Premature deagriculturalisation? Land inequality and rural dependency in Limpopo province, South Africa*

Abstract

Cross-national regressions reveal abnormally low agricultural workforce shares, given GNP, in developing countries that had historically concentrated land into large capital-intensive farms. We argue that such deagriculturalisation was premature, since its concomitant labour shedding has undesirable outcomes. In a new South African survey, a large proportion of rural households (and working-age persons) was ‘dependent’, relying for income almost wholly on either migrant remittances or pensions. A separate group (with less poverty and unemployment) relied mainly on local, including own-farm, income. The group was heavily over-represented in one of the three regions, where many more households had significant land.

Notes

1. Among unwanted side-effects, deficient rural earning opportunities have stimulated urban slum growth (and associated unemployment and crime) – but not emigration. South Africa is a net immigrant country, partly because of sluggish agricultural earning opportunities elsewhere in Africa. Recent sharp falls in agricultural employment suggest that rural dependency, although a long-run effect of land distributions associated with apartheid, may have intensified since its demise. Total paid farm employment fell only slowly in 1985–93 (from 1.32m to 1.14m). However, from 1994 to 2001, total employment in agriculture, fisheries and forestry – including self-employment – dropped by some 66 per cent (Vink and Kirsten, 1999).

2. In South Africa, 15 per cent of farmland is divided among about a million mainly African farm operators (mostly part time), leaving 85 per cent of land with some 60,000 white commercial farmers outside the surveyed (smallholding) areas (NDA 2001; Department of Land Affairs, 2002). About 2 per cent of land in large white-owned farms in 1994 had been redistributed up to the end of 2001, only about 5 per cent of it through DLA projects and the rest through private transactions.

3. This paper derives from EU-supported research on the impact – on fertility, migration, and thus and otherwise agro-environmental sustainability – of land and asset size and distribution in selected rural drylands of Limpopo province, Rajasthan, and Botswana.

4. Such healthy RNF growth may also be happening in parts of Sub-Saharan Africa (Reardon, 1997; Bryceson, 2000; Reardon and Barrett, 2000) where land is less unequal than in Limpopo.

5. Bryceson defines ‘deagrarianisation’‘as a process of occupational adjustment, income-earning reorientation, social identification and spatial relocation of rural dwellers away from strictly agricultural-based modes of livelihood’, whereas ‘“depeasantisation” represents a specific form of deagrarianisation in which peasantries lose their economic capacity and social coherence, and shrink in size’, Bryceson (2000: 1–2).

6. A specific-factors model, with agricultural capital unimportant, becomes less plausible as development proceeds. Empirical H-O models explaining agriculture's trade share include Leamer (1984), Lal and Myint (1996), Wood (1994).

7. One imperfection is insufficient: e.g., non-tradability of (family) labour, gives heterogeneity in farm size – large families have large farms – but all physical ratios are scale-invariant.

8. See above sources and Binswanger et al. (1995). Booth (2002: 85) shows ‘post-Green-Revolution’ Indonesian farm income/ha. steadily falling from Rp718,000 on holdings below 0.1 ha to Rp23,000 above 4.5 ha. Land productivity falls less sharply than the labour–land ratio, since labour productivity rises.

9. However, some endogenous differences – smaller, labour-intensive farms doing more to upgrade their land – are implied by higher yields on smaller holdings on land of similar region, type or quality (Berry and Cline, 1979; Lipton, 1993; Binswanger et al., 1995).

10. By symmetry, the release of labour from agriculture might raise non-agricultural output by lowering wages. Such effects may be assumed to be small in labour-surplus economies.

11. We include all countries with available data. To allow for variations in country size we weight observations by the square root of total workforce. If countries may be viewed as aggregates of independent regions, then the error variance decreases with country size suggesting this initial heteroscedasticity correction (Blanchet, 1988).

12. These results might have been distorted by cross-country variation in the share of extractive industry in non-agricultural GDP, but this variable had no statistical significance when included in our equations.

13. The unweighted means are 0.82 and 0.56.

14. See, for example, Bennett and Powell (1999), Reynolds (2003).

15. The proportion of households selected differed across villages. Unless otherwise stated, all statistical calculations use weights inversely proportional to the probability of a given household having been selected.

16. We depart in one respect from the administrative regional sub-division: on geographical and agro-ecological grounds, Zebediela sub-district is included in our ‘South’.

17. African tenancy had little impact on land access in Limpopo. On white farms it was prohibited, though this did not prevent some ‘labour tenancy’. In African areas tenancy is minimal, due to customary law.

18. At PPP exchange rates, sample income per adult equivalent is 50 per cent higher in Limpopo than in Rajasthan. Using the official rural poverty line in Rajasthan, household poverty incidences are respectively 20 per cent and 29 per cent.

19. Henceforth we omit the qualifiers; our surveys are restricted to rural dryland areas in both cases and additionally to former homeland areas in Limpopo.

20. Our ‘income’ measure is imperfect. We estimate production for own-consumption using household reports of the fraction of needs so met. Farm income is defined as production for own-consumption plus sales, i.e., without subtracting input purchases, on which data were unavailable.

21. This risks double-counting if smallholders earn wages on each other's farms (rather than on large commercial farms, as happens to an unknown extent in Limpopo), so the totals given are likely to be overestimates.

22. Estimated employment in agriculture, including self-employment, is only 22 per cent of all rural employment in Limpopo (about 44 per cent in Rajasthan). The employment numbers, estimated from individuals who reported a sector of work, must be treated with caution; some individuals may have failed to report a sector of work, even though they work on their own land.

23. Access to common grazing land is not taken into account in our measure of landholding. However, absence of cropland is strongly associated with lack of livestock, and therefore with lack of benefits from common grazing.

24. The null of equality of income per household between landed and landless households cannot be rejected in Limpopo, but is rejected in Rajasthan at the 1 per cent level.

25. The asset value and share estimates must be treated with caution; see Table 2, note (c). In the Limpopo ex-homeland area, land is rarely traded; the assumed value (1500 Rand/ha) is based on market prices for comparable land in adjacent commercial farming areas (Department of Land Affairs, 2002a). Livestock aside, other asset valuations in Rajasthan and Limpopo are by household respondents.

26. Our poverty line is the rural expenditure per person needed in Rajasthan to achieve the Indian poverty line (Dr S. Sharma, Delhi School of Economics, personal communication, using 1999–2000 round, National Sample Survey). We converted this into Rand using the PPP exchange rate. In both Rajasthan and Limpopo we used sample mean adult equivalents per household to convert the per person line into a per-aeq line, which for Limpopo is 1594 Rand/aeq/yr. Lacking expenditure data for Limpopo we had to ignore below-poverty-line differences between expenditure and income.

27. In Rajasthan too, there is a strong negative association between poverty risk and landholding.

28. Compare Rajasthan, where the poor, of whom only 7 per cent are landless, derive a high share of income from farming and a correspondingly low share from wages.

29. Depth of poverty, however, is significantly lower for pension-dependent households (defined below; estimated mean income per adult equivalent among the poor is 23 per cent below the poverty line) than for others (around 50 per cent).

30. The concentration index is analogous to the Gini. For instance the associated Lorenz curve for, say, largestock has households ordered by wealth on the x-axis and the cumulative share of total largestock on the y-axis.

31. As for gender, the female/male ratio is 2.10 in MD households as against 1.34 for FRs. However, this is probably an effect of selective male migration, not a cause of MD status.

32. The weak association may also be partly due to offsetting effects of much higher unemployment among both households with pensions (which have high poverty incidence) and with migrancy income (which have little poverty) than among households with neither.

33. A rank correlation test is more appropriate than a linear one given the highly bunched nature of the data (many zeroes). The correlation for local income against (pensions plus remittances) was – 0.44 (p = 0.000). The same tests on our Rajasthan sample give a very small positive, insignificant correlation between local income and pensions and small negative correlations between local income and, respectively, remittances (rho = −0.12, p = .003) and pensions (rho = −0.10, p = 0.02)

34. As one would expect (because the definitions of PD and MD select households that have little local income and accordingly low employment of residents), the male unemployment rates in the PD and MD households that together account for 59 per cent of all households are even higher, at 93 per cent and 92 per cent.

35. The poor have a higher share in dwellings, consistent with the presence of capital market imperfections that may prevent even the landed poor from escaping poverty through more effective land use.

36. The 1995 Central Statistical Services's Household Survey confirms this picture (Gyekye and Akinloye, 2001)

37. Both living in West and landholding are strongly associated with both the possession of inanimate farm assets and livestock and the quantity possessed.

38. Point estimate 1.6 per cent (two sample households).

39. Including teachers but not parastatal employment.