequality has been growing over the last decades. This thesis analyses the possible
relationship between trade openness and income inequality and compares OECD countries
with developing countries. Classical economic theory predicts a significant negative
relationship between trade openness and income inequality, but more recent researchers
are finding other factors to be more important for determining income inequality. In this
research, I use an unbalanced panel dataset of 59 countries. A fixed effects regression is
used to estimate the model parameters. The results show a positive effect of trade in OECD
countries. But no significant effect in developing countries. Technological advancement
seems to be of more importance.
This document is written by student Idse Kuipers who declares to take full responsibility
for the contents of this document.
I declare that the text and the work presented in this document are original and that no
sources other than those mentioned in the text and its references have been used in
creating it.
Table of contents
INTRODUCTION ... 4
LITERATURE REVIEW ... 6
M
ANY RESEARCH
,
ON DIFFERENT VARIABLES
. ... 6
METHODOLOGY ... 10
T
HE VARIABLES
... 10
T
HE DATA
... 15
T
HE REGRESSION MODEL
... 16
T
HE HYPOTHESES
... 18
EMPIRICAL ANALYSES ... 20
D
ESCRIPTIVE STATISTICS
... 20
C
ORRELATION STATISTICS
... 21
C
ONCLUSION AND PRACTICE
,
REJECT OR ACCEPT THE HYPOTHESES
... 27
CONCLUSION ... 27
REFERENCE LIST ... 29
APPENDIX ... 32
All country ... 32
OECD countries ... 41
Developing Countries ... 50
Introduction
“The distribution of wealth is too important an issue to be left to economists, sociologists, historians,
and philosophers. It is of interest to everyone” (Piketty, 2014, p.2). The 8 richest people dispose of
the same amount of money as the poorest half of the whole world (3.500.000.000 people) (Novib,
2017,. Income inequality is growing in a rapid pace and has negative effects; between countries and
within countries. Not only material but also psycho-social differences. People start to compare and
distrust each other and the government. It is the cause of migration.
Income inequality is a phenomenon the world has been dealing with since the old ages.
The world has been globalizing and the gap between rich and poor has been growing.
Some people believe that inequality is inevitable. Nothing can be done about it. Others believe that
it is naturally decreasing by free economics that lead to equilibrium levels. By ignoring that
something can be done of both camps challenge researchers to find the answers and solutions.
The subject of this paper is to shed light on the mechanisms responsible for income
inequality. If there is a better understanding of what causes income inequality, governments can
adjust their policy on it. The problem of income inequality must be taken very seriously, because of
the problems and instability it will cause.
According to Jaumotte (2013), trade openness and technological progress are widely
regarded as two of the main drivers of recent economic growth. Economic growth, in turn, is not
divided equally. All drivers of economic growth divide income differently. The exact effects are
ambiguous and show different results for countries belonging to The Organisation for Economic
Co-operation and Development (OECD) and developing countries. Most empirical research suggests a
positive effect of trade on the diminishing of income inequality. Trade openness reduces the gap.
The first research on trade and economic growth division between high-skilled and low-skilled labor
by Stolper and Samuelson (1941) supports this. OECD countries are high-skill-labor abundant and in
developing countries, low-skill-labor is the norm.
This creates an opportunity for trade, exchanging their abundant production factor. It is
interesting to compare these countries. What effect does the difference of the standard of labor
have on average income and the difference in income inequality? What type of country profits most
of the trade? What type of country distributes their wealth the most equal? What type of country
grows fastest? These questions can be answered by looking at data. The objective is to measure and
understand the relationship between trade and wealth; the central question in this paper is: what is
the effect of trade openness on the income inequality in developing countries compared to the
OECD countries?
In order to answer this question, it is important to first elaborate on earlier research. How
did major philosophers and economists write about this issue? This will be presented in the
literature review. After creating some understanding of the matter, the methodology chapter of this
paper will deal with the most important factors necessary for the research; what are the variables
and the corresponding data come from. The next chapter will give a broad description of the model
used analyzing the data and the regression output in the empirical analyses. Finally, I will provide an
answer to the central question and give a proposal for further research in the conclusion.
In this thesis, the effect of the level of trade openness on the level income inequality for
developing countries and OECD countries is analyzed. The relationship between trade openness and
income inequality will be tested by analyzing an unbalanced panel data set of OECD countries and
developing countries. The regions exist of respectively 30 and 29 countries with a time series from
1994 until 2014. This is the most recent time series available. For these regions plus the total data
set as a control region a fixed effects regression will be done. Income inequality will be regressed
against trade openness, foreign direct investment, technology and other control variables. For each
region control variables, will be added one by one to the simple model to generate the full model of
8 variables. The theoretical relationships of these variables to income inequality are explained in the
methodology section. This recent data set with the distinct comparison between OECD countries and
developing countries is what distinguishes this research from other research.
The results of the regression show that trade has a mildly positive relationship with income
inequality for OECD countries, but no significant effect in developing countries. The main findings are
that income inequality is mostly influenced by research and development. This is in contrast with the
expectation of the classical economic theory. Recent literature also predicts the effect of
technological advancement. The null hypotheses are rejected in this research, but to generalize
these findings further research is needed. The distinct difference between OECD and developing
countries is difficult to compare because of insufficient data for developing countries. Better data
series would improve the insight in the effect of economic environment. The new data series show
that, in a new era of technological advancement, the main causes of income inequality change. The
gain of this knowledge is of utmost importance when policy decisions are made considering the
income distribution.
Literature review
The cause of income inequality has been the subject of many researchers. Economists
and philosophers look for the connection between growth and income inequality. In this
section, the issue of income inequality and how it has evolved over time will be keynoted.
The main results will be addressed.
Many research, on different variables.
Distinguished economic philosophers like Ricardo and Marx already thought about
the division of wealth in the 18
thcentury. Ricardo observed that increasing demand for one
product can increase the costs of another. Population growth increases the demand for
food. Instead of increasing wages for farmers, the growing demand for land resulted in an
accrue of the prices of arable land. The landowners gain more than the farmers. Marx
observed the same effects related to the industrial revolution. And the increase in industrial
products resulted in higher profits for the industrialists, much bigger than a gain for the
factory workers. They observed how the income inequality grew between capital owners,
first landowners and later factory capital owners on the one hand, and the labor force, first
farmers and later factory workers on the other. Their predictions about the future were
pessimistic. Growing income inequality was inevitable. Although their theories were valued
and gave a great insight for further economics, their predictions did not turn out the way
they expected. However, Current researchers observe a widening of the income inequality
between high-skilled labor and low-skilled labor.
There were some changes in the world that Karl Marx did not expect. One was
globalization. Now that factories increased productivity, trade became more profitable.
New research was executed on trade and economic growth. Stolper and Samuelson (1941)
confirm and add to earlier economic theories from Hecksher and Ohlin. Hecksher and Ohlin
concluded that trade increases economic growth. Stolper and Samuelson found that if you
divide labor in high-skilled and low-skilled it has a different effect. They observed total
economic growth in combination with growing income inequality. By using the right
distributive matters economic growth can be more equally distributed. Together they
formed the HO-SS model to analyze two-factor trade and found the new division within
economic growth. They concluded that the loss of one factor of production is always less
than the gain from the other factor of production. Redistributive measures, specific
subsidies, for instance, can result in more equally distributed economic growth. In
developing countries, unskilled labor is abundant, so an export economy will emerge. In
developing countries, the increase in low-skilled labor increases wages and decreases
income inequality. (Meschi, & Vivarelli, 2009)
:
“Growth is a rising tide that lifts all
boats.” (Piketty 2014 p.11) Stolper and Samuelson added trade as a factor of economic
growth and connect it to income inequality.
The first empirical research on income inequality took place in the USA. It was Simon
Kuznets (1955), who gave definition to all the measures of income inequality and economic
growth. He used data, previously not available before, covering a period of 35 years, from
1913-1948. He found that when a capitalistic country is growing, income inequality rises
only to a certain level. Once the capitalistic country is so more developed income inequality
starts to decrease again. This resulted in “Kuznets curve”, still widely known. He was the
first using empirical evidence in a debate that had been going on for decades. The
pessimistic predictions of David Ricardo and Karl Marx are proven wrong.
Kuznets does mention to read his analysis with caution, for any exogenous shocks on
political front, like the two World Wars (1914-1918 and 1939-1945) could have influenced
his research. This weakened his conclusion and further research was needed. But only in the
late 20
thcentury, when income inequality again started to increase again, researchers
started to investigate the issue once more. Numerous researchers started to analyze
different countries. Piketty started on a program with many others to make an extension of
the Kuznets hypotheses. First for the United States of America and later for France (2013).
Piketty brought new knowledge about the division of wealth in the developed countries.
Though these researchers build the fundamental basis of research on income
inequality, the most of their theories are outdated and proven wrong by more recent
research. Researchers nowadays are more careful with generalizing the results of their
research. Another difference is that they have different approaches to the subject. In this
paper, I am interested in comparing income inequality within developing countries to
income inequality within OECD countries.
Branko Milanovic (2005) was the first who gave insight in the income inequality over
the whole world. He specifically compared emerging countries with the developed world.
Also interesting is his conclusion that financial trade has been growing. Though this shows
insignificant effects in his results. The new theory is that financial trade is a typical form of
high-skilled labor. People, who own capital in combination with the skills to trade, benefit
strongly of growing financial trade. Most of the capital in concentrated in the OECD. This
means it widens the gap between the rich countries and the poor. Branko Milanovic added
the factor of financial trade as an explanatory variable of income inequality.
More recently Jaumotte (2013), Meschi and Vivarelli (2009) also compared
developing countries with developed countries. They add technological advancement to be
a driver of economic growth. This is an interesting addition. They see developing countries
grow much faster because they profit from the innovative development of the western
countries went through. This positive effect is however in conflict with the next conclusion
of Meschi and Vivarelli (2009). They suggest that income distribution in developing
countries is worsened by trade with high-income countries. What they see is that a part of
the labor force in underdeveloped countries is picking up the advantages that high-skilled
technology brings. A majority of the population is not benefitting. The gap between have
and have not in these countries is widening.
Rafael Reuveny (2003) finds that income inequality is reduced by democracy. In a
democracy, the large numbers of low-skilled laborers have the same voting rights as a small
group of high skilled laborers. This results in more redistributive measures. Politicians have
to take into account the needs and demands of the majority, resulting in a more social
distribution of wealth. Coincidentally Sanjeev Gupta e.a.(2002) find evidence that corruption
is correlated to income inequality. Politicians who fight corruption and stimulate
redistributive measures, thus decrease income inequality.
Milanovic finds, in 2005, strong evidence stating that openness to trade in
developing countries increases income inequality. In the OECD, with its higher average
income levels, openness to trade decreases income inequality. This is an interesting
contradiction. Clarke (1995) finds that income inequality is negatively correlated with
growth in developing countries, but only has a small effect. Jaumotte (2013) finds that
financial trade is positively correlated to economic growth, but also positively correlated
with income inequality.
Branko Milanovic (2016) explains inequality is not the same as poverty. Poverty is, in
principle, reduced by economic growth. No such simple relationship exists between
economic growth and inequality. Li et al (1998) found an ambiguous effect of income
inequality on economic growth. But the gross domestic product (GDP) growth rate has a
positive effect on the income inequality. Lundberg and Squire (2003) found that income
inequality depends on economic growth and the other way around. Meaning that trying to
influence one factor will affect the other simultaneously.
Piketty (2014) investigates income inequality over the long run and tries to
generalize his findings. He concludes that inequality does not follow a predictable path.
Technology creates income inequality unless education levels can catch up with the rapid
progress. Whether the one or the other force is stronger depends on the institutions and
policies that countries choose to adopt.
Overall we can see that earlier research finds many different results. Where the
period of 1960 until the mid 1990’s show significant positive effects of trade on income
inequality (Lunberg and Squire 2003). Branko Milanovic (2005) finds a negative result of
trade openness. Meschi and Vivarelli (2009) find a positive result for trade openness in the
same time period. They conclude that technology alters the effect of trade. The more recent
decades show that technology on its own has impact on income inequality (Jaumotte 2013).
Piketty (2014) claims technology and education together are the main forces over the long
run but he does not show much empirical results. Since all researchers claim that different
factors are the most important in determining income inequality, this research will control
for most of these acclaimed influences. Jaumotte (2013) has the most comparable research.
Her method controls for most of the factors that have a supposed impact on the income
distribution according to earlier research. The proxies are different, the time series are more
recent and the developing country set consists of more extreme low income countries, but
the line of thought is similar to Jaumotte’s. The recent developments in the world ask for a
more complete research.
However, the world develops, more variables evolve that result in a faster growing
economy. Economic growth will per definition decrease poverty. This means that even poor
people profit from economic growth. The variables that cause economic growth, however,
all have a different effect on income distribution. The main variables are trade, financial
trade, technology and political regime. For the remainder of my thesis, I will follow the
methods of Jaumotte. Her paper is of high influence in the research world.
Methodology
This chapter provides the empirical framework and methodology of the research.
Instead of focusing on income inequality within a specific country, this paper focusses on
the comparison between developing countries and OECD countries. What is the relation
between income inequality and trade openness? Other factors that also influence income
inequality are discussed. The first section will elaborate on variables used and how they
affect income inequality according to the literature. In the second subsection, I will explain
how the eventual data set was established. In the third subsection, the regression model
used will be explained and justified.
In this paper, I will do a panel data research on developing countries compared to
OECD countries. Income inequality is the dependent variable, represented by the Gini
coefficient on the income distribution. On the right side of the equation, a set of growth
variables will be tested on significant positive or negative contribution to income inequality.
This model is based on the model of Jaumotte (2013). Her research presents a clear
distinction between developed and developing countries. Jaumotte is the first who
incorporates trade finance and technology together as an effect on income inequality. Her
variables are roughly the same. The data set to explain the variables, however, are different.
It consists of a different country sample.
The variables
According to economic theories and earlier research, the variables I used have some
level of influence on income inequality. Some variables with a significant effect might be
omitted. Those are either not accounted for, or the data were not available. In the end, I
used the most relevant and complete data to fit the model. The variables described are
almost complete sets that have the right time series for the right countries I want to
compare.
Variables
Explanation
Gini
The Gini coefficient, measuring income inequality.
Trade
The exports and imports as a percentage of GDP, measuring
trade openness.
FDI
Foreign Direct Investment as a percentage of GDP, measuring
financial trade openness.
R&D
Research and development as a percentage of GDP, proxy for
technological advancement.
Education
Government expenditure on education as a percentage of GDP,
proxy for the diffusion of knowledge.
Unemployment
Unemployment rate, percentage of total labor force
Government
expenditure
General government total expenditure, percentage of GDP
GDP growth
GDP per capita growth (annual %).
Inflation
Consumer prices (annual %)
Empirical evidence on the income distribution in developing countries is scarce and the
quality can be questioned. (Perottie, 1995) Thus researchers try to find different ways to
measure the income distribution. There is, however, no perfect way. Some have thorough
measurement ways but incomplete data, others have bigger data sets but simpler general
measuring tools.
The Gini Coefficient from Corrado Gini (1912) measures the distribution of income
and provides a measurable unit to income inequality. This is between 0 and 100. 0 meaning
perfect equality, everyone earns the same income. 100 meaning total income inequality,
only 1% of the people have all income. It is possible to use data from household surveys, or
the data from income tax registers. These both give a picture of the distribution of income.
According to Jenkins (Jenkins, 1991), it is best to measure social values in the most
consistent way. In this data set both income and consumption surveys are used. Some
countries switch from consumption to income surveys, countries change to one general,
most convenient and correct way of measurement. In the transition period, they use both
sources for a couple of years. This inconsistency is solved by using country fixed effects to
correct. Gaps in the data are linearly interpolated. This results in a data set of more than
1000 usable Gini data. Atkinson (1970) fought the Gini coefficient as a measurement of
income inequality. The properties of the Gini are unlikely to be accepted. But most
researchers, like Lundberg and Squire (2003) Li and others (1998), Jaumotte et all (2013) use
the Gini coefficient. A vast majority of available data on income inequality is measured by
the Gini coefficient.
Trade stands for the openness of a country. Trade openness is measured by imports
and exports as a percentage of GDP. In this case the only importance is how a certain level
of trade possibly influences the income distribution (Milanovic, 2005). The general theory is
that more openness decreases income inequality. As described in the literature review, the
HO-SS model explains how trade increases economic growth and how high-skilled and
low-skilled profit in a different way. Since the division of high-low-skilled and low-low-skilled is
substantially different in developing countries than OECD countries, trade should have a
different effect in developing countries than OECD countries. This is because OECD countries
are high-skill abundant and developing countries (DC) are low-skill abundant. This results in
a trade flow, consisting of high-skilled products from OECD to DC, and low-skilled products
from DC to OECD in reverse. This increases the demand for low-skilled labor in DCs, wages
go up and income inequality goes down. In the OECD, demand for high-skilled labor goes up,
high-skilled wages go up and low-skilled wages go down; the income inequality increases.
Though some of the assumptions of the HO-SS model are outdated, the connection between
trade and income inequality trough high-skilled and low-skilled labor is central to this issue.
More recent Branko Milanovic (2005) concluded that though trade may increase economic
growth and lift absolute poverty it had no favorable or neutral impact on income
distribution for developing countries.
Trade is an indicator for globalization. But globalization is not merely the result of
the flow of goods and services. Trade in the form of capital can also be globalization. The
implies it originates from high-skilled labor market and flows to DC. Goods are mostly
produced in dominantly low-skilled developing countries and find, as a result of trade
openness, their way easily to rich countries. According to the HO-SS model, this would mean
that a big supply of high-skilled labor would mean more advantage on financial trade.
Financial trade is mostly practiced by the rich. Only when the financial institutions are strong
and provide easy access to money for lower income groups it favors income equality. The
general theory suggests a positive effect of financial trade on income inequality.
The influence of technological advancement is more complicated. Technological
advancement is measured by expenditure on research and development (R&D) as a
percentage of GDP. This is the main source of the development and innovation. Where R&D
increases the high-skilled demand it increases the skill gap and accordingly income
inequality. The problem with using R&D as a proxy for technological advancement is that
most innovation comes from OECD countries. This technology is spread over the world by
trade. That is why the effect of technological advancement in developing countries is
actually underestimated. In developing countries, research and development is low but
technological advancement is high. For example in African countries most people own a
cellphone, they skipped landlines. Those countries skip a stage of development and lift on
the speed of R&D in developing countries. Data on ICT stock, as a percentage of total stock,
would capture the technological advancement in developing countries. Only ICT stock is
more complex to find. Therefore, I chose R&D as an alternative. Since it is measured as a
part of GDP, it is easy to compare with the other data and corrects for economic growth. In
short; Technology is high-skilled labor intensive. Causing both countries to increase demand
for high-skilled labor. Which increases income inequality. But the spread of technology also
increases the productivity of the low-skilled labor. Which decreases income inequality.
The main forces that determine the level of labor income inequality in the long-run
are, according to Goldin (2009) education and technology. These forces need equal growth
to maintain income equality. Higher level of education creates a bigger supply of high-skilled
labor. While technological development leads to a bigger demand for high-skilled labor.
whether education develops faster or technology occurs faster will cause income inequality
to respectively decrease or increase (Piketty, 2014).
Consider education as factor. Piketty (2014) called the spread of knowledge the main
divergence force. An increase in education implies an increase in the supply of skilled labor,
a decrease in the relative skilled/unskilled wage differential, and an overall decrease in
income inequality (Meschi and Vivarelli, 2009). I used government expenditure to indicate
the level of education in a country. Note that this measurement should be read with caution
because private expenditure on education is not included.
Another variable is inflation. This decreases real wages and effects the lower income
groups more than higher. Thus, inflation increases income inequality (Lundberg and Squire
2003). Inflation is added to capture the macroeconomic influence on income inequality.
The unemployment variable is taken as a percentage of the total labor force.
Unemployment is usually bigger at the bottom of the income distribution. This would mean
bigger unemployment rations bigger income inequality. In terms of high-skilled and
low-skilled, in a country with high unemployment rate, the high-skilled will be the first to get a
job.
GDP per capita growth, the annual percentage, is a proxy for economic growth.
Lundberg and Squire (2003) found that income inequality depends on economic growth and
the other way around. Meaning that trying to influence one factor will affect the other
factor simultaneously.
Another tool to use is "dummies". In order to exclude time trend effects, I used
"year" and "country dummies". Adding each year with a value of 0 or 1. So that every year
specific effect is caught in the dummy.
Total government expenditure affects the income distribution because the
expenditure is mostly on subsidies for weaker sectors and public goods. This increases the
income of the bottom part of the income distribution. This is a proxy for redistributive
policy.
Variables for democracy and corruption are omitted because measurement of
democracy strength is subjective and measurements of corruption are biased. These control
variables are insignificant in for example Branko Milanovic’s (2005) research.
The data
This subsection will elaborate on the origin and modification of the data to create
transparency. Most of the data I used are originating from the World Bank national accounts
data files. All data together form an unbalanced panel set. Meaning multiple observations
over multiple countries in multiple years.
For the Gini coefficient, the World Bank POVCAL data is used. The POVCAL is a
database of 176 available countries. After selecting a sample of developing and OECD
countries with sufficient data, a sample of 59 countries is left. I aimed for a time series from
1990 till 2014, but some data are incomplete. For example, before 2004 OECD countries are
not covered by the POVCAL. To extend this data to 1990, I used data from the World Income
Inequality Database (WIID). The WIID consists of a collection of all measurements on
inequality. For consistency, I used one GINI coefficient within a country. Remaining gaps of
maximal 4 years, I interpolated linearly. Some variables are less complete. All data are put in
logs, except inflation and GDP growth per capital (annual%) to make sure negative
observations are not dropped.
When adding more control variables to the model “Stata”, a complete, integrated
statistical software package, takes out the incomplete years. The trade variable, or trade
openness, is measured by imports as a % of GDP, plus exports as a % of GDP. Trade covers
the whole time-series. Financial trade is also downloaded from the World Bank. This data
set also provides a good coverage of the time series 1990-2014. R&D as a percentage of GDP
from the World Bank only has data starting from 1996.
The countries used in the regression are countries of which at least 2 years per
country cover all the variables used. To get a clear comparison between OECD and
developing countries, all countries are selected by either being part of the OECD or having
lower (lower-middle) average income. The World Bank classification is used to separate
these groups on Gross National Income (GNI): (L) Low-income economies ($1,025 or less)
and (LM) lower-middle-income economies ($1,026 to $4,035). OECD, members are drawn
from the member list at
https://www.oecd.org
. OECD countries are also listed by the World
Bank as high-income-economies ($12,476 or more)
The comparison between OECD and developing countries is interesting because
these country groups differ on many levels. OECD are high-skill abundant, and developing
countries are low-skill abundant. The difference in the amount of capital stock is huge.
Average of $142,24 billion in high-income countries and $17,924 billion in low-income
countries (World Bank, 2014). The technological development is mostly placed in OECD
countries but the market for technology is growing in developing countries. Even though the
differences, or because of the differences, the trade between these country groups has
been growing over the last decades and so is the income inequality within and between
these countries. The countries used are listed below. These are picked based on sufficient
available data. For the developing countries, the data set is contaminated with
measurement errors and therefore the data is not as complete as the OECD dataset. The
countries are listed below.
OECD countries (high-income-economies) (29)
Australia, Austria, Belgium, Canada, Czech Republic, Chili*, Denmark, Estonia, Finland,
France, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Latvia, Luxembourg,
Mexico*, Netherlands, Norway, Poland, Slovak Republic, Slovenia, Spain, Sweden,
United-Kingdom, United-States.
*indicating countries that recently joined the OECD.
Developing countries
Armenia, Bangladesh, Benin, Bolivia, Burkina Faso, Burundi, Cambodia, Democratic Republic
of Congo, El Salvador, Ethiopia, Gambia, The, Honduras, Lao People's Democratic, Republic,
Indonesia, Malawi, Mali, Moldova, Mozambique, Niger, Pakistan, Philippines, Rwanda,
Senegal, Sierra Leone, Togo, Uganda, Vietnam, Zambia
The regression model
In this subsection, the econometric cross country model is constructed to find the
relationship between Trade and income inequality for OECD countries compared to
low-income countries. Before, we have established the variables that are included in the model
based on economic theory. As explained the dependent variable will be the income
inequality expressed in the Gini coefficient, GINI. The independent variable central to this
paper is trade or trade openness. Financial trade openness is the second independent
variable. Also, a series of control variables will be added to avoid biased coefficients. The
adjustments and modifications needed for econometric testing will be explained.
𝐺𝐼𝑁𝐼(it) = α + β
1𝑇𝑟𝑎𝑑𝑒 𝑜𝑝𝑒𝑛𝑛𝑒𝑠𝑠
𝑖𝑡
+ β
2𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑜𝑝𝑒𝑛𝑛𝑒𝑠𝑠
𝑖𝑡+ β₃ 𝑅&𝐷
𝑖𝑡+ β
4𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛
𝑖𝑡+ β
5𝑈𝑛𝑒𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡
𝑖𝑡+ β
6Govermet Exp
𝑖𝑡+ β
7𝐺𝐷𝑃 𝑔𝑟𝑜𝑤𝑡ℎ
𝑖𝑡+ β
8𝐼𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛
𝑖𝑡+
𝑖+ ∂
𝑡+ µ
𝑖𝑡where the subscript i is to indicate the country and t being the time series. The term
𝑖is a
full set of country dummies, ∂
𝑡represents a full set time dummies, and µ
𝑖𝑡captures all the
omitted factors. These extra dummies are added to capture the fixed effects of countries
and years. This way shocks in a certain year, or other influences that don’t change much
over time are accounted for. By expressing the model in this way, a causal direction is
assumed. The variables affect the Gini and not the other way around. Also, all variables are
contemporaneous, meaning there is no lag assumed (Allison, P. D. 2009). Last all variables
are logged except the variables that contain negative observations, Inflation and GDP
growth.
To get a clear comparison between OECD and developing countries the regressions
will be done three times, for three different regions. Once for the whole data set. Once for
all OECD Countries and once for developing countries. For each region, 7 regressions are
done. One summary model regression with only trade and FDI as an independent variable. A
full model with all control variables. In between one by one all control variables are added
to increase the prediction value of the model.
I have chosen a fixed effects regression because this gives the most efficient results
and this method matches the best with the panel data set. A fixed effects model treats
unobserved differences between countries as a set of fixed parameters that can either be
directly estimated or partial out of the estimating equations (Allison, 2009). First, the
requirements for fixed effects regression. The dependent variable must be measured for
each country for at least 2 years. These data need to mean the same but change over time.
This is the case for the data set. Including the fixed parameters in the model indicate using
the fixed effects model.
For empirical prove, I conducted a Hausman test for the full model of all regions.
H0: Random effects is appropriate
H1: Fixed effects is appropriate
Hausman test for all countries region
chi2(26) =
66.55
Prob>chi2 =
0.0000
Hausman test for OECD countries region
chi2(26) =
113.91
Prob>chi2 =
0.0000
Hausman test for Developing countries region
chi2(26) =
30.68
Prob>chi2 =
0.1999
For both all countries and OECD countries the H0 is rejected. For the developing countries,
the H0 is not rejected but this is probably due to the reduced observations in the full model.
Because the countries drop to 6 observed countries it may be possible that the within-group
variation is not big enough. Therefore, a random effects regression should be pulled. Using
Random effects opens up to omitted variable bias. Because the model is probably much
more complicated than conducted in this research and the identifying assumption* is
assumed, I have chosen to use the fixed effects regression throughout all the regions and
models, thereby excluding the omitted variable bias.
*Identifying assumption: Unobservable factors that might simultaneously affect the
left-hand side and right-left-hand side of the regression are time-invariant.
The hypotheses
To understand how the central question will be answered a clear hypothesis is formulated.
What is the effect of trade openness on the income inequality in developing countries
compared to the OECD countries? So, the null hypotheses, based on the basic economic
theory, is that trade decreases income inequality in developing countries compared to an
increase in income inequality in OECD countries. The alternative hypotheses are that trade
increases income inequality in developing countries, like the results of Branko Milanovic
(2005). Trade has a positive effect on income inequality in OECD countries and trade
openness is positive for all country groups. Shortly: the rich are getting richer and de poor
don’t profit that much.
Three regressions will be done, so the hypotheses also exist of 3 parts.
H0: the β of trade openness is smaller than 0 for developing countries
the β of trade openness is bigger than 0 for OECD countries
the β of trade openness is unequal to 0 for all countries
H1: the β of trade openness is bigger than 0 for developing countries
the β of trade openness is bigger than 0 for OECD countries
Empirical Analyses
Descriptive statistics
First of all, let us look at some summary statistics.
OECD country
Variables
Obs
Mean
Std. Dev.
Min
Max
Country
621
15.37198 8.447855
1
30
Year
621
2002.089 6.857787
1990
2014
Gini
620
32.38892 7.394461
19.49
57.25
Trade
621
91.45078
53.4828 19.73534 352.9038
Foreign direct investment
590
5.349092 19.56188 -57.42675 430.6151
Research and development
443
1.689207 0.9216993
0.25923 4.40744
Education
483
5.241816 1.172526
1.92723 8.61797
unemployment
613
8.188155 4.183025
1.256
27.466
Government expenditure
567
43.85837 8.876292
18.708
68.313
GDP growth
613
2.113933 3.349974 -14.57305 18.62113
Inflation
591
4.903648 23.39716 -4.479938 555.3812
Developing country
variables
Obs
Mean
Std. Dev.
Min
Max
Country
398
15.03015 8.222602
1
29
Year
398
2002.608 6.511539
1990
2014
Gini
398
41.16154 8.549592
23.9
65.76
Trade
398
68.63267 31.84574 18.88983 169.5345
Foreign direct investment
388
3.278112 3.428731
-2.75744 32.41621
Research and development
93
0.250253 0.162252
0.00544 0.87295
Education
205
9.446889 7.093168
0.99855 26.69957
unemployment
168
8.57897 5.901036
2.4
38.4
GDP growth
373
2.905516
4.43955
-30.7128 14.44751
Inflation
360
9.679322 17.33336
-8.48425 183.312
Some results stand out, first of all, the independent variable, the Gini, in OECD countries this
is about 10 percentage points lower than developing countries. The standard deviations also
differ with more than 1. This is as expected, the income inequality is higher in developing
countries. The other important notice is that research and development is the galling factor
in the developing country dataset. This is probably because there is no real data on research
and development since research and development is less important in developing countries.
Most innovation takes place in OECD countries which explains the 1 percentage point
difference between OECD and DC's. The total number of observations is less for developing
countries but this should not matter for the regression.
Correlation statistics
In this section the results of the empirical regressions will be discussed, results are shown
for the regions models separately. The regions are all countries, OECD countries, and
Developing countries (DC). For every region, the regressions run for the full model to the
summary model dropping one control variable at the time. To check if there is no
correlation between variables a correlation matrix is shown below.
OECD
Trade
FDI
R&D
Educ
Unemp
Gov. EXP
GDP
growth
Trade
1
FDI
0.344
1
R&D
-0.2836
0.007
1
Educ
-0.0588 -0.0037 0.5379
1
Unemp
-0.1647 -0.1277 -0.1918 -0.2237
1
Gov. EXP
-0.1396 -0.0327 0.5602 0.4869
0.1659
1
GDP growth
0.1424 0.0877 -0.3133 -0.2192
0.0409
-0.3244
1
Inflation
0.0975 0.0427
-0.288 -0.1201
-0.0003
-0.0932 0.1834
Developing
countries
Trade
FDI
R&D
Educ
Unemp
Gov. EXP
GDP
growth
Trade
1
FDI
0.2668
1
R&D
0.2933
0.453
1
Educ
0.6668 0.4116
0.5616
1
Unemp
0.013 0.2023 -0.1245 -0.3618
1
Gov. EXP
0.7155 0.4466
0.6455 0.9564
-0.3124
1
GDP growth
0.1692
0.206 -0.0348
-0.1
0.2721
-0.1217
1
Inflation
0.1985 -0.0396
0.3875
0.237
-0.2528
0.2457 -0.0314
The correlation of research and development with education has a clear causation. A higher
level of education creates more high-skilled labor. This, in turn, leads to more research and
development. The correlation between R&D and government expenditure is possibly
because governments fund some research and development projects. Because R&D is
slightly correlated with the other variables it could be wise to take R&D out of the model.
But since the effect of research and development is of interest for this thesis. The other
variables will be dropped in order of theoretical importance. Overall the correlation levels
of the developing countries are higher, this is most likely due to the limited data points. So,
taken that in consideration only general government expenditure is highly correlated with
education. This is due to the fact that education is measured as government expenditure on
education as a percentage of total government expenditure.
The results of the models for each region are presented in the tables below. For
each group, the independent variables in the summary model are the same. The results
show that trade is positively related to income inequality in OECD countries and negatively
related in developing countries. This is in line with the economic theory. Though the effects
are only significant when most control variables are dropped. For Developing countries
trade is, trough out all the models, insignificant. This implies that the income of both poor
and rich grow from trade by the same relative percentage. The rise of the Gini coefficient is
due to other factors than trade.
FDI is not significant for any of the samples. This implies that the profit from financial trade
is equally divided over the income groups. Both in developing countries and OECD countries.
A reason for this could be that foreign direct investment creates easy excess to financial
capital.
Again analyzing the coefficients, we notice contradiction between the results and earlier
economic theory. Trade is not significant adding the control variables. Though the prediction
levels of the model increase R-squared.
All countries
Summary
model
Dropping
Variables
Full
model
VARIABLES
1
2
3
4
5
6
7
Trade
0.0236
0.0897***
0.0360
-0.00389
-0.0103
-0.0114
-0.00622
(0.0189)
(0.0287)
(0.0343)
(0.0368)
(0.0373)
(0.0374)
(0.0377)
Foreign Direct Investment
-2.67e-05
-2.69e-05
-0.000356 -0.000434 -0.000452 -0.000464 -0.000486
(0.000172) (0.000146) (0.000473) (0.000469) (0.000469) (0.000470) (0.000471)
Research and development
0.0255**
0.0159
0.00878
0.00779
0.00872
0.00615
(0.0129)
(0.0169)
(0.0199)
(0.0202)
(0.0203)
(0.0204)
expenditure on education
0.0141
0.000710
0.0430
0.0429
0.0458
(0.0351)
(0.0405)
(0.0477)
(0.0478)
(0.0483)
unemployment
0.0598*** 0.0583*** 0.0590*** 0.0640***
(0.0143)
(0.0154)
(0.0155)
(0.0159)
Total government expenditure
-0.00144
-0.00124
-0.00116
(0.00149)
(0.00153)
(0.00154)
GDP per capita growth
0.000799
0.00102
(0.00131)
(0.00132)
Inflation
0.00240
(0.00150)
Constant
3.414***
3.088***
3.262***
3.301***
3.329***
3.321***
3.259***
(0.0786)
(0.122)
(0.167)
(0.180)
(0.183)
(0.183)
(0.188)
Observations
977
527
413
395
385
385
382
R-squared
0.056
0.127
0.142
0.190
0.176
0.177
0.183
Number of country1
59
45
39
35
35
35
35
Standard errors in parentheses
There are several reasons this might not be the case in this empirical analysis. First of
all, it is possible that extending the data set with two decades would show other results
because of the larger variations in trade and income inequality. Statistically speaking a series
of 1991 until 2014 is big enough for testing but perhaps the relationship has weakened
because the influence of trade has decreased over the last couple of years and other
variables are now stronger forces of economic growth.
Developing countries
Summary
model
Dropping
variables
Full
model
VARIABLES
1
2
3
4
5
6
7
Trade
-0.00973
-0.00619
0.0102
0.0129
-0.0690
-0.103
0.0429
(0.0290)
(0.0672)
(0.0799)
(0.130)
(0.152)
(0.171)
(0.205)
Foreign Direct Investment
-0.00145
0.00442
0.00938
0.0109
0.0114
0.0110
0.0192
(0.00179)
(0.00490) (0.00648) (0.00983) (0.0110)
(0.0114)
(0.0130)
Research and development
0.0684*** 0.0602*
0.152**
0.127
0.144
0.122
(0.0183)
(0.0344)
(0.0696)
(0.0761)
(0.0856)
(0.0857)
expenditure on education
-0.0645
-0.233
-0.295
-0.336
-0.397
(0.0713)
(0.138)
(0.200)
(0.222)
(0.223)
unemployment
0.151
0.0572
0.0272
0.00941
(0.133)
(0.145)
(0.162)
(0.159)
Total government expenditure
0.00470
0.00989
0.00285
(0.0108)
(0.0153)
(0.0160)
GDP per capita growth
0.00301
0.00106
(0.00607) (0.00615)
Inflation
-0.00958
(0.00777)
Constant
3.714***
3.869*** 3.823*** 3.607*** 4.214*** 4.387***
4.310***
(0.120)
(0.288)
(0.387)
(0.846)
(0.912)
(1.003)
(0.983)
Observations
388
93
64
46
42
42
42
R-squared
0.122
0.509
0.571
0.708
0.753
0.758
0.788
Number of country1
29
15
10
6
6
6
6
Standard errors in parentheses
Second, it could be possible that the data set of developing countries has too many
missing values, this creates multi-correlation between the variables. This is somewhat
corrected in the reduced model. Still, the data from developing countries are less reliable
than the OECD data. Household surveys are difficult in poor parts of developing countries
and the tax system will have some flaws in correctly predicting the lower part of the income
distribution.
OECD countries
Summary
model
Dropping
Variables
Full
model
VARIABLES
1
2
3
4
5
6
7
Trade
0.0423*
0.0823**
0.0659*
0.0362
0.0417
0.0337
0.0519
(0.0241)
(0.0330)
(0.0398)
(0.0409)
(0.0415)
(0.0417)
(0.0432)
Foreign Direct Investment
-2.91e-05
-7.36e-05
-0.000574 -0.000580 -0.000560 -0.000596 -0.000615
(0.000139) (0.000131) (0.000436) (0.000432) (0.000433) (0.000433) (0.000434)
Research and development
-0.0759*** -0.0605** -0.0577** -0.0591** -0.0567** -0.0546**
(0.0223)
(0.0246)
(0.0244)
(0.0253)
(0.0254)
(0.0255)
expenditure on education
0.118**
0.0783
0.104*
0.0987*
0.107*
(0.0461)
(0.0480)
(0.0561)
(0.0561)
(0.0572)
unemployment
0.0394*** 0.0421*** 0.0451*** 0.0519***
(0.0146)
(0.0156)
(0.0157)
(0.0163)
Total government expenditure
-0.00103
-0.000537 -0.000641
(0.00150)
(0.00154)
(0.00155)
GDP per capita growth
0.00212
0.00231
(0.00147)
(0.00148)
Inflation
0.00279*
(0.00165)
Constant
3.234***
3.076***
2.939***
3.044***
3.023***
3.029***
2.907***
(0.101)
(0.140)
(0.201)
(0.203)
(0.206)
(0.206)
(0.217)
Observations
589
434
349
349
343
343
340
R-squared
0.207
0.194
0.207
0.226
0.216
0.222
0.230
Number of country1
30
30
29
29
29
29
29
Standard errors in parentheses
Third, measuring income inequality by the Gini coefficient may be more sensitive to
changes in the middle of the distribution. Changes in the upper and lower part of the
distribution are less visible in the Gini coefficient (Kakwani, 1980). Therefore, the Gini
coefficient may underestimate the changes in income inequality. The bad measurement of
inequality may cause bias of the variables towards zero. If the income inequality is
measured by quintiles of 20% income groups, the lower and upper-income groups can be
weighted. This would give a better insight into the specific distribution of economic growth
through globalization.
Last, it might be possible that the effect of democracy and corruption are
underestimated. These are theoretically speaking strong factors that could influence the
income distribution but also have influence on trade openness and FDI.
If we look at the regression output, we notice that research and development is
significant for the smaller models of developing countries (0.0684***) and for all models in
OECD countries (-0.0759*** model 2) ( -0.0546** model 7). These results show that
research and development has a positive effect on income inequality in developing
countries. Research and development increases the demand for high-skilled labor. Since
developing countries have relatively little high-skilled labor the already high wages increase
and the low-skilled labor is replaced by more productive technology and low-skilled wages
decrease. This explains the positive effect of R&D on the Gini in developing countries. In the
OECD country group, a 1% increase in R&D causes a 0.08% decrease in income inequality.
This is hard to fully explain with the same reasoning as above. Though OECD is high-skill
abundant, an increase in demand would cause income inequality either way. Perhaps the
low-skilled group profits more from the increased productivity. Since the educational level
of low-skilled labor is much higher than in developing countries. This is in line with the
reasoning of Piketty (2014) that income inequality is caused by the race between
technological advancement and education. Where the gap between the two is smaller in
OECD countries.
Although the positive effect of education is contradicting the divergence force of
knowledge diffusion. This is perhaps because of the use of the wrong proxy for education
which underestimates private expenditure on education.
Conclusion and practice, reject or accept the hypotheses
The regressions don't show very strong results but let's have a look at the benchmark
model, which is the second regression. Here trade, financial trade, and technological
advancement are regressed, the basis of Jaumotte’s research. We see that in fact, for OECD
countries, trade has a positive effect on the Gini, 0.0823**(5% significance level). When
trade increases with 1% the Gini will increase with 0.08%. For all countries, this effect is
almost the same only of a significance level of 1%. For developing countries, there is no
significant effect. Trade might actually be equally divided over the poor and the rich.
Observing this implies that the null hypotheses are rejected. As explained before some
flaws in the model and the dataset may be the cause of non-significant effects. It is possible
that the focus on trade is outdated. Though many researchers agree that globalization is a
driver of economic growth it may be that the more recent data sets show that trade is in
fact not that important anymore. Perhaps the effect of technological advancement, which is
a recent development of the 21
stcentury, is underestimated in this central question and is
of much more importance. Jaumotte (2013) already introduced technological advancement
to be an important control variable. This research agrees that technological advancement is
of much more influence. Leaving the classical economic theory behind and accepting the
theory of Jaumotte (2013). Piketty introduced education but in this research education has
no effect.
Conclusion
In this research, I have tried to find the relationship between trade openness and income
inequality and compare this effect in OECD countries with developing countries. The
literature review gives a short look into the ambiguity about this topic. General consensus is
far from reached. With the unique dataset, including the most recent decades of 59
countries, this research will give new insight and perhaps lead to a greater understanding of
the relationship between trade openness and income inequality.
The answer to the central question is clear but not very satisfying. In general, we find
that the benchmark model for all countries trade has a positive effect on income inequality.
In OECD countries, trade openness also has a positive effect on income inequality. For
developing countries there is no significant effect of trade on income inequality. The
classical literature suggested significant relationships between trade and income inequality
but it seems like the findings are more in line with new findings and modern economic
theory. The findings suggest that research and development is a more important driver of
income inequality than trade. Seeing trade as a manner of distributing technological
advancement rather than directly affecting income inequality. A reason for this could be
that trade divides growth quite equally and both rich and poor profit equally. Technology is
actually the factor that creates a skill-gap and therefor an income-gap. This is in line with
recent research, where technological development and education play large roles in income
inequality. The role of education is not clear in this research so technological advancement
is the number driver of income inequality.
It is important to look critically at the way of measurement. The comparison
between OECD countries and Developing countries is difficult because the data set for
developing countries is too small to create a strong reliable model. This is due to
inconsistent and insufficient measurement points of income in developing countries. The
data search is growing and perhaps the future can give more complete datasets of the
whole world.
In further research, it would be necessary to focus on technology, education and
political systems and policies. This is a more modern view on the matter. Putting these
factors central to the income inequality question means it is important to find the right way
of measurement for both developing and OECD countries. Another interesting question
would be to research the between country income inequality.
Reference list
- Allison, P. D. (2009). Fixed effects regression models (Vol. 160). SAGE publications.
-
Angeles, L. (2010) An alternative test of Kuznets’ hypothesis J Econ Inequal 8: 463.
- Aradhyula, S., Rahman, T., & Seenivasan, K. (2007). Impact of International Trade on Income
and Income Inequality. Portland: American Agricultural Economics Association.
- Atkinson, A. B. (1970). On the measurement of inequality. Journal of economic theory, 2(3),
244-263.
- Blumenstock, J.(2016)
Fixed Effects Models, Mimeo
Available at:
http://www.jblumenstock.com/files/courses/econ174/FEModels.pdf
Last accessed: 17/02/17”
-
Ceriani, L. & Verme, P.(2012) The origins of the Gini index: extracts from Variabilità e
Mutabilità (1912) by Corrado Gini, Journal of Economic Inequality (2012)
10, 3,
pp 421–443
- Ceriani, L., & Verme, P. (2012). The origins of the Gini index: extracts from Variabilità e
Mutabilità (1912) by Corrado Gini. The Journal of Economic Inequality, 10(3), 421-443.
-
Clarke, G. R.G. (1995) More evidence on income distribution and growth. Journal of
development economics
- Cornia, G. A. (2015). Income inequality in Latin America, UNU-Wider
-
Deininger, Klaus, and Lyn Squire. "A New Data Set Measuring Income Inequality." The World
Bank Economic Review 10 (1996): 565-591
- Dollar, D., & Kraay, A. (2002). Growth is Good for the Poor. Journal of economic growth, 7(3),
195-225.
-
Goldin, C. D., & Katz, L. F. (2009). The race between education and technology. Harvard
University Press.
- Gupta, S., Davoodi, H., & Alonso-Terme, R. (2002). Does corruption affect income inequality
and poverty?. Economics of governance, 3(1), 23-45.
- Heckscher, E. F., & Ohlin, B. G. (1991). Heckscher-Ohlin trade theory. The MIT Press.
- Jaumotte, F., Lall, S., & Papageorgiou, C. (2013). Rising income inequality: technology, or
trade and financial globalization?. IMF Economic Review, 61(2), 271-309.
- Jenkins, S. (1991). The measurement of income inequality. Economic Inequality and Poverty,
3-31
-
Kuznets, S, 1955, “Economic Growth and Income Inequality,” American Economic Review,
Vol. 45, No. 1, pp. 1–28.
- Lundberg, M., & Squire, L. (2003). The simultaneous evolution of growth and inequality. The
Economic Journal, 113(487), 326-344.
- Meschi, E., & Vivarelli, M. (2009). Trade and income inequality in developing
countries. World development, 37(2), 287-302.
- Milanovic, B. (2005). Can we discern the effect of globalization on income distribution?
Evidence from household surveys. The World Bank Economic Review, 19(1), 21-44
- Milanovic, B. (2011). Worlds apart: Measuring international and global inequality. Princeton
University Press.
-
Novib, O. (2017). An economy for the 99%,
Oxfam briefing paper
- Perotti, R, 1996, “Growth, Income Distribution, and Democracy: What the Data Say,” Journal
of Economic Growth, Vol. 1, No. 2, pp. 149–87.
- Piketty, T, 2003, “Income Inequality in France, 1900–1998,” Journal of Political Economy,
Vol. 111, No. 5, pp. 1004–42
-
Piketty, T., & Goldhammer, A. (2014). Capital in the Twenty-First Century. Cambridge
Massachusetts: Harvard University Press.
- Piketty, T., & Saez, E. (2014). Inequality in the long run. Science, 344(6186), 838-843.
- Piketty, T., and E. Saez, 2003, “Income Inequality in the United States, 1913–1998,”
Quarterly Journal of Economics, Vol. 118, No. 1, pp. 1–39
- Reuveny, R., & Li, Q. (2003). Economic openness, democracy, and income inequality: an
empirical analysis. Comparative Political Studies, 36(5), 575-601.
-
Sailesh K. Jha The Kuznets curve: A reassessment, World Development Volume 24, Issue 4,
April 1996, Pages 773-780
-
Stolper, W.F., and P.A. Samuelson, 1941, “Protection and Real Wages,” Review of Economic
Studies, Vol. 9, No. 1, pp. 58–73
Appendix
All country
SUMMARY:
Correlation matrix:
Inflationc~u 951 6.711474 21.42169 -8.484249 555.3812
GDPpercapi~l 986 2.413386 3.816255 -30.71279 18.62113
Generalgov~i 881 36.04717 13.31418 8.738 68.313
lnunemp 781 1.987651 .4957844 .2279321 3.648057
lneduc 688 1.707547 .5425073 -.0014511 3.284647
lnrnd 536 -.0094558 1.054006 -5.213976 1.483294
lntrade 1,019 4.275369 .5184768 2.938624 5.866195
FDIGDP 978 4.527476 15.3747 -57.42675 430.6151
lntrade 1,019 4.275369 .5184768 2.938624 5.866195
lnGini 1,018 3.549506 .2371977 2.969902 4.186012
Year 1,019 2002.291 6.726223 1990 2014
Country 0
Variable Obs Mean Std. Dev. Min Max
Inflationc~u 0.1624 1.0000 GDPpercapi~l 1.0000 GDPper~l Inflat~u Inflationc~u 0.0563 0.0083 0.0563 -0.3673 -0.1123 0.0128 -0.2319 GDPpercapi~l 0.1304 0.0822 0.1304 -0.2735 -0.1960 0.0947 -0.2971 Generalgov~i 0.0532 0.0260 0.0532 0.7206 0.6355 -0.0050 1.0000 lnunemp -0.1573 -0.1156 -0.1573 -0.2358 -0.2761 1.0000 lneduc 0.1585 0.0501 0.1585 0.5615 1.0000 lnrnd -0.0831 0.0569 -0.0831 1.0000 lntrade 1.0000 0.3351 1.0000 FDIGDP 0.3351 1.0000 lntrade 1.0000 lntrade FDIGDP lntrade lnrnd lneduc lnunemp Genera~i (obs=382)