Author/s: Katharine Hall & Winnie SambuDate: October 2016
This indicator shows the number and proportion of children living in households where children were reported to go hungry “sometimes”, “often” or “always” because there wasn’t enough food.
Section 28(1)(c) of the Bill of Rights in the Constitution gives every child the right to basic nutrition. The fulfilment of this right depends on children's access to sufficient food. This indicator shows the number and proportion of children living in households where children are reported to go hungry “sometimes”, “often” or “always” because there isn’t enough food. Child hunger is emotive and subjective, and this is likely to undermine the reliability of estimates on the extent and frequency of reported hunger, but it is assumed that variation and reporting error will be reasonably consistent so that it is possible to monitor trends from year to year.
The government has introduced a number of programmes to alleviate income poverty and to reduce hunger, malnutrition and food insecurity, yet 2.3 million children (12%) lived in households where child hunger was reported in 2014. There was a significant drop in reported child hunger, from 31% of children in 2002 to 16% in 2006. Since then the rate has remained fairly consistent, suggesting that despite expansion of social grants, school feeding schemes and other efforts to combat hunger amongst children, there may be targeting issues which continue to leave households vulnerable to food insecurity.
There are large disparities between provinces and population groups. Provinces with relatively large numbers of children and high rates of child hunger are KwaZulu-Natal (19%), Western Cape (14%) and the North West (15%), which together have over a million children living in households that report having insufficient food for children. The Northern Cape (18%) has a relatively low child population but has the second highest rate (18%) of child hunger. These provinces consistently reported high rates of child hunger throughout the past decade, although the proportion of children experiencing hunger has declined substantially in all provinces over the period. The Eastern Cape has had the largest decrease between 2002 and 2014, with reported child hunger having reducing by 37 percentage points over the 13-year-period. Limpopo has a large rural child population with high rates of unemployment and income poverty, yet child hunger has remained well below the national average, reported at 4% in 2014.
Hunger, like income poverty and household unemployment, is most likely to be found among African children. In 2014, some 2.1 million African children lived in households that reported child hunger. This equates to 14% of the total African child population, while relatively few Coloured (8%) children lived in households where child hunger was reported, and the proportions for Indian and White children were below 3%.
Although social grants are targeted to the poorest households and are associated with improved nutritional outcomes, child hunger is still most prevalent in the poorest households: 21% of children in the poorest quintile go hungry sometimes, compared with 1% in the wealthiest quintile of households. The differences in child hunger rates across income quintiles are statistically significant.
There are no significant differences in reported child hunger across age groups. However, close to 800,000 children aged less than five years are reported to have experienced child hunger. Young children are particularly vulnerable to prolonged lack of food, which increases their risk of nutritional deficiencies such as stunting. Inadequate food intake compromises children’s growth, health and development, increases their risk of infection, and contributes to malnutrition. Stunting (or low height-for-age) indicates an ongoing failure to thrive. It is the most common form of malnutrition in South Africa and affects 25% of children under five.1
The ‘hunger’ question in the General Household Survey provides notoriously weak data. Child hunger is emotive, subjective and estimates of frequency unreliable – particularly since the presence and frequency of ‘child hunger’ is reported by one adult in the household. It is assumed, however, that reporting error will be similar in each year of data collection, so that it is possible to report trends even if proportions for a single year are questionable. For this indicator the 5-point scale is collapsed into a dichotomous (yes/no) variable.
The data are derived from the General Household Survey2, a multi-purpose annual survey conducted by the national statistical agency, Statistics South Africa, to collect information on a range of topics from households in the country’s nine provinces. The survey uses a sample of 30,000 households. These are drawn from Census enumeration areas using multi-stage stratified sampling and probability proportional to size principles. The resulting estimates should be representative of all households in South Africa.
Person and household weights are provided by Stats SA and are applied in Children Count – Abantwana Babalulekile analyses to give estimates at the provincial and national levels. Survey data are prone to sampling and reporting error. Some of the errors are difficult to estimate, while others can be identified. One way of checking for errors is by comparing the survey results with trusted estimates from elsewhere. Such a comparison can give an estimate of the robustness of the survey estimates. The GHS weights are derived from Stats SA’s mid-year population estimates. For this project, weighted GHS population numbers were compared with population projections from the Actuarial Society of South Africa’s ASSA2008 AIDS and Demographic model.
Analyses of the ten surveys from 2002 to 2011 suggest that some over- and under-estimation may have occurred in the weighting process:
§ When comparing the weighted 2002 data with the ASSA2008 AIDS and Demographic model estimates, it seems that the number of children was under-estimated by 5% overall. The most severe under-estimation is in the youngest age group (0 – 9 years) where the weighted numbers of boys and girls yield under-estimations of 15% and 16% respectively. The next age group (5 – 9 years) is also under-estimated for both boys and girls, at around 7% each. The difference is reduced in the 10 – 14-year age group, although boys are still under-estimated by around 1% and girls by 3%. In contrast, the weighted data yield over-estimates of boys and girls in the upper age group (15 – 17 years), with the GHS over-counting these children by about 5%. The pattern is consistent for both sexes, resulting in fairly equal male-to-female ratios of 1.02, 1.01, 1.03 and 1.01 for the four age groups respectively.
§ Similarly in 2003, there was considerable under-estimation of the youngest age groups (0 – 4 years and 5 – 9 years) and over-estimation of the oldest age group (15 – 17 years). The pattern is consistent for both sexes. Children in the youngest age group are under-estimated by as much as 16%, with under-estimates for babies below two years in the range 19 – 30%. The results also show that the over-estimation of males in the 15 – 17-year age group (9%) is much more severe than the over-estimation for females in this age range (1.4%), resulting in a male-to-female ratio of 1.09 in this age group, compared with ratios around 1.02 in the younger age groups.
§ In the 2004 results, all child age groups seem to have been under-estimated, with the under-estimate being more severe in the upper age group (15 – 17 years). This is the result of severe under-estimation in the number of girls, which outweighs the slight over-estimation of boys in all age groups. Girls are under-estimated by around 6%, 8%, 8% and 12% respectively for the four age bands, while over-estimation in the boys’ age bands is in the range of 2 – 3%, with considerable variation in the individual years. This results in male-to-female ratios of 1.10, 1.11, 1.12 and 1.14 for the four age groups.
§ In 2005, the GHS weights seem to have produced an over-estimate of the number of males and an under-estimate of the number of females within each five-year age group. The extent of under-estimation for girls (by 7% overall) exceeds that of the over-estimation for boys (at 2% overall). These patterns result in male-to-female ratios of 1.06, 1.13, 1.10 and 1.13 respectively for the four age groups covering children.
§ The 2006 weighting process yields different patterns from other years when compared to population estimates for the same year derived from ASSA2008, in that it yielded an under-estimation of both females and males. The under-estimation of females is greatest in the 0 – 4 and 5 – 9-year age groups, while the under-estimation of males is in the range 3 – 10% in the 5 – 9 age group and 1 – 6% in the 10 – 14-year age group. This results in male-to-female ratios of 1.09, 0.99, 0.96 and 1.00 respectively for the four age groups covering children.
§ The 2007 weighting process produced an over-estimation for boys and an under-estimation for girls. The under-estimation of females is in the range of 4 – 8% while the over-estimation for boys is in the range of 1 – 5%. This results in male-to-female ratios of 1.07, 1.06, 1.08 and 1.06 respectively for the four age groups covering children.
§ In 2008, the GHS weighted population numbers when compared with ASSA2008 over-estimated the number of boys aged 10 and over, in the range of 3% for the 10 – 14 age group, and 8% for the 15 – 17 age group. The total weighted number of girls is similar to the ASSA population estimate for girls, but this belies an under-estimate of female babies below two years (by 7 – 8%), and an over-estimate of young teenage girls. The GHS 2008 suggests a male-to-female ratio of 1.03 for children aged 0 – 4 years, which is higher than that of the ASSA2008 model.
§ A comparison of the GHS and ASSA for 2009 suggests a continuation of the general pattern from previous years, which is that GHS weights result in an under-estimation of children in the 0 – 4 age group (especially infants), and an over-estimate of older children. In 2009 the under-estimation in the 0 – 4 age group ranges up to 4% for boys and 5% for girls. In the 15 – 17 age group, the GHS-weighted data produce population numbers that are 7% higher than ASSA for boys, and 3% higher for girls. The male-to-female ratios in 2009 are in keeping with those in ASSA2008, with the exception of the 15 – 17 age group where the GHS-derived ratio is higher, at 1.08, compared to 1.00 in ASSA.
§ In 2010, the GHS weights again produce an underestimation of children in the 0 – 4 age group and an over-estimate of children aged 15 – 17 years. For the middle age groups, and for the child age group as a whole, there is less than 1% difference in the estimates from the two sources. For the 0 – 4 age group the under-estimate is lower than previously, at 2%, but for the oldest age group there is an over-estimate of 5%. The male-to-female ratios are similar across the two sources, although the ratio is 1.00 for all but the 0 – 4 age group in ASSA as against 1.01 for the youngest age group in ASSA and for all age groups in the GHS.
§ A comparison of the GHS2011 to ASSA2008 (projected to 2011) suggests an under-estimation of children below two years and an over-estimation of children aged 14 – 17 years in the Stats SA survey. This pattern holds for both boys and girls. The under-estimation is particularly pronounced for babies under a year, at 8%. The male-to-female ratio for all children under 17 is 1.00 in ASSA, and 1.01 in the GHS.
The apparent discrepancies in the ten years of data may slightly affect the accuracy of the Children Count – Abantwana Babalulekile estimates. From 2005 to 2008, consistently distorted male- to-female ratios means that the total estimates for certain characteristics would be somewhat slanted toward the male pattern. This effect is reduced from 2009, where more even ratios are produced, in line with the modelled estimates. A similar slanting will occur where the pattern for 10 – 14-year-olds, for example, differs from that of other age groups. Furthermore, there are likely to be different patterns across population groups.