Effect of Corruption, Inequality, and Unemployment on Poverty in South African Perspective: Evidence from Quantile Regression Model

1. Introduction

“Do not accept a bribe for a bribe blinds those who see and twist the words of the innocent” Holy Bible, as cited by (Šumah 2018). Corruption is an issue affecting the world’s economies and institutions, and it influences poverty in emerging markets. Some studies have established its relationship with poverty (Salahuddin et al. 2020; Justesen and Bjornskov 2014). Corruption may divert resources intended to help the poor alleviate poverty. The corruption committed by government officials may fall disproportionately on the poor, thereby amplifying poverty.

South Africa is the victim of a high level of poverty, unemployment, and high inequality. Therefore, the objective of this study is to evaluate the effect of corruption on poverty in South Africa. Evaluate the role of inequality measured by the Gini coefficient on poverty. Lastly, to evaluate the impact of unemployment on poverty in SA.

At the beginning of 2025, more than 800 million people worldwide live in poverty or below the international poverty line (Alfani et al. 2021). This gives a glimpse of COVID-19 role in amplifying poverty; however, at the same time, some authorities were driven out of power because they misused funds that were supposed to help people (Mudau 2022). The poverty has its roots in rural areas in Africa, (Alfani et al. 2021) noted that Africa hold two-third of the world poorest population (Silwal et al. 2020). In the same demographic, 40% of those people live in transitory poverty, 60% of whom are chronically poor, especially in remote locations, whereas in urban areas, more than 50% of those living in poverty are children under the age of 15 (Beegle and Christiaensen 2019).

In the history of colonization, developed economies colonized less-developed economies with knowledge, slavery, and trade (Acemoglu and Robinson 2017, Ocheni and Nwankwo 2012).

Today Africa hold two-third of the world poorest population (Silwal et al. 2020, Alfani et al. 2021). Some factors contributing to poverty include various forms of corruption by government representatives. For example, the average unemployment rate has not changed since the post-apartheid era began in 1994, and it has been recorded above 30% after COVID-19. Food Poverty stood at 17.6% following the 2023 revision of the poverty lines (Stat SA 2024).

South Africans imagined the post-apartheid period as a game changer, where blacks and whites owned assets and wealth equally. However, white people remain in the top positions Human Right Commission (2023) based on assert ownership (see Figure 1 below). Simultaneously, blacks remain highly dependent on the government’s social security benefits. John (2021) validated that corruption remains systematically intact in societal structures.

Figure 1. Assert Ownership according to Races.

Source: Statistics South Africa: the figure shows average asset score between 2009 (in blue) and 2015 (in black). It shows a trend, blacks are at the bottom end, and whites on top end. The vertical axis, shows the population group, horizontal axis shows the values.

Therefore, Figure 1 can be interpreted in two ways concerning poverty and corruption. This simply means that assertions based on white people remain distributed in the same channel even in the post-apartheid period. This means that black people remain the majority living in poverty, especially absolute poverty. Second, policies such as BBBEE aimed at uplifting black people have triggered a highly unequal society among blacks (Koelble 2022). Those in power exercise corruption to attain public assertions, while those living in poverty remain in the cycle of poverty.

The Palma ratio compares the income expenditure of the 10% richest population divided by the poorest 40% of the population, as indicated in Figure 2 below. In 2015, the top richest people spent 7.9 times more than the bottom 40% population. This means that the income for the richest 10% could be distributed among the poorest 40% of the population, thus making them wealthier. The year 2015 marked 21 years of democracy and fell under the NDP program towards 2030.

Figure 2. Palma ratio.

Source: Statistics South Africa: in the figure there is definition of Palma ration, figure of ten individuals representing the total population of SA, the blue guy represent the richest 10% of the population, the grey individuals sample indicates middle income, and the marron individuals represent bottom lower income 40% population.

Education is well established in South Africa, but graduates produced annually are not equally absorbed in the labor market (Macginty 2024, Mseleku 2022, Graham et al., 2019). Moreover, it is becoming normal for people to pay bribes or sexual enslavement to attain employment systematically (Ntshangase 2024).

Justesenb and Jørnskov (2014) argued that poor people, through their reliance on public services, are more likely to pay bribes. This means that poor people in high societal positions are more likely to be greedy in satisfying themselves first. In this way, the poor remain poor, which could also be a reason they have been emerging for the killing of ward councillors, especially in KZN province (Nomarwayi et al. 1938, Matamba and Chwayita 2024). People perceive that at a lower level of government, they are at the forefront of using state funds for their benefits.

The literature provides insights into the connection between corruption and poverty. For example, Salahuddin et al. (2020), Rahayu (2012), Cabral (2017), Oliveira da Silva et al. (2022) and Eshun and Baah (2019) find that corruption causes poverty. On the other hand, Riley and Chilanga (2018), Adebayo (2013), and Ünver and Koyuncu (2018) indicated that poverty is the cause of corruption, contrary to earlier studies. Therefore, this study investigates the effect of corruption on poverty from the South African perspective.

Most studies relied on mean-based estimators that assume a homogeneous effect between the variables of interest. On the other hand, heterogeneity in the determinants of poverty has received limited attention in the literature. To differ from the other studies that focus on homogeneous effects of corruption on poverty through the use of mean-based approaches. The study look at poverty using quantile regression that put the dependent variable into different quantiles, from lower, middle and higher quantile. Moreover, individuals’ levels of deprivation differ; for example, there are low-, middle-, and high-income individuals.

The South African literature raises the same issues; however, Vinutali and Saba (2024) excluded the influence of unemployment from their analysis of the effect of corruption on poverty. Salahuddin et al. (2020) used time-series data from 1991-2016; the study does not separate the possible influence of the regime change in 1994, when democracy was introduced. Jonck and Swanepoel (2016) used the cross-sectional data. Therefore, this study investigates the effect of corruption on poverty from the South African perspective. Corruption may exist within the system, making it weak and fueling poverty. The study is interested in poverty as a daily deprivation. Poverty can be measured objectively using a poverty line to show the number of people living below a certain threshold. Poverty may also serve as an indicator of deep corruption within a certain economic structure; for this reason, SDG 1 states that “end poverty in all its forms”. Therefore, this study examines poverty in the form of corruption or corruption-induced poverty.

The remainder of this paper is structured as follows: the second section covers the literature review, followed by the methodology, discussion of the results, and last section covers conclusion.

2. Literature review

The Pareto Economic Disability (PED) theory posits that poverty is the product of the dysfunctional structure of the economy that allows for corruption to persist (Adeleyi 2016). The theory presents a case for bidirectional causality that may run in both directions, from both poverty and corruption. Koechlin (2022), underlines that instruments to fight corruption, such as the rule of law, transparency, accountability, and integrity, are preconditions for sustainable development. That means corruption reduction may lead to less poverty. On the other hand, Chetwynd et al. (2003) note that economic models indicate that poverty may result from economic growth that is hampered by corruption. That means corruption may divert resources intended to fight poverty. The theory under consideration will be used to address the first objective, namely: to evaluate the effect of corruption on poverty.

The connection between poverty and inequality is explained by the poverty-growth-inequality theory. The theory asserts that poverty emanates from changes in economic growth and inequality. According to Oktay and Algan (2022), income inequality exacerbates poverty in one form or another. The theory in question will address the second objective: to evaluate the impact of inequality on poverty in SA.

The connection between poverty and unemployment is put forward by the Keynesian school of thought. According to Davis and Sanchez-Martinez (2014), poverty is mainly caused by unemployment, especially if the economy is underperforming. The theory also put forweard the role of the government in providing public goods and stimulating the economy to boost employment. The theory in question addresses the study’s final objective: to evaluate the impact of unemployment on poverty in SA. The following discussion explains the definition of poverty and corruption.

Chetwynd et al. (2003) clusters poverty definitions into four groups: income poverty, material deprivation, Sens Capability deprivation, and multidimensional deprivation. Income poverty refers to the income or expenditure required by an individual or household to meet their basic needs. Material poverty, on the other hand, could be attributed to a lack of household needs that, in the South African case, are provided by the state, such as access to water, shelter, education, health, and other services.

Capability deprivation is the lack of growth-related factors that improve an individual’s ability to meet his or her needs. The high unemployment rate in SA is one example of how people are deprived of employment.

The last definition put forward is that poverty is multidimensional (Kakwani and Son 2025). This definition entails that poverty varies from one person to another because they may lack different things. Some may suffer from education poverty, health poverty, income poverty, and others. The Apartheid system in South Africa was another example of a poverty-inducing system that people experienced. Therefore, it challenges researchers to examine poverty across different dimensions, measures, causes, and consequences. The four definitions of poverty will be used to explain the connection between the variables in the results section.

The following views of poverty are derived from the classical school, which views poverty as a social problem that requires society to confront it. Keynesian theory further outlines the need for government intervention through fiscal policy, in which the rich are taxed, and income is redistributed to lower-income individuals to close the poverty gap. (Strauss 1957) have noted that, during the redistribution process, corruption may emerge to divert resources for personal gain, thus amplifying poverty.

Šumah (2018) derives corruption from the word corruptus which means corruption, abuse of the high-power position for personal gain. This is normally related to the abuse of funds by trusted high-profile officials for their gains, influenced by greed, unethical conduct, and other weakening factors of the system. Aidt (2003) lists several elements of corruption, such as efficient corruption, where the transaction between the two parties that should run from one party (official) to the other (beneficial) ends up benefiting a third party that was not supposed to benefit. For example, Šumah (2018) highlights the issue of bribery as an open employment post ends up being given to an individual who bribes the employer instead of hiring the best candidate. The second element is when non-beneficiaries are given the power to make final decisions; this is called the principle of benevolent corruption. The third element involves the non-beneficial principle of corruption, meaning that officials are supposed to distribute resources for public benefit. Instead, they introduce policies that permit them to abuse the private sector by extracting funds illegally from them. For example, selling public goods is characterized as community goods for all. Corruption inertia is the fourth element, which means that corruption occurs over time because of spill-over from one office to another.

Šumah (2018) argues that corruption is amplified by a lack of professional ethical practice and corruption, which in turn harms economic growth, investment, job creation, and assistance programs meant for the poor. It reduces the tax income for the state at the same time that state officials have a way to operate above the law they formulated and, lastly, the quality of education.

Both poverty and corruption are multidimensional and difficult to measure. According to Yang (2017), poverty may be measured through income. Chetwynd et al. (2003) underscore that the burden of corruption falls on the poor and state that it diverts resources. In that manner, the poor bear the burden of external costs resulting from corruption. This means the system meant to help the poor is like a transaction between two parties (the system and the poor). Then an agent (officials), through corruption, steps in as a second party (beneficiary), making the poor a third party that suffers from deprivation. The discussion above is based on one side of the theory that poverty is driven by corruption, and it will also inform the specification of the model.

Empirical literature begins with a relationship that involves the causal effect between corruption and poverty. There is limited evidence in the broad literature and South African context. Considering the following example, Salahuddin et al. (2020) investigated the nexus between the two aforementioned variables in South Africa. The secondary data from 1991 to 2016 were regressed using the ARDL model. The study found that corruption amplifies poverty in the long run, whereas globalization reduces it. This study provides a glimpse of the advantages of globalization in the economy. Similarly, Rahayu (2012) noted that corruption, bribery, fraud, and nepotism become daily life experiences. The study found that corruption is the root cause of poverty in the economy and not vice versa. The study based its findings on the Granger causality test between variables over time. The study follows the same pattern, treating poverty as a dependent variable. Therefore, the aforementioned studies above guide the selection and specification of the model for this study.

Based on the primary data from the Lilongwe area, the study undertook in-depth interviews were conducted. It found that the incidence of food insecurity in the region is mainly caused by corruption the author named it the “political culture” following the cash-gate scandal revealed in 2013 September (Riley and Chilanga, 2018). In Nigeria, Adebayo (2013) shared similar sentiments and corrupted proper insults. The primary data analysis provides another angle on how corruption exacerbates poverty from the perspective of the research informants.

In the USA, Dincer (2008) notes that corruption causes both poverty and inequality. The study relied on various measures of inequality, such as the Gini index, standard deviation, coefficient of variation, and Atkinson index, while poverty was sourced from the US Census Bureau. Justesenb and Jørnskov (2014) argued that poor people, through their reliance on public services, are in the trap of paying bribes. Public servants are equally confident that people will pay bribes because they are a monopoly of the services. The study’s findings were derived from a multiple regression model, and the data were extracted from 18 different economies.

Other findings confirm that poverty causes corruption. For example, Pierre (2020) indicates that corruption, slavery, international aid, and unemployment are the root causes of poverty in Haiti. On the other hand, Mantzaris and Pillay (2019) validated that corruption equally harms economic growth and international trade. This study is based on the fact that the loss of a good leadership style, ethical conduct, group greediness, and political institutions are root sources of corruption. Contrary to Šumah (2018), who in the theoretical discussion indicated that religious organizations where protestants dominate have less corruption stimulus packages. Ayub (2013) added that not only is economic growth negatively affected by corruption, but community resources are also diverted, as the theory highlighted earlier. Resources highlighted by the author include education, health, electricity, water, and sanitation. All these resources are valid in the current study for living income. Vahideh et al. (2010) indicate that poverty and corruption have bidirectional causality. The study used the panel GMM and data stretched from 1997 to 2006, and poverty was measured using the Human Poverty Index (HPI). The studies above have identified a causal effect running from poverty to corruption. The study, however, only focuses on how corruption causes poverty, because the central theme of this study is poverty. The objective is to explain how corruption-inducing poverty is informed by SDG 1.

Other studies validated the relationship between corruption and income inequality. For example, Ullah (2022) indicated that corruption has a positive effect on inequality because income often goes to government officials, who are expected to distribute wealth evenly. The study uses the GMM approach and indicates the robustness of the results obtained, irrespective of changing the equation specification. Regarding poverty and unemployment, most studies in the literature concur with Keynesian theory, which holds that unemployment causes poverty. The studies, such as Ngubane et al. (2023), in South Africa through the NARDL model, Gelle et al. (2021), used OLS and Poisson correlation approach in Somalia, Surawan (2022), through the use of the OLS model in the Sumbawa agency. The studies outlined above informed the selection of the unemployment rate as a control variable in the study.

3. Methodology

The following discussion describes the four important building blocks of a model: the nature of the data, theoretical model, empirical model, and diagnostic tests. The study relies on secondary data that consist of numerical values to make economic sense through a quantile regression model. This means that the study relies on the positivism approach, in which quantitative models are used to display certain meanings displayed by the data.

4. Discussion of the results

The discussion of the results begins with the analytical behavior of the variables used in the study. As previously indicated, the nexus between poverty and corruption is indicated by the control of corruption in the economy, as ranked by the World Bank. The control of corruption in Figure 1, located in the upper-left corner, shows a downward slope up to 2012. In the preceding periods, it remained constant and decreased dramatically in other periods such as 2014, 2018, and 2020. The most notable point A is explained by the changes in the government structure in South Africa, which is cited as the most corrupt in the post-apartheid regime (Georgieva and Krsteski 2017). On the other hand, point B highlights corruption related to the abuse of state funds, such as the provisioning of food parcels and the Department of Health (John 2021).

Poverty on the same Figure 1, but on the upper-right corner, depicts a downward trend from the beginning of the period up to 2010. This trend indicates that in the post-apartheid period, all policies were like spears set to destroy poverty, especially among Black people. Following the global financial crisis, this trend started to increase slightly from 2011 to 2018. The number of people living in poverty has been equally amplified by the COVID-19 pandemic due to many factors, such as unemployment. For example, Ngubane et al. (2023) noted that poverty increases irrespective of whether unemployment increases or decreases, and the study was undertaken using the NARDL approach.

At the bottom right of Figure 3, the Gini coefficient shows the degree of inequality. Before 2010, the level of inequality hovered between 0.70 and 0.80. This indicates that it took time for the post-apartheid system to reduce inequality in the SA. Prior policies, such as RDP, GEAR, and ASGISA, failed to reduce the level of inequality in the economy (Jarstad 2021). Equally important, in the post 2010 period the inequality level remained at approximately 0.60. However, in other periods, such as 2010 and 2011, inequality further decreased dramatically, but it failed to remain at a low level in the long run. This could be explained by the fact that the periods around 2010 were concentrated by high capital formation in preparation for and during the FIFA World Cup. During this period, there was ongoing labor-intensive construction, the transport industry benefited equally, and the overall tourism industry opened jobs and entrepreneurial opportunities.

Figure 3. Data sample of the variables.

Most data on the unemployment rate for South Africa date from 2000 onwards; in the same year, it appears that it was hovering below 24%. This Figure looks similar to that of the second quarter of 2020 during the pandemic, which decreased to 24%, showing a promising trend. However, one questions the issue of unemployment decreasing by 5% between the first and second quarters of 2020. This leads back to the definition of the unemployment rate, which is the percentage of people aged between 15 and 64 years who are actively looking for employment and cannot find one. The important word in the previous definition is “actively” and it can be used to explain the unemployment in 2000 and 2020Q2. Lack of education and high concentration of black people in informal employment for unskilled and semi-skilled labor made the percentage of people who were actively looking for employment look lower in the Figure but remained very high in the population. Concerning informal employment, consider that a person looking for a survival job in rural areas, such as construction, is normally done by people considered unskilled. This type of employment is not accounted for by the formal unemployment rate.

The same issue could relate to 2020Q2, where people were not allowed to move, look for jobs, or operate business. Therefore, the unemployment rate decreased because of the lockdown. However, the GDP growth decreased exponentially. Unemployment increased sharply after the lockdown, when people were allowed to move around and search for lost jobs. Ngubane et al. (2023) indicated that employers in South Africa are reluctant to hire, even when economic activities are in the right condition.

The second part of the results briefly outlines the descriptive statistics table indicated in Table 1. It consists of measures of the central tendency, dispersion, and total number of variables. The percentage of corruption ranges from -2.5 (poor control of corruption) to +2.5 (highly effective control of corruption). The minimum and maximum values for corruption in South Africa range from -0.42 and 0.86 respectively. The highlighted values indicate that the positive values of poverty control are less than 1 and are not extreme values desired by society at large. The median value is 0.06, whereas the mean is 0.13, indicating that dealing with corruption in South Africa is not at the heart of the state. John (2021) indicated that the media has underreported the incidence of corruption. The standard deviation from the mean is (0.25) higher than the mean, indicating that the data were not stationary. The kurtosis of the data is Platykurtic because it is less than three, which indicates the spread of the data away from the average value or point of asymmetry.

Moving to the next variable of interest, poverty spreads are skewed to the left (lower values). This is shown by the median value (36) being lower than the mean value (38) and closer to the lower end (32). However, the data according to kurtosis is mesokurtic, close to 3. This could be the case because poverty is represented by food poverty, and in SA, fewer people go to bed without food (thanks to a social security grant). Moving to the control variables, the Gini coefficient, and unemployment. The former ranged from 0 (highly equal society) to 1 (highly unequal society). Therefore, it is clear that 0.5 is moderate, but values below it are highly desirable for most economies. On the other hand values above 0.5 are not desirable; in SA, the mean is 0.71 and the median is 0.7. This means that South Africa is a highly unequal society (Koelble 2022), with the majority of people remaining unemployed, social grant dependent, and living in rural areas without adequate infrastructure. The data are playtokurtic a few points away from three; however, the standard deviation is very small at 0.05, showing that most values hover around the average value.

The unemployment rate mean (25), median (24), and standard deviation (2.87) do not best describe the behaviour of the data. In this case, the average value could be related to the natural rate of unemployment in South Africa and the data are equally positively skewed. However, kurtosis and skewness may be relevant to this juncture. For example, the data are leptokurtic (5) away from 3, meaning that the data are highly peaked. Society is unwilling to experience a high unemployment rate. This means that the South African democratic system, economic system, political system, technological development, and other sectors tend to favor unemployment instead of making it an enemy.

The stationarity test results are shown in Table 2. It is important to trace the nature of the data used in the study and access its behaviour through an accredited pre-test. The Gini coefficient is stationary at levels because, in the descriptive statistics table, it has to mean reverting values over time. Unlike the remaining three variables (poverty, corruption, and unemployment), these were found to be stationary after taking the first difference. This is because the data for the latter variables show different trends over time, and can easily respond to shocks. This is an extension of simple OLS (see the methodology section).

Moving to the core results of this study. The results in Table 3 were divided into three parts. The first part indicates the results at the 25^th quantile, followed by those of the 50^th or median quartile, and finally those of the 75^th quantile. In the first quantile, a 1% increase in corruption led to a 0.19% increase in poverty, and the effect was statistically significant at the 5% level. The results support the PED theory, which posits that corruption influences poverty (Adeleyi 2016). Corruption undertaken by high-profile people indirectly affects those living below the food poverty line. The findings of this study are in line with those of Rahayu (2012), Ikharehon (2019), and Salahuddin et al. (2020). However, if corruption is undertaken by local government officials, it can directly affect people living below food poverty levels. For example, during the COVID-19 pandemic, food parcels were issued to poor families because the R350 grant was not sufficient to afford household food (John 2021). Therefore, the distributors were only concerned with people they were close to them, such as relatives, friends, and friends. Therefore, poverty is defined as the deprivation of basic needs and a state of being voiceless or deprived of important social decisions. The same condition occurred because poor people were not represented, and nobody stood and spoke on their behalf to obtain adequate food. The Gini coefficient has a greater coefficient in all quantiles than the coefficients that represent corruption and poverty. A 1% increase in the Gini coefficient leads to a 0.22% increase in poverty; however, this effect is statistically insignificant. Equally important unemployment has an insignificant effect on poverty, as indicated by (Dahliah 2021). The constant term indicates that if all variables are held constant, poverty increases by 3.36% and is statistically significant. This means that poverty remains positive because of the economic structure, education, and other factors that contribute to poverty in South Africa.

In the 50^th quantile, all the variables are statistically significant. A 1% increase in corruption led to a 0.29% increase in poverty; the results are consistent with PED theory. This means that corruption is still harmful to people living at the lower end of the poverty spectrum. This means that corruption is still harmful to people living in the lower-bound forms of poverty. People living in lower-bound poverty are those who do very low-paying jobs and equally depend on government child support or elderly grants (Ngubane, Mndebele, and Kaseeram 2023). For example, people who work for rich people perform casual work, such as gardening, cooking, and babysitting. They normally do not have formal certificates for their work, and are deprived of recognition by trade unions. Therefore, they continue to live under the employer’s mercy for their survival. These people normally participate in saving schemes such as stokvel to purchase huge assets once a year or help one another to build modern housing. The form of corruption they experience is a lack of formal recognition, underpaid, working extra unpaid hours, denied holidays, treated unfairly, they are not respected in society, and ignored.

Regarding the Gini coefficient, a 1% increase in inequality leads to a 1.03% increase in poverty. The results align with the poverty-growth-inequality theory, which holds that inequality exacerbates poverty alongside slow growth. South Africa has a highly unequal economy, and inequality dominates the other variables that exacerbate poverty. In contrast to the previous example of men and women working for richer families, they remain underpaid. This scene provides a full picture of the South African reality; those who are rich have better-paying jobs with many benefits for workers and family members. On the other hand, poor domestic workers are paid less, work more hours, do more than what is required by the job description, and without benefits. Koelble (2022) noted that BBBEE triggered South Africa, which is highly unequal. In this way, inequality suppressed the poor in support of the rich, which is what Blacks thought would end in the post-apartheid period. However, history repeats itself, and Koelble (2022) argued that inequality is not racial. The unemployment rate has a lesser effect on the lower bound level of poverty (0.01%). This case does not necessarily imply the truth because the previous example of domestic workers is not taken into account when unemployment rates are calculated. Second, the data inconvenience for the low unemployment rate depicted in Figure 1 above, where unemployment was recorded at 24% in 2020Q2, could be the reason. Moreover, this indicates that this type of unemployment could be best described by the primary data collected by the survey. The constant term has decreased slightly because, in this case, we discuss people who are relatively exposed to the working environment.

Moving to the 75^th quantile regression results. This quantile represents people who hold value and assert that they are still living in poverty or deprived of other essentials to maintain a quality living income benchmark. A 1% increase in corruption leads to a 0.40% increase in poverty; the results are statistically significant at the 1 percent level. The results are in line with PED theory and Yusuf et al. (2014); although the results were based on causal effects and did not specify the sign, the result also relates to those obtained in previous quantiles. These results could relate to young people- exposed to some benefits like NSFAS–who look like rich people through the clothes they wear, high-profile cell phones, and study gadgets in public figures. At the same time, the same asserts are short-lived because of frictional unemployment, which lasts longer after their studies. In addition, in the end, they remain without shelter of their own and do less-paying jobs that are unrelated to their studies (youth experience the highest unemployment rate in SA). Equally important since cultural norms are becoming less important, most young people bear the burden of having children before they finish schooling, which makes their lives much more complicated in post-study periods.

Regarding inequality, a 1% increase in the Gini index leads to a 1.03% increase in poverty, and the results are statistically significant at the 5% level. The Gini index is still dominant in the upper quantile and is related to the findings of Lakner et al. (2022). The results could be equally related to the example of young people highlighted above. Statistics indicate that young people experience high unemployment rates in SA, which implies that they experience high inequality among themselves and other age groups. For example, the so-called “connected people” can get jobs after the University completion period. That is, they can be offered jobs by relatives or friends, be bribed to get employed, or have an affair with HR to find employment. This is the reason behind individuals’ resistance to reporting incidences of poverty (Clemente and Calca, 2023). These conditions are dominant because it is widely known that one can neither deny employment nor find a job easily. These people tend to get good-paying jobs, opening a wide gap between them and their unemployed counterparts. Sometimes, elderly people are important for their experience, which makes it easier for them to find and change jobs.

Moving to unemployment-causing poverty, a 1% increase in unemployment leads to a 0.02% increase in poverty, and the results are statistically significant at the 5% level. These findings are consistent with the Keynesian theory of unemployment and poverty those of (Meo et al. 2023). The small impact of unemployment on poverty might indicate that unemployment might cause poverty through other variables such as being unemployed and young, family size, and not doing any formal/informal occupation. For example, Dahliah (2021) confirmed that unemployment causes poverty through economic growth. Meanwhile, little influence of unemployment on poverty is depicted in the fall figures depicted in Figure 1 above by the graph of unemployment, which was 24% in the second quarter of 2020. Hence, in real life, we are fully aware that at that time, many people were not employed and yet willing to work, but were hindered by lockdown. Hence, in that period, the definition of unemployment that involved the word actively seeking employment was still active. However, the truth holds that people who are not part of the unemployment percentage are there, some are ready to work, and some are illiterate and look for jobs for unskilled labor.

Table 4 is out of the last table in this study, and depicts the post-diagnostics tests and goodness of fit. The Pseudo R-squared increases with quantiles from 24.60 on the 25 quantile, 33.32 on the 50^th quantile to 50.74 in the 75^th quantile. This implies that variations up to 50% are explained by the selected dependent and control variables. This implies that other factors that influence poverty (such as gender, race, age, and others that could work in the surveyed data) were not included in the study because of the objective of the study. Equally important, the Quasi-LR-stat is 38.51 in the lower quantile, 48.28 for the middle quantile, and 93.19 for the upper quantile. All the highlighted statistics are statistically significant at the 1% level, implying goodness of fit for the overall model. A normality test was then performed. The J-B test is statistically insignificant for all quantiles, implying that the data are normally distributed across disciplines. The asymmetric plot for the first quantile is 5.03, 5.03 for the middle quantile, and 15.18 for the last quantile and is statistically significant for all quantiles. This implies that the data have a linear relationship with the variables, as established by the study. This further implies a vibration of the estimated results. The slope equity test is 69.25 for the 25^th, 61.47 for 50^th and 62.28 for the last quantile (75^th), all of which are statistically significant. This indicates that the slope differs across the periods. This supports the reason behind the use of quantile regression as an extension of the simple OLS regression model.

The graph in Figure 4 is used to compare the results of the OLS with those of the quantile regression. The bold line should hover around the shaded region and within the two parallel dotted lines. To examine the Gini coefficients, the quantile process hovers around the two bounds indicated by parallel lines. This means that the results obtained are similar to those obtained from the simple OLS. For Corruption, the values on the lower quantile are relatively similar to those of the OLS model, but as the quantiles increase, they move to the upper and lower bounds. For unemployment rates, the small coefficients obtained indicate that both the values in the lower and upper quantiles are out of bounds.

Figure 4. Quantile processes.

Source: Stata.

5. Conclusion

This study has established a connection between corruption and poverty in South Africa and the world. The results of the quantile regression model for all quantiles confirm that corruption causes poverty. Based on the results, the study recommends that policymakers make corruption one of the objectives of the economy. The study recommends that policymakers implement measures to reduce corruption to ensure that the poor receive intended attention. South African authorities should implement strategies to reduce unemployment rates and bridge the inequality and poverty gaps.

A limitation of the study is the potential endogeneity between poverty and corruption; future studies may use an instrumental variable quantile regression model to address this issue. The study was limited to the quantile regression model, and future studies can use other models, such as the moderation model, to examine the effect of one variable on other influential variables. Based on the study’s overall results, policymakers should invest in youth development in entrepreneurship and education, and support women’s cooperatives and small businesses to reduce reliance on state services and address high inequality and unemployment rates.