Job Creation and Destruction in South Africa

Literature › Works

Andrew Kerr; Martin Wittenberg; Jairo Arrow (2013). Job Creation and Destruction in South Africa. SALDRU Working Paper Series, Number 92. Southern Africa Labour and Development Research Unit, University of Cape Town.

Type	working paper
Scope	Post-1994 South African economy
Method	descriptive quantitative
Period studied	2005–2011
Regions	South Africa
Themes	labour markets wages, manufacturing industrial policy, inequality poverty unemployment, methods data sources, transition post apartheid
Citation key	kerr2013a

Argument

Using Statistics South Africa's Quarterly Employment Survey (QES) enterprise panel for 2005-2011, the paper gives the first firm-level panel analysis of job creation and destruction in South Africa. It finds gross job reallocation of about 20% a year, close to US and OECD levels, which it reads as little evidence that labour regulation makes hiring and firing rigid. Small firms destroy more jobs than they create, so larger firms have higher net job creation, and net creation in manufacturing is negative. The authors argue that undercounted firm births are unlikely to overturn the size result, so the National Planning Commission's expectation that small and medium firms will drive job growth is doubtful. The method follows Davis, Haltiwanger and Schuh, using average firm size and quarterly and annual flows.

Evidence and method

Confidential Statistics South Africa QES enterprise-level panel (three overlapping samples, 28 quarters, about 28,599 private enterprises, covering roughly 45-55% of private non-agricultural, non-mining employment) plus the business register and response codes. Compared with published international job-flow studies (Davis et al., Haltiwanger et al., Hijzen et al., Sandefur and others). Weighted descriptive tables and relative standard errors.

Findings

• Annual gross job creation and destruction average around 10% each, so about 20% of jobs turn over each year. (p. 13)

  Job creation rates are about 9.9 % on average across the waves, whilst the job destruction rates suggest that, on average, around 10.5% of existing jobs are destroyed in a 12 month interval.

• Reallocation is similar to US manufacturing and below transition and Latin American economies, despite South Africa's reputation for rigidity. (p. 13)

  This job reallocation figure is very similar to those calculated for US manufacturing plants by Davis et al. (1996), and is surprising given that South African firms rank labour regulation as a key regulatory constraint to business

• Enterprise deaths account for about 27% of job destruction, more than in the UK or US studies. (p. 13)

  enterprise deaths contribute an average across all waves of 27% to total job destruction.

• Births contribute only about 11% of job creation, which is a sign of undercounting births in the QES. (p. 13)

  Births contribute around 11% to job creation in the QES.

• Manufacturing shows persistent net job destruction and declining employment. (p. 14)

  Job creation is lower than job destruction in 15 of 24 quarters for which we can calculate job creation and destruction statistics in manufacturing

• Formal manufacturing employment fell about 7% between June 2006 and December 2011. (p. 15)

  between June 2006 and December 2011 there was a decline in formal manufacturing employment, as measured in the QES, of about 7%.

• The smallest firms destroy far more jobs than they create, while the largest firms are net creators. (p. 15)

  the category of smallest firms contributed about 75 000 jobs to yearly gross job creation, but around 110 000 jobs to yearly gross job destruction.

• Even large corrections for births would not make small firms net creators, so the NPC's SME-led job growth expectation is questioned. (p. 16)

  to generate even zero net job creation in the smallest firms, births would have to contribute around 45% of job creation, which is unlikely.

• South Africa's firm size distribution is dominated by large firms, and informal micro-enterprises are largely missing. (p. 10)

  the median worker in South Africa works in an enterprise with 140 employees, whilst when limiting the sample to manufacturing the median worker works with 155 other employees.

• The authors expect that the true reallocation rates are higher than reported because of data limitations. (p. 14)

  the average gross reallocation rate could be between 1 and 2 percentage points higher if we could observe job creation and destruction rates for all firms in all waves.