Unemployment in SA: building a spatio-temporal understanding

Literature › Works

Gina Weir-Smith; Fethi Ahmed (2013). Unemployment in SA: building a spatio-temporal understanding. PositionIT (March 2013), 'Processing' technical section.

Type	journal article
Scope	Post-1994 South African economy
Method	descriptive quantitative
Period studied	1991–2007
Regions	South Africa, Eastern Cape, KwaZulu-Natal, Limpopo, North West, Gauteng, Western Cape
Themes	inequality poverty unemployment, methods data sources, demography urbanisation
Citation key	weirsmith2013

Argument

The article argues that unemployment in South Africa has a persistent, statistically significant spatial pattern that provincial or national figures obscure. To compare census-based unemployment rates for 1991 to 2007 it tackles the modifiable areal unit problem (MAUP), proportionally aggregating 1991 and 1996 magisterial-district data to 2005 municipal boundaries. Standard-deviation mapping and Moran's I show that unemployment is clustered in every year, that clustering rose between 1991 and 2007, and that the same provinces stay badly off. The authors conclude that interventions should be targeted locally, and that Stats SA should release small-area data.

Evidence and method

HSRC 1991 census data for magisterial districts and enumeration areas; Census 1996 and 2001 and Community Survey 2007 from Stats SA, at 2005 municipal boundaries; Global Insight modelled labour-market series for comparison. Narrow (official) unemployment definition throughout.

Findings

• Official unemployment at national level rose then eased slightly between 1996 and 2010. (p. 1)

  the figure for unemployment (strict/official definition) has risen from 21,0% in 1996 to 25,8% in 2001 and slightly decreased to 25,2% in 2010

• Areal interpolation of enumeration areas produced large errors where populations were small. (p. 3)

  A difference of up to 75% in the results was experienced in areas of small populations while differences in areas with larger populations were less than 1%.

• Because EA-level interpolation left gaps, the authors aggregated magisterial-district data to the 2005 municipal boundaries. (p. 3)

  the census data was aggregated proportionally from the magisterial districts to the municipal boundaries of 2005.

• Nongoma (KwaZulu-Natal) had the highest unemployment among the municipalities examined, persistently. (p. 4)

  The municipality with the highest unemployment was Nongoma in KwaZulu-Natal. This municipality showed continuous high levels of unemployment in 1996, 2001 and 2007.

• Interpolated census figures diverge sharply from Global Insight, especially for metros; the censuses also had notable undercounts. (p. 5)

  Both the 1996 and 2001 censuses reported a high undercount of between 10,7% and 17,64%.

• More municipalities fell into the worst-off category over time. (p. 5)

  increasingly more municipalities were grouped in the worse off unemployment category over time period under investigation.

• Unemployment is spatially clustered in every year and clustering intensified by 2007 (Moran's I 0,61 in 1991 to 0,71 in 2007). (p. 5)

  In fact the spatial clustering of unemployment increased between 1991 and 2007 and the Moran’s I value was 0,71 for the 2007 data.

• Provincial unemployment shows regional persistence, with four provinces above 40% in 2001. (p. 5)

  KwaZulu-Natal, Limpopo, North West and Eastern Cape recorded unemployment rates of over 40% in 2001.

• The authors infer that interventions should be spatially targeted rather than national. (p. 5)

  unemployment interventions should be focused at local areas like municipalities or at least at a provincial level instead of nationally.