Who are the Robots Coming For? The Evolving Task Content of Employment in South Africa

In this paper, we examine how the task content of employment in South Africa, a developing country, has evolved in the post-apartheid period. By investigating the South African labour market’s evolving task content, we are able to assess whether there is evidence of increased utilisation of automation and other 4IR technologies. We find that the South African labour market in the formal private sector has undergone a pattern of relative de-routinisation through a relative contraction in routine manual jobs and an expansion of non-routine cognitive analytical jobs. In absolute terms, although employment within all task content component groups grew over the period, non-routine jobs experienced far greater rates of jobs growth relative to routine jobs. Despite representing just 4% to 6% of workers, employment in non-routine cognitive analytical jobs more than doubled. Employment in routine jobs, which represents most workers (75% to 81%), also grew, but at a much slower rate. In relative terms, the share of routine manual jobs shrunk significantly over time, while those of all other task content component groups grew, especially non-routine cognitive analytical jobs. Most of these changes occurred between 2000 and 2010. This aggregate pattern of relative de-routinisation is driven by similar trends in the mining, manufacturing, construction, transport, storage and communication, and community, social and personal services industries. We also observe this pattern in both small and large firms alike. We find similar evidence when considering trends in annual entries into employment; that is, while the number of recent entries has grown within all task content components, those into nonroutine cognitive (analytical) jobs has grown the fastest. While we do not find evidence of relative de-routinisation when considering employment exits, this does not necessarily invalidate our previous findings, given data comparability concerns across the survey instruments. Finally, we document considerable variation in the demographic and labour market characteristics of workers across these groups of occupations.