The failure of early warning in Malawi in 2001: time to rethink international famine early warning systems

Malawi offers a clear example of the difficulties plaguing international early-warning systems. This article outlines some of the changes that are required if these problems are to be addressed.

In October 2001, Save the Children-(UK) (SC-UK) predicted that, unless the price of maize in Malawi was lowered by early 2002, there would be famine. This turned out to be correct. Increasing rates of child malnutrition were observed from late 2001 into early 2002. In February 2001, in Salima in central Malawi, the mortality rate was 2.8/10,000 children under five per day. Yet SC UK’s prediction was dismissed. In November 2001, a multi-agency assessment by the Famine Early Warning Systems Network (FEWSNET) reached essentially the opposite conclusion. Only in February/March 2002, by which time crisis was undeniable, was there any articulated international recognition that a serious problem existed.

In the famine literature, much has been made of the failure of donors to respond to famine predictions. Much less has been said about the institutional and technical problems of providing reliable and convincing warnings. While the Malawi case is far from unique, it should have been straightforward. A considerable amount of information was available about the state of the economy, and there was no impediment to obtaining more data. All agencies with a declared role in global early warning or food-related information were represented, along with the major donors, and the prediction was technically undemanding. How then was it possible that the major food security agencies could not agree on what was likely to occur?

Different systems, different results

The literal answer to this question is that there is no agreed ‘best practice’ in early warning. Each agency obtains different information, analyses it differently and so arrives at different results. The regional FEWSNET in Harare, in conjunction with the Southern African Development Community (SADC), provides routine reports on Malawi’s weather, food production and prices; in Malawi itself, FEWSNET operates an information system with the Ministry of Agriculture. The Food and Agriculture Organisation (FAO) is primarily concerned with crop production and aggregate food supplies, and technical advisers from the UN and donors do their own rapid assessments. SC-UK uses what is known as the household economy approach (HEA).

Current technical approaches to early warning have been largely shaped by two events. The first was the 1974 World Food Conference, which marked the first articulated attempt to set up an international early warning system. The conference proceedings were overshadowed by the drought and famines in the Sahel in 1968–73, and in Ethiopia in 1973. These had largely escaped international attention. The conference proposed a global system of nutritional surveillance. The idea was to monitor events using indicators already available or cheaply collected by governments. In practice, these indicators were reduced to meteorological and crop data, and the prices of staples, livestock and major commodities. Anthropometric data and a variety of other indicators were proposed, but proved impossibly expensive to collect. The assumption was that changes in indicators, for instance a fall in production followed by an increase in the price of staples, would give early warning of an approaching problem.

The second influential event was the publication in 1981 of Amartya Sen’s theory of exchange entitlements, which for the first time put the phenomenon of famine on a firm theoretical basis. The theory of exchange entitlements provided an explanation for the observed features of all famines. According to Sen, famine is the result, not of levels of aggregate or local food supply, but the inability of people to acquire enough of this food to meet their needs. Food may be abundant and cheap, but some people may starve because they do not have the cash to purchase it. Food may be in marginal supply, but where access is reasonably equal, there may be no starvation.

Sen’s theory does not diminish the importance of food supply, but it does show that it is a poor guide to outcome. While most food crises in Africa have been preceded by crop failure, few crop failures have led to starvation. The basic mechanics of survival are now well understood. In much of rural Africa, even the poor may be able to maintain their food entitlement, sometimes in the face of severe and persistent shocks, by storing food, selling their labour, livestock and other assets, gathering wild foods and by gifts and non-market transfers.

A direct corollary of Sen’s theory is that predicting famine requires an understanding of people’s entitlement – unless we know how people get their food, we cannot understand the effect of a shock. The theory showed that indicator-based surveillance alone was insufficient for famine early warning. For example, an observed fall in food-crop production will have little direct impact on the food access of people who depend on the market; an observed rise in prices will not affect people who meet their entire food needs through their own production. Indicators can be interpreted only with knowledge of the underlying economy.

The requirements of early warning

In practical terms, early-warning systems must achieve two things.

1. They must be able to identify a shock, such as crop failure, price rises, the closure of markets or some event or combination of events which might possibly cause a famine.
2. They must be able to assess the likely effect of this shock on people’s ability to obtain sufficient food.

In Malawi, the first of these requirements was met. In 2001, the maize harvest fell by up to a third because of flooding, and from July 2001 prices steadily increased, reaching approximately five times normal levels in early 2002. These changes were monitored and regularly reported by the regional FEWSNET. The problem came with the second step: assessing the effect of the maize failure and price rise on people’s food access. Two predictions were made: the one by SC-UK in October 2001, and the multi-agency assessment the following month.

The SC-UK prediction

The SC-UK prediction arose by chance. SC-UK gathered information on the rural economy as part of a training exercise in vulnerability analysis using the HEA. This allows an estimate to be made of the effect of a shock on household food access. In this case, the prediction was straightforward. The poor in Malawi are very poor. In much of the south, poor people depend for most of their income on a small amount of cultivation, and on ganyu – chiefly day paid agricultural employment. Before the crisis, poor households in Mchinji district obtained approximately half of their food from agriculture, and half from labour paid in cash or food. Around 40% of total annual cash income was spent on food, and the remainder on fuel, soap and other essentials. During the period between November and February, when there are no household stocks, virtually all household income is devoted to food. People live on a day-to-day basis, are usually paid in maize flour or grain and have no cash for other essentials.

The fall in domestic maize production forced more people onto the labour market at an earlier than usual date: as ganyu is in limited supply, labour income fell. The increase in the price of maize relative to household income from labour led to a reduction in the ability of households to purchase maize. As household assets are minimal – at best a goat and often only some household goods – wild foods are hardly available and there is little help between households.

SC-UK estimated the food deficit for the poor group at 30%. As this was actually concentrated in the shorter pre-harvest period, it was proportionally more severe. A confident prediction could be made that poorer households would suffer a catastrophic failure of food access starting at the end of 2001, and accelerating into 2002 until the first maize from the new harvest in February/March.

The multi-agency assessment

The multi-donor rapid assessment in November 2001 conceded that there was a problem, but appears to have had difficulty in defining the aggregate food supply and assessing people’s ability to acquire this.

In assessing aggregate supply, the report acknowledged that maize production did not do well, but concluded: ‘Other food crops are playing a big role in supplementing and substituting for maize’. These other crops were chiefly roots and tubers. A subsequent report commissioned by SC-UK showed that the estimated national supply of these crops was exaggerated. They were chiefly produced in the north and the lakeshore areas, mostly for domestic consumption, and any surpluses were not available elsewhere.

In assessing people’s food access, a main finding of the report was that ‘groups of people depend on other sources, especially the market, for food. Fortunately, general observations indicate that people in the southern region are more enterprising and engage in various activities to generate income’.

Lessons for early warning from Malawi

The difficulty with early warning in the poorest countries is no longer technical. Remote sensing has steadily improved, and government systems for the routine collection of data on weather, crops and prices now cover most famine-prone countries. Outstanding technical problems are of a relatively detailed kind, for example relating to rapid data collection. Using existing techniques, we can now confidently detect any significant food crisis. The challenge is to find an effective institutional framework to ensure that these techniques are routinely applied.

The central difficulty lies in developing systems that can acquire information on livelihoods at a sufficient level of accuracy and geographical disaggregation to permit surveillance data to be interpreted in terms of food needs. This requires a step change in the skills and resources required to run early-warning systems. Livelihood systems, if developed independently by external agencies, are costly, chiefly in the consultant support needed for fieldwork to establish and maintain baseline descriptions. Even if these resources could be found, this does not seem to be an efficient use of them except under very special circumstances, such as in south Sudan where there is a huge annual expenditure on food aid and relief, and where the relative costs of information are small. In most countries, famine is rare and maintaining complex and expensive systems exclusively for this purpose is difficult to justify. It could be argued that the money might be better invested in relieving the desperate poverty in which many rural Africans live.

A solution to this might be found in two parts. First, livelihoods information and the techniques used to predict starvation have value well beyond early warning, and the costs can more easily be justified in those terms. This information can be used to improve the selection and management of emergency interventions; in some situations, for instance, it can obviate the need for food aid entirely by reducing household non-food costs. This information is also likely to have value for the wider understanding and management of poverty. Most economic decisions are currently taken in complete ignorance of actual livelihood patterns.

Second, the costs can be much reduced and the development benefits increased if external agencies concentrate their efforts on supporting national systems. Considerable progress has already been made in some countries, such as Mozambique, and there are encouraging signs that, through the SADC, this may now happen more widely in Southern Africa.

It is clear that the current involvement of multiple organisations, each with its own interests, cannot meet the basic challenges of early warning. Agency coordination is a national function: only governments can adequately coordinate external agencies. Paradoxically, those countries that depend most heavily on external assistance have the least capacity to control its use. In some countries, agencies coordinate where there is a common interest in doing so. If there is to be progress in famine early warning more generally, a common interest needs to be found.

John Seamanis Technical Director, Food Security and Livelihood Unit, SC-UK.

References and further reading

FEWSNET Malawi/ECFSP/WFP, Final Report: Rapid Food Availability Assessment for Malawi, November 2001.

Save the Children (UK), The Household Economy Approach: A Resource Manual for Practitioners, 2000.

Save the Children (UK), A Final Report of the Findings of a Household Economy Assessment of Mchinji District, October 2001.

Amartya Sen, Poverty and Famines: An Essay on Exchange Entitlements (Oxford: Clarendon Press, 1981).