Human capital inequality and electoral outcomes

15 August 2017
Publication Type: Policy Brief
Economic Theme: Public Finance
JEL Code: D31, J15

Inequality is a problem which has beset South Africa for a long time with the country being on record as having one of the highest levels of inequality in the world. However, little is known about horizontal inequality between different groups in South Africa. Given the level of racial and ethnic diversity in South Africa, it is important to have a good grasp of the dynamics of group inequality in the country.

Historically, inequalities in South Africa were the product of legalized discrimination enforced as part of apartheid. Hence, racial inequality remains the most politically salient aspect of income inequality in South Africa. Racial wage inequality persists even after systematic discrimination was removed with political liberalization in the mid-1990s, with the differences in years of schooling between white groups and other groups explains almost half of the remaining wage inequality. In general, the high incidence of poverty among black Africans in the post-apartheid period can be attributed to the accumulation of mainly pre-labour market disadvantages. According to Gradin, quality of education as well as quantity of schooling seems to put blacks at a disadvantage judging from the rise in the importance of unobservables in explaining poverty differentials by race.

In our recent paper we explore the patterns and trends of vertical and horizontal human capital inequality in South Africa since 1994. Human capital here is defined as years of schooling. We use data from the census of South Africa in 1996, 2001, and 2011 to compute five different measures of horizontal inequality, and one measure of vertical inequality. We compute the group co-efficient of variation, the group weighted Gini index, the Group weighted Theil index, crosscuttingess, and crossfractionalization. We use three different measures of schooling to compute inequality measures based on age. We compute for individuals 25 years or older, for individuals 15 years or older, and for individuals below 15 years of age. Finally, individuals are grouped by race, language, gender, and rural or urban location to calculate horizontal inequality. Inequality measures are computed both at the national level and for each municipality.

Patterns and trends

Interesting patterns show up across all three measures of schooling. On average human capital inequality, both horizontal and vertical, appears to be falling over time. However, the fall is not universal across the country. There are areas where human capital inequality is stagnant or has actually increased. The spatial patterns of horizontal inequality however depend on how they groups are defined. When individuals are grouped by race, horizontal inequality appears to be higher in the western parts of the country. When individuals are grouped by language, horizontal inequality appears to be higher in the north eastern parts. This distinction highlights the importance of group definition when measuring horizontal inequality.

Within group vertical inequality also reveals some interesting findings. Inequality appears to be highest within the black group as a whole, and even higher within the black linguistic groups individually, compared to other groups. However vertical inequality across all groups appears to have been falling since 1996. Vertical inequality within the male and female groups appears to be almost identical. Finally, inequality between urban residents appears to be much lower than inequality between rural residents.

Horizontal and vertical inequality on electoral outcomes

Various authors have suggested that inequality, both horizontal and vertical, can have effects on political systems. In unstable countries, for instance, high horizontal inequality is associated with the lower likelihood of civil wars The argument is that high crosscuttingness makes building broad based coalitions more difficult. On the other hand, high or increasing vertical inequality, being an undesirable outcome, could serve as an incentive to mobilize for regime change.

Although the possibility of civil war in South Africa is very low, the same kind of dynamics could play out at the ballot box. However, instead of rebel leaders, political parties are the groups which try to dethrone leading parties. They have to mobilize enough support to overthrow the leading party. Is it the case, as with civil wars, that horizontal inequality makes building such support more difficult? With regards to vertical inequality, do citizens punish incumbent political parties for rising inequality, or does inequality cement the dominance of political parties as argued by Boix and Stokes, and Ziblatt? Their argument is that in many cases in societies with high levels of inequality, groups at the top of the spectrum have greater incentives to monopolize political power.

To explore this question we combine our panel of inequality measures with a panel of electoral outcomes, measured as the share of votes by the winning party and the winning margin between the top two parties, over the same period. We find that horizontal inequality, using all measures, does not matter much for electoral outcomes. The exception is when individuals are grouped by language. The suggestion is that language and ethnicity might be a factor in intra party politics.

However vertical inequality appears to be strongly correlated with electoral patterns. This strong association remains regardless of what measure of electoral outcome is used, and despite what external factors are controlled for. Municipalities with higher levels of vertical inequality appear to be the least competitive electorally. The result is consistent with the arguments made by Boix and Stokes, and Ziblatt (2008). Perhaps human capital inequality does play a role in shaping electoral outcomes. Our study leaves room for more research to get a better understanding of the relationship between the two.

Research Brief 124
1 August 2017