Heterogeneous relationships between income levels and associated correlates in Gauteng province, South Africa: quantile regression approach

Figures & data

Figure 1. Income distribution in Southern Africa: Share of deciles in income distribution. Source: UNU-WIDER (2017) for various years in each country – Angola (2008), Botswana (2009), Lesotho (2010), Madagascar (2012), Malawi (2010), Mauritius (2012), Mozambique (2008), Namibia (2010), South Africa (2011), Swaziland (2009), and Zambia (2010).

Figure 2. Gauteng province and its surroundings – forming the wider Gauteng city-region.

Table 1. Description of model variables.

Variable	Measurement scale	Description
Income	Continuous	Per capita income
Race	Nominal	0 for Africans/blacks. 1 if coloured. 2 if Indian/Asians. 3 if white
Gender	Nominal	0 for household head is female. 1 if household head male
Age cohort	Nominal	0 for Adult (35–64 years). 1 if youth (18–34 years). 2 if Pensioners (65+ years)
Marital or relationship status	Nominal	0 for those in a relationship. 1 for those not married or in relationship
Education level	Continuous	Total number of years of school/college attendance
Business ownership	Nominal	0 for not owning business. 1 for those who owns a business
Household location	Nominal	0 for Ekurhuleni and the rest of the 8 metros taking a value from 1 to 8, accordingly.
Migration status	Nominal	0 for born in Gauteng. 1 for born in another province in South Africa. 2 born outside South Africa
Disability	Nominal	0 if household head has some disability. 1 for no disability
Sector	Nominal	0 if working in primary sector. 1 if working in secondary or tertiary sector

Figure 3. Income distribution by quantiles in QoL IV survey data.

Table 2. Univariate statistics of model variables,

Variable	n	Min	Max	Mean	Standard deviation
Per capita income (Rands)	8567	29	409,600	5208	16 579
Race	14 847	0	3	0.73	1.25
Gender	14 847	0	1	0.58	0.49
Age cohort	14 932	0	2	0.46	0.54
Marital or relationship status	14 847	0	1	0.17	0.38
Education level	14 847	0	18	12.31	3.20
Business ownership	14 847	0	1	0.11	0.32
Household location	14 847	0	8	1.50	1.72
Migration status	14 847	0	2	0.46	0.66
Disability status	14 847	0	1	0.97	0.17
Sector	11 129	0	1	0.91	0.28

Figure 4. Per capita monthly income in Gauteng.

Table 3. Quantile regressions results.

Variable	Pooled OLS	tau = 01	02	03	04	05	06	07	08	09
Coloured	–0.079	–30.72	–85.13	–76.46	62.86	–200.11	–302.86	–413.18	–451.85	–1783.39
Indian/Asian	–0.045	–13.63	–6.10	35.10	–52.47	–204.55	–94.29	–91.11	–261.11	–1113.59
White	0.087	31.23	32.83	81.63	155.84	114.29	102.86	79.37	307.41	537.94*
Male	0.242***	46.02***	97.43***	171.70***	294.03***	439.26***	702.86***	1013.02***	1553.70***	2037.86***
Youth	0.051	–11.36	17.43	35.37	49.35	–12.01	14.29	135.71	468.52	1307.57**
Pensioners	0.162	149.35	185.67	286.26	191.17	137.04	28.57	–121.43	231.48	3319.64
Relationship status	0.222***	26.74	95.01**	185.31***	248.83***	360.74***	562.86***	713.33***	1483.33***	2334.89**
Education	0.177***	28.41***	66.05***	108.84***	168.83***	237.62***	320.00***	435.87***	581.48***	898.48***
Business Owner	–0.051	–10.20	4.26	6.12	12.99	–16.88	–120.00	4.76	81.48	–427.26
Johannesburg	0.087	29.33	71.19**	103.13***	151.17**	263.33***	182.86	360.16**	246.30	62.12
Tshwane	0.065	33.83	53.95	67.76**	112.21	204.55**	97.14	280.79	518.52	-224.35
Emfuleni	0.165	83.31	138.89	243.95**	331.95**	279.37	340.00	280.79	970.37	1483.23
Lesedi	0.186	-73.86***	31.61	138.50	62.86	367.14	1217.14**	1153.97	1331.48	2646.87
Midvaal	0.078	39.77	–6.39	125.17	150.65	279.37	145.71	352.38	–83.33	1213.51
Merafong City	0.130	103.19**	95.59	103.13	224.42	330.05	200.00	360.00	570.37	–354.55
Mogale City	–0.033	85.58**	–13.95	–117.55	–74.81	97.78	–182.86	45.71	-70.37	–837.36
West Rand	0.313**	158.68***	276.42***	312.93**	499.22**	496.35***	345.71	518.73	1714.82**	352.94
Born outside Gauteng	0.056	–1.16	19.08	32.38	62.34	81.38	102.86	135.71	272.22	224.35
Foreign born	0.249***	43.70	113.556	216.05***	330.39***	466.40***	1014.29***	1098.57**	2016.67***	3233.36***
Disability	0.187	73.27***	133.03**	170.61	280.00**	418.20***	654.29***	867.94	1187.04**	1134.75
Sector	–0.177***	–73.34**	–57.05	–149.93**	–296.62***	–320.16**	–557.14***	–344.92	–644.44	–2206.05

*** and ** indicate significance at 1%, and 5%, respectively.

Notes: Pooled OLS model was estimated with log-transformed per capita income as the dependent variable. In quantile regressions estimation, per capita income was used, since there was no need to transform the dependent variable (Koenker & Hallock, 2001).

Figure 5. Varying coefficients for race variables, with 95% confidence intervals. Blacks was the reference group.

Figure 6. Quantile regression results for pensioners and youth compared to adult (reference group) – with 95% confidence intervals.

Figure 7. Varying estimate values for other selected explanatory variables (with 95% confidence intervals each).

UNU-WIDER, 2017. World income inequality database (WIID3.4), January 2017. UNU-WIDER, Helsinki www.wider.unu.edu/database/world-income-inequality-database-wiid34. Accessed 28 October 2018.

 Google Scholar

Koenker, R & Hallock, KF, 2001. Quantile regression. Journal of Economic Perspectives 15, 143–56. doi: 10.1257/jep.15.4.143

 Web of Science ®Google Scholar