Sensitivity of Educational, Maternal, and Mental Health Outcomes from Income Redistribution Fluctuations in South African Children

I. INTRODUCTION

Some papers argued that parent income positively impacts children's education outcomes (Chevalier et al., 2013). Therefore, income redistribution is expected to advance children's education outcomes, assuming the education system prioritizes early childhood education (Neuman & Powers, 2021). Other authors reject this view and argue that household income and financial resources are associated with early school dropouts (Alcaraz, 2020). Therefore, the idea that income redistribution advances children's education outcomes may not be accurate.

These inconsistencies necessitate more international studies investigating the intersection between income redistribution, education outcome, and mental health outcome - especially in developing countries in light of the increasing cases of poor mental health in the adolescent population (Twenge et al., 2021).

1.1 Income redistribution in South Africa

In 2019, the South African government published the assessment report of income redistribution between the haves and have not (StatSA, 2019).

According to the report, income redistribution has improved during 2006, 2009, 2011, and 2015 where data samples were collected. It remains to be seen how the improvement of income redistribution advances educational, maternal, and mental health outcomes concurrently. This paper is designed to fill this gap and analyze the sensitivity of South African children's educational, maternal and mental health outcomes from income redistribution fluctuations.

So far, South Africa has utilized the social security system to advance income redistribution (Satumba, Bayat, & Mohamed, 2017). Other papers show that social grants in South Africa advance the ability to read, write, and attend school (Mostert & Vall, 2020). However, it is unknown if the accumulative income from redistribution programs improves children's academic failure in primary and secondary learning phases.

Another angle that remains open for investigation is how sensitive academic failure is from income redistribution fluctuations. Meaning, do high-income redistributions yield a better reduction in academic failure than low-income redistributions. One would expect that high-income redistribution will significantly avert academic failure, considering that parent income impacts children's education positively (Chevalier et al., 2013). This angle of literature remains unknown in South Africa.

Another question worth exploring is whether high-income redistribution begets more considerable improvement in mental health outcomes (depression) in both girls and boys comparatively.

The mental health of the adolescent population has not been studied comprehensively in South Africa (Buckley et al., 2020 and Somefun & Fotso, 2020) despite its influence on education outcomes (Andersen, Holm, & Cote, 2021).

Nevertheless, one would expect that income redistribution to improve depression levels of disadvantaged children.

Another question worth exploring is how income redistribution affects the maternal health of children (teenage pregnancy). Currently, it is documented that poverty causes teenage pregnancy, forcing girls to drop out of school (Jochim, Cluver, & Meinck, 2021). Therefore, if one accepts this proposition, income redistribution should avert teenage pregnancy and poverty, thus improving education outcomes. Once again, this intersection remains unknown in South Africa.

1.2 Evolution of income redistribution in South Africa

According to Statistics, South Africa income redistribution improved with the share of income owned by the poorest 40 percent of the population increasing from 7.8 percent in 2006 to 9.3 percent in 2015 (StatSA, 2017) (See Figure 1).

Such improvement was credited to the Republic of South Africa Constitution, which states that all citizens have the right to appropriate social assistance from the government if they cannot provide for themselves.

{"image_source":{"path":"images/3a0f388de8fadf5a4dcf7c67f057f6d3d09708773f473cabfef2c1e4ffdb1328.jpg"},"content":"","chart_caption":[{"type":"text","content":"Figure 1: Evolution of income redistribution in South Africa"}],"chart_footnote":[]}

South Africa has utilized the seven social protection grants to advance income redistribution, which assisted the disadvantaged communities with household necessities (Maitra & Ray, 2003). Given the improved income redistribution from 2006, 2009, 2011, and 2015—the current study seeks to analyse the sensitivity of educational, maternal, and mental health outcomes arising from income redistribution fluctuations of these periods.

The current study seeks to answer the following questions: (1) What impacts do income redistributions exert on children's mental health outcomes (depression) in both low and high-income redistribution periods? (2) What impacts do income redistributions exert on children's education outcome (failure rate) in both low and high-income redistribution periods? (3) What impacts do income redistributions exert on a maternal health outcome (teenage pregnancy) in both low and high-income redistribution periods? (4) Are there any gender-based differences in the impact of income redistributions? This study is designed to answer these interrelated questions and unlock the role of income redistribution in fostering educational, maternal, and mental health outcomes in South African disadvantaged communities.

II. MATERIALS AND METHODS

The author uses a recently released database, the General Household Survey (GHS), provided by Statistic South Africa during 2006, 2009, 2011, and 2015. The GHS is an annual survey of approximately 120,000 individuals from more than 20,000 households. These yearly surveys aim to provide information on any changing trends in the composition of South African families. The GHS sampling procedure involves explicit stratification by province and urban and non-urban areas within each province. Household units are drawn under this stratification. For each household unit, individual characteristics are presented, including age, gender, educational outcomes, household income levels, health outcomes, and other general socioeconomic status variables.

The GHS structure questions for a binary response. For example, senior individuals in the household are asked if members of the household suffer from depression. The general reply is either yes, the member has episodes of depression, or no the member has normal moods without depression. The author uses these binary variables to quantify the changes in mental health outcome (depression), educational outcome (academic failure), and maternal health outcomes (teenage pregnancy) during low-income redistribution periods (the year 2006) versus high-income redistribution periods (the year 2009, 2011, 2015) as depicted in Figure 1. The author selected these year periods in line with the inequality report published by Statistic South Africa in 2019 (StatSA, 2019).

For quality purposes, all Statistic South Africa datasets are reviewed at regular intervals by the Statistician-General and the head of the relevant organ of state to ensure that they remain relevant and of specified quality. Therefore, the current binary variables reflect the genuine legal reality of the South African social outcomes, which are worth to be used for international scholarship.

The author utilized these cross-sectional datasets to understand the sensitivity of educational, maternal, and mental health outcomes from income redistribution fluctuations. The outcomes analysed are all dummy variables reported by the head of the household when asked about the education and health outcomes of primary school cohorts (age 7-13 years) and secondary school cohorts (age 14-18 years).

The GHS also includes information on households benefiting from the social security system.

However, the author did not want to compare households benefiting from social security to those not benefitting. These two groups of households can be different in many additional dimensions (for example, income, education background, or information barriers), which can influence the exogenous aspect of the model and inflate the possible outcomes. The author did not use an Ordinary Least Squared Model (OLS) for these reasons. Instead, the author opted for a Two-stage Least Squared Model (2SLS) – similar to the analysis done previously in South Africa (Mostert, 2020 & 2021). In this analysis, the author instrument income redistribution with the cohorts as presented below.

$Income-re{d}_i^a={\alpha }_2+{\beta }_{2}Trea{t}_i^a+\psi RaceFE+\xi ProvinceFE+\delta Age+\mathrm{\wp }MotherEducationFE+FatherEducationFE+\sigma Femal{e}_i$

In the first equation, Y is one of the outcomes for individual ${}_i$ at age $\alpha$ (7-13 year cohort or 14-18 year cohort), and "~ Income-red" is the predicted benefit from income redistribution. The regression includes race fixed effects (which capture the different racial groups in South Africa), province fixed effects (including nine provinces in South Africa), age fixed effects (which capture age differences for primary and secondary cohorts), and a dummy for females (for the regressions where the author estimates the effects for both males and females). Furthermore, the author also includes controls (fixed effects) for the parent's educational level, which may influence education outcomes (Alcaraz, 2020).

In the second equation (which corresponds to the first stage regression of the 2SLS), income redistribution is estimated as a function of the treatment dummy variable, which identifies the cohorts whose parents held any government social security grants and were satisfied with social welfare services. This group of cohorts was classified as the treatment group for income redistribution. Cohorts whose parents did not benefit from any social security grant and were dissatisfied with the social welfare services did not benefit from income redistribution, thus considered the control group. The author opted to analyse these cohorts in households earning less than US $ 500 per month. In South Africa, a significant share of disadvantaged households still does not benefit from the social security system, despite suffering from poverty. The most common causes for failure to qualify for social grants are lack of identity documents, fraud, and being outside the eligibility requirements of the social security system.

The author focus on these critical cohorts: 7-13 year cohort and 14-18 year cohort - which correspond to both primary and secondary phase of schooling in South Africa. The author also includes parent educational categories as proxies for socioeconomic status, which may influence education and health outcomes. The first stage regression also controls for gender and race and province fixed effects. Thus, the race fixed effects account for any trend across cohorts' racial groups. The province fixed effects account for any baseline difference across provinces.

In all model estimations, one needs two assumptions to be fulfilled: first, the instrument has to be relevant in explaining the probability of being treated, and this will be corroborated by the F-test of the first stage equation; and second, the exclusion restriction needs to hold, that is, the instrument should not influence the primary outcome directly through any channel other than treatment. In this case, this assumption means that differences in depression, academic failure, and teenage pregnancy between the treated and control groups during these cross-sectional periods (the years 2006, 2009, 2011, and 2015) can only be due to income redistribution.

For example, there is no reason to believe that control cohorts should have different educational, maternal, and mental health outcomes than the treated cohorts observed from the same year, same regions, and from the same household income category (less than $ 500 per month).

There is no other event in South African history that explains any difference in educational, maternal, and mental health outcomes that would affect precisely the treated cohorts but not the control cohorts. For this reason, the author is confident that income redistribution can explain the differences between the treated cohorts and control cohorts.

III. RESULTS

3.1 Descriptive analysis

Table 1 presents the exact difference between the treated and control groups. The author noted that individuals who benefitted from income redistribution (treated group) recorded a low probability of depression, academic failure, and teenage pregnancy than those not treated with income redistribution (control group).

3.2 Results of two-way 2SLS model

The tables present a comparative picture of how income redistribution averted depression, academic failure, and teenage pregnancy. The tables also show the sensitivity of these outcomes from income redistribution fluctuations during 2006, 2009, 2011, and 2015.

When analysing the 2SLS model results, the author noted that the first stage regression's F-statistic is very large, pointing towards the instrument's strong validity (See Table 2). Thus, in Table 2, income redistribution grants proxied by the cohort instruments are determinants for improving depression outcomes.

For example, cases of depression were reduced by 4.32 percentage points in 2006. The mean for depression is 23.15 percent, in line with other mental health studies conducted in South Africa (Somefun & Fosto, 2020). Therefore income redistribution averted depression cases by 18.66 percent in 2006. Similarly, cases of depression were reduced by 7.52 percentage points in 2015.

The mean for depression is 21.03 percent. Therefore income redistribution averted depression cases by 35.76 percent. Table 2 confirms that the higher the share of income redistributed to the poor, the better the cases of depression in the primary school cohorts.

The study then examines whether income redistribution positively impacts depression outcomes in the secondary school cohorts. Table 3 shows a 2.32 percentage point reduced cases of depression in 2006. The mean for depression is 28.13 percent. Therefore income redistribution averted depression cases by 8.25 percent in 2006.

Similarly, cases of depression were reduced by 3.31 percentage points in 2015 for the secondary school cohorts. The mean for depression is 22.33 percent. Therefore income redistribution averted depression cases by 14.01 percent. Once again, the higher the share of income redistributed to the poor, the better the cases of depression. Table 3 also shows that the impact of income redistribution on depression outcomes is lower in the secondary school cohorts than in the primary school cohorts.

The study then examines whether income redistribution positively impacts education outcomes, considering that it positively impacts depression. The idea behind this analysis is that good mental health should improve education outcomes. The author found in Tables 4 and 5 that income redistribution significantly improves academic failure in primary and secondary school cohorts. For example, income redistribution reduces academic failure by 9.86 percentage points in 2006, implying a 25.85 percent decline in academic failure for primary school cohorts (See Table 4).

The mean for academic failure was 56.13 in 2006 in the secondary school cohort. This mean estimation is in line with a study conducted in South Africa where it was reported that the number of learners enrolled in grade 1, only half make it to grade 12 - with higher dropouts recorded in the secondary phase of learning (Modisaotsile, 2012). Income redistribution reduced academic failure by 7.32 percentage points in 2006, implying a 13.04 percent decline in academic failure for secondary school cohorts (See Table 5). Both Table 4 and 5 demonstrate that the higher the share of income redistributed to the poor, the better the cases of academic failure.

***denote significant p-value at <0.05.Coefficients in brackets represents standard errors. Note: The results are from a 2SLS model. In the first stage equation, the dependent variable is the probability of benefiting from income redistribution. At the same time, the instrument is a dummy variable equal to 1 for the primary school cohorts who benefited from the social security grants system and o for the primary school cohorts who did not benefit from the social security grants system. In the second stage regression, the dependent variable is a dummy variable of "failing to complete school grade." Both regressions include race, province, age, and gender fixed effects—source: General Household Survey (GHS) provided by Statistic South Africa.

The study then examines whether income redistribution positively impacts maternal health outcomes of children, considering that it positively impacts depression and academic failure. The author believes that improvement in income, better mental health, and exemplary academic achievements are positive incentives for keeping young girls focused on their school curriculums and discouraging teenage pregnancy.

Table 6 shows a 0.17 percentage point reduced cases of teenage pregnancy in 2006. The mean for teenage pregnancy is 4.94 percent, in line with the reported percentage by the Department of Basic Education in South Africa (Panday et al., 2009).

Therefore income redistribution averted teenage pregnancy by 3.44 percent in 2006. Similarly, cases of teenage pregnancy were reduced by 0.32 percentage points in 2015. The mean for teenage pregnancy in 2015 was 4.71 percent. Therefore income redistribution averted teenage pregnancy by 6.79 percent. Once again, the higher the share of income redistributed to the poor, the better the cases of teenage pregnancy. Indeed government policy can avert teenage pregnancy (Chung, Kim, & Lee, 2018).

***denote significant p-value at <0.05.Coefficients in brackets represents standard errors. Note: The results are from a 2SLS model. In the first stage equation, the dependent variable is the probability of benefiting from income redistribution. At the same time, the instrument is a dummy variable equal to 1 for the secondary school cohorts who benefited from the social security grants system and 0 for the secondary school cohorts who did not benefit from the social security grants system. In the second stage regression, the dependent variable is a dummy variable of "reported pregnancy." Both regressions include race, province, age, and gender fixed effects—source: General Household Survey (GHS) provided by Statistic South Africa.

The paper then focuses on gender-based differences in the impact of income redistribution on depression and academic failure. The study repeated the same regressions for the male population group and the female population group separately. In Table 7, the author discovered that income redistribution improves depression and academic failure more for boys than girls.

IV. DISCUSSION

The current paper discovered that income redistribution has a positive role, concurrently averting depression, academic failure, and teenage pregnancy. This intersection is essential, considering the extensive evidence showing the positive societal outcomes arising from educational, maternal, and mental health improvement.

The impact of income redistribution is more robust for boys than girls, similar to the evidence reported by other papers (Mostert & Vall, 2020).

Girls are more susceptible to depression (Twenge et al., 2021) than boys, which explain the lower impact of income redistribution in the girls' populations. Furthermore, income redistribution is more robust for the younger primary school cohorts than secondary school cohorts. These results reflect the expensive nature of raising children, with costs increasing as children grow older – diminishing the purchasing power on income redistributions.

The study also finds that the educational, maternal, and mental health outcomes are sensitive to income redistribution fluctuationswith big improvements recorded during the peak of income redistribution.

4.1 Policy implications

The study's results have significant policy implications. The current Covid-19 lockdown policies and on-going economic recession threaten the South African social security system.

The government imposed austerity measures that lowered grant pay-outs widened the existing income inequalities (Makinana, 2021). The combination of these factors will reverse the income redistribution gains and potentially worsen educational, maternal, and mental health outcomes. South Africa may need to improve income redistribution throughout the lockdown period or face a possibility of an inequality crisis that will rob the developmental outcomes of disadvantaged children.

4.2 Robustness test

In this section, the author provides robustness checks and additional results to reinforce the current estimation's validity. Tables 1A in the Appendix show the results of the OLS regressions for both the primary and secondary school cohorts. The variable of interest is now the variable in the survey that identifies the receipt of the grant at the household level. The study included parent education, race, age, and province fixed effects. As explained above, there are reasons to believe that this is not a randomly assigned program. Therefore, the OLS estimation may overestimate or underestimate the income redistribution effects on education and health outcomes. Many unobserved variables may directly affect the education and health outcomes, such as household income from relatives, access to the necessary information, and proximity to school and healthcare facilities. Indeed, the results in Tables 1A are all substantially bigger in magnitude than the baseline results of the 2SLS models presented in Tables 7, and this is consistent with the OLS model overestimating the actual effects of the income redistribution results.

Finally, the study ran some placebo regressions in which the study "pretended" that control (unaffected) cohorts were treated with fake income redistribution. Thus, the author excluded the cohorts genuinely affected by the income redistribution from the sample. The author then assigned all males as treated cohorts affected by the fake reform and used females as comparison cohorts. The author then ran the same 2SLS.

One can see in Table 2A in the Appendix that the F-test of the first-stage regression is extremely low (which suggests that the instrument is not relevant). The treatment variable is not significant in any of the periods analysed. Therefore, the results of these placebo tests analysing the effects of the fake income redistribution reinforce the validity of the study's identification strategy and provide additional evidence of the fulfilment of the exclusion restriction criteria as any cohort-specific events that are not captured by the model that could be biasing the main results should also provide significant results in these placebo tests.

4.3 Limitations

The paper acknowledges that the current binary variables do not capture all outcomes linked to these households. There is no advance information on the determinates of educational, maternal, and mental health outcomes in the Statistics South Africa data. Thus, the paper interprets the results as providing evidence of substantial improvement in these outcomes attributed to income redistribution while not capturing other qualitative changes that may further explain the improvement of these outcomes. For example, academic success can also be driven by teaching style and incentives for teaching (Camelo & Ponczel, 2021).

Unfortunately, the Statistics South Africa data does not contain information on teaching styles and incentives for teaching, which should have been controlled in the analysis. Nevertheless, the author believes such omission will not significantly influence the current estimation considering that parent income is still a competent primary driver of better education outcomes.

V. CONCLUSION

Income redistribution plays a crucial role in advancing disadvantaged children's educational, maternal, and mental health outcomes. For example, income redistribution averts depression, academic failure, and teenage pregnancy concurrently. Higher-income redistributions intensify these outcomes.