Rejoinder to Simkins on Trends in Homeland Income Distribution and Poverty; and Reply to Dr. McGrath's Criticism¶
McGrath, M.D.; Simkins, C.E.W. (1985). Rejoinder to Simkins on Trends in Homeland Income Distribution and Poverty; and Reply to Dr. McGrath's Criticism. Second Carnegie Inquiry into Poverty and Development in Southern Africa, Post Conference Series No. 4. Southern Africa Labour and Development Research Unit / Second Carnegie Inquiry, Cape Town.
| Type | working paper |
| Scope | Economics and economic history of apartheid |
| Method | mixed |
| Period studied | 1960–1980 |
| Regions | South African homelands (bantustans), Venda, KwaZulu, Kangwane, Transkei, Ciskei, Bophuthatswana, metropolitan regions |
| Themes | bantustans removals influx, inequality poverty unemployment, mining migrant labour, methods data sources, demography urbanisation |
| Citation key | mcgrath1985 |
Argument¶
McGrath argues that Simkins's simulated homeland income distributions for 1960, 1970 and 1980 rest on poorly documented and unverified assumptions (equal incomes within income types, independence of income sources, constant remittance and commuter patterns, an urban poverty line). He concludes, without sensitivity analysis, that we do not know what happened to homeland poverty and distribution since 1960, though harder evidence on metropolitan incomes does not conflict with the broad direction of Simkins's results. Simkins replies that his results are a provisional first approximation, that most homeland households improved in real terms, and that a pessimistic alternative would imply implausibly extreme inequality. He concedes errors on agricultural employment, and that his method likely understates mid-range inequality, but defends his remittance assumptions, uniform poverty line and the choice of no sensitivity analysis.
Evidence and method¶
Textual critique of Simkins's 1984 Carnegie paper and its assumptions; Bureau of Market Research (BMR) surveys (Pietermaritzburg 1976, Venda 1977, Kangwane 1978, minimum living level reports); BENSO statistical surveys (1976, 1980, 1982); 1980 Population Census; South African Statistics 1980; McGrath 1984 Carnegie paper; a chi-squared test; Simkins's own simulation tables (points per adult equivalent, GNP per capita growth, rich/poor income shares).
Place in the debates¶
- Trends in homeland incomes, inequality and poverty 1960-1980 (rising incomes for most vs deepening poverty) (see Living standards, health and social policy). McGrath says the trends are unknowable from Simkins's simulations, though metropolitan evidence is compatible with Simkins's broad direction. Simkins argues that over two-thirds of homeland households improved rapidly in real terms and that inequality rose.
- Poverty-line choice: uniform urban Minimum Living Level vs separate rural MLL (see Data reliability and econometric method). McGrath says the urban MLL exaggerates poverty by ignoring rural-urban living cost differences. Simkins says a uniform convention is defensible because the rural MLL leaves non-market production unvalued.
- Validity of simulation and statistical testing (chi-squared fit; sensitivity analysis) (see Data reliability and econometric method). McGrath says the simulation is untestable and not robust. Simkins says the chi-squared test was misapplied and uninformative for a synthetic first approximation, and that sensitivity analysis is complicated in a stochastic model.
- Role of migrant remittances and commuter incomes in homeland household welfare (see Living standards, health and social policy). McGrath says remittance and commuter assumptions are arbitrary and probably vary with correlated wages. Simkins argues that homeland households benefited through remittances and commuting from integration with the metropolitan economy.
Findings¶
-
McGrath says survey data for the 1960s and 1970s cannot support actual homeland distributions, so the simulated results can at best capture a trend and cannot be tested empirically. (p. 4)
the income survey data which exists for the 1960s and 1970s is inadequate for making estimates of the actual distribution of homeland incomes.
-
The assumption that everyone receiving a given income type gets the same amount is challenged, with evidence of wide earnings and remittance variation, which would widen simulated inequality. (p. 6)
BMR data also shows that remittances received vary considerably across areas. For example, in Venda in 1977 average cash remittances received varied between R161 in Sibasa to R227 in Dzanani.
-
Remittances may be negatively correlated with local wages, so actual distributions could be more equal than simulated. (p. 7)
If these forms of income are negatively related the actual distributions will be more equal than has been simulated.
-
The chi-squared comparison of the simulated 1980 distribution with BMR data shows a significant difference at the 95 per cent level (McGrath), though Simkins later disputes the degrees of freedom and method. (p. 10)
A Chi-squared test, however, shows that they are significantly different at the 95 per cent confidence level.
-
Excluded informal non-agricultural income is substantial relative to pensions and marketed output in the homelands. (p. 11)
For 1977 non-market production excluding agriculture in the homelands has been estimated at approximately R82 million,5'which is 80 per cent of the value of pensions paid in the homelands in that year
-
McGrath's hard evidence shows slower African modern-sector employment growth than population growth and widening urban-rural inequality, not contradicting Simkins's broad direction. (p. 12)
that the growth rate of African modern sector employment (excluding domestic service) between 1960 and 1980 was slower than the growth rate of the African population.
-
McGrath's overall verdict is that without sensitivity analysis the actual trends in homeland poverty and distribution are unknown. (p. 13)
one is forced to conclude that we really do not know what has happened to poverty and income distrib-ution in the homelands since 1960.
-
Simkins's core position is that most homeland households gained in real terms, a minority lost, and relative inequality rose. (p. 18)
The core of my position is that easily the majority of. homeland households (over two-thirds) have improved their position rapidly in real termS.
-
Simkins's reduction-to-absurdity calculation: if the bottom half gained nothing, the rich-to-poor income ratio would rise from 2.79 to 8.62 rather than the estimated 3.88. (p. 20)
Rich 50% 73,6 79,5 89,6 2,79 3,88 8,62 Poor 50% 26,4 20,5 10,4
-
Simkins accepts that his 1980 homeland agricultural employment estimate was too low (413,600 vs 767,200), which overstated agricultural-income inequality. (p. 23)
total agricultural employment is 767 200, which I certainly" now accept as the best estimate. My adjustment was in the right direction but too cautious.
-
Simkins attributes rising homeland incomes mainly to integration into the wider South African economy through remittances, commuting and modern-sector jobs, and secondarily to state transfers. (p. 21)
mainly because of the integration of the homelands in the wider South African economy and secondarily because of state policy in respect of unrequired payments.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Simkins's simulated homeland household income distributions for 1960, 1970, 1980 (Carnegie Conference Paper No. 7 / Development Southern Africa 1(2)) | secondary literature compilation | 1960, 1970, 1980 | The object of McGrath's critique and Simkins's reply: a synthetic household file and simulated income distributions from which poverty and percentile income growth are estimated. |
| Bureau of Market Research (BMR) income survey data, incl. 1976 Pietermaritzburg report (Report No. 50.12, Table 12m) | household or labour survey microdata | 1976 | Illustrates wide variation in earnings by occupation, undermining Simkins's assumption that all people with a given income type earn the same. |
| BMR 1978, Income and Expenditure Patterns of Black Households in Venda, Report No. 64 (Table 3.1) | household or labour survey microdata | 1977 | Shows remittances vary across districts and are negatively correlated with wages, contradicting Simkins's independence assumption. |
| BMR 1980, Income and Expenditure Patterns of Black Households in Kangwane 1978, Research Report No. 82 (Table C1) | household or labour survey microdata | 1978 | Used to show remittances negatively correlated with wages in three of four districts. |
| BMR homeland income distribution (Simkins's comparison distribution for 1980) | household or labour survey microdata | 1980 | Benchmark against which the simulated 1980 distribution is compared via chi-squared test and Lorenz curves. |
| BMR 1974 and 1975 Minimum Living Level reports (Research Reports No. 42 and 44) | commission or official report | 1974-1975 | Supply urban and rural MLL figures used to criticise Simkins's uniform urban poverty line. |
| South African Statistics 1980 (Table 6.2) | official statistical series | 1960-1970 | Apparent source of Simkins's 1960 and 1970 pension recipient proportions, which he overrode for 1970. |
| 1980 Population Census estimates of homeland agricultural employment | population census | 1980 | Simkins adjusted census subsistence agricultural employment upward (doubling men, +50% women) as it implied implausible productivity growth. |
| BENSO 1976, Black Development in South Africa (p.47 footnote on commuters) | official statistical series | 1970 | Footnote indicating males were ~80% of commuters in 1970 versus Simkins's 71%, implying understated commuter incomes. |
| BENSO 1980, Statistical Survey of Black Development 1980 (Tables 48, 50, 52, 103) | official statistical series | 1977-1980 | Used for homeland non-market production, income composition (77% from commuters, remittances, agriculture) and GNP per capita growth comparison. |
| BENSO, Statistical Survey of Black Development 1982, Parts I and II (Part I Table 16; Part II Table 8) | official statistical series | 1980-1982 | Provides 1980 census agricultural employment for all homelands (767 200), replacing Simkins's estimate. |
| McGrath 1984, Carnegie Conference Paper No. 269 (pp.6, 26-27) | secondary literature compilation | 1960-1980 | Supplies 'hard' evidence on income growth, inequality and poverty (62% poverty in Black states 1975; 30% urban) used to assess Simkins. |
| South African consumer price index | official statistical series | 1960-1980 | Used by Simkins to inflate incomes to constant 1980 prices; McGrath questions appropriateness. |
| Hansard 1984, No. 12, Col. 5356 | parliamentary debates | 1984 | Cited to show the Prime Minister referred approvingly to Simkins's simulations. |
| Frontline article (Helen Zille, July 1984) and Development Southern Africa review (Coetzee and van der Kooy) | newspapers periodicals | 1984 | Quoted to illustrate the range of reception of Simkins's results. |
| Anthropological literature, Surplus People Project results, Transkei material, UCT work in progress | secondary literature compilation | n/a | Simkins says his individual remittance assumptions were based on judgement informed by these. |
Inference¶
Quantification: Descriptive tables.
McGrath's critique is a desk audit of the assumptions of Simkins's stochastic microsimulation, using BMR survey tables and BENSO statistics to show assumptions are doubtful, plus a chi-squared test of fit. Simkins replies with argument, a reductio ad absurdum inequality calculation, comparison with BENSO growth rates, and Lorenz curve comparison, while conceding the agricultural employment error.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Simkins's results rest wholly on unverified assumptions and cannot be empirically tested because adequate 1960s/1970s income survey data do not exist. (p. 4) | McGrath's reading of Simkins's paper; no direct source test possible. | Asserted |
| Survey data show that incomes and remittances vary within income types and that remittances are negatively correlated with wages, so Simkins's equal-receipts and independence assumptions are doubtful. (p. 4) | BMR Pietermaritzburg 1976, Venda 1978 and KaNgwane 1980 survey tables, cited for isolated examples. | Illustrative or anecdotal |
| Simkins's simulated 1980 distribution differs significantly from the BMR homeland distribution (chi-squared), though Simkins argues the test was misapplied and uninformative. (p. 7) | Comparison with BMR data; McGrath's chi-squared (X2=20.69), disputed by Simkins who says it used percentages and wrong degrees of freedom. | Partial data |
| Using an urban MLL for all homelands exaggerates the level of poverty though perhaps not the trend. (p. 7) | BMR 1974/1975 MLL reports giving urban vs rural minimum living levels; McGrath 1984 poverty estimate. | Partial data |
| Simkins's homeland agricultural employment for 1980 was too low; the correct figure is 767,200 rather than 413,600, so inequality in agriculture was overstated. (p. 20) | BENSO 1982 Parts I and II using 1980 census figures for all homelands. | Systematic data |
| Over two-thirds of homeland households improved rapidly in real terms; the pessimistic alternative would require implausibly large rises in inequality (richer/poorer ratio of 8.62 versus 3.88 estimated). (p. 15) | Simkins's simulation results compared with BENSO real GNP per capita growth rates (Tables 1 and 2). | Partial data |
| Hard evidence (McGrath) on metropolitan urbanisation, income growth and inequality is not in conflict with the broad direction of Simkins's results, but without sensitivity analysis we do not know what happened to homeland poverty and distribution. (p. 13) | McGrath 1984 pp.26-27 plus the assumption audit. | Partial data |
The author's caveats¶
McGrath notes that income survey data exist for only four homelands in 1980 and are inadequate for the 1960s and 1970s. Simkins concedes his results are provisional, a first approximation, and that the agricultural employment figures were too cautious. Both acknowledge that BMR and Simkins measure different things and that BMR data are subject to sampling error.
Gaps in the evidence¶
The debate lacks representative household income surveys for 1960 and 1970 covering the homelands, and any sensitivity or Monte Carlo runs under alternative assumptions on remittances, commuter wages and pensions. Data on the effects of boundary changes and removals, region-specific price indices and rural poverty lines, and informal incomes are also missing. Some of this (1980 census details, BMR surveys) existed then, but historical household distributions cannot be reconstructed with certainty.
Replicability: Data poorly documented.