1. Introduction

Income inequality has long been a topic of research in the fields of economics. In an economy, income can be generated in production by factors of production including land, labor, capital and entrepreneur. Meanwhile, income will be distributed to different firms and industries [1].

However, the distribution of income generated by these factors of production is far from even. In different firms and industries, the income-earning capabilities could differ significantly. For instance, high-tech industries with abundant capital and advanced technology tend to generate much higher incomes compared to labor-intensive industries. This disparity in income within and between firms and industries gradually accumulates, leading to the issue of income inequality. In addition, on the perspective of society as a whole, income inequality could lead to serious consequences such as poverty, higher crime rates, poorer health level and so on, which make it vital to tackle [2].

Income inequality can be shown by Gini coefficient, a concept introduced by Italian statistician Corrado Gini in 1912, it is a measurement of income inequality which uses numbers between 0 and 1 inclusive to reflect the degree of income inequality in an economy. For example, a Gini coefficient of 0 represents perfect equity, meaning that the income is equally distributed in an economy. Meanwhile, a value of 1 represents maximal inequality [3].

The following paragraphs listed three possible factors of income inequality: inflation, demographic changes, and life expectancy. Firstly, inflation represents a general rise in price level in an economy. Inflation reduces the purchasing power of money. When prices surge due to inflation, workers' real wages are likely to decline if their nominal wages do not the same pace with the rising price. Low-income workers are often more vulnerable in this situation. They usually work in jobs with less bargaining power, such as those in the service or unskilled labor sectors. To compare, high-income earners, especially those that are highly demanded, may be more successful in getting more wage to offset inflation. This can widen the income gap between low and high-income groups [4]. Past researches have shown similar results. Bulír found that inflation generally increases income inequality, with a more significant impact in hyper-inflationary countries, and the effect of low inflation on improving income equality is more prominent in low-and high-income countries [5]. Monnin showed that there is a U - shaped relationship between long - run inflation and income inequality in developed economies, where inequality decreases as inflation rises until it reaches about 13% and then increases [4].

Secondly, rapid population growth, especially in developing countries, leads to a significant increase in the labor supply, and when the growth of the labor force exceeds the job vacancies, oversupply of labor occurs [6]. This drives down wages for low-skilled labor, widening the income gap. Immigration can also impact income inequality. Immigrants may take low-skilled and low-paid jobs in host countries, reducing job opportunities for domestic labors. In past researcher, Odusola et al. discovered that whether demographic changes affect income inequality remained ambiguous. At the bivariate level, variables like higher fertility rates and population growth rates are associated with lower inequality. However, at the multivariate level, when control variables are introduced, the population variables often have no effect on inequality [7].

Thirdly, life expectancy is the average number of years a person is expected to live. When life expectancy is longer, individuals are more likely to invest in long-term education and skills training. Researchers have investigated the impact of life cycles on incomes. Strong associations between infant mortality rate and income inequality are shown [8]. Chetty et al. found that higher income was associated with greater longevity throughout the income distribution and inequality in life expectancy increased over time. The reasons behind this are mainly due to health issues like smoking and obesity [9].

To further explore the complex issue of income inequality, this paper will use the method of multiple linear regression to identify the key factors that influence this economic phenomenon. Through this analysis, a deeper understanding of the mechanisms behind income inequality in different economic contexts can be gained, providing a basis for coming up with more targeted and effective policies to address this persistent economic and social problem.

2. Methods

2.1. Data Source

The datasets used in this paper are available on the Kaggle website. The first multiple linear regression involves the use of three datasets, focusing on China, and the second one involves six, expanding the subject to the international scale.

2.2. Variable Selection

The datasets contain a large amount of data, those that are selected mainly divide into three categories: demographic trends (net migration, change in population), population health (birth rate, life expectancy), and economic indicators (Gini coefficient, change in Gross domestic product (GDP), inflation rate). The list of variables and their definitions are shown below in Table 1.

Table 1: List of Variables

2.3. Method Introduction

This paper uses multiple linear regression to explore the factors influencing income inequality, using the Gini coefficient as a measure. This method analyses the relationship between a single dependent variable and various independent variables [10]. The formula of multiple linear regression is:

\( Y={β_{0}}+{β_{1}}{X_{1}}+{β_{2}}{X_{2}}+{β_{3}}{X_{3}}+…+{β_{n}}{X_{n}}+ϵ \) (1)

Where \( {β_{0}} \) is a constant term, X is the variable and ϵ is the error term.

This paper is going to conduct two multiple regression analysis. The first one focuses on China, covering the period from 2001 to 2020. The second one expands the scope to the entire world. 30 developed countries in 2020 are selected for the study. The purpose is to investigate whether different results will emerge compared to the analysis of China.

3. Results and Discussion

3.1. Linear Regression Model for China

The relationships among variables are also investigated. As shown in Figure 1 below. Figure 1 shows the correlation between variables. According to the variable correlation heatmap, birth rate is highly related to percentage increase in population and inversely related to life expectancy. However, it can be noticed that the correlation coefficient of all variables listed are approximately 0, indicating that each of them are not strongly correlated to Gini coefficient.

Figure 1: Variables Correlation Heatmap

Given these low individual correlations, it is hypothesized that combining these variables might produce a different result. Therefore, linear regression is going to be conducted. This analysis aims to explore whether the combined effect of these variables on the Gini coefficient is more significant than their individual effects.

Table 2: Regression Coefficient Table for China

Note: * represents 0.1 significance, ** represents 0.05 significance level

Table 2 shows the regression coefficient of all variables. The p values for X₁, X₂, X₄, X₅ and X₆ do not exceed 0.05, indicating that they are in close relationship with the Gini coefficient. Variables with non-zero coefficients show that some factors exacerbate income inequality while others mitigate it. For instance, X₁ and X₅, with positive coefficients, suggest that increase in net migration rate and population growth are associated with a rise in the Gini coefficient. On the contrary, X₂ and X₆ have negative coefficients, meaning that a higher birth rate and longer life expectancy are related to a lower Gini coefficient.

The \( {R^{2}} \) value is 0.659, indicating a 65.9% level of fitting. Base on the data, the multiple linear regression equation is shown:

\( Y=83.204+6.580{X_{1}}-2.508{X_{2}}+0.890{X_{3}}+2.738{X_{4}}+337.297{X_{5}}-0.693{X_{6}} \) (2)

3.2. Linear Regression on International Level

Although the regression analysis of China's data has listed several possible factors of income inequality, it is insufficient to rely solely on this result to summarize the influencing factors. Therefore, another linear regression model will be constructed using data from 30 developed countries. The results from the two models will be compared to see if the previous findings are applicable globally.

Figure 2: Pearson Correlation Chart

Figure 2 shows the Pearson Correlation Chart of each variable. From it, net migration and life expectation has the main contributions to changes in Gini coefficient with correlation coefficient of approximately 0.4. While others show a relatively lower correlation with Gini coefficient.

Table 3: Regression Coefficient Table for Selected Countries

Note: ** represents 0.05 significance level

Table 3 shows the regression coefficient of all variables for the selected countries. Base on the data, the multiple linear regression equation is shown:

\( Y=0.824+0{X_{1}}-0.003{X_{2}}+0.005{X_{3}}-0.008{X_{4}}-3.987{X_{5}}-0.004{X_{6}} \) (3)

Only variable X1 has p value smaller than 0.05. This indicates that in the international sample, fewer variables have a significant impact on income inequality. According to the Beta value, only X3 shows a positive relationship with the change in Gini coefficient, which is different to the previous result. Additionally, the \( {R^{2}} \) value of 0.526 is lower than that of the Chinese model, suggesting that the model is less able to explain the variation in the Gini coefficient in this international context.

3.3. Comparison Between Two Models

For China, the variable correlation heatmap presented earlier indicated that there were no significant individual relationships between each variable. However, the results from the regression model shows that the combined effect of these variables on the Gini coefficient is substantial. This finding shows that seemingly unrelated variables can interact and jointly impact economic phenomena.

While comparing the two models, the one for China not only shows a stronger relationship between Gini Coefficient and each variable, but also a higher level of fitting. One possible reason for this difference is that the international model does not include several other crucial factors. In developed economies, factors such as technological innovation, globalization, and tax policies might have a more significant impact on income inequality, and these were not considered in the current model.

4. Conclusion

This study employed multiple linear regression to explore the factors influencing income inequality, using data from China from 2001 to 2020 and 30 developed countries in 2020. The findings reveal significant differences in the relationships between variables and income inequality in the Chinese and international models.

In the Chinese model, despite the low individual correlations between variables and the Gini coefficient as shown in the correlation heatmap, the regression results indicate that multiple variables, including net migration rate, birth rate, inflation rate, percentage increase in population, and life expectancy, are closely associated with the Gini coefficient. In contrast, in the international model, only the net migration rate has a significant impact on the Gini coefficient. The lower R² value suggests that the model is less capable of explaining the variation in income inequality compared to the previous one. Therefore, what is observed in the Chinese context may not be true on the international level, and vice versa.

Nevertheless, the study has limitations. High Variance Inflation Factor (VIF) values for many variables in the Chinese regression model, such as birth rate and life expectancy, imply strong collinearity among independent variables. This collinearity reduces the accuracy of assessing each variable's contribution. Moreover, the economic theories used rely on idealized assumptions such as perfect competition, overlooking real-world complexities such as information failure and monopolies. As a result, there is a difference between the theoretical analysis and real-world situations.

Future research could focus on addressing the collinearity issue. For example, by using more advanced statistical techniques or selecting variables more carefully. Additionally, incorporating real-world complexities into the theoretical work would enhance the understanding of income inequality and its underlying economic mechanisms, contributing to the development of more effective policies to mitigate it.

In conclusion, while this study has laid a foundation for understanding income inequality, there is still much more to be done. Through these findings, researchers can contribute to the development of more comprehensive and effective strategies for reducing income inequality both in China and globally.