Urbanisation, economic growth, and income inequality:
Kuznets’ theory revisited after 60 years in South Korea
Name: Sumin Lee Student Number: 11982209
Faculty of Economics and Business Economics Supervisor: Kees Haasnoot
Submission Date: 15th July 2021
Statement of Originality
This document is written by Student Sumin Lee 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.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract: The aim of this work is to assess the Kuznets curve that was introduced in The American Economic Review in 1955 and see if the hypothesis is verified using the time
series data of South Korean economy of 1965-2018. Using quadratic regressions and linear regressions with structural breaks, the research shows that to some extent, the Kuznets
hypothesis is valid in South Korea’s Economy.
Keywords: Kuznets curve, Income inequality, Urbanisation, South Korea, Economic development
JEL classification: O10; 053; R0
Table of Contents
Introduction ... 5
Theoretical Framework ... 8
The Kuznets Hypothesis... 8
Previous Studies ... 9
Background ... 11
Methodology and Data ... 14
Models ... 14
Measurement of Inequality... 16
Measurement of Income and Urbanisation ... 17
Graphical Representations... 17
Descriptive Statistics ... 21
Results ... 22
Main Findings ... 22
Discussion ... 24
Concluding remarks ... 26
References ... 28
Appendix ... 31
Until a few decades ago, South Korea was praised by many economists for achieving both growth and equality (Ahn, 2016). However, over the last couple of decades, its equality level has been deteriorating. Statistics show that since early 1990s, the Gini coefficient has shown an upward trend in which it marked its unprecedented rate in 2007 by 0.313 and since then, it rapidly accelerated reaching to 0.34 in 2018 (SWIID, 2020).
In the aftermath of the WWII, the Korean War which lasted from 1950 to 1953 virtually destroyed the majority of infrastructure and hierarchy within the country which helped the country to start in relatively economically equal state. From the late 1950s and to early 1960s, most people suffered from absolute poverty, and it was not until 1960s when modern economic growth embarked in South Korea and the shift from agrarian society to industrialised economy in the nation started (Yeon et al, 2016).
In the beginning of the 1960’s, South Korean GDP per capita was just a little over
$1,500 (2011 US Dollars) which is incomparable to half a century later in 2010 when it reached over $31,500. In 2018, the nation’s GDP per capita was estimated to be around $37,928 (Maddison Project Database, 2020). With an average growth rate percentage of 5.6, now Korean economy is one of the world’s largest as it stands 10th highest in nominal GDP rank and 26th in GDP per capita rank.
During the transition, Korean economy was always accompanied by urbanisation. In 1960, only less than one third of the population resided in urban areas, however in less than two decades, the number went up to half of the population. In fact, it only took 3 decades for the most of its population to move to urban (Appendix 1). As of 1988, South Korean urban rate surpassed 70% and that was on a level to the urban rate of first world countries of the time. In less than four decades since the modern industrialisation has started in the country, its urban rate has reached to over 80% in 2002 and the rate has been steady since then.
Historically, industrialisation which created rapid economic growth, has led to urbanisation. It drove higher demands of labour in urban areas where infrastructure for mass manufacturing was located. In the modern era, these labour-intensive manufacturing factories are often replaced by technology-industry hubs that led to requiring for educated and skilled workers that can create such technology which intensifies income inequality in the labour market (Ahn, 1997).
In recent decades, a lot of attention has been paid to the relationship between economic growth and income inequality. In the 1990s, total global inequality declined for the first time since early 1800s. Indeed, cross country income inequality has declined in the last 25 years.
However, income inequality within countries has risen, which is the form of inequality that people face more in their everyday lives. This can be accredited to rapid economic growth in emerging economies in Asia (United Nations, n.d).
Simon Kuznets in his article that was published in American Economic Review in 1955, reviewed the correlation between the phase of a country’s economic development and income inequality. Further, he tried to find out what factors determine the secular level and trend of income inequalities. This is done by studying three countries in particular; United States, Germany, and England, as that were ‘presently’ developed countries of the time when the paper was published. Kuznets proposed a graphical representation of an inverted U-shape that draws the relationship between urbanisation and income inequality. However, as Kuznets has admitted himself in the paper, the research lacks in empirical evidence due to scarcity of relevant data and socio-political context of the time to confirm his hypothesis. Nevertheless, Kuznets’ studies became one of the most renowned work in the field of economic growth and inequality and widely discussed in many subsequent papers. However, Kuznets’ hypothesis is still a subject of debate today as the findings show heterogenous results.
This paper aims to join the live discussion by assessing the Kuznets’ curve by using the data at our disposal today. This will be done by analysing modern data of South Korean demographic transition, economic growth, and income inequality. The reasons why South Korea is an interesting economy to conduct a research are as follows. First, the country has experienced both early stage and later stage of economic growth within the last half a century.
Thus, income distribution for all critical stages of the nation’s economic development that is needed to test the Kuznets curve and the turning point is relatively well recorded (Bowman, 1997). Secondly, South Korea is poor in natural resources that the economy inevitably had to rely on its human resources for its economic development that income plays an immense role in the nation’s economy (Chiang, 2018). Furthermore, immediately following WW2, Korean government went through a massive land distribution. This is a good proxy to wealth equality since 75% of the population were engaged in agriculture back then (Bowman, 1997). Lastly, the country started with a relatively low income inequality in the beginning of the nation’s modern urbanisation and industrialisation. This was not only lower among other
underdeveloped countries but also lower compared to industrialized countries (Koo, 1984). The presence of the Kuznets curve will be estimated in two steps. In the first step, using non-linear regressions, income growth rate and urbanisation rate will be regressed in regard to the Gini Index. In the second step, we test for structural breaks in the regressions and use separate linear regressions for the ranges that are divided based on the structural breaks identified in the Gini regressions on income growth and urban rate.
The rest of the paper is structured as follows. In section 2 related literatures and background information will be reviewed. Section 3 presents methodologies and data including graphical representations descriptive statistics. It is followed by section 4 which addresses the results of regressions and section 5 where the acquired results will be analysed.
Finally, section 6 will conclude the findings of the paper and give suggestions for the future research.
The Kuznets Hypothesis
In 1955, Simon Kuznets postulated a relationship between economic development and inequality. He did this by documenting both cross-country and time series data. The famous Kuznets’ inverted bell-shaped curve represents that in the earlier phase of economic development, inequality widens, and in the later phase, it narrows down again (Figure 1).
According to Kuznets, the key elements underlying the mechanism of income inequality are concentration of savings and migration to the city.
According to Kuznets, there are two factors that determine the disparity of income in the beginning of economic growth. Firstly, the concentrated savings of the individuals of the higher income population will cumulate over time. This will translate into a fuelling proportion of asset ownership among a small wealthy population while cheap labour holds down wages that leads to income inequality. Secondly, an economy faces structural transition in which households migrate from agricultural sector to industrial sector. This movement has two sub- components of inequality: between sector inequality and within sector inequality. The agricultural sector is usually lower in average per capita income than the industrialized sector, as labour productivity is higher in the latter, featuring between sector inequality. Furthermore, the income inequality for the rural population is somewhat narrower than that of urban. Thus, increasing weight of the urban population means an increasing share for more unequal distribution which explains within sector inequality. That is, if the transition occurs from a sector with both a low mean and low variance of income to high mean and high variance income, then inequality will increase. However, if the workers move from a sector with low mean income but higher variance to a sector with a higher mean income but lower variance in income then the inequality will not necessarily increase.
Once early economic development phase proceeds and eventually the majority of workers locate in the city with only a small portion remains in countryside, the trend is reversed.
Then, the poor rural population will join the richer industrial sector and, within the industrial sector, those who started at the bottom will move up the ladder and earn higher wages, offsetting the widened income inequality. In other words, human capital accumulation occurs, and workers are incentivized by differentials. Thus, at the later stage of economic development, there is a negative correlation between economic growth and income inequality. Kuznets’
studies show this upward movement in the developed countries in the West occurred in the 19th century, peaked around in the end of the century or in the beginning of the new century and the moved downward in the 20th century.
In accordance with Kuznets, this could explain the differences in income inequality across countries as well. He thought that other countries will repeat the pattern of these ‘older’
developed countries. He specifically looked at three underdeveloped countries India, Ceylon (Sri Lanka) and Puerto Rico after the WW2. He finds that these underdeveloped countries are somewhat more unequal than developed countries during the period. Reasons for this, he predicts, is associated with a much lower level of average income.
Kuznets’ findings were based on the limited data that covers only the narrow part of the world and there is a high tendency of a wide margin of error. Indeed, Kuznets concluded the paper by stressing out that his findings are more of a theoretical implication than a complete empirical research such that more data should be available to confirm it.
Figure 1. Kuznets’ Curve
Kuznets’ theory was tested numerical times by many economists following his publication. Plentiful of earlier researchers that conducted cross-country studies on existence of the Kuznets curve have confirmed his theory (S. Ahluwalia, 1976, Lindert and Williamson, 1985, Papanek and Kyn, 1986). Some of the recent studies have also found an evidence of the
Kuznets’ curve including Younsi and Bechtini(2018), who explored annual panel data of BRICS and approved Kuznets’ findings. Martinez-Navarro et al (2020) made a unique approach in verifying the theory as they took a variable other than the log-GDP such as the hdi or the proportional contribution of the agricultural sector on GDP which showed a robust result.
Unlike many other studies that did the research at the time, S. Robinson (1976) pioneered in investigating Kuznets’ work in the presence of within-sector inequality. He supported the explanation with a simple economic assumption that economy is divided into two sectors with different income distributions as measured by log variance of income.
A number of the Eastern countries began to develop, and researchers started to look into development of Asian countries in correlation to economic inequality. H. Oshima specifically investigates the Gini trends in Asian countries from 1950s to 1980s. In his paper Kuznets’ curve and Asian income distribution trends (1992) he confirms that income distribution trends in most Asian countries resemble the Kuznets theory. However, he finds that the peak that is depicted in the curve is reached earlier in Asian economy than those of the West.
That is, when the peak is reached in Asian countries, still a majority of workers are in agricultural sectors rather than industrial sectors.
Using both provincial-level panel data and time series data, et al (2012) extensively examined the relationship between urban-rural income disparities and development in China for the period of 1978 to 2006. Authors were able to identify the Kuznets curve in the Chinese economic development. This research was succeeded by Cheng and Wu in 2014 focusing more on urbanisation where they also presented an evidence of the inverted-U relationship in their studies.
However, critics argued that the Kuznets’ U-shape is a by-product of historical differences between countries, and not from progression in the individual country’s development. They were concerned of the presence of cross-sectional dependence that panel data might obtain. Saith (1983) questioned the validity of the U-hypothesis found in cross- country econometric studies conducted by various researchers including that of Ahluwalia. He argued that these studies were questionable in methodological premises and had defective statistics. He remarked that although long-run series of homogenous data are limited, inter- temporal studies would more critically and accurately evaluate the presence of the Kuznets’
hypothesis. This is because using a time series data of an individual country will avoid
distinguishable differences in size, historical heritage, and the timing of their industrialisation process among countries as they interplay among economic growth.
In 1990s, data for inequality became more available than in the past and a number of researchers actively tried to replicate Kuznets’ findings based on these newly available data.
Klaus Deininger and Lyn Squire (1998) used the updated inequality data, i.e. the Gini coefficients and income shares of population quintiles of as many places in the world as possible for around four decades. Although they find strong relationship between growth and reduction of poverty, they failed to find a systematic link between growth and change in aggregate inequality. Higgins and Williamson (1999) report an evidence of the inverted U- pattern only when conditioning for other variables. Other researchers such as Anand (1993) and Baymul and Sen (2019) who based their research on a large data of both intertemporal and cross country did not confirm correlation between growth and inequality that align with the Kuznets hypothesis.
Irma Adelman and Sherman Robinson (1978) in their book Income Distribution Policy in Developing Countries: A Case Study of South Korea, collected the scattered evidence and constructed income distribution data for 1964 and 1970. They concluded that South Korea avoided a negative income distribution consequence from economic growth. It is further supported by Kirk Bowman in 1997 as both data and the country experts repudiate the classic theory.
Simon Kuznets hypothesises that industrialising nations undergo an increase and subsequent decrease in income inequality. His theory is not yet conclusive as papers that later replicated his research have shown conflicting results. Hence, this paper will evaluate applicability of the Kuznets’ hypothesis for South Korea’s economy through examining and quantifying the available data.
Definition of urban and trend of urban demographics of South Korea.
Although there are discrepancies in the classification of ‘urban’ between countries, the connotation of the word generally refers to areas related to cities. The majority of Asian countries define urban areas based on administrative criteria such as economic functions,
infrastructure, or services available and a few define it based on size or density of population.
By country economists, ‘urban’ population is referred to any amount of population living in designated cities as the definition is used in Republic of Korea (United Nations, 2012). In terms of occupational characteristics, urban areas are where residents are predominantly engaged in secondary and service sectors rather than primary sectors (Kang, 1998).
UN DESA’s World Social Report 2020 describes urbanisation as a determinant of the future of inequality. Urbanisation has the potential to fuel growth but also when planned poorly, has a tendency for increase in inequality and social exclusion. Data supports that inequality is higher in urban areas than rural areas and this polarity within cities is not just economic but also spatial and social (United Nations, 2020).
According to the government’s statistics, Seoul, the nation’s capital city itself holds around 10 million people. Including its neighbouring towns and metropolitan cities, the population of capital region counts around 25 million which is almost half of South Korea’s population. Other metropolitan cities, cities with more than 1 million inhabitants, include Busan (3.4 mil), Incheon (3.0 mil), Daegu (2.5 mil), Gwangju (1.5mil), Daejeon (1.5mil) and Ulsan (1.2mil). The capital city, its neighbouring cities and 6 metropolitan cities are home for more than two thirds of the country’s population (KOSIS, 2020).
There are usually two factors for urbanisation: i) rural-urban migration and ii) growth of rural areas to urban areas and. In the early stage of urbanisation in Korea, proportional increase of the urban areas was the main reason for urbanisation (Kang, 1998). From 1960 to 1965, about 5% of the rural population headed to cities of which 70% moved to Seoul. From 1965 to 1970, around 14% of rural population moved to cities and 61% of them moved to Seoul.
According to surveys, the main reasons for new immigrants from other parts of the country to the capital were “job transfer” and “business” and some minor reasons included “education”
and “conditions of living”. However, in the later phase of urbanisation, since 1980s, as many of the rural population has moved to urban areas, the process of urban growth was quickly replaced by the installation of new cities in rural areas (Asia Society, n.d.)
Looking at the historical background of government planning of urbanisation, South Korean urban administrative development plans can be divided into three phases.
1960-1980: The Quickening Period (Beginning of urban development) 1980-2000: The Growth Period (Development of new towns)
2000-present The Maturation Period (Heightening the value of cities)
Determinants and Trends of Income Distribution in South Korea
Although there is no reliable data to infer the distribution of income in Korea before mid-1960s due to social and historical conditions, it is believed that its initial distribution of wealth and financial assets was one of the most even in the capitalist world. The reasonings behind this assumption include widespread absolute poverty, confiscation of wealth after liberation that once was controlled largely by Japan or that was illegally accumulated during the colonial period and monetary reforms after the war (Adelman & Robinson, 1978).
There is a general agreement that the income distribution of Korea has become somewhat more equitable from the mid-60s through the early 70s and then worsened significantly during the mid-70s. However, there are two contradictory estimations since the 1980s. The National Statistics Office estimates steady and substantial improvement since the 80s and other studies estimate a rapid deterioration in the late 80s and upturn in the early 90s.
Nonetheless, various private nationwide surveys through early 1990s showed that Korean people did not feel that income has been equitably distributed nor has witnessed this improvement (Ahn, 1997).
South Korea’s income distribution has seen to be determined not just by the level of economic development, but also by its extensive dependency on foreign exports as well as a large extent by major industrialisation policies that the governed adopted for rapid export expansion since early 1960s (Koo, 1984). Indeed, from the early 1960s, private sector-led exports began to grow with labour intensive products. Since 1970, these labour-intensive products were replaced by heavy and chemical industrial products that are more skills intensive.
Although, opinions vary, openness to trade and increase in production of skill-intensive export goods also might affected the nation’s income inequality. Further, Chang and Lee (2012) predicts that population aging due to increase in the elderly population and changes in family structure such as increase in single parent families can also be factors for a rise in income inequality, although the major driving force in income inequality since the mid-1990s has been increased earnings dispersion.
Methodology and Data
As the model aims to check if economic development has resulted in the rise of income disparity in the beginning of the development phase and then the fall in the later phase, the dependant variable is income inequality. For independent variables we look at income and urban rate.
Model 1. We first consider income as an independent variable. Since the graphical representation that we assume here is non-linear, for the accuracy, we add a square variable to the model.
Inet= α + β1*GDPt + β2*GDPt2 + Ɛt
Hypothesis: β1>0 and β2<0.
Model 2. We use the same model as above, however, we replace urban rate as an independent variable instead of income in this model.
Inet= α + β1*Urbt + β2*Urbt2 + Ɛt
Hypothesis: β1>0 and β2<0.
In models 3 and 4, we first check for structural breaks using the supremum Wald test.
If the test recognises any shifts, we divide the time series accordingly and using a linear regression, we regress the variable for each range. The Wald test constructs a test statistic for possibility of a structural break that is unknown. It uses the maximum, an average or the exponential of the average of the tests computed at each possible break date (Stata, n.d.). A structural break in a time series refers to an abrupt change at a point of time. Detecting structural changes in the data will give insights where, if there was any, significant changes occurred in the data. This will give more insights to the poolability of our time series data (Baltagi, 2001). .
Inet= α + β1*GDPt + Ɛt [for years 1965-(Y1-1)]
Inet= α + β2*GDPt + Ɛt [for years Y1-(Y2-1)]
Inet= α + β3*GDPt + Ɛt [for years Y2-(Y3-1)]
Inet= α + βi+1*GDPt + Ɛt [for years (Yk-2018)]
Hypothesis i) A structural break does not exist.
Hypothesis ii) If tests detect one or more structural breaks in years Y1…Yk, βi>0 and βi+1<0 for any i (1⩽i⩽k, k=1,2,3…)
Inet= α + β1*Urbt + Ɛt [for years 1965-(Y1-1)]
Inet= α + β2*Urbt + Ɛt [for years Y1-(Y2-1)]
Inet= α + β3*Urbt + Ɛt [for years Y2-(Y3-1)]
Inet= α + βi+1*Urbt + Ɛt [for years (Yk-2018)]
Hypothesis i) A structural break does not exist.
Hypothesis ii) If tests detect one or more structural breaks in years Y1…Yk, βi>0 and βi+1<0 for any i (1⩽i⩽k, k=1,2,3…)
16 Table 1. Dependent and independent variables
Inet Gini Coefficient of South Korea
GDPt GDP per capita of South Korea (in $1,000)
GDPt2 Square of GDPt
Urbt Urbanisation rate of South Korea
Urbt2 Square of Urbt
Measurement of Inequality
There are a variety of ways to measure economic inequality, however, commonly wealth, income, and consumption are used. It is a known fact that wealth is notorious to track and thus data for wealth distribution are less available. Since consumption reflects typical income and data on income are more available, we use income distribution, especially discretionary income as the metrics of inequality. Income is defined as household disposable income in a particular year. It is comprised of earnings, self-employment and capital income and public cash transfers (OECD, 2021). Where taxes and other mandatory chargers are deducted from the total personal income is disposable income. In South Korean economy, the gap between market and disposable income is not large and inequality defined as disposable income inequality is driven by market income inequality (Choi, 2013). In the model, income inequality is expressed through the Gini index that is calculated through the population’s disposable income. The Gini index or Gini coefficient is an indicator that is used to measure the extent to which income is equally distributed within the population. The index ranges from 0 to 1 where the former indicates perfect equality, and the latter represents perfect inequality.
The Gini index is said to be more sensitive to the changes of the middle of the income distribution than other indicators such as the Theil Index or the Atkinson Index (Solt, 2020)
Total national income inequality includes within region inequality; that is within urban
and within rural and between region inequality; that is urban-rural inequality. To check existence of the Kuznets curve in a national economy, information of inequality in smaller scale region must have been very helpful. However, statistical information was only available for the national-level and bigger provincial-levels, which included both big cities and rural areas, making it difficult to track urban-rural income disparities. Thus, fifty-two yearlong national level income inequality in Gini coefficient was used for income distribution measurement.
Although there were a few researchers that worked on to calculate the Gini Coefficient of South Korea in the past years such as Choo (1992), Ahn (1992, 1995), Yoon (1998), and Whang and Lee (1998), they do not provide consistent data that cover for the entire period that the research objects. Further, the non-verifiability and absence of official estimates make it difficult to secure reliability of the data. Nevertheless, their research gives critical insights to the trends of the income inequality in the nation. In Appendix 2, Gini coefficient calculated by the aforementioned authors can be compared alongside to the SWIID data. In this paper, we use the Gini Coefficient data based on disposable income that were retrieved from SWIID.
Measurement of Income and Urbanisation
For an income indicator, real gross domestic product (GDP) per capita was used. It is calculated by dividing the national GDP by its population and it represents the country’s economic output per person. Data for GDP per capita were drawn from the Maddison Project Database which collates historical economic statistics. In data set, US Dollars in 2011 is taken as a unit of currency. For the ease of reading, GDP per capita is divided by $1,000 and is expressed in thousands in the regression.
The level of urbanisation is given by urban rate, which counts the urban population out of the total population. These data are provided by the World Development Indicators, the primary World Bank data collection.
Figure 2 which depicts the recursive cumulative sum plot of the Gini index shows a pattern that resembles more of a U-shape, which is opposite to what Kuznets had predicted.
Figure 2. Recursive cumulative sum plot of the Gini coefficient of South Korea (1965-2018)
It can be seen from Figure 3, in a glance, that GDP per capita has grown roughly in a linear line, but again, the income inequality does not show an inverted-U shape. The Gini index more or less represent no particular pattern that in relation to the income growth level just by referencing the graph.
Figure 3. Two-way graph of income growth and income inequality in South Korea (1965-2018)
When urban rate is considered as an explanatory variable, the graph aligns more with the Kuznets’ assumption. From the graph, it can be seen that the rapid urbanisation has started from the beginning of the given time slot until approximately late 1980s when it started to slow down. Looking at the Gini coefficient during the time where rapid urbanisation happened, a pattern that resembles an inverted-U shape is visible with the peak in 1980 (Figure 4). When the Gini Index started to increase was concurrent to the time when urbanisation started to slow down. At this point, we can see that most of the population has already moved to urban by around 75%. In other words, during urbanisation until most of the population had moved to urban, inequality level shows somewhat what Kuznets had theorised.
Figure 4. Two-way graph of urbanisation and income inequality in South Korea (1965-2018)
In Figure 5, structural breaks in the regression of the Gini on GDP per capita that are detected in the supremum Wald test are shown in yellow vertical lines on its two-way graph.
The first vertical line that marks 1979, is close to the tip of the Gini of the first half of the time series. The second vertical line in 1990, roughly speaking, crosses the lowest point of the Gini of the whole range. Lastly, the third vertical line of 2012 crosses where it is close to the beginning of the income inequality overshoot of the economy,
Figure 5. Two-way graph of income growth and income inequality with structural breaks
Structural breaks in the regression of Gini on urban rate detected in Wald tests are shown in yellow vertical lines in Figure 6. Although not accurately the same, the two lines that illustrate structural breaks of the urban rate regression resembles second and third lines of the GDP regression. The first vertical line in Figure 6 crosses in 1991 which is one year later than the second line of the Figure 5 and is closer to the bottom tip of the nation’s Gini index. This line was estimated in the above paragraph as where urbanisation slowed down, the Gini coefficient has started to increase. The second vertical line in in Figure 6 is in 2008 and it is four years earlier than the third structural break of the income regression shown in Figure 5 and is closer to where the Gini has started to move up higher than its history has ever been.
Figure 6. Two-way graph of urbanisation and income inequality with structural breaks
Table 2 provides a summary of descriptive statistics of the data used in the test for years between 1965 to 2018. It can be seen that the Gini coefficient was in range between 0.293 and 0.34 with average of 0.309. Looking at the minimum and maximum values of GDP per capita, amount of income growth can be checked. The year that marked for the maximum GDP per capita at $37,928 is roughly 20 times of the minimum at $1,917 and this explains the large number for its standard deviation. As for the urbanisation rate, we can see that the maximum value is around 2.5 times larger than that of the minimum.
Table 2. Summary of descriptive statistics of variables
urb 54 .6690602 .1645548 .32351 .81936 gdp 54 16.79337 11.75489 1.917428 37.92761 gini 54 .3089444 .0094238 .293 .34 Variable Obs Mean Std. Dev. Min Max
The results of regressions of models 1 and 2 are shown in columns (1) and (2) of Table 3. Starting at the income variable, there is a negative effect of GDP per capita and a positive effect of the square of GDP per capita on the Gini coefficient. Both coefficients are statistically significant at p-value of 0.001, although the effect of GDP on the Gini seems to be smaller than the square of GDP. Drawing from the coefficients, we can see the function depicts a shape that is more of an U-shape. As for the urbanisation, it shows a negative coefficient for the variable urban rate and negative coefficient for the square of urban rate, although, they are not statistically significant. The results of both models 1 and 2 reject the hypothesis of existence of the Kuznets curve.
Table 3. Regression coefficients of model 1 & 2
Using the Wald test, statistics rejected a null hypothesis of no structural break for GDP per capita and detected a break in 1990. In addition, in order to find other possible structural breaks before and after this date, the same test was conducted, but this time in range of 1965 to 1989 and 1990 to 2018. Within years from 1965 to 1989, the test showed 1979 as another
* p<0.05, ** p<0.01, *** p<0.001 t statistics in parentheses
N 54 54 (172.52) (14.75) _cons 0.312*** 0.323***
(1.26) urb2 0.0791 (-1.00) urb -0.0767 (8.12) gdp2 0.0000542***
(-5.96) gdp -0.00151***
gini gini (1) (2)
structural break and years between 1990 to 2018, it showed 2012 to be a structural break. Thus, years 1979, 1990 and 2012 will be used as the basis dates to divide the time series data in model 3. For the regression of urban rate in South Korea, a null hypothesis is again rejected and a structural break in 2008 was recognised. The same method was used as GDP regression and there was another structural break before 2008 in 1991 and as there was not enough observations for after the break, we find no further break between 2008 and 2018. Hence, for model 4, years 1991 and 2008 will be used as the basis dates to divide the time series.
For ease of reading, Model 3 and 4 are revised with structural breaks in the following:
Inet= α + β1*GDPt + Ɛt (for years 1965-1978) Inet= α + β2*GDPt + Ɛt (for years 1979-1989) Inet= α + β3*GDPt + Ɛt (for years 1990-2011) Inet= α + β4*GDPt + Ɛt (for years 2012-2018)
Inet= α + β1*Urbt + Ɛt (for years 1965-1990) Inet= α + β2*Urbt + Ɛt (for years 1991-2007) Inet= α + β3*Urbt + Ɛt (for years 2008-2018)
In Table 4, results of Model 3 are represented. Columns (1), (2), (3) and (4) each show the regression of the Gini on GDP for periods between 1965 to 1978, 1979 to 1989, 1990 to 2011 and 2012 to 2018 respectively. Except for column (2), the coefficients in all the other columns showed positive numbers. This can be interpreted as for the period between 1965 to 1990, there was an inverted-U shape graph of income inequality with a peak in1979. In terms of hypothesis, in phase 1 and 2, the result does approve the Kuznets’ hypothesis. All the coefficients and constants in model 3 could be seen as qualitatively impacting on the dependent variable.
24 Table 4. Regression coefficients of model 3
Table 5 presents the coefficient specifications of model 4. Regression of the Gini on the urbanisation rate in years between 1965 and 1990 can be found in column (1), years from 1991 until 2007 in column (2) and from 2008 to 2018 in column (3). The coefficients of urban variable are positive in the first phase and the second phase, and negative in the last phase, however, the coefficient of the first phrase is not statistically significant. From looking at these coefficients, one can conclude that there was a Kuznets curve in the second and third phase which is years between 1991 and 2018, marking the turning point in 2008.
Table 5. Regression coefficients of model 4
To begin with, the results in the previous section can be summed up as follows. The Kuznets curve does not appear in South Korea’s economy when applying the regression to the entire sample frame. The non-linear regressions of the Gini index on both income and urban levels that are depicted in Models 1 and 2 did not show an inverted-U curve, but rather
* p<0.05, ** p<0.01, *** p<0.001 t statistics in parentheses
N 14 11 22 7 (318.89) (546.83) (84.04) (10.78) _cons 0.297*** 0.319*** 0.277*** 0.132***
(9.06) (-19.24) (9.05) (15.82) gdp 0.00226*** -0.00125*** 0.00124*** 0.00550***
gini gini gini gini (1) (2) (3) (4)
* p<0.05, ** p<0.01, *** p<0.001 t statistics in parentheses
N 26 17 11 (102.13) (1.11) (7.91) _cons 0.303*** 0.0393 4.888***
(1.23) (7.49) (-7.39) urb 0.00671 0.335*** -5.587***
gini gini gini (1) (2) (3)
‘unturned-U’ curves. Results of Model 2 showed a steeper graph than that of Model 1, which can be interpreted as the increase of urban rate affected income disparity more than the increase of income itself. In Models 3 and 4, when the time series were broken down into a few ranges after testing for the structural breaks, Kuznets curve relationships were shown in both models.
In Model 3, with the turning point to be in 1979, there was an evidence of the Kuznets curve when looking at years between 1965 and 1990. The peak was a lot later in Model 4, as it was in 2008 and the sample range was between 1991 and 2018. Even though all the coefficients of Model 3 were statistically significant, the coefficient of the first phase and the constant of the second time phase of Model 4 were not statistically significant. Referencing to Figure 6, one can see that it is because there were both upward and downward movement of the Gini within the range before the structural break that might have resulted in such variations.
An interesting observation was made while examining the structural breaks in each model. In Models 3 and 4, we looked for possible structural breaks in both linear regressions of the Gini on GDP and urban rate, where the shifts discovered in each model were somewhat coherent to each other. The placements of two structural breaks detected in Model 4 were close to the breaks that were already observed in Model 3, although there was an additional break in Model 3. This structural shift is further supported by the vertex of the parabola of Model 1. The result of the regression supports a tipping point to be when GDP is at $13,930 and the Gini is 0.30. It is a consistent finding as Korean GDP per capita moved passed that point in years between 1990 and 1991, which is where the structural breaks were detected performing a supremum Wald test. The vertex of regression of Model 2 was at when the urban rate was at 0.48 and the Gini index at 0.31. Looking at the historical data, South Korean urban rate reached 0.48 in year 1975. This is close to the additional shift found in Model 3 with four years of difference which was not recognised as a break in a Wald test of Model 4. This shows that the identified structural breaks in the models were fairly homogenous with first shift being in mid to late-1970, second one in 1990 or 1991 and the last cleft in couple of years before or after 2010. These estimates give certainty to structural breaks that give a clue in which points you can base on in examining the movements of the regressions.
Regarding the reliability of data used in the paper for analysis, there are a few limitations that need to be pointed out. First, depending on researchers and methodologies they use, values of the Gini coefficient vary (Appendix 3). Although, the SWIID where the Gini
coefficient data were drawn from tries to provide the source data that cover all the population.
However, in the recent article that documents the database of SWIID, Solt (2020) has stated that this rule was somewhat relaxed to data for earlier dates of South Korea due to absence of its figures. In other words, early series of the Gini coefficient retrieved from SWIID do not reflect within rural income distribution nor urban-rural income distribution as they only cover the urban population, which is a major drawback of this research.
This further indicates that there was no available data that could compare the Gini coefficient of rural population and urban population during the nation’s urbanisation process which is the second drawback of this research. There were a couple of early researchers that tried to estimate urban-rural gap such as Choo where in his paper he predicted farm and non- farm income inequalities along with national income inequalities based on surveys conducted on a limited sample.
In terms of methodology, this paper evaluated the Kuznets’ hypothesis with both quadratic model and linear model with structural breaks. The quadratic model is better fitting to Kuznets’ original idea and his method, but its result does not accurately explain the trend when the curve exists in a limited part of the sample or outside of the sample. The structural breaks model is farther from Kuznets’ approach, but it avoids issues with extrapolation better as it can capture observations that the quadratic models might have overlooked. Nevertheless, the linear regressions ended up with seven observations of structural breaks as a result, and the process of establishing multiple structural breaks was somewhat imperfect.
This study presented a time series analysis on income inequality of Korean economy for the years 1965-2018 and discussed its trends in regard to the theoretical stance of Simon Kuznets on relation of urbanisation, economic development and income inequality. This was done by using quadratic regressions and linear regressions with structural breaks. From the analysis, we cannot generalise the hypothesis of Kuznets inverted-U curve on the whole data set. However, we may safely interpret that in specific time ranges, the Kuznets effect was identified. In summary, the research showed that to a certain extent, the Kuznets’ hypothesis
was valid in Korean Economy.
Kuznets’ theory argues that as economy of a nation starts to develop, urbanisation spurs and income inequality increases until most of the population resides in urban when income disparity starts to decrease. The literature indicates the transformation of agrarian economy to an industrial economy, but it is suggested that it can also apply to a maturing economy with new high-productivity industry (Ikemoto & Uehara, 2000). It has been 65 years since Kuznets published his article and the process of economic development of recent days are different from when Kuznets observed it in many aspects. “Is the Kuznets’ hypothesis still valid and is it valid to every economy?” Although Kuznets’ findings have been studied extensively by many researchers and a number of empirical observations have been examined, there is still no one answer to this question. At least for the case of Korea, we can give a primitive answer. In history of South Korean economy, you can see both of what Kuznets has predicted and what he has not. If you only look at the first half of the time series, a Kuznets curve is there reaching its peak faster than the Western countries that underwent its development earlier than Korea.
However, when you look at the whole series, you will see that, shortly after, the graph changes its direction and moves up again which is something that was not explained in the Kuznets’
Does this ‘N’ curve or an upward trend after income disparity had improved only appear in Korean economy, or does it appear in other nations’ economies too? To answer this, more intra country research will be needed to further confirm both the Kuznets’ hypothesis and the additional findings. Also, more detailed data will draw clearer picture of the trends that we aimed to examine and reveal more precise relationships between the variables. For example, this study leaves out the impact of urbanisation on within-region and between-region inequality due to lack of adequate data to study it. Further, estimations of inequality measures vary and taking alternative ways to take up the analysis such as using a Pareto distribution, a log normal distribution or even the Gini coefficients with different estimation methods will also be needed for the accuracy of the studies.
Adelman, I., and Robinson, S. (1978). Income Distribution Policy in Developing Countries: A Case Study of Korea, Stanford University Press.
Ahluwalia, S. M., (1976), "Income Distribution and Development," American Economic Review, 66, 2, 12-35.
Ahn, C. Y. (2016). Rising inequalities in South Korea. East Asia Forum Economics, Politics and Public Policy in East Asia and the Pacific.
Ahn, K. (1992), “Trends in the Size Distribution of Income in Korea,” Korean Social Science Journal, Vol. 18, 27-47.
Ahn, K. (1995), “Economic Development and Income Distribution in Korea,” Journal of Economic Development, Vol. 1, No. 1, Korea Development Economics Association, 53-76 (in Korean).
Ahn, K. (1997), "Trends in and Determinants of Income Distribution in Korea," Journal of Economic Development, 22, 1, 27-56
Anand, S., and S. M. Kanbur (1993), "The Kuznets Process and the Inequality-Development Relationship," Journal of Development Economics, 40, 25-72.
Baltagi, B. H. (2021). Econometric Analysis of Panel Data. Springer International Publishing AG.
Baymul, Ç., & Sen, K. (2019). Kuznets revisited: What do we know about the relationship between structural transformation and inequality? Asian Development Review, 36(1), 136–
Bowman, K. S. (1997). Should the Kuznets effect be relied on to induce equalizing growth:
Evidence from post-1950 development. World Development, 25(1), 127–143.
Chang, J., & Lee, B.H. (2012). Increasing Income Inequality and Policy Options in Korea. e- Labor News , No. 129 .
Cheng, W., & Wu, Y. (2014). Understanding the Kuznets Process: An Empirical Investigation of Income Inequality in China 1978-2011. Monash University Department of Economics Discussion paper. 12-14.
Chiang, M. H. (2018). Post-Industrial Development in East Asia: Taiwan and South Korea in Comparison. Springer Singapore Pte. Limited.
Choi, K. (2003), "Explaining Income Inequality in Korea," Korea Development Institute, July 2003, manuscript.
Choi, K. (2013). Korea’s Income Inequality: The Trend and Major Issues. Fiscal
Sustainability and Innovative Welfare System 2013. KDI Journal of Economic Policy Conference. KDI KAEA. 282-338
Choo, H. (1992), “A Comparison of Income Distribution in Japan, Korea and Taiwan,” in Making Economies More Efficient and More Equitable: Factors Determining Income Distribution, eds. Toshiyuki Mizoguchi, Kinokuniya Company and Oxford University Press.
Choo, H., & Yoon, J., (1984) “Size Distribution of Income in Korea, 1982: Its Estimation and Sources of Change,” Korea Development Review, 6, 2-18 (in Korean).
Deininger, K., & Squire, L. (1998). New ways of looking at old issues: Inequality and growth.
Journal of Development Economics, 57, 259–287.
Higgins, M. J., & Williamson, J. G. (1999). Explaining Inequality the World Round: Cohort Size, Kuznets Curves, and Openness. National Bureau of Economic Research Working Paper 7224:57
Ikemoto, Y., & Uehara, M. (2000). Income Inequality and Kuznets Hypothesis in Thailand.
Asian Economic Journal, 14(4), 421–443.
Kang, S. (2001). Globalization and Income Inequality in Korea: An Overview . FDI, Hyman Capital and Education in Developing Countries , 18–31.
Kang, M. (1998). UNDERSTANDING URBAN PROBLEMS IN KOREA: CONTINUITY AND CHANGE. Development and Society, 27(1), 99-120.
Koo, H. (1984). The political economy of income distribution in South Korea: The impact of the state’s industrialization policies. World Development, 12(10), 1029–1037.
KOSIS: Korean Statistical Information Service. Population Census (2020). [Data file].
Retrieved from. https://kosis.kr/eng/.
Kuznets, S. (1955). Economic Growth and Income Inequality. The American Economic Review, 45(1), 1-28.
Kwon , T. H. (n.d.). Population Change and Development in Korea. Asia Society.
Lee, Jong-woo, and Seong-hyeon Whang (1998), “The Problems of Income Distribution and Related Policy Issues in Korea”, The KDI Journal of Economic Policy, Vol.20, No.1, 2.
Lindert, P. H., & Williamson, J. G. (1985). Growth, equality, and history. Explorations in Economic History, 22(4), 341–377
The Maddison Project: Maddison style estimates of the evolution of the world economy. A new 2 0 2 0 u pd at e (2 0 2 0 ) https://www.rug.nl/ggdc/historicaldevelopment/maddison/release
Martínez-Navarro, D., Amate-Fortes, I. & Guarnido-Rueda, A. (2020). Inequality and development: is the Kuznets curve in effect today?. Econ Polit 37, 703–735
OECD (2021), Income inequality (indicator). https://data.oecd.org/inequality/income- inequality.htm
Oshima, H. (1992). Kuznets' Curve and Asian Income Distribution Trends. Hitotsubashi Journal of Economics, 33(1), 95-111.
Papanek, G. F., & Kyn, O. (1986). The effect on income distribution of development, the growth rate and economic strategy. Journal of Development Economics, 23(1), 55–65.
Robinson, S. (1976). A note on the U hypothesis relating income inequality and economic development. American Economic Review, 66(3), 437–440.
Saith, A. (1983). Development and distribution: A critique of the cross-country U-hypothesis.
Journal of Development Economics, 13(3), 367–382.
Solt, Frederick. (2020). Measuring Income Inequality Across Countries and Over Time: The Standardized World Income Inequality Database. Social Science Quarterly. 101. 1183- 1199
STATA 14. (n.d): Tests for structural breaks in time-series data.
SWIID: The Standardized World Income Inequality Database (2020). [Data file]. Retrieved from. https://fsolt.org/swiid/swiid_source/
United Nations. (n.d.). Inequality – Bridging the Divide.
United Nations (2020) The Challenge of Inequality in Rapidly Changing World. UN DESA’s World Social Report 2020. https://www.un.org/development/desa/dspd/world-social- report/2020-2.html
United Nations Publication. (2012). World Urbanization Prospects The 2011 Revision . https://www.un.org/en/academic-impact/united-nations-projects-increase-worlds-urban- population
World Bank. (n.d.). World Bank Indicators: Urban Population (% of total). [Data file].
Retrieved from https://databank.worldbank.org/source/world-development-indicators Yoon, Gil Jung (1997), Analysis of Inequity in the Korean Economy, Seoul: Bak Yung Sa. (in
Yeon, J.-I. Y., Pyka, A., & Kim, T.-yoo. (2016). Structural shift and increasing variety in Korea, 1960-2010: Empirical evidence of the economic development model by the creation of new sectors.
Younsi, M., & Bechtini, M. (2020). Economic Growth, Financial Development, and Income Inequality in BRICS Countries: Does Kuznets’ Inverted U-Shaped Curve Exist? Journal of the Knowledge Economy, 11(2), 721–742.
Zhang, H., Chen, W., & Zhang, J. (2012). Urban–rural income disparities and development in a panel dataset of China for the period from 1978 to 2006. Applied Economics, 44(21), 2717–2728.
Appendix 1. Korean urban transformation
Photo1: Daegu, Bukgu district (1950s) Photo 2: Daegu, Bukgu district (2021) Photo1 Source: Wonshik Jang
Appendix 2. The Estimated GINI coefficients from different authors along with the SWIID data Gini* Gini** Gini*** Gini**** Gini****
1965 0.303 0.3365 0.3439 1966 0.303 0.3287
1967 0.303 0.3647 1968 0.303 0.3458
32 1969 0.302 0.3469
1970 0.302 0.3125 0.3322 1971 0.303 0.3074
1972 0.304 0.3121 1973 0.306 0.3676 1974 0.307 0.3823 1975 0.308 0.3769
1976 0.309 0.3899 0.3908 1977 0.31 0.378
1978 0.31 0.3699 1979 0.311 0.3752
1980 0.311 0.3567 0.366 1981 0.311 0.3572
1982 0.31 0.3766 0.3574 0.393 1983 0.309 0.3736
1984 0.309 0.3804 0.351
1985 0.308 0.3803 0.384
1986 0.307 0.3771 0.3368 0.34 1987 0.306 0.3777
1988 0.304 0.384 0.327 0.365 1989 0.302 0.4127
1990 0.299 0.4017 0.3226 0.3
1991 0.296 0.4017 0.302 0.365 1992 0.293 0.4013 0.287
1993 0.293 0.3883 0.89
1994 0.294 0.3797 0.363
1995 0.296 0.3845
1996 0.299 0.288
1997 0.303 0.282
1998 0.308 1999 0.31 2000 0.31 2001 0.311 2002 0.311 2003 0.31 2004 0.31 2005 0.31 2006 0.31
33 2007 0.313
2008 0.314 2009 0.316 2010 0.314 2011 0.316 2012 0.314 2013 0.317 2014 0.32 2015 0.325 2016 0.332 2017 0.338 2018 0.34 Gini*:SWIID
Gini**: Ahn(1992,1995) Gini***: Choo(1992)
Gini****:Yoo(1998) (Source: Kang, 2001) Gini*****:Whang and Lee(1998)
Appendix 3. The Gini Index Estimates from various sources for South Korea (1965-2018)