The Effect of Income Inequality on Human Development

John Leal
March 22, 2021

ECON 3161: Dr. Shatakshee Dhongde

Abstract:
Income inequality is how unevenly income is spread across a population. This paper looks at the effects
of income inequality on human development through analyzing cross sectional data with multiple linear
regression models. The dependent variable in this study is a country’s Human Development Index (HDI)
in 2018. The primary explanatory variable is a country’s Gini coefficient from 2010 – 17. This paper
hypothesizes that higher levels of income inequality will have a negative effect on human development as
this often results in a misallocation of resources. The models support this hypothesis, but the exact degree
of the ceteris paribus effect is still not fully clear.
I. Introduction
Human development is about the robustness of human life. According to the prominent economist
Mahbub ul Haq (1999), there is two-dimensional approach to evaluating human development:
augmenting human abilities and fostering human development. The former dimension refers to factors
like standard of living, life expectancy, and education levels. The latter dimension refers to factors like
social equality, human rights, sustainability, and political stability. While there is no single, definite
answer to what is human development, it is generally agreed upon that it extends beyond mere economic
factors and reaches across a wider range of indicators.

For economic inequality, there are often two primary factors that economists examine: the distribution of
income and the distribution of wealth. Unlike human development, there is much more debate on this
topic as to what the effects are of higher economic inequality and to what severity those effects are. This
paper focuses on income inequality. For income inequality according to Kuznets’ (1955) Hypothesis,
income inequality starts off at lower levels for pre-industrial societies and increases as industrialization
occurs. From this point, Kuznets (1955) thought that income inequality levels would begin to lower as
governments adjusted their taxing and welfare programs accordingly. However, income inequality has
risen across the globe since the mid-late 20th century, so this issue presses on with importance now more
than ever. It is critical to not only examine income inequality but analyze its effects on human
development as well.

This paper will uncover the relationship between human development and income inequality through
regressing several pertinent variables for both simple and multiple linear regression models. I hypothesize
that the relationship between human development and income inequality will be negative. My thought for
this relationship is that income inequality actually stunts human development as the greed of the few is
detrimental to the many. Income inequality can lead to improperly allocated resources, and education
programs, health centers, social initiatives, and more all come at a cost. These resources are the catalysts
for human development, so without them, a country’s human development is hindered.

II. Literature Review

Mo (2000) studies the effect of income inequality on the economic growth using the Gini coefficient to
represent income inequality and GDP growth to represent economic growth. In addition to those two
variables, several regression models include other variables such as population growth, average years of
schooling for adults over 25, ratio of private investment to GDP, political instability, and others. All
variables are percentage changes. In all the models, the coefficients for the Gini variable are negative

1
meaning that income inequality generally had a negative effect on economic growth. Of all the models,
income inequality had the largest impact on economic growth when only factoring in schooling years and
population growth; a one percent increase in the Gini coefficient signaled a .2162 percent decrease on
GDP growth. Mo (2000) concludes with the strong statement that income inequality significantly
decreases GDP growth. He goes further and states that income inequality has a direct negative effect on
the growth rate of productivity. He suggests that this may be due to an increased lack of trust among the
social classes. This distrust leads to higher transaction costs and lower gains from cooperation in addition
to many other reasonings.

Haseeb, Suryanto, Hartani, and Jermsittiparsert (2020) research the connection between globalization,
income inequality, and human development in Indonesia. To do this, they use a monthly data set from
1990 to 2016 and perform partial and multiple wavelet coherence across several periods ranging from 1-
month to 32-month time periods. Globalization is measured through “foreign direct investment and trade
… to the GDP ratio” for this study. The coherence between all three variables has significant power
regions in shorter time periods, specifically one to twelve months. Their overall analysis of the variables’
co-movements across several coherence models leads to several findings. First and foremost, the
researchers find that these co-movements change over different time periods. This signals that the short
run, medium run, and long run will all behave differently. Furthermore, increasing globalization will
increase income inequality. This increase in income inequality decreases human development in turn
which they suggest might be due to increased inequality in schooling and labor markets.

Mellor and Milyo (2003) examine the lagged effect of income inequality on public health. They analyze
results at both the individual and state-wide level on census data and Current Population Survey data. For
the individual analysis, they use a probit model with zero representing poor and one representing fair and
lagging the Gini data between five and nine years. Four of the five values had a negative sign indicating
that in increase in income inequality led to a decrease in health. However, these results were relatively
inconclusive as all values were statistically insignificant when holding individual characteristics and mean
state income constant. For state-wide analysis, they lag the Gini data 10 and 20 years across a 30-year
time frame. This is a multivariate analysis, and the dependent variables are mortality from cardiovascular
disease, malignant neoplasms, homicides, and accidents. When controlling for mean income, age
composition, and decade effects, the models suggest that there is a .53 percent increase and .69 percent
increase for having any of the four causes of death when lagging for 10 and 20 years respectively.
However, all these values are also not significant. This means that the effect of income inequality on

2
public health is limited to none, but the authors stress the importance of not putting too much weight into
these results as there are a large number of outside factors too.

There is quite a bit of research that examines the effects of income inequality on a wide range of factors
such as economic growth, public health, education levels, and more. However, this paper will look into
the effects of income inequality on a composite of these factors – human development. The limited
research into the relationship between these two variables is not on international data sets, so the trends
identified in those papers may not hold true for the rest of the world. Using an international data set
provided by the United Nations will allow a more holistic understanding of how income inequality affects
human development. Furthermore, this paper will include numerous secondary explanatory variables to
the models: poverty rate, population growth percentage, savings rate, political instability, and
Organisation for Economic Co-operation and Development membership. These additional explanatory
variables and expanded data sets will allow this paper to offer deeper insights into the relationship
between income inequality and human development as has not been done before.

III. Data
The data is collected from multiple sources: The United Nations Development Programme (UNDP), the
World Bank, the World Health Organization (WHO), and the Organisation for Economic Co-operation
and Development (OECD). All of the data is cross-sectional data, and five of the seven datapoints are
from 2018. The two datapoints not from 2018, the Gini coefficient and poverty rate, are collected from
the most recent available data in a span ranging from 2011 to 2018. This data has a sample size of 189,
and each observation represents a country. The data sources note that the following countries are missing:
Democratic People’s Republic of Korea, Monaco, Nauru, San Marino, Somalia, and Tuvalu. Only
missing 6 countries should not significantly skew the results since this is a minor portion of countries
(approximately 3%).

The dependent variable selected was the Human Development Index (HDI) from the UNDP data set. The
HDI is a composite measurement of three dimensions: health, education, and standard of living. This
measurement is for countries and yields a value ranging from zero to one with higher indices indicating a
stronger level of human development; in this paper, HDI has been multiplied by 100 so the values range
from 0 to 100 for cleaner analysis. For health, HDI examines life expectancy at birth. For education, it
examines mean years of schooling for adults older than 25 and expected years of schooling for children.
For standard of living, it examines gross national income (GNI) per capita; however, it uses the logarithm
of this value in order to highlight a “diminishing importance of income” (Mahbub ul Haq 1999). For all

3
three measurements, they are indexed and published under the names: health index, education index, and
GNI index. These three indices are indexed together to yield the final result – the HDI. This statistic is the
golden standard for measuring human development as it is commonly used across research because of its
quantitative nature.

The primary explanatory variable selected was the Gini coefficient. The Gini coefficient is a measure of
income distribution and represents income inequality; it ranges from 0 to 100. A value of 0 represents a
perfect distribution of incomes per household and a value of 100 represents perfect inequality. This
statistic was chosen as it is also very commonly used to measure income inequality. The slight drawback
is that it measures solely income, not wealth. A scatterplot, Figure 3.1, of the two primary variables
displays a negative and mild correlation between the two variables.

Figure 3.1

Additional explanatory variables were added in to support the models and strengthen the goodness of fit.
These are meant to help highlight a more accurate ceteris paribus effect of income inequality on human
development from the Gini coefficient and HDI respectively.

The first secondary explanatory variable is the proportion of the population below the international
poverty line. The international poverty line is $1.90 income per day for adults; it is set by the World Bank
and most recently updated in 2015. This is meant to emphasize the effects of poverty on human
development to support income inequality further. This variable is predicted to have a negative coefficient

4
since higher poverty levels intuitively would lead less access to education and healthcare and thus lower
human development. The second explanatory variable is the population growth as a percentage of the
country from 2017 to 2018. This year was selected since this matches the year of which the HDI data
comes from. This is meant to shed light on the effects of population growth on human development as
uncontrollable growth can lead to unprepared national infrastructure and government programs. Because
of these possibilities, this variable is predicted to also have a negative coefficient. The third secondary
explanatory variable is the gross domestic savings rate as a percentage of GDP and was chosen due to its
prominence as a consideration in well-cited literature. This variable is predicted to have a positive
coefficient since higher savings allows for stronger income security which in turn should boost GNI per
capita – a factor of HDI. The fourth secondary explanatory variable is a measure of political instability
from the Worldwide Governance Indicators. It is a normally distributed measure with a target mean of 0
and standard deviation of 1. This measure attempts to capture a country’s level of political stability and
absence of politically motivated violence. This variable is included as political instability is a factor of
theoretical human development, and it is predicted to have a positive coefficient as lower political
instability fosters human development. The fifth and final secondary explanatory variable is a dummy
variable to identify countries that are a member of the OECD. The country receives a 1 if they are a
member and 0 if they are not a member. This is to examine how income inequality may affect globalized
nations differently. OECD countries are predicted to have much higher HDI levels as they are developed
countries with strong international trade presences.
Variable Description Units Year Source
HDI Country’s Human Development Index Percentage 2018 UNDP
2010 –
Gini Country’s Gini coefficient Percentage UNDP
17
Percentage of population below 2011 –
povRate Percentage WHO
international poverty line 18
popGrow Population growth of country Percentage 2018 World Bank
Gross domestic savings as percentage
saveRate Percentage 2018 World Bank
of GDP
Country’s political stability and Standard
polInstab absence of politically motivated Normal 2018 World Bank
violence Distribution
Whether or not a country is in the 1 if member
OECD 2018 OECD
OECD 0 if not
Table 3.1

5
 Variable Observations Mean Std. Deviation Min Max
HDI 189 71.34 15.08 37.66 95.34
Gini 151 38.12 7.69 25 63
povRate 140 12.25 18.86 0 77.6
popGrow 187 1.29 1.15 -1.80 4.92
saveRate 161 20.14 16.75 -50.69 60.70
polInstab 188 -.08 .97 -2.99 1.54
OECD 189 .19 .39 0 1
Table 3.2

Table 3.1 displays a summary of the variables. Table 3.2 displays the variable’s descriptive statistics.

In order to find an unbiased estimation of the parameter coefficients in ordinary least squares regression,
the following assumptions were made with supporting evidence for the multiple linear regression (MLR)
models. MLR 1 – the model is linear in parameters. All models from section IV are linear in parameters,
so this assumption holds. MLR 2 – the model is based on random sampling. Data was collected from
every country possible, meaning no countries were purposely excluded. Therefore, this assumption holds.
The list of countries included can be found in Appendix A. MLR 3 – no exact linear relationships
between explanatory variables. All of the explanatory variables were carefully selected to ensure there is
no perfect collinearity. The table with all of the correlation coefficients between the variables proving this
fact can be found in Appendix B, so this assumption holds. MLR 4 – the error term has an expected value
of zero given any explanatory variable. This assumption is difficult to concretely prove true since the
exact population model is unknown. This paper will assume MLR 4 but consider the possibility of
misspecified models and unobserved factors in every regression analysis. MLR 5 – error has the same
variance given any explanatory variable. This assumption is also difficult to concretely prove true since
the exact variance of error is unknown. This paper will assume MLR 5 but consider the possibility of
heteroskedacity due to variance depending on unobserved factors beyond the explanatory variables
examined in the models. This completes the five Gauss-Markov assumptions. MLR 6 – the population
error is independent of the explanatory variables and normally distributed about 0. Several variables
actually break this assumption as HDI and Gini take on values only between 0 and 100 and OECD is
either 0 or 1. However, this paper will continue forward and assume MLR 6 in order to provide more
robust analysis while not forgetting the fact that MLR 6 is technically not satisfied. With MLR 1 through
MLR 6 having been considered, this satisfies the Classical Linear Model assumptions.

6
IV. Results
All of the following regression models have been created in STATA and can be found in more detail in
Appendix C.

Model 1
This is a simple linear regression model to show the ceteris paribus effect of income inequality on human
development without additional variables. The dependent variable is HDI and the explanatory variable is
Gini.
Model 1: 𝐻𝐷𝐼 = 𝛽0 + 𝛽1 (𝐺𝑖𝑛𝑖) + 𝑢
From the STATA output, the coefficients are as follows.
Estimated Equation 1: 𝐻𝐷𝐼 = 102.25 − .84 (𝐺𝑖𝑛𝑖)
As predicted, the coefficient on the Gini has a negative sign meaning that increased income inequality as
measured through the Gini coefficient causes a decrease in human development as measured through the
HDI. More specifically, this model can be interpreted as a 1% increase in a country’s Gini coefficient
leads to a .84% decrease in a country’s HDI. Here, the coefficient has a p-value of 0.00, so it has a
significance at the 1% level and even the .1% level. This means that the Gini coefficient is very likely to
at least have some effect on HDI. The nearly-zero p-value leads to a tight 95% confidence interval for this
coefficient: [-1.14, -.54]. This means that there is a 95% chance that the coefficient falls within this range
based on the simple regression. The intercept in this model of 102.25 may be interpreted as what the HDI
value would be in the absence of any income inequality – a 0 for the Gini coefficient. This is only
theoretical though as it is essentially impossible for a perfectly even distribution of income. While this
model uses 151 observations which is a strong amount of the total sample size (189), it only has an R-
squared value of .1687. This means that the model explains 16.87% of variation in HDI, so the correlation
between these two values is weak. This simple linear regression model is likely not fully explaining the
ceteris paribus effect of income inequality and human development due to missing additional key
explanatory variables. This can be a starting point for understanding the basic relationship between
income inequality and human development.

Model 2
This model is a multiple linear regression model. The dependent variable is HDI and the explanatory
variables are Gini, povRate, popGrow, saveRate, and polInstab.
Model 2: 𝐻𝐷𝐼 = 𝛽0 + 𝛽1 (𝐺𝑖𝑛𝑖) + 𝛽2 (𝑝𝑜𝑣𝑅𝑎𝑡𝑒) + 𝛽3 (𝑝𝑜𝑝𝐺𝑟𝑜𝑤) + 𝛽4 (𝑠𝑎𝑣𝑒𝑅𝑎𝑡𝑒)
+ 𝛽5 (𝑝𝑜𝑙𝐼𝑛𝑠𝑡𝑎𝑏) + 𝑢
From the STATA output, the coefficients are as follows.

7
 Estimated Equation 2: 𝐻𝐷𝐼 = 82.85 − .18 (𝐺𝑖𝑛𝑖) − .31 (𝑝𝑜𝑣𝑅𝑎𝑡𝑒) − 3.29(𝑝𝑜𝑝𝐺𝑟𝑜𝑤)
+ .23 (𝑠𝑎𝑣𝑒𝑅𝑎𝑡𝑒) + 5.88 (𝑝𝑜𝑙𝐼𝑛𝑠𝑡𝑎𝑏)
As predicted, the coefficient on Gini still has a negative sign along with povRate and popGrow. This
means that increased income inequality as measured through the Gini coefficient, increased poverty
levels, and increased population growth all have a negative impact on human development. More
specifically, this model can be interpreted as a 1% increase in a country’s Gini coefficient leads to a .18%
decrease in a country’s HDI. A 1% increase in a country’s poverty level leads to a .31% decrease in a
country’s HDI. Thirdly, a 1% increase in a country’s population growth leads to a 3.29% decrease in a
country’s HDI. It is interesting to note that the coefficient on Gini increased from Model 1, meaning that
the Gini coefficient may have less of an impact on HDI when there are other factors considered. On the
other hand, the coefficients on saveRate and polInstab both have a positive sign, meaning increased
savings rates and increased political stability lead to higher levels of human development. To specify, a
1% increase in a country’s savings rate leads to a .23% increase in a country’s HDI. Lastly, a 1-point
increase in a country’s political stability leads to a 5.88% increase in a country’s HDI. This coefficient is
so large since a 1-point increase on a normal distribution can be a significant increase in stability. If the
company moves from a rating of 0 to 1, the country is actually improving to be more stable than an
additional 34% of countries – a very large amount. Another interesting piece to note is that all of four of
the secondary explanatory variables have a p-value of 0.00. This means all four of these variables are
statistically significant at the highest level and therefore almost undoubtably have an effect on HDI. As
for Gini, t-statistic is -1.93, so it is significant on a two-sided test at the 10% level. The coefficient has a
.056 p-value, so it is nearly significant at the 5% level but not quite. This significance drops from Model 1
which may mean that income inequality has less of an impact on human development than initially
thought when several other factors are held constant. Additionally, the 95% confidence interval for Gini is
tighter but closer to zero than in Model 1 with it now being [-.36, 0.00]. This tighter interval may mean
that the exact ceteris paribus effect of income inequality on human development is closer to being
uncovered. Furthermore, this model uses 129 observations, less than Model 1, but still enough
observations for sufficient analysis. It has a much higher R-squared value of .7889 compared to Model 1’s
R-squared value of .1649. Again, this means that Model 2 explains 78.89% of variation in HDI, so the fit
of this model is strong. This multiple linear regression model is a step in the right direction to explaining
the ceteris paribus effect of income inequality and human development, but it may be overfitting the
model as the R-squared value is quite high.

8
A summary of all regression models can be found below in Table 4.1.

Dependent Variable: HDI

Explanatory
Model 1 Model 2 Model 3 Model 4
Variables

-.84*** -.18* -.03 -1.10

Gini
(.15) (.09) (.08) (.67)

-.31*** -.32*** -.32***

povRate
(.05) (.04) (.05)

-3.29*** -2.73*** -3.20***

popGrow
(.74) (.63) (.74)

.23*** .17*** .23***

saveRate
(.05) (.04) (.05)

5.88*** 3.65*** 5.67***

polInstab
(.84) (.78) (.85)

10.97***
OECD
(1.56)

.01
Gini2
(.01)

102.25*** 82.85*** 74.60*** 100.75***

Intercept
(5.92) (3.48) (3.18) (13.26)

Observations 151 129 129 129

R-squared .17 .79 .85 .79

Table 4.1
Significant at * 10% ** 5% *** 1%

9
V. Extensions
All of the following regression models have been created in STATA and can be found in more detail in
Appendix C unless otherwise stated.

Model 3
This model is a multiple linear regression model. The dependent variable is HDI and the explanatory
variables are Gini, povRate, popGrow, saveRate, polInstab, and OECD. OECD is a dummy variable
taking on the form 1 if the country is a member of the OECD and 0 if not.
Model 3: 𝐻𝐷𝐼 = 𝛽0 + 𝛽1 (𝐺𝑖𝑛𝑖) + 𝛽2 (𝑝𝑜𝑣𝑅𝑎𝑡𝑒) + 𝛽3 (𝑝𝑜𝑝𝐺𝑟𝑜𝑤) + 𝛽4 (𝑠𝑎𝑣𝑒𝑅𝑎𝑡𝑒)
+ 𝛽5 (𝑝𝑜𝑙𝐼𝑛𝑠𝑡𝑎𝑏) + 𝛽6 (𝑂𝐸𝐶𝐷) + 𝑢
From the STATA output, the coefficients are as follows.
Estimated Equation 3: 𝐻𝐷𝐼 = 74.60 − .03 (𝐺𝑖𝑛𝑖) − .32 (𝑝𝑜𝑣𝑅𝑎𝑡𝑒) − 2.73(𝑝𝑜𝑝𝐺𝑟𝑜𝑤)
+ .17 (𝑠𝑎𝑣𝑒𝑅𝑎𝑡𝑒) + 3.65 (𝑝𝑜𝑙𝐼𝑛𝑠𝑡𝑎𝑏) + 10.97 (𝑂𝐸𝐶𝐷)
This model differs from Model 2 in that a dummy variable OECD has been introduced. This dummy
variable allows for a comparison of regressions for countries that are a member of OECD against
countries that are not. In this model, the dummy variable has been brought in as an intercept change.
From the estimated equation, countries that are a member of the OECD have, on average, 10.97%
increase in HDI compared to other countries. This is quite a large shift, yet it makes sense as these
countries emphasize international cooperation and have strict standards for membership. Additionally, all
of the coefficients on the explanatory variables, except povRate, have a smaller magnitude than in Model
2 indicating that these factors have less of an impact when considering whether or not the country
embraces international cooperation to the extent of membership in OECD. All of the secondary
explanatory variables have p-value of 0.00, so they are all statistically significant at the 1% level.
However, the most interesting change in this model is OECD’s effect on Gini. The coefficient increases to
-.03 meaning that a 1% change in a country’s Gini coefficient leads to a mere .03% decrease in a
country’s HDI. On top of that, this coefficient’s standard error is .08, so it has a t-statistic of -.33. This
incredibly near-zero t-statistic leads to Gini no longer being statistically significant at any relevant level
which may signal income inequality’s limited to no impact on human development. This model has an R-
squared value of .85 which means a very strong goodness of fit.

Model 4
Now, this paper will examine an alternate functional form of Model 2. The change in functional form
comes in to play as having a term for the Gini coefficient squared. Gini2 is simply the country’s Gini
coefficient squared. This is to better examine the marginal effect of income inequality on human

10
development. Additionally, this new model will explore Kuznet’s hypothesis as countries in a later phase
of development have lower income inequality levels per Kuznet’s hypothesis. This model is a multiple
linear regression model. The dependent variable is HDI and the explanatory variables are Gini, povRate,
popGrow, saveRate, polInstab, and Gini2.

Model 4: 𝐻𝐷𝐼 = 𝛽0 + 𝛽1 (𝐺𝑖𝑛𝑖) + 𝛽2 (𝑝𝑜𝑣𝑅𝑎𝑡𝑒) + 𝛽3 (𝑝𝑜𝑝𝐺𝑟𝑜𝑤) + 𝛽4 (𝑠𝑎𝑣𝑒𝑅𝑎𝑡𝑒)

+ 𝛽5 (𝑝𝑜𝑙𝐼𝑛𝑠𝑡𝑎𝑏) + 𝛽6 (𝐺𝑖𝑛𝑖2) + 𝑢
From the STATA output, the coefficients are as follows.
Estimated Equation 4: 𝐻𝐷𝐼 = 100.75 − 1.10 (𝐺𝑖𝑛𝑖) − .32 (𝑝𝑜𝑣𝑅𝑎𝑡𝑒) − 3.20 (𝑝𝑜𝑝𝐺𝑟𝑜𝑤)
+ .23 (𝑠𝑎𝑣𝑒𝑅𝑎𝑡𝑒) + 5.67 (𝑝𝑜𝑙𝐼𝑛𝑠𝑡𝑎𝑏) + .01 (𝐺𝑖𝑛𝑖2)
In this model, all of the coefficients from previous models have consistent signs. However, the new
coefficient, Gini2, has a positive sign indicating that there is a diminishing marginal effect of income
inequality on human development. As income inequality grows larger and larger, its effect on human
development is supposedly not as strong. The R-squared value is .79 which equals the goodness of fit
from Model 2. However, one change from Model 2 is the fact that Gini and Gini2 have p-values of .102
and .164 respectively. This indicates that neither variable is significant at the 10% level, though Gini is
incredibly close. This lack of significance may be due to collinearity since the two variables have a
relatively high correlation. To test whether in fact these two variables in combination have any
significance at all on human development, this paper will perform an F-Test to test the joint significance
of Gini and Gini2.
H0: 𝛽1 = 0, 𝛽6 = 0
H1: H0 is false
The STATA output of the unrestricted model can be found in Appendix D. From this, the F-Statistic was
calculated to be 2.73. The critical value for F2, 129 at 10% is 2.30, and the critical value for F2, 129 at 5% is
3.00. The calculated F values is not within the rejection region for 5%, but it is within the rejection region
for 10%. Thus, we reject the null hypothesis at the 10% significance level. This means that the two
variables have a joint significance at the 10% level. Together, these measures of income inequality likely
do have some impact on human development – a new finding compared to a simpler t-test. Again, they
may have been separately insignificant due to multicollinearity, but this F-test allows us to examine them
without that concern.

VI. Conclusions
After analyzing all four models, our initial hypothesis was correct in that income inequality has a negative
effect on human development as the coefficients were all negative. However, it is important to consider

11
the fact that the Gini coefficient did not have statistical significance when the dummy variable to
represent countries that are a part of OECD was present. Also, without OECD but including a squared
term for Gini, there was not an independent significance on those terms, but there was indeed joint
significance for Gini and Gini squared after analyzing the results of the F-test. This may mean that across
all countries, there is a trend between income inequality and level of human development, but perhaps
when accounting for additional factors and holding them constant, that may lead to income inequality not
truly have a causal effect on human development. However, this fact is still uncertain.

On the other hand, all of the secondary explanatory variables had strong significance across all four
model. This indicates that poverty rates, population growth, savings rates, and political instability are
quite likely factors to be included in the population model for estimating a country’s human development
index. As expected, higher levels of poverty and population growth had an adverse effect on human
development, and high levels of savings and political stability had a beneficial effect on human
development. These variables were brought into the models to better uncover the ceteris paribus effect of
income inequality on human development. Since the coefficient on the Gini variable still varied from
model to model, yet the exact effect is still unclear.

After understanding the results of the models in this paper, there may still be future research that brings in
additional secondary explanatory variables. Factors like sustainability, social equality, and human rights
have not fully been accounted for, despite being a piece of human development for many prominent
researchers in the field. Additionally, it may be interesting to analyze how income inequality moves with
human development over time. A time series analysis could be utilized on similar data sets in order to
understand if the two variables have a relationship even over a long period of time. Overall, income
inequality is an issue that persists across all modern societies – both developed and underdeveloped –
while human development is a critical measure of how we can make the world a safer, more tolerant, and
more equitable place for all to live.

12
VII. References

“Gross Domestic Savings (% of GDP).” The World Bank, databank.worldbank.org/source/world-

development-indicators

Haq, Mahub ul. 1999. Reflections on Human Development. Oxford: Oxford University Press.

Haseeb, Suryanto, Hartani, and Jermsittiparsert. 2019. “Nexus Between Globalization, Income Inequality
and Human Development in Indonesian Economy: Evidence from Application of Partial and
Multiple Wavelet Coherence.” Social Indicators Research 147: 723–745.
doi:https://doi.org/10.1007/s11205-019-02178-w.

“Inequality Adjusted Human Development Indicator.” United Nations Development Programme,

data.un.org/DocumentData.aspx?q=inequality&id=423.

Kuznets. 1955. “Economic Growth and Income Inequality.” The American Economic Review 45 (1): 1 –
28. https://www.jstor.org/stable/1811581.

Mellor, Milyo. 2003. “Is Exposure to Income Inequality a Public Health Concern? Lagged Effects of
Income Inequality on Individual and Population Health.” Health Services Research 38 (1): 137 –
151. doi:https://doi.org/10.1111/1475-6773.00109.

Mo. 2003. “Income Inequality and Economic Growth.” International Review for Social Sciences 53 (3):
293 – 315. doi:https://doi.org/10.1111/1467-6435.00122.

“Political Stability and Absence of Violence/Terrorism.” The World Bank,

databank.worldbank.org/source/worldwide-governance-indicators

“Population Growth (Annual %).” The World Bank, data.worldbank.org/indicator/SP.POP.GROW.

“Proportion of Population Below the International Poverty Line.” The World Health Organization,
who.int/data/maternal-newborn-child-adolescent-ageing/indicator-explorer-new/mca/proportion-
of-population-below-the-international-poverty-line

13
VIII. Appendix
Appendix A: List of countries included in datasets.
Afghanistan Djibouti Lebanon Saint Kitts and Nevis
Albania Dominica Lesotho Sao Tome and Principe
Algeria Dominican Republic Liberia Saint Lucia
Andorra Ecuador Libya Saint Vincent and the
Angola Egypt Liechtenstein Grenadines
Antigua and Barbuda El Salvador Lithuania Samoa
Argentina Equatorial Guinea Luxembourg Saudi Arabia
Armenia Eritrea Madagascar Senegal
Australia Estonia Malawi Serbia
Austria Eswatini (Kingdom of) Malaysia Seychelles
Azerbaijan Ethiopia Maldives Sierra Leone
Bahamas Fiji Mali Singapore
Bahrain Finland Malta Slovakia
Bangladesh France Marshall Islands Slovenia
Barbados Gabon Mauritania Solomon Islands
Belarus Gambia Mauritius South Africa
Belgium Georgia Mexico South Sudan
Belize Germany Micronesia (Federated States Spain
Benin Ghana of) Sri Lanka
Bhutan Greece Moldova (Republic of) Sudan
Bolivia (Plurinational State Grenada Mongolia Suriname
of) Guatemala Montenegro Sweden
Bosnia and Herzegovina Guinea Morocco Switzerland
Botswana Guinea-Bissau Mozambique Syrian Arab Republic
Brazil Guyana Myanmar Tajikistan
Brunei Darussalam Haiti Namibia Tanzania (United Republic
Bulgaria Honduras Nepal of)
Burkina Faso Hong Kong, China (SAR) Netherlands Thailand
Burundi Hungary New Zealand Timor-Leste
Cabo Verde Iceland Nicaragua Togo
Cambodia India Niger Tonga
Cameroon Indonesia Nigeria Trinidad and Tobago
Canada Iran (Islamic Republic of) North Macedonia Tunisia
Central African Republic Iraq Norway Turkey
Chad Ireland Oman Turkmenistan
Chile Israel Pakistan Uganda
China Italy Palau Ukraine
Colombia Jamaica Palestine, State of United Arab Emirates
Comoros Japan Panama United Kingdom
Congo Jordan Papua New Guinea United States
Congo (Democratic Republic Kazakhstan Paraguay Uruguay
of the) Kenya Peru Uzbekistan
Costa Rica Kiribati Philippines Vanuatu
Côte d'Ivoire Korea (Republic of) Poland Venezuela (Bolivarian
Croatia Kuwait Portugal Republic of)
Cuba Kyrgyzstan Qatar Viet Nam
Cyprus Lao People's Democratic Romania Yemen
Czechia Republic Russian Federation Zambia
Denmark Latvia Rwanda Zimbabwe

14
Appendix B: Correlation coefficients between each variable.

HDI Gini povRate popGrow saveRate polInstab OECD

HDI 1.00

Gini -.42 1.00

povRate -.77 .40 1.00

popGrow -.64 .36 .62 1.00

saveRate .42 -.13 -.29 -.03 1.00

polInstab .63 -.19 -.39 -.35 .23 1.00

OECD .69 -.38 -.36 -.35 .30 .52 1.00

Appendix C: STATA regression outputs for all models.

Model 1

15
Model 2

Model 3

16
Model 4

Appendix D: STATA output for unrestricted model

17