Exploring Seasonal Poverty Traps: The ‘Six-Week Window’ in Southern Malawi

Abstract

Conventional wisdom in Malawi holds that seasonal food deficits force smallholders to hire out their labour to buy food during the critical first six weeks after planting, thereby reducing maize yields on their own fields and reinforcing poverty. This model was tested empirically for the Blantyre Shire Highlands using evidence from a panel survey and household case studies. Results showed no significant relationship between the supply of hired labour (ganyu) and the timeliness of weeding for maize, and that timely weeding was not a significant determinant of household maize deficits. This puzzle is explained by the diversification of the rural economy that has allowed households to develop alternative livelihood strategies to cope with maize deficits. Livelihood diversity helped poor households avoid the seasonal poverty trap and also helps explain the paradox of why the poorest region in Malawi escaped the 2001–2002 famine.

Acknowledgements

The authors wish to thank two referees for helpful comments. Research was jointly funded by the Department for International Development (DFID) and the Ministry of Agriculture and Irrigation (MOAI) of the Government of Malawi. However, the views expressed are those of the authors and are not necessarily those of the DFID or MOAI.

Notes

1. Literature reviewed by Gill (1991) and Sahn (1989) contained no empirical tests of the archetypal poverty trap outlined by Chambers. A search of mainstream journal literature from the 1990s found articles that analysed seasonality and risk, seasonal labour constraints, and seasonality and wages but none that analysed how labour allocation might reinforce food deficits among rural households.

2. One Indian village study did find evidence of a draught power constraint because owners would not hire out their animals for fear of losing caste by working for others (Bliss and Stern, 1982: 102–103). In this case, cultural factors prevented the emergence of a rental market for draught power. In Africa, the main constraint on rental markets is the short time available for land preparation, which limits the returns to renting-out tractors or oxen (Pingali et al., 1987: 68, 156). Seasonal poverty traps in African agriculture may therefore be more common where tillage depends on oxen rather than the hoe.

3. The poor harvest refers to the 1996–1997 crop season in the study area, when continuous heavy rain delayed weeding, flooded dambo fields, and leached out fertiliser nutrients.

4. Ganyu in this context refers to agricultural employment paid in cash or kind. In practice ganyu is a blanket term that covers a wide range of wage work, including non-farm jobs (Peters, 1996).

5. Households averaged 12 days/month of ganyu between October and February. On farms between 0.25–0.99 ha, this accounted for less than one-quarter of available labour days (Leach, 1995: 19). Even if household labour supply was reduced by illness or if crop management activities were compressed by late rains farms of 0.5 ha (which as a proportion of their labour time supplied the most ganyu) would have more than enough labour both for ganyu and for their own fields (Leach, 1995: 29). This study was made in the 1993–1994 crop year. The supply of ganyu may have risen since then in response to the collapse of smallholder credit and higher fertiliser prices.

6. Supply and demand for hired labour is also determined by the price of maize and wage rates for ganyu. We assume that the price of maize is the same for all households during the six week window. The supply response of ganyu to changes in maize price could be estimated if we had data for more than one year. We collected data on earnings from ganyu but not on wage rates.

7. The data was collected in 1998 before the 2001–2002 famine and when all rural households were allocated Starter Packs of free fertiliser. It is fair to ask whether this might reduce the relevance of our conclusions. Yet the evidence does not suggest these had a significant effect on the market for ganyu, at least in the Shire Highlands. PRA in adjacent Zomba district found a structural shift in the market after the fertiliser price hike in 1996, with sharply reduced demand and increased supply (Devereux et al., 2007: 28). This is inconsistent with evidence of a rise in wage-rates for kuwerenga ganyu in our study area between 1997–1999 and in other areas between 1998–2001 (van Donge et al., 2001: 34). The introduction of Starter Packs in 1998 might have reduced supply among households that used ganyu to purchase fertiliser or hybrid seed but also increased demand for ganyu to weed fertilised maize, so the net effect is indeterminate. Free fertiliser under the Targeted Inputs Programme (2000–2001) had no discernible effects on either supply or demand for ganyu (Levy and Barahona, 2001: 24). The impact of the 2001–2002 famine is discussed in the text and footnote 21 below.

8. The 605 households in the four villages were screened to meet the project's socioeconomic objective to work with resource-poor smallholders, including women. Households without agriculture as their primary source of income and without cultivable land were excluded from the sample frame. Of the 74 households that participated in research trials, 30 male-headed and 30 female-headed households were randomly selected for survey. A matching sample of 30 male-headed and 30 female-headed households was then randomly selected from households that did not participate in these trials (Orr et al., 1997).

9. ‘Days’ refers here to the number of days that households participated in ganyu and not to standard working days of a fixed duration.

10. In 2004–2005, 51 per cent of households in Chiradzulu district grew cassava; the proportion for the southern region was 25 per cent (GoM, 2005a: 98).

11. Rainfall records for Kamphonje estate, Matapwata EPA, show that the six week window period between 27 November 1998 and 14 January 1999 had 24 rainy days, including six days where daily rainfall exceeded 35 mm/day. Records for Matambo estate, Mombezi EPA, for the same period show a total of 19 rainy days including nine days where rainfall exceeded 35 mm/day. We are grateful to the owners of these estates for giving us access to this data.

12. We used the mid-point in the rate of payment, namely 7.5 tambala for three planting stations. The recommended planting density for hybrid maize is 44,000 plantings stations per hectare (MoALD, 1994: 45), giving a total payment of 733 MK/ha. On light soils, the labour requirement for weeding and banking has been estimated at 170 hours/ha (Werner, 1987: 170) which gives a rate of 4.31 MK/hour or 26 MK/day. From personal observation, farmers usually planted more densely than the recommended rate. In theory, this would increase labour's potential earnings from weeding, but employers would also reduce the rate of payment to compensate for easier weeding.

13. This strategy might also backfire. In H8, Mai M.'s husband – her third - generally made himself scarce when there was fieldwork to be done. One morning in October 1999 we arrived to find his belongings dumped outside the front door. He left for town and did not return. Divorce – village style.

14. Livelihood mapping has shown that ganyu is an important source of cash income for ‘middle’ as well as ‘poor’ households in the Blantyre Shire Highlands (MVAC, 2005: 11).

15. In the first six weeks after planting, on days when they participated in ganyu, men did an average of 4.4 hours/day of ganyu labour compared to 3.2 hours/day for women (p = 0.036). Those aged 6–14 did ganyu for 2.2 hours/day, those aged 15–49, 4.3 hours/day, and those aged 50 and over, 2.3 hours/day (p = 0.000). These figures are un-weighted means of labour time and exclude ganyu labour on estates that required significant amounts of travel time.

16. ‘Orphan’ is defined here as children aged 14 or less with a least one parent dead. Nationally, the share of children in this age group with one or both parents dead was 8.5 (Benson et al., 2002: 37). The Integrated Household Survey (IHS) for 2004–2005 gives a figure of 13 per cent for the southern region (GoM, 2005a: 16). One third of all households in the southern region included orphans (GoM, 2005b: 91). Simulations based on the 1997–1998 IHS show that adding one child to a household that already had children reduced income per capita by 19 per cent (Mukherjee and Benson, 2003: 352).

17. Our 1997–1998 panel survey showed that mean time of application for first top-dressing was 4.4 weeks after planting for a sample of 80 plots that received fertiliser. Only 10 plots received a second top-dressing. The recommendation for hybrid maize is to apply a basal dose soon after emergence and one top-dressing not later than three weeks after emergence or about four weeks after planting (MoALD, 1994: 46).

18. Substituting cassava for maize flour to make porridge (nsima) is a popular coping strategy for maize deficits in southern Malawi (Devereux, 1999: 48).

19. The Starter Pack Scheme, introduced in 1998–1999, gave each rural household enough inorganic fertiliser for 0.1 ha. As a result, the area planted to maize by the survey households that was fertilised rose to 87 per cent (Orr et al., 2002: 268). Analysis of variance showed no significant difference in the mean timing of second weeding between fertilised plots (5.31 weeks after planting) and unfertilised plots (5.26 weeks after planting) in this season (p = 0.863).

20. Mortality from the 2001–2002 famine was unofficially estimated at 1000–3000 (Devereux, 2002: 70). No information is available on regional distribution. Deaths were presumably concentrated in Salima and Mchinji districts in the central region, which were the earliest affected and where maize prices reached their highest levels between November 2001 and April 2002.

21. In contrast to the central region, maize production in Blantyre ADD was ‘normal’ in 2000–2001 (FEWSNET, 2002). Revisits made to the 15 case-study households in November 2003 found that of the 11 households with food deficits in 2002, only two believed they had been ‘badly affected’ in 2001–2002 (Orr and Orr, 2003: 20). In short, the south escaped because the maize deficit in 2001–2002 was routine and households had effective strategies for coping with high maize prices. This was bad news for NGOs eager to be seen doing something about the food crisis. Oxfam's local office in Mulanje district refused to request food aid but was overruled by headquarters in the United Kingdom. By contrast, the Integrated Food Security Project in Mulanje funded by GTZ rejected the need for food aid in 2001–2002, and Germany did not participate in the relief programmes with other donors (Orr and Orr, 2003: 25).

22. Similarly, the Targeted Inputs Programme (2000–2001) discovered that giving free fertiliser to poorer households did not reduce the supply of ganyu since ‘so many basic needs (including food) remain unsatisfied that it is still necessary for people to go for ganyu’ (van Donge et al., 2001: 35). The same lesson was learned from a review of targeted food security projects a decade earlier (Simler, 1993).

23. Between 1998 and 2001, an estimated 85,000 households benefited from public works programmes. This represented only seven per cent of the 1.2 million households in Malawi living below the poverty line (Benson et al., 2002: 76).