The gendered effects of the Covid-19 crisis in South Africa: Evidence from NIDS-CRAM waves 1–5

ABSTRACT

There is substantial evidence in the global literature that women have been disproportionately affected by the Covid-19 crisis and associated lockdowns. In this paper, we use five waves of the National Income Dynamics Study - Coronavirus Rapid Mobile Survey (NIDS-CRAM) to track the gendered effects of the pandemic in South Africa during the first year of the crisis. We focus here on the impact in the labour market and on unpaid care work in the home. We find that, relative to men, women were more severely affected in both spheres, yet they benefited less from the government income support provided to unemployed and furloughed workers. The evidence from the NIDS-CRAM survey from the first year of the pandemic therefore suggests that gender inequality in South Africa has risen, undoing some of the gains of the last two decades.

KEYWORDS:

• Gender
• employment
• childcare
• Covid-19
• South Africa

JEL:

• J16
• J21
• J13

Acknowledgements

The authors would like to thank Nic Spaull as well as the other NIDS-CRAM team members for their support and collegiality. Special thanks also to the reviewers on our working paper series, Reza Daniels, Murray Leibbrandt, Nic Spaull and Ingrid Woolard, as well as the two anonymous reviewers of this paper.

Disclosure statement

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

Notes

1 The much longer working paper version of this paper (Casale and Shepherd 2021b), which contains additional detail about the gendered impacts of the crisis over the first year and more disaggregated data analysis, is available at www.cramsurvey.org. More detail on the results from the first wave of NIDS-CRAM specifically, namely on the impact of the initial strict lockdown in April 2020, can be found in Casale and Posel (2020; 2021).

2 For a fuller discussion on why the results from the NIDS-CRAM survey and QLFSs over the 2020–2021 period would differ, see Daniels et al in this issue.

3 Some of the Wave 1 figures presented in this paper will be marginally different from those displayed in Casale and Posel (2020; 2021), which described gendered findings from the first wave of the data. This is because updated weights for the earlier waves were subsequently released by the NIDS-CRAM team (see Ingle et al 2021) and are used here. However the gendered patterns and trends remain the same.

4 In keeping with ILO definitions, a person is defined as employed in the survey if they had worked in ‘any kind of job’, had done ‘work for any profit or pay, even if just for an hour or a small amount’, had been engaged in ‘any kind of business such as selling things - big or small – even if only for one hour’, or if they had a job/activity that they would return to in the next 4 weeks.

5 As an additional check, we restricted the sample to those who remained in the panel across all five waves, and analysed average hours worked among all the employed, regardless of whether they stayed employed over the whole period or not. We found a very similar series to that shown in Figure 3 for the unbalanced panel. Women’s hours worked fell from 35.4 h in Feb 2020 to 33.0 h in March 2021 or by 2.4 h per week (compared to 35.3 h and 33.4 h in Figure 3 or 1.9 h per week), while men’s increased from 39.1 h to 40.9 h (compared to 38.8 h and 40.2 h in Figure 3). This suggests that the gender differences in hours worked are not being driven by the varying sample of individuals appearing in the panel over time (if anything, the gender differences are slightly more pronounced when limiting to the balanced panel).

6 Due to space constraints, we do not explore earnings among the employed. Changes in earnings are more complex to assess using NIDS-CRAM, as the baseline question on earnings in February 2020 was not asked in the same detailed manner as for the remaining months. As with hours worked, it is likely that earnings changes are being driven to some extent by compositional issues. Casale and Shepherd (2021b) track mean earnings by gender over the first year of the pandemic and find that, among the employed, mean earnings fluctuated considerably and changes in earnings were generally in the opposite direction to changes in employment. Because job losses hit harder at the lower end of the earnings distribution (Casale and Posel 2020), the changes over time are likely reflecting that those who retained employment were a more select sample of higher-earning individuals. For a fuller discussion of the trends over the period, see Casale and Shepherd (2021b) and for a detailed analysis of changes in the gender wage gap between February and June 2020, see Hill and Köhler (2020).

7 When asked who was looking after these young children (0-6 years) now at home, around two-thirds of women said they were looking after these children themselves compared to 14-25% of men across the various waves (Casale and Shepherd 2021b).

8 Similarly large increases were found in the UK during the first lockdown, with parents reporting the equivalent of an extra 40 h a week on childcare, almost a full workweek’s worth of additional care (Sevilla and Smith 2020).

9 In NIDS-CRAM Wave 2, only 4.9% of respondents (4.0% for women and 5.9% for men) reported employing a domestic worker/childminder in June 2020, higher than the April 2020 figure of 3.3% (2.5% for women and 4.3% for men), but still down from 8.0% in February 2020 (7.4% for women and 8.7% for men).

10 NIDS-CRAM was unable to collect detailed school attendance information per child per day, but there is some suggestion that the majority of children living in the sampled households had returned to school in some capacity by the Wave 3 interviews, and by the Wave 5 interviews, proportions attending were estimated to be between 93-94% (Mohohlwane et al 2021, Shepherd and Mohohlwane 2021).

11 Wills et al (2021) estimate that only 31% of children aged 0–6 years were attending an ECD centre using the Wave 3 NIDS-CRAM data, down from about 42% pre-Covid. One of the main reasons cited for children not returning was that the centre had not reopened (with inability to afford fees another important factor).

12 Collecting information on time use in a one-shot question that relies on recall over the phone is not ideal (in typical time use surveys, individuals are asked to keep a time diary and to respond in half-hour slots). It is therefore possible that respondents either over- or under-stated the hours spent on childcare. Assuming women and men have similar reporting biases over time, one should pay more attention to the gender differences in childcare hours and the relative changes over time, than the point estimates of hours reported.

13 It is reassuring that the figures reported in NIDS-CRAM are very close to those reported by SASSA in their monthly grant report to Nedlac on 30 June 2020. According to that report, as at 27 June 2020, a total of 2 712 963 successful applicants had been paid, 35% (or 951 972) of whom were women.

14 The gender gaps in Casale et al (2021) are for the age group 15 to 65, while the gaps quoted here based on the NIDS-CRAM data are for adults 18 and older. However, when we recalculate the gaps for the population aged 18–65 in NIDS-CRAM, the figures barely change.

15 After some delay, the government did re-introduce several income support measures in August 2021, specifically the UIF-TERS for affected industries and the SRDG, with caregivers eligible from September 2021. But it is not yet clear if a more permanent solution to the holes in the social protection system will be implemented.

Additional information

Funding

Funding for the broader NIDS-CRAM project was provided by the Allan & Gill Gray Philanthropy Fund, the Federated Employer’s Mutual Education Fund, and the Michael & Susan Dell Foundation.