Trends in household formation and living alone in South Africa, 1995–2011

ABSTRACT

In South Africa, households were formed at about twice the rate that the population grew be-tween census 1996 and census 2011 and the number of single-person households ballooned by 150%. Reweighted household survey data shows a surge in household formation in the late 1990s was driven by prime-aged and older women and Black African men, likely connected to new freedoms afforded to these groups after the transition to democracy. Household formation steadied in the 2000s, hiding variation in who formed what types of households. Astonishing growth in the rate at which South Africans live alone was led by Black African men, a group historically associated with circular labour migration. Women instead are heading up complex households including children. These changes connect to long-term marital decline. By 2011, most female heads were never-married and the growing majority population group of never-married adults increased their rate of household formation.

KEYWORDS:

• Households
• household formation
• living alone

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 From 4.55 people in census 1996 to 3.57 in census 2011 according to own calculations using the census 10% samples.

2 Male and female strict unemployment rates increased from 17% and 25%, respectively, in 1994 to 26.8 and 31.5% in 2019 (Casale et al. 2021)

3 I use the term ‘Black’ to apply to groups of the population classified as ‘African’, ‘Coloured’ (people of mixed-race heritage from the Cape and also associated with a distinct cultural identity); and ‘Indian/Asian’. These are apartheid-era classifications which Statistics South Africa still uses to collect demographic data owing to the continued importance of understanding, quantifying and monitoring the legacy of apartheid in the post-apartheid era. The fourth population group category is ‘white’.

4 The 1980 census excluded the independent homelands of Transkei, Bophuthatswana, and Venda.

5 Another recent example is the study by Posel & Hunter (2022) who plot rates of living alone, although their study is also not specifically focused on rates.

6 OHS data exist for 1993 and 1994 but I exclude these years. I exclude 1993 because the 1993 survey had a different sampling frame to the later surveys in that it excluded the ‘independent’ homelands of Transkei, Bophuthatswana, Venda, and, Ciskei. OHS 1994 also differed to later surveys in that it oversampled whites and undersampled Black Africans. These differences proved pertinent when trying to recalibrate the survey weights and so I omit this year for quality concerns.

7 The 2004 Master Sample used for GHS 2005–7 was stratified at the district council level, although StatsSA caution that the data is not representative at this level and more recently released versions of the GHS for 2002–7 do not include a district council variable (DataFirst 2015).

8 Exceptions are that the 1996 and 1998 October Household Surveys only surveyed about 16 000 and 20 000 dwelling units, respectively, due to budget constraints.

9 The headship rate for white women in 2011 was 24% household-weighted, but 19.3% person weighted. In the same year, the headship rate for Black African men was 45.6% person-weighted and 50.1% household-weighted. In both cases, 95% confidence intervals do not overlap.

10 See Figure 5 in Thornton & Wittenberg (2022b).

11 The release of the 2022 census data should provide the benchmark information to update the survey weight soon. This will depend on a quality assessment of the data, however, as the enumeration of the 2022 census was challenged by the conditions of the Covid-19 lockdown.

12 The African and Indian respondents from KwaZulu-Natal province in the 1993 South African Living Standards Survey were re-interviewed in 1998 for the KwaZulu-Natal Income Dynamics Study. From this two-wave panel, Klasen & Woolard (2008) find 96% of household heads or spouses who were alive and resident in 1993 were still head or spouse in 1998. The few ‘demotions’ from the position of head had an average age of 67.

13 This age cut-off is chosen because it corresponds with the age that one can legally start working in South Africa.

14 Black Africans are by far the majority population group representing 80% of the population, or 41 million people, in 2011. This means the trend for Black Africans largely determines the overall trend. The other race categories make up the following percentages in 2011: Coloured (9%), Asian/Indian (2.5%) and White (9%).

15 Close inspection of the 1996 and 2001 census reveal this was driven by a sizeable decline in household headship rates of white men aged less than 35 years.

16 Practical and conceptual data constraints on the measurement of circular labour migration makes it difficult to ac-curately quantify a national-level trend and therefore how this would link up with the figures in Table 1. Most national household surveys do not ask questions allowing migrant-destination households to be identified. Some allow for migrant-origin households to be identified and Posel (2020)’s assessment of these data reveals a considerable decline in the share of households that are labour-sending between 1993 and 2014. However, Posel (2020)’s figures will omit labour-sending households in South Africa’s neighbouring countries which are the origin of a notable share of migrants (Ginsburg et al. 2016; Wittenberg & Collinson 2007) and the trend is complicated by whether migrants are temporary, permanent, or somewhere in between. Some migrants to urban areas do not intend to return permanently to their rural homes, but nevertheless consider themselves members of those households (Bank et al. 2020) pointing to the conceptual difficulty of defining households in South Africa.

17 These include the outlawing of marital rape (Karimakwenda 2020); legislation preventing domestic violence (Vetten 2013); and, the legalisation of abortion (Althaus 2000).

18 See Appendix Figure A1.

19 See also Appendix Figure A1.

20 The labour market statistics use 1996 as the base year because employment numbers look too high in 1995 across the board as noted in Wittenberg (2014). This is visible in Panel A of Figure 4.

21 These changes over time hold true even if the sample is divided into those of working-age and pension-age (although levels change so that working-age heads are more likely to be employed than pension-age). Also note that household heads and people living alone are more likely to be employed than non-heads in general.

22 Women historically qualified for pension-receipt upon turning 60 years of age whilst for most of the period men only became eligible at age 65. However, this changed between 2008 and 2010 when age-eligibility for men was brought down to 60 as well (van der Berg et al. 2010).

23 This age restriction is used to maintain consistency with the rest of the analysis, although it must be acknowledged that it is illegal to marry under the age of 18 in South Africa.

24 The response ‘living together’ which should identify cohabiting spouses is absent from the 2002 to 2004 GHS questionnaires.

25 Indeed, by 2011, this very large group must be quite heterogeneous. For now, I bluntly treat them as one group, but try to at least account for age in this section.

26 This pattern holds if the sample is divided into working-age and pension-age.

27 This might be more contentious for those aged 15–18 years who I have classified as adults to be consistent with the rest of the paper. The results are robust to adjusting the age cut-off to 18 years.

Additional information

Funding

This research was completed whilst I was receiving funding from DataFirst and the African Centre of Excellence in Inequality Research.