Skip to content

Forming Representative Household and Factor Groups for a South African SAM

Literature › Works

Kalie Pauw (2005). Forming Representative Household and Factor Groups for a South African SAM. PROVIDE Project Technical Paper Series, 2005:2. PROVIDE Project, Elsenburg.
Type report
Scope Post-1994 South African economy
Method descriptive quantitative
Period studied 2000–2000
Regions South Africa, Western Cape, Eastern Cape, Northern Cape, Free State, KwaZulu-Natal, North West, Gauteng, Mpumalanga, Limpopo
Themes methods data sources, inequality poverty unemployment, bantustans removals influx, demography urbanisation, labour markets wages, education human capital
Citation key pauw2005

Argument

The paper documents how representative household groups (RHGs) and factor groups were built for the PROVIDE Social Accounting Matrix (SAM) for CGE modelling. Groups should reproduce socio-economic stratification, be relatively homogeneous and be recognisable for policy purposes. Households are disaggregated by province, race, farming status, homeland residence, sex and education of the head, and income, giving 162 groups. Factors are disaggregated by province, race and occupational skill. Alternative groupings (income only, race-urban/rural-income, adult-equivalent income) are discussed and tested with Hotelling's T2 on expenditure patterns. The paper is a technical data-construction note; it argues that race and homeland location remain analytically necessary because of apartheid's legacy.

Evidence and method

Statistics South Africa's Income and Expenditure Survey 2000 (IES) merged with the Labour Force Survey of September 2000 (LFS); Agricultural Survey 1996; summary statistics tables of household income and wages by group; comparison with other SAMs (Tanzania, Malawi, Botswana, Zimbabwe, Bangladesh).

Findings

  • The SAM household accounts are disaggregated on seven dimensions, driven partly by apartheid's history. (p. 12)

    The household accounts of the PROVIDE SAM are disaggregated by province, race, farming/non-farming, homelands/non-homelands, gender of the head of the household, education level of the head of the household, and income of the household.

  • A total of 162 household groups were formed; sparse cells forced mergers (e.g. Coloured with Asian). (p. 13)

    A total of 162 household groups were formed.

  • The paper links migrant labour to the prevalence of female-headed households and 'fractured families'. (p. 13)

    The country has a very large share of female-headed households who are left to run the household while their husbands search for work on mines and in the cities.

  • Inter-racial income differences are very large in 2000: white adult-equivalent income is about seven times African income. (p. 43)

    White households earning on average more than twice as much as Asians, four times as much as Coloureds and about seven times as much as Africans.

  • Former homelands remain home to a large population share despite a small land share. (p. 60)

    The former homelands areas constitute less than 13% of the total land area of South Africa, but is still today home to 27.1% of the population and more than one third of all Africans (IES 2000).

  • Poor households spend a far higher share of spending on food; expenditure patterns differ significantly between poor and non-poor. (p. 43)

    Poor households spend an average of 54.2c on food per R1.00 spent compared to 30.0c of non-poor households.

  • Rural-urban income and household-size gaps are large, with rural areas dominated by African households. (p. 46)

    In South Africa the average adult equivalent per capita income in rural areas is only R7,438 compared to R25,986 in urban areas.

  • Agricultural households are a small share of the population and much poorer than non-agricultural households. (p. 64)

    Agricultural households earn an average income of R22819 compared to non-agricultural households' average income of R46587.

  • Factor groups are formed by province, race and occupation, with skill categories aggregated from LFS occupation codes. (p. 34)

    Typically, factor codes 1 to 3 were grouped together as highly skilled workers, codes 4 – 5 as skilled, codes 6 – 8 as semi-skilled and codes 9 – 11 as unskilled.

  • Hotelling's T2 tests showed Asian expenditure patterns resembled White ones in several provinces, informing merging decisions. (p. 48)

    Hotelling's T2-test revealed that expenditure patterns of Asian households in the Western Cape, Northern Cape, Freestate and North West are a closer match with White household expenditure patterns than Coloured or African households.