An early warning system is defined as a tool for rationalising the allocation of food aid. Established to help donors, UN agencies and other aid actors to plan their interventions, EWS are systems for the collection and analysis of data relating to people’s access to food. In theory, they enable the prevention of food crises by means of early intervention or, at the very least, attenuate the effects of such situations (Buchanan-Smith & Davies, 1995; UNHCR, 1996).

Famine early warning systems are exclusively oriented towards eliciting an appropriate response in terms of food aid (Devereux, 2000). In this respect, Buchanan-Smith and Davies object to the qualification of ‘early’ warning system which implies that it is more concerned with ‘saving lives’ when crises arise than with intervention at an earlier stage in order to preserve people’s ‘living standards’ (Buchanan-Smith & Davies, 1995). Ultimately, the success of an EWS is measured “in terms of lives saved and not in terms of livelihoods protected” (Devereux, 2000).

A. ORIGINS AND ACTORS

1. From the Indian Famine Codes to the WTO

Some authors locate the origin of early warning systems in the Indian Famine Codes drawn up by the British colonial regime in 1880, following the catastrophic famines that afflicted India in 1876 and 1877. The Codes were highly ambiguous, for they were designed both to prevent famine and to strengthen imperial power (Devereux & Howe, 2004; de Waal, 2000). The colonial regime was “embodied by centralised administration, state-produced statistics and, especially, by the progressive application of the Famine Codes from 1880 onwards” (Fourcade, 2003, p.313). The Codes were prompted by a parliamentary commission which disregarded Britain’s share of responsibility for the aggravation of food shortages in India. These shortages, however, were above all else the result of the trade deficit that Britain imposed on its colonies, the development of agricultural exports and the high taxes levied on the peasants.

According to the Codes, famine was essentially the product of natural causes and could be described by means of specific technical indicators such as rainfall, cereal prices and mortality rates Reminder, the first Ethiopian EWS were modelled on the Famine Codes (Buchanan-Smith & Davis, 1995).In theory, the monitoring of these indicators would trigger, if the need arose, preventive measures such as the provision of funds and subsidies to populations exposed to food shortages. In practice, the Codes proved unable to counter the famines that occurred between 1896 and 1908. Policy-makers ignored the warnings issued by local administrators, and aid money was diverted to the war effort on the Afghan border. The application of the Codes was confined to measures involving compulsory labour and the liberalisation of cereal markets (Davis, 2003).

The first modern early warning systems emerged from the 1975 World Food Conference, which was organised to formalise procedures for the allocation of aid. The initial EWS, the Global Information and Early Warning System (GIEWS), was established by the FAO in 1975. A decision-making tool for donors, the United Nations and other food aid institutions, the GIEWS could be likened to a large-scale system for the analysis of agricultural data. A number of reforms designed to improve the system’s accuracy and speed were initiated after the African famines of the 1970s.

Between 1985 and 1990, a further eight EWS were established in the Sahel and the Horn of Africa. Influenced by the ‘theory of entitlement’ developed by Amartya Sen, these systems were based on multiple indicators, including socio-economic data (Devereux, 2000). In 1985, USAID set up its own EWS: the Famine Early Warning System (FEWS). Non-governmental organisations subsequently launched more localised information systems, based on multi-indicator models but adapted to a small scale, which facilitated the fine targeting of beneficiaries (Buchanan-Smith & Davies, 1995; Devereux, 2000). In 1996, The World Food Summit launched a new programme in which partners set up information and mapping systems to pinpoint food insecurity and vulnerability (SICIAV) (Devereux, 2000).

The WTO now plays an increasingly important role in the allocation of food aid. At the Uruguay Development Round in 1994, it recommended “establishing a level of food aid commitments sufficient to meet the legitimate needs of developing countries during reform programmes”. These “legitimate needs” were not clarified and are still highly ambiguous (Konandreas & Sharma et al, 2000). On the other hand, the decisions taken at the end of the Doha Development Round (December 2005) moved towards a strengthening of the mechanisms for assessing needs, and indeed the standardisation of the evaluation methods upon which international bodies (WTO, FAO) rely when authorising food aid donations. At the conclusion of discussions on the elimination of subsidies (regarded as barriers to trade), WTO members agreed that certain forms of aid, including emergency aid, should be maintained. The authorisation of food aid is therefore heavily dependent on the assessment processes employed by aid organisations, whose evaluation methods and results are considered reliable.

“This, it seems to me, comes down to an empirical and practical question as to whether, as a matter of performance in the field, these organisations ‘get it wrong’. If the answer is that they do, or that some of them do, there might be understandable reticence to include their appeals as a standard. But if the answer is that they don’t, why would inclusion of an appeal from them be a problem?... It seems hardly credible to deny their expertise and role.”

Chair’s Reference Paper, Committee on Agriculture, Special Session, Export Competition, 11 April 2006 (WTO, 2006)

The WTO’s concern to distinguish aid for acute situations from aid for chronic situations highlights the importance of targeting procedures (such as early warning systems), which must now ensure that structural food insecurity is not confused with any crisis situations that may arise.

In the end, the various institutions involved have tended to devise their own tools for collecting and processing data, as this enables them to defend the independence of their assessments (see the appendix for a summary of the principal EWS). The FAO system, GIEWS, is held to be of higher quality than national early warning systems, while FEWS, set up by USAID, is claimed to produce more reliable results than GIEWS. But they all draw their information from the same sources -governmental agricultural data and statistics- and there is ultimately little difference in the results they produce. The existence of different EWS demonstrates the desire of the participants (USAID, WFP-FAO, NGOs, etc.) to act independently, but as yet no one has found a suitable method which would enable them to produce a truly original analysis and stand apart from their competitors.

B. A TECHNOLOGICAL SMOKESCREEN: EWS AS PRESENTED ON THE INTERNET

Despite their proliferation, most EWS have many points in common. Therefore, they tend to convey the impression that they are effective, rational, scientific and technical instruments. A brief ethnographic look of EWS as they are presented on the internet gives us the opportunity to penetrate this ‘technological smokescreen’ and reveal the central role of pragmatism behind the ‘high-tech’ façade. The Global Information and Early Warning System (GIEWS) and Famine Early Warning System (FEWS) websites are particularly useful, as they exemplify the “theoretical constructions and practical methods” on which EWS are based (Le Robert, 1986). For GIEWS, see http:/www.fao.org/giews/english/index.htm. For FEWS, see htpp:www.fews.net

As soon as we log on to these sites, we are presented with a mass of polymorphous information: maps, evaluation mission reports and more succinct monthly bulletins accompanied by tables, graphs and photographs. The emphasis is on the warning function: the GIEWS map of the world indicates areas at risk and provides links to the latest special reports. The FEWS site lists countries according to the severity of the crises they face. Monitoring indicators may be accessed by consulting files and bulletins that record specific agricultural conditions. Other sections describe the more methodological aspects and there are also brief descriptions of the mechanisms of institutional cooperation on which the systems are based.

1. A “SCIENTIFIC ENDEAVOUR”

Drawing on a combination of scientific norms, early warning systems claim to be objective and universal. Their modelled, standardised and mapped approaches, mathematical and statistical processing and use of advanced information technology are all indicative of a “scientific endeavour” (Pestre & Dahan, 2004, p.15). According to its designers, the GIEWS has invested in “innovative methods for collecting, analysing and disseminating information, making full use of the revolution in information technology and the advent of computer communications (GIEWS internet document  See http:/www.fao.org/giews/English/index.htm, p. 2).

In other words, EWS claim to belong to the world of ‘techno-science’, to an “institutionalised corpus dedicated to the systematic development ... of scientific and technical applications”. Their techno-scientific attributes place them in a “culture of emergency and permanent mobilisation” inherited from the Second World War and perpetuated by the Cold War (Pestre & Dahan, 2004). This culture, stimulated by a “faith in technology which is invariably rooted in the myth that science won the last war,” relies on the belief that “coordinated and concerted techno-scientific action ... will overcome any difficulty that may arise” (Pestre & Dahan, 2004, p.12).

2. INSTANTANEOUS ACCESS TO GLOBAL KNOWLEDGE

The websites offer a direct reading of a planet that has been mapped comprehensively. The FEWS logo is a terrestrial globe. The GIEWS site opens with a world map depicting ‘global food shortages’; flashing lights draw our attention to countries at risk. Links invite us to access a ‘hunger map’ or another portraying world trade in agricultural commodities, while a continuous banner reveals a list of ‘countries in crisis’, in the style of CNN newsflashes. The FEWS site opens with a cartographic depiction of countries according to the degree of food insecurity they face. By selecting GIEWS ‘News’ or ‘Latest Publications’ or a particular country (FEWS), we can also travel - navigating by means of a click - from one catastrophic situation to another, taking in countries ravaged by locusts, floods, bird flu, etc. Almost instantaneously, the sites furnish us with a global picture of crisis situations, while suggesting that each case is treated in a systematic, comprehensive and rigorous fashion. The information reflects the desire to act on a global scale.

More detailed information comes in the form of special evaluation reports arranged by country (FAO), news/information bulletins (FEWS) and the FAOSTAT statistical databases recording global harvests for each type of cereal, recipient countries, donors, etc. These documents are always accompanied by maps - zones containing endangered or beneficiary populations, satellite rainfall patterns, vegetation density - which enhance the browser’s ability to visualise the problem.

3. THE “MATHEMATISATION OF REALITY” I have borrowed the term from Dominique Pestre, who uses it in the introduction to Les sciences pour la guerre (Pestre & Dahan, 2004, p. 33).

This systemic body of knowledge provides answers to any type of question relating to global food security (“How much food is the world producing?” “Where are food interventions most needed”, etc.) See http:/www.fao.org/giews/English/index.htm by means of the “most up-to-date and accurate information”. Presented as an aid to decision-making, EWS offer fact-based, comprehensive information, principally in the form of figures, charts and graphs. In this respect, they hark back to the “promotion of information ... based on calculation and predictability, on a statistical approach” typified by the computer tools and modelling that evolved from wartime technology (Pestre & Dahan, 2004).

The GIEWS site leads us directly to the statistical sub-department known as FAOSTAT (“Statistics for a better world”). In my view, the raw information in the FAOSTAT databases - and the complete lack of commentary - constitutes a kind of statistical matrix which exemplifies a particular way of interpreting and expressing global issues. Every report is based on these data. “Drawing on over twenty years of time-series statistics, GIEWS officers continuously update and analyse data on food production” (GIEWS, See footnote 5. p.10). We are later informed that “GIEWS has developed a Computer Workstation for data management and early warning analysis, ranging from crop monitoring using up-to-date satellite images to estimating food import requirements.”

The condensed style in which FAO evaluation report summaries are written presents the main statistics - agricultural production and percentage variations, vulnerable populations, totals of food aid required - in a form that is little different from raw data tables: “Aggregate cereal and pulse production in Ethiopia from the 2005/06 ‘meher’ season is forecast at 17.2 million tonnes, about 14% above the previous year’s revised estimates” (FAO Special Report Ethiopia, February 2006). The titles of the FEWS bulletins are more eloquent, combining crude figures with the language of urgency in a style familiar from newspaper headlines and televised bulletins: “Ethiopian regions confronted with a pre-famine” (FEWS 9/12/2002); “Another 600,000 people will need food aid in 2003” (FEWS 29/08/03); “Ethiopia needs emergency aid” (FEWS 11/05/04); “Ten million people will need humanitarian aid in 2006” (FEWS 26/1/06), etc.

The narrative mode of the evaluation reports and information bulletins is strictly technical, as the function of these documents is the provision of statistics and, occasionally, descriptive commentary. In addition, they are accompanied by tables, charts, coloured graphs indicating commodity prices, percentage columns indicating beneficiary populations, etc., which constitute information in its proper sense, rather than its illustration.

4. A ‘NON-POLITICAL’ TOOL

The fact that EWS are firmly rooted in the world of science gives them a kind of political immunity. As Alvarez notes in his article Sciences in the Development Dictionnary, “This knowledge technique is so reliable that the expertise acquired is non-negotiable in terms of its practical application. The indispensable expertise that science claims to offer is deliberately distanced from the political arena ... It is never the outcome of bargaining or choice” (Alvarez, 1996). This observation echoes that of Alex de Waal, who interprets the use of EWS as an evasion of the political responsibilities incumbent upon governments and international agencies when contemporary food crises occur. In the 1980s, the World Bank was promoting neo-liberal reforms (promoting austerity, privatisation, etc.) at the very time when African countries were suffering the effects of drought. While structural adjustment programmes were unable to halt the economic decline of these countries, they did take food crises into account via the World Bank’s formulation of the concept of food security in 1986 and the notion of ‘adjustment with a human face’ advanced by UNICEF in 1987. These new concepts reinforced the tendency to regard famine as a non-political issue.

Thereafter, famine prevention and response were depicted as essentially technical issues, the province of a “citadel of experts”, whose technocratic discourse contributed, according to Alex de Waal, to the mystification and bureaucracy surrounding famine prevention and aid (de Waal, 1997). De Waal argues that the suppression of the “political bite”, the reduction of famine to a matter of charity and humanitarian intervention, will ultimately defuse any mobilisation at local level. He concludes that “international humanitarian bureaucracy itself is the greatest obstacle to the prevention of famine in Africa” (de Waal, 1997). De Waal pushes his argument to extremes, suggesting that EWS would be more effective if they shifted from the apolitical field of technology to the political arena where, with the support of professionals such as journalists, businessmen, organizations and administrators, they would become a “political trigger” for action (de Waal, 2000).

The internet presentation of the GIEWS constantly refers to the various partners involved in its operation and furnishes a list of agencies, sub-departments of multi-lateral agencies, bilateral partnerships, multilateral donors, NGOs, etc. The many references to the institutional network on which EWS rely are implicit reminders that the role of such systems is largely political, if only because they bring together all the actors operating in the aid sphere.

5. CUTTING-EDGE TECHNOLOGY AND EMPIRICISM WITH A HUMAN FACE

The GIEWS presentation contains several pages on the use of satellite images and information software. We learn that a “crop-monitoring system using near real-time satellite images” has been established in countries where “continuous reliable information” is hard to obtain. The system compares current meteorological data with historical averages and is complemented by a “Normalised Difference Vegetation Index” (NDVI) that indicates the extent of vegetation cover. The processing of this information is matched with the assessment of national cereal output, the analysis of maps and satellite images, the management of agency dispatches and other tasks.

However, it appears that it is sometimes necessary to fall back on more empirical methods. There is also a page about rapid evaluation missions, which are described as a “short-term expedient” in countries where accurate information is not readily available. Their aim is to “verify the reliability of official data by assessing crops and interviewing farmers”. The results should be dispatched to decision-makers within ten working days. From the description it seems that these missions, whose reports are published on the website, contain all the characteristics of an empirical investigation, and are a far cry from the high technologies to which so much space is devoted.

Although the EWS is presented as an abstract system (impersonal technology, disembodied teams of experts), there is no lack of images of peasants tilling their fields and herders tending their animals. Possibly concerned by the dryness of its content, the GIEWS site offers a gallery of photographs featuring people and places - dignified peasants, market stalls, graceful and smiling women - that were probably encountered during evaluation exercises. There are also more humorous allusions to the world of humanitarian aid, like the photograph of an African stall with a sign saying “I need donation my businness”. Finally, the browser is invited to admire the inevitable family photograph, the group shot of GIEWS experts, a reminder that there are real men and women working within this seemingly disembodied system. The photographs also enhance the illusion that most of the work is conducted in the field, and indeed that the data comes directly from peasants. The cover of the GIEWS document juxtaposes an enormous globe, peasants toiling in the fields and ears of corn, which illustrates the desire to establish an immediate relation between global information and local information, with the latter extending to the smallholder and indeed to the micro-level- the quality of an ear of corn.

EWS therefore relate to a techno-scientific culture in which decisions are governed by the rational application of science and technique. They are depicted as a resolutely modern technical artefact which use cutting-edge technology; process comprehensive, quantitative information; encapsulate it succinctly; disseminate it through the internet and, finally, are firmly embedded in the global community. The institutional documents do not question the rationality of this technical artefact, but there is an implicit acknowledgement that exceptional circumstances may call for empirical techniques. However, the combination of a disembodied approach and personalised touches (photographs) is just one indication that EWS are ultimately more of a patchwork, a mixture of high technology and pragmatism.

C. TECHNICAL AND POLITICAL USES OF EWS

We should therefore venture beyond the technological and scientific smokescreen of EWS. In doing so, we shall be confronted with questions concerning both their effectiveness and the political issues that crystallise around them. As tools for famine prevention, what are their real capabilities? As political tools, what issues arise from their use?

1. A POOR PREDICTIVE CAPABILITY

The authors who have made in-depth studies of EWS posit three types of technical restriction (Buchanan-Smith & Davies, 1995; Devereux, 2000; Pillai, 2000).

• The predominance of agricultural data. The information produced by EWS is filtered in accordance with their primary objective: to elicit an appropriate response in terms of food aid. Consequently, agricultural production and food deficit indicators take precedence over other socio-economic indicators. Nutritional data helps to measure the impact of an existing food crisis, but it is of limited use when devising preventive measures. In no case does it permit a fine targeting of beneficiary populations. EWS targeting focuses on the assessment of an agricultural deficit and the number of beneficiaries in a given area. EWS are an upstream tool for working out the allocation of aid tonnages according to geographical and administrative boundaries; they are disconnected from the later downstream operations, whether these concern the sharper targeting of beneficiaries or effective methods of aid distribution.

• The inadequacy of the food balance sheet approach. The WFP and FAO calculate the number of people suffering from a calorific deficit by means of global food balance sheets. The food balance sheet weighs domestic supply (production, imports, food aid, stocks) against domestic utilization (consumption by humans and livestock, seed, agro-industrial uses, waste). The quantities destined for human consumption are converted into kcal per day. The number of people suffering from food shortage is calculated by comparing the required minimum with the quantity available. Besides relying on an average energy requirement that takes no account of regional disparities, this method reduces food security to a matter of aggregated food availability. It does not help to assess the difficulties of access to food, which are at the root of shortages in many countries. Moreover, it ignores non-agricultural incomes and survival strategies (Devereux, 2002; European Commission, 2002).

• Poor quality data. The two principal sources of quantitative data on agricultural production are agro-meteorological models of agricultural yields based on satellite images, and harvest/pre-harvest evaluation missions that attempt to estimate future output by multiplying planted areas by their theoretical yield. According to J.-P. Minvielle, an expert on food security, data of this sort is riddled with inaccuracies which, when accumulated, often lead to diametrically opposed results (a surplus or shortage in the cereal balance sheet). Analysing the ‘information stream’, Minvielle goes back to the source of statistical production to highlight the weaknesses in the field measurements taken by technical agents. When added to other variables, this data produces uncertain statistics (Minvielle, 1994). Since the 1980s, projections of food aid requirements derived from projections of global food availability have been affected by the differences in sources and methods, which have caused final estimates to double (Webb, 2003).

As a consequence, it is impossible to predict with any great accuracy the occurrence of food insecurity, or to determine the exact amount of food aid required. Buchanan-Smith and Davies conclude that “prediction will always be closer to art than to science ... Decision-makers must learn to live with uncertainty ... to adapt their response system instead of waiting for a definitive forecast” (Buchanan-Smith & Davies, 1995).

2. UNCERTAINTY, THE IRREDUCIBLE COMPONENT OF HUMANITARIAN AID

Aid actors must therefore adapt to a context dominated by two forms of uncertainty: the future availability of aid (which is inevitably conditioned by the trade and foreign policies of donor countries), and the local food security context (which EWS should help to clarify). Decision-makers, however, seek hard facts on which to base decisions concerning the allocation of thousands of tonnes of food and the survival of entire populations. Furthermore, decision-making is subject to deadline constraints arising from emergency situations, the agricultural and climatic calendars, administrative delays and many other factors.

Aldo Benini argues that uncertainty is an irreducible component of the environment in which aid agencies work. Besides the unknown factors arising from the external context of the action, agencies are hampered by complex internal procedures, which further exacerbate uncertainty. For example, the procedures they employ to evaluate needs result in a flood of inaccurate data. Uncertainty is the “nemesis” of agency information systems (Benini, 1997).

Benini agrees with the analyses of the sociology of organisations, that stress the central importance of managing uncertainty in relations between organisational actors. In Le pouvoir et la règle, Erhard Friedberg posits that all collective projects are inevitably subject to uncertainty, which dogs attempts to define problems as much as efforts to develop solutions to them. As information is always incomplete, awareness of possible courses of action is always fragmentary. In addition, a project’s achievement will always be affected by a series of unpredictable events (Friedberg, 1997).

Supporting their argument with concrete examples, Buchanan-Smith and Davies note that the enhanced predictive capability of EWS has not led to more effective responses. The authors believe that the real issue here concerns the constraints that affect the operational decision-making process, rather than the quality of the data and information. There is a ‘missing link’ between EWS data and its use with regard to an effective response. Poorly adapted institutional procedures aside, Buchanan-Smith and Davies argue that the bureaucratic structures and procedures established by governments and agencies are inappropriate in terms of crisis response. The principal constraints they cite take a variety of forms: the lack of human and financial resources available to local government, the reliance of NGOs on donors, the rigidity of bureaucratic procedures, the poor coordination between field and office and the constraints of budgetary calendars. All these factors are exacerbated by the chronic lack of institutional memory. the quest for certainty is probably the main reason for the dysfunctional nature of the aid system (Buchanan-Smith & Davies, 1995).

Buchanan-Smith and Davies argue that bureaucrats tend to be risk averse. The uncertainty inherent in the context of action thus reinforces the “evasion of responsibility”. Decisions are delayed until bureaucratic procedures - writing reports, justifying decisions, etc. -ensure the formal degree of certainty demanded by their institution. For members of these bodies, the bypassing of standard procedures and precipitation of action constitute higher risks than respecting procedures. The authors suggest that even if donors had a vast amount of information from various sources at their disposal, they would still not have the accuracy level demanded by “risk-averse” bureaucrats. The quest for certainty creates delays; large quantities of aid are not pledged until the harvests are almost over, which is usually in January (for the Horn of Africa). The aid is thus never distributed when it is most needed. Buchanan-Smith and Davies conclude their discussion with a quote from Field: “The quest for certainty as the key to a decision converts an ‘early warning’ into a ‘late warning’” (Field 1993, quoted p.36).  The quest for certainty may explain delays in allocating aid, but it cannot explain the failures to target aid properly, an issue I raised in the introduction.

3. EWS AS A PRESCRIPTIVE TOOL

EWS bear comparison with another tool, modern statistics, whose origins have been studied by Desrosières in La politique des grandes nombres. The emergence of statistics has been associated with the construction, unification and administration of modern states (Desrosières, 2000). Similarly, it is not unreasonable to suggest that EWS have contributed to the rise of the humanitarian aid regime, the rationalisation of the allocation of donations, the definition of the roles played by multilateral institutions and to relations between other actors in the system, such as donors and donor countries. Like modern statistics, the creation and use of EWS present us with a paradox: their claim to autonomy is based on objective, universal values, yet their authority can only be exercised through participation in the sphere of action, in decision-making and the transformation of a global situation.

The statistical tool was subjected to the “tension between a descriptive and a prescriptive perspective” (Desrosières, 2004, p.14). In the case under review, the tension takes the form of a radical disequilibrium: in a context dominated by uncertainty, prescription and the urgent need for decisions take priority over descriptive qualities.

4. THE ROLE OF EWS IN INSTITUTIONAL AGREEMENTS

As we have noted, some institutional actors have set up their own EWS, although every system draws its data from the same sources. This means actors have resorted to a single, error-prone measuring device.

Friedberg’s sociological approach immediately positions itself in a political field in which the various groups involved jockey for power. Given the uncertainty that characterises their working environment, the groups which control the definitions of problems (food deficits, beneficiary numbers) will dominate negotiations and cooperative efforts relating to the allocation of aid. Among other things, power hinges upon the control of uncertainty and the ability to define a situation. Uncertainties become power as soon as one is in a position to control the evaluation tool. In the present case, power plays between actors or groups of actors revolve around evaluation methodologies, for these are the means by which problems and solutions can be quantified and qualified.

This analytical perspective enables us to formulate two hypotheses to explain the strategies developed by aid actors. In the first strategy, each group seeks to establish its own EWS because it has a tendency to “monopolise control of the uncertainties (Friedberg, 1997), to mark itself out by its ability to define a situation and thus achieve power and autonomy. In the second strategy, the actors draw from the same pool of errors and arrive at similar results because they are using the “logic of alliances”. At the negotiating stage, as the imprecision of the instruments for measuring needs becomes increasingly apparent, discussions will tend to become more political and will contribute to the “logic of agreements” between actors (Friedberg, 1997). Friedberg himself uses the term when referring to the work of Lucien Karpik who, in an essay entitled L’économie de la qualité (describing the strategies lawyers use to build a client base), claims that when a situation contains an “irreducible element of uncertainty and unpredictability, action must amount to more than the application of objective expertise and knowledge, it requires an apt choice of tactics and the conclusion of alliances... this is part of the art of strategy (Karpik, 1989, p.198). Karpik refers to “institutional alliances” (p. 199). Returning to our own field, the proliferation of EWS does not necessarily generate conflict; in fact it fosters agreement - given that all concerned are aware of the errors, limits and uncertainties of such systems - and contributes to alliances between agencies. Consensus appears to be based on the tacit acceptance of uncertainty and the inability to do better! Donors regard WFP and FAO evaluations as the most reliable EWS. Conducted during the growing season, they provide a comprehensive data set from which aid requirements can be directly quantified; they “affix the international seal of credibility” (Buchanan-Smith & Davies, 1995).

5. EWS AS AN EXPERT SYSTEM

We have seen EWS are distinguished by the plurality of actors, levels of action and decision-making; by the length of the information chain and use of sophisticated technologies; and, finally, by an investment in form that is indispensable if decision-makers are to coordinate results that will facilitate the interpretation of an abstract situation cut loose from concrete, measurable realities. In this respect, EWS belong to the category of ‘expert systems’ described by Anthony Giddens in The Consequences of Modernity (Giddens, 1994). Giddens argues that the development of modern institutions is determined by a distinction between space and place, with such institutions favouring relations with an absent other, a person with whom one never comes face to face. Place thus becomes “increasingly phantasmagoric: the various social settings are completely penetrated and shaped by very distant social influences” (p. 27).

Giddens goes on to describe two ‘disembedding mechanisms’: the creation of ‘symbolic tokens’ and the establishment of ‘expert systems’. The latter are defined as “systems of technical accomplishment or professional expertise that organise large areas of the material and social environment in which we live today” (p. 35). Like ‘symbolic tokens’, they act as “guarantees of expectations across distanciated time-space. This ‘stretching’ of the social system is achieved via the impersonal nature of ... technical knowledge” (p.36). An expert system rests on trust - “the feeling of security justified by the reliability of a system” (p. 41), which constitutes a “fundamental notion of the institutions of modernity”. Trust derives from lack of information, faith in the efficiency of a system and the validity of principles of which we know nothing. Faith in the authenticity of the expert system that is EWS are based largely on their investment in form being able to foreground their technological, scientific and statistical attributes.

D. THE EWS IN ETHIOPIA

The Ethiopian National Early Warning System, the first of its kind in Africa, was established in 1976. Now a veritable institution, it forms part of a context in which food insecurity constitutes a structural element of Ethiopian political culture.

1. FOOD INSECURITY AS A STRUCTURAL ELEMENT IN ETHIOPIA

Food insecurity is reflected as much by the historical recurrence of famines (and their current severity) as by its political consequences and the institutions developed to deal with it. Ethiopia suffers from a chronic and increasing food deficit, the result of generally inadequate agricultural output and very high population growth. Production is rising at 1.2% per annum, whereas population growth borders on 3% (Webb & Von Braun, 1994). Per capita production having fallen due to the repeated famines of the 1970s and 80s, the consequent decapitalisation of the peasantry, the disastrous effects of the Derg’s collective agricultural policies and the years of war, recovered in the 90s and reached a level comparable to that of the 60s (around 190kg per person per annum), but is still highly fragile, as shown by the crises of 1998, 2000 and 2003.

Moreover, the history of Ethiopia is marked by the recurrence of famine. Ninth- and twelfth-century legends of the early Christian saints refer, in apocalyptic terms, to the catastrophic famines provoked by divine wrath. A chronological analysis of medieval manuscripts and travellers’ accounts written between the thirteenth and nineteenth centuries suggests that famines occurred every ten to thirty years (Pankhurst, 1885 and 1990). These narratives sometimes credit Ethiopian saints (Tekle Haymanot) and emperors (Lebna Dengel, 1508-1540) with almost Christ-like gifts, including the miraculous ability to multiply food stocks (Pankhurst, 1985).

In more recent times, famines have directly destabilised the course of political life in Ethiopia. The widespread droughts and rinderpest epidemics that ravaged the country between 1888 and 1892 were the most severe, laying waste to entire regions, causing mass migrations and killing thousands of people. Having devastated Tigray, famine also wrought changes in the regional hierarchy, wearing down the Tigrayan ruling dynasty and allowing an Amhara dynasty to seize imperial power (Davis, 2003; Pankhurst, 1985). In 1974, famines struck Wollo and precipitated the collapse of Hailie Selassie’s regime, which had ignored a crisis so grave that it had claimed the lives of between 40,000 and 200,000 people. During the famines of 1984-1985, aid was used to further a massive programme to resettle dissident populations in camps (the ‘villagisation’ policy) (de Waal, 1997; Jean, 1986). The current government is still grappling with the problem of food insecurity. Since taking power in 1991, it has had to deal with a series of crises, particularly in 1993, 1998-99, 2000 and 2003.

This historical overview would be of little interest but for the fact that politicians always refer to historical precedents when a crisis occurs; the evocation of past events is constant. The socialist Derg regime made symbolic use of the 1974 famine to enhance its legitimacy, promising that it would eradicate famine (de Waal, 1997). The preface to Pankhurst’s study of the 1889 famine was supplied by a government body (the RRC), and suggested that the regime’s new policies would put an end to the tragedies of the feudal era (Pankhurst, 1985). In 2001 and 2003, under the present government, the prime minister and certain aid actors regularly invoked the spectre of the 1984 famine in their appeals to public opinion and the international community. The discourse during crisis has made systematic use of the 1984-85 famines, sometimes by means of television documentaries that mix archive and contemporary footage. As the term ‘famine’ has no terminological or technical criteria (Devereux & Howe, 2004), it facilitates the shift to an emotional register which is partly sustained by images of previous crises.

2. CENTRALISED MANAGEMENT OF INSECURITY

The total amount of international aid - including food aid - delivered to the Ethiopian government has risen to 10% of the country’s GDP. Between 1984 and 1994, Ethiopia received food donations equivalent to 10% of national output, an annual aid budget in itself (Planel, 2005).

Governmental institutions and national programmes designed to manage food shortages constitute one of the pillars of the Ethiopian political system (Weissman, 2001). Like the policies of the Derg regime (1974-1991), those of the present party of government, the Ethiopian People’s Revolutionary and Democratic Front (EPRDF), accord a central place to the problem of food insecurity. The management of food insecurity was institutionalised in 1974 with the creation of the Relief and Rehabilitation Commission (after restructuring in 1995, the RRC became the Disaster Preparedness and Prevention Commission (DPPC); this in turn became an agency (DPPA) in 2004). The government set up an early warning system in 1976, and established emergency food stockpiles in 1982. The national institutions are complemented by United Nations agencies, bilateral partnerships (USAID, EU, etc.) and NGOs. Almost 300 international aid organisations, as well as around 100 national NGOs, are working in Ethiopia.

The current EWS stems from the prototype developed during the Derg regime. A classic top-down detection and response system, it received financial support from donors and was managed by the RRC. Early Warning and Planning Services (EWPS).Its failure to avert the 1984-85 famine may be ascribed to causes that were essentially political in nature. In the eyes of the donor community, Ethiopia was of secondary importance in the Soviet sphere of influence, while the government was more concerned with celebrating the tenth anniversary of military rule than with relaying the RRC’s appeals for a massive injection of aid.

The defiance and suspicion that typified relations between the West and Mengistu’s socialist regime prompted donors to set up their own EWS to run alongside the system maintained by the RRC. The government, keen to obtain western aid, tolerated this duplication and allowed donors to direct aid to NGOs rather than to the RRC. Relief and Rehabilitation Commission evaluations were regarded as “dramatic in the extreme”. But the WFP and FAO evaluations accepted by donors relied on the same sources as those of the RRC, and contained much the same information. UN agencies were heavily involved in checking the findings of the government EWS and affixing the “international seal of approval” which lent credibility to national data.

The National Disaster Prevention and Preparedness Strategy (NDPPS) developed in 1989 signalled a new approach. The strategy attempted to link emergency aid to development projects designed to reduce vulnerability in future droughts. The measures to be applied were laid down in an Emergency Code that was modelled on the Indian Famine Codes.

In 1993, the transitional government approved a revised version of the Emergency Code, which became known as the Directives for Disaster Prevention and Management. The new version announced the government’s determination to emancipate itself from free aid, which was described as a form of dependency. The DPPM included a set of mechanisms which would both satisfy short-term food requirements and promote the achievement of long-term development goals. The Employment Generation Scheme (EGS), based on ‘work-for-food’ or ‘work-for-cash’, was the main component. In theory, 80% of aid would be distributed in return for work, with the remaining 20% reserved for the elderly, pregnant women and the disabled.

Since 2003, and especially since 2005, evaluations have been required to distinguish between ‘chronic’ and ‘urgent’ food insecurity. This approach amends the previous strategy and takes its inspiration from the Safety Net programmes proposed by the principal donors (WFP, USAID and the EU). In theory, it provides three to five years of aid to groups regarded as being moderately affected. Assistance of this kind (money, foodstuffs) should enable the recipients to rebuild their domestic economy and achieve financial independence.

The new approach was superimposed on the government’s massive ‘villagisation’ scheme, the gradual transfer of more than two million people from arid, economically weak zones to more prosperous areas with greater land availability.

While some spontaneous resettlement had already taken place, with people moving from Oromiya (west and east Harerge) to Bale, it was not until late 2002 that the government launched a pilot scheme in Tigray and Oromiya (donors played no part in its financing).

Larger schemes involving populations from Amhara, the Southern region (SNNPR) and Tigray were initiated between January and April 2003. Theoretically, the programme rests on the willingness of populations to resettle; the provision of land, tools, fertilisers and plough oxen; and finally on the provision of infrastructure to meet individual and collective needs.  Without going into detail regarding the circumstances of resettlement, which varied according to the site and the organizational abilities of the local officials tasked with settling the newcomers, two important points should nevertheless be noted. First, the programme launched in 2003 suffered from a lack of preparation. Second, there were two competing assistance programmes: food aid and villagisation. Observers claimed that the preparation, registration and transportation of the first candidates were rushed. Little time had been spent on transport logistics or on establishing appropriate infrastructure for the new arrivals. In some cases, land allocation was delayed because district officials had no plan for it. The allocation of pairs of plough oxen and credit facilities did not meet the needs of the peasants, who had either not received any information or had interpreted it incorrectly before their departure. At the time of the field work (in Wag Hemra) for this study, most of the first wave (those who had resettled at the beginning of 2003) had returned, citing as their main reasons the incidence of malaria, the lack of medical facilities and the complex credit arrangements for obtaining plough oxen (Hammond & Dessalegn, 2003).

Villagisation is a supposedly ‘voluntary’ process, but it goes without saying that food assistance programmes are components of the ‘coping mechanism’, the resort to survival strategies. Because beneficiary populations saw food aid as a rival attraction, the villagisation programme got off to a slow start. Between January and February 2003, the organisations responsible for distributing food aid were forced to wait until the ‘volunteers’ for resettlement had been registered (Hammond & Dessalegn, 2003).

3. THE LIMITS OF THE EWS IN ETHIOPIA

The EWS relies on various government departments such as the Ministry of Agriculture (MOA), the Central Statistical Authority (CSA) and the National Meteorological Agency (NMSA). Its principal source of data is the MOA, but since 2005, it has increasingly relied on data supplied by the CSA.

Numerous irregularities have been detected in the functioning of the Ethiopian EWS (Pillai, 2000). ‘Targeting errors’ have come to light, notably in relation to the distribution of aid between the northern and southern regions, between woredas suffering shortages and those that have sufficient stocks, and between rich and poor people living in the same woreda. There are no correlations between food shortages in a particular area and the amount of aid allocated to it. Food aid reaches a mere 22% of the people affected by food insecurity, either because their woreda has not been targeted or because their family has not been targeted (Clay & Molla et al, 1999).

Even though the reliability of the data leaves much to be desired, strenuous efforts are made to shape it into a convincing format. The countless files heaped up in government offices are thus transformed into concise reports adorned with columns of figures, photographs and eye-catching multi-coloured charts. Furthermore, the use of such up-to-date presentation styles conveys the impression that the system is based on a rigorous and effective method of data collection. They clearly help to enhance the legitimacy of the institutions involved in the management of food security in Ethiopia.

4. HUMANITARIAN ACTORS AND THE QUEST FOR INDEPENDENCE

Until 2004, all NGO and international agency activity was controlled by the DPPC. The government could not escape partnerships with the WFP and NGOs, for they provided most of the resources - food, funds, logistics (vehicles, computer hardware and software, etc.) - to keep the food aid programmes rolling. In practice, the foodstuffs provided by the WFP and other donors became the de facto property of the government as soon as it entered the country. The aid went directly to the DPPC, which used it as it saw fit, but was still accountable to the donors.

The overall management of food insecurity and aid rested principally with the DPPC national committee, which was responsible for all national decisions relating to the prevention and management of risks. The committee, an inter-ministerial body presided over by the prime minister, grouped together the ministers of finance, health and economic development. This structure was reproduced at every level of the administrative hierarchy - region, zone, woreda, kebele and gott. A woreda is similar to a district. A kebele is the smallest administrative unit. A gott is the equivalent of a village. The DPPC-dominated structure also involved other departments, such as agriculture, at the various levels.

The DPPC, a centralising body, was directly responsible for producing regular reports on the current state of food security in Ethiopia. For government agents, the collection of information relating to food security was a near permanent activity, continuing throughout the year and occurring at every level of the hierarchy. It took place according to two methodologies. The first consisted of writing monthly reports at village administrative level, an activity that was repeated at each level of the pyramid up to the apex (Addis Ababa), where a final compilation of data was produced. The second proceeded by means of the ‘field trips’ that agents undertook from time to time. Weighed down by cumbersome procedures and the sluggish transmission and processing of data, the ‘timely’ warning system generated a vast amount of paperwork, some of which was extracted from reports published two or three months late. But it could also be galvanised into action by a simple telephone call or radio message when a crisis broke out locally!

Some aid actors questioned the data produced by the DPPC and the reliability of the EWS. Donors suspected that estimates of people requiring aid were too high, and tended to reduce the numbers systematically. In 1994, the decentralisation of the early warning system triggered competition between the regions, which all attempted to maximise the aid flow (Maxwell, 2002).

International actors have therefore developed their own EWS. Donors are especially reliant on FAO agricultural production estimates and WFP evaluations of needs. An exhaustive inventory has revealed that at one stage government bodies, international agencies and NGOs were conducting some thirty evaluations between them (Standford, 2002). The authors of a major USAID study confined their field of inquiry to early warning systems and came up with no more than sixteen (Risk and Vulnerability in Ethiopia, Lautze & Yacob et al, 2003). Once again, it turned out that the principal source of information was MOA crop data, which was used by the national EWS and contributed to WFP, FAO and FEWS evaluations. Finally, in 2002, an attempt was made to restore the confidence of donor agencies by setting up ‘multi-agency’ teams to share the evaluation task. These teams are composed of experts who normally work for the government, international aid agencies and NGOs (Maxwell, 2002). Their evaluations provide the official data on which to base the future allocation of aid in Ethiopia.

E. THE CONSENSUS MILL

Like other EWS, the Ethiopian version is a tool designed to describe a situation of food insecurity and to elicit a response - food and/or financial assistance - from international actors. Given the system’s reliance on erroneous data and inadequate methodologies, its ability to produce accurate and independent information is in fact extremely limited. On the other hand, the investment in form has resulted in the construction of a technical instrument whose claims to technological and scientific objectivity mask, to a large extent, the methodological gaps in its descriptive function. But while it is technically prone to error, the EWS nevertheless facilitates consensus in the sense that it enables decision-makers to reach agreement. Given the scale of the investment in form, decision-makers can fall back on the supposed validity of this technical instrument when justifying their actions. The EWS thus fosters cooperation between the actors involved in the aid system by bringing them together around a single set of results.

The next step is to explore the modalities of this consensus, a task which requires us to examine in much greater detail the processes through which EWS produce their data. As J.P. Minvielle suggests, a socio-anthropological approach to internal decisional procedures is appropriate here. If we are to comprehend the “true circumstances” under which data is produced, we must open the “methodological black boxes” of institutional information production (Minvielle, 1994).