In response to COVID-19 most governments used some form of lockdown policy to manage the pandemic. This required making iterative policy decisions in a rapidly changing epidemiological environment resulting in varying levels of lockdown stringency over time. While studies estimating the labour market effects of lockdown policies exist in both developed and developing countries, there is limited evidence on the impact of variation in lockdown stringency, particularly in developing countries. Such variation may have large heterogenous effects both on aggregate and between worker groups. In this paper, we estimate the causal effect of lockdown stringency on employment probabilities, adopting a quasi-experimental design on unique labour force panel data from South Africa. South Africa is a useful case study given its upper-middle-income status and relatively small informal sector, thus serving as an example to a variety of developing and developed country economies. We find that the negative employment effects of the country’s lockdown policy were driven by effects on the informal sector. Furthermore, we observe important effect heterogeneity by employment formality as the stringency of the country’s lockdown regulations changed over time. We find that more stringent lockdown levels negatively affected informal, but not formal sector employment, while less stringent levels negatively affected formal, but not informal sector employment. From a policy perspective, evidence of such heterogeneity can inform decisions around the optimal targeting of support as the pandemic progresses and lockdown policies are reconsidered.