This appendix gives a very brief account of the approach followed in this study. For a more extensive treatment, the reader is referred to Appendix 2 in Alexandratos (1995). The final part of this appendix discusses briefly some of the problems with the data and the assumptions, and explains why only one scenario was designed for this study.
In projecting the likely evolution of the key food and agricultural variables, a "positive" approach has been followed, aiming at describing the future as it is likely to be (to the best of our knowledge at the time of carrying out this study), and not as it ought to be from a normative point of view. The study therefore does not attempt to spell out actions that need to be taken to reach a certain target (for example the World Food Summit target of halving the number of chronically undernourished persons by no later than 2015) or some other desirable outcome sometime in the future. The second overriding principle of the approach followed in this study was to draw to the maximum extent possible on FAO’s in-house knowledge available in the various disciplines present in FAO, so as to make the study results represent FAO’s “collective wisdom” concerning the future of food, nutrition and agriculture.
The quantitative analysis and projections were carried out in considerable detail in order to provide a basis for making statements about the future concerning individual commodities and groups of commodities as well as agriculture as a whole, and for any desired group of countries. For this reason the analysis was carried out for as large a number of individual commodities and countries as practicable (see Appendix 1). Another reason for the high degree of detail has to do with the interdisciplinary nature of the study and its heavy dependence on contributions provided by FAO specialists in the different disciplines. Such contributions can find expression only if the relevant questions are formulated at a meaningful level of detail. For example, a useful contribution can only be obtained from crop production experts, if questions of yield growth potential are addressed separately for maize, barley, millet and sorghum, not for coarse grains as a group, and preferably disaggregated in terms of agro-ecological conditions because, say, irrigated barley and rainfed semi-arid barley are practically different commodities for assessing yield growth prospects. Moreover, statements on, and projections of, land and water use cannot be made unless all the major crops are accounted for. For example, cereals account for about 55 percent of the total harvested area of the developing countries. An analysis limited to cereals would not provide a sufficient basis for exploring issues of land scarcities and possibilities of expansion in the future.
The variables projected in the study are (i) the demand (different final and intermediate uses), production and net trade balances for each commodity and country; and (ii) key agro-economic variables, i.e. for crops: area, yield and production by country and, for the developing countries only, by agro-ecological zone (irrigated and rainfed with the latter subdivided into dry semi-arid, moist semi-arid, subhumid, humid land and fluvisols/gleysols); and for livestock products: animal numbers (total stock and offtake rates) and yields per animal.
A significant part of the total effort is devoted to the work needed to create a consistent set of historical and base year data. For the demand-supply analysis, the overall quantitative framework for the projections is based on supply utilization accounts (SUAs). The SUA is an accounting identity, showing for any year the sources and uses of agricultural commodities in homogeneous physical units, as follows:
Food (direct consumption) + Industrial non-food uses + Feed + Seed + Waste = |
The database has one such SUA for each commodity, country and year (1961 to 1999). The data preparation work for the demand-supply analysis consists of the conversion of the approximately 330 commodities for which the primary production, utilization and trade data are available into the 32 commodities covered in this study, while respecting SUA identities (see the Note on commodities in Appendix 1). The different commodities are aggregated into commodity groups and into “total agriculture” using as weights world average producer prices of 1989/91 expressed in “international dollars” derived from the Geary-Khamis formula as explained in Rao (1993). The growth rates for heterogeneous commodity groups or total agriculture shown in this study are computed from the value aggregates thus obtained.
A major part of the data preparation work, undertaken only for the developing countries, is the unfolding of the SUA element “production” (for the base year only, in this case the three-year average 1997/99) into its constituent components of area, yield and production that are required for projecting production. For crops, the standard data in the SUAs contain, for most crops, also the areas (harvested) and average yields for each crop and country. These national averages are not considered by agronomists to provide a good enough basis for projections because of the widely differing agro-ecological conditions in which any single crop is grown, even within the same country. An attempt was therefore made to break down the base year production data from total area under a crop and an average yield into areas and yields for rainfed and irrigated categories. The problem is that such detailed data are not generally available in any standard database. It became necessary to piece them together from fragmentary information, from both published and unpublished documents giving, for example, areas and yields by irrigated and rainfed land at the national level or by administrative districts, supplemented by a good deal of guesstimates.
No data exist on total harvested land, but this can be obtained by summing up the harvested areas reported for the different crops. Data are available for total arable land in agricultural use (physical area, called in the statistics "arable land and land in permanent crops"). It is not known whether these two sets of data are compatible with each other, but this can be evaluated indirectly by computing the cropping intensity, i.e. the ratio of harvested area to arable land. This is an important parameter that can signal defects in the land use data. Indeed, for several countries the implicit values of the cropping intensities did not seem to make sense. In such cases the harvested area data resulting from the crop statistics were accepted as being the more robust (or the less questionable) ones and those for arable area were adjusted in consultation with the country and land use specialists (see Alexandratos [1995] for a discussion of these problems).
The bulk of the projection work concerns the drawing up of SUAs (by commodity and country) for the years 2015 and 2030, and the unfolding of the projected SUA item “production” into area and yield combinations for rainfed and irrigated land and, likewise, for livestock commodities into the underlying parameters (number of animals, offtake rates and yields).
The overall approach is to start with projections of demand, using Engel demand functions and exogenous assumptions on population and GDP growth.1 Subsequently, the entry point for the projections of production is to start with provisional projections for production for each commodity and country derived from simple assumptions about future self-sufficiency and trade levels. There follow several rounds of iterations and adjustments in consultation with specialists on the different countries and disciplines, with particular reference to what are considered to be “acceptable” or “feasible” levels of calorie intakes, diet composition, land use, (crop and livestock) yields and trade. Accounting consistency controls at the commodity, land resources (developing countries only), country and world levels have to be respected throughout. In addition, but only for the cereal, livestock and oilseeds sectors, a formal flex-price model was used (FAO World Food Model; FAO, 1993b) to provide starting levels for the iterations and to keep track of the implications for all variables of the changes in any one variable introduced in the successive rounds of inspection and adjustment. The model is a partial equilibrium model, composed of single commodity modules and world market feedbacks leading to national and world market clearing through price adjustments. It is emphasized that the results of the model projections (whether the single Engel demand functions or the flex-price model) were subjected to many rounds of iterative adjustments by specialists on countries and of many disciplines, particularly during the phase of analysing the scope for production growth and trade. The end product may be described as a set of projections that meet conditions of accounting consistency and to a large extent respect constraints and views expressed by the specialists in the different disciplines and countries.
It should be emphasized here that the projections presented in this study are definitely not “trend extrapolations”, whether the term is used to denote the derivation of a future value of any variable by simple application of its historical growth rate to the base year value (exponential trend) or the less crude notion of using time as the single explanatory variable in functional forms other than exponential, e.g. linear, semi-log, sigmoid, etc. For one thing, projecting all interlinked variables on the basis of estimated functions of time is a practical impossibility; for another, projecting any single variable at its historical growth rate (which could be negative, zero or very high) often leads to absurd results. Therefore, the term “trend” or “trend extrapolation” is not appropriate for describing these projections.
Summary statements of the methodologies of supporting and complementing analyses are given in the main body of this report: for example, the approaches followed for estimating the number of chronically undernourished people (Box 2.1), for deriving estimates of land with potential for rainfed agriculture (Box 4.1), for estimating water requirements in irrigated agriculture (Box 4.3), and for deriving projections of fertilizer consumption (Section 4.6).
The significant commodity and country detail underlying the analysis requires the handling of huge quantities of data. Inevitably, data problems that would remain hidden and go unnoticed in work conducted at the level of large country and commodity aggregates come to the fore all the time. Examples of typical data problems are given below.
Data reliability. When revised numbers become available in the successive rounds of updating and revision of the historical data, it is not uncommon to discover that some of the data were off the mark, sometimes by a very large margin. It may happen therefore that changes projected to occur in the future have already occurred in the past. A typical case is presented in Chapter 2, Box 2.2. There the point is made that the revisions of the population (downwards) and food production data (upwards) in Nigeria implied that in the previous (1995) edition of this study, the projections were based on a food security situation in Nigeria that was worse than the one actually prevailing, assuming the new revised data are nearer reality. Another example: at the time of writing of this report (mid-2002), the latest available revised trade data for Namibia show significant rice imports that were not present in the previous data. They lead to an increase by 50 kg of the per capita food consumption of cereals for the base year 1997/99 of this study. The revised total food consumption is 2600 kcal/person/day, up from the 2090 kcal/person/day before the revisions. As a result, the revised estimate of undernourishment for 1997/99 is 9.5 percent of the population, down from the 33 percent before the data revisions.
Obviously, there is not much that can be done about this problem as errors in historical data become apparent after completing the projections. Such changes in the historical data also bedevil attempts to compare in any degree of detail the projections of earlier editions of this study with the actual outcomes for the latest year for which data are available. The comparisons occasionally shown in Chapters 3 and 4 are for the developing countries as a whole: net cereal imports (Figure 3.7) and production and yields of wheat, rice, maize and other coarse grains (Figure 3.12 and Box 4.4). As such, they are not greatly influenced by significant revisions in the historical data of individual countries.
Unbalanced world trade. A second data problem relates to the large discrepancies often encountered in the trade statistics, i.e. world imports are not equal to world exports. Small discrepancies are inevitable and can be ignored but large ones pose serious problems since in the projections exporting countries must produce export surpluses equal to the net imports of other countries. For example, the sugar exporters had net exports of 32.3 million tonnes in 1997/99 while importers had net imports of only 28.5 million tonnes, leaving a world imbalance of 3.7 million tonnes. In the projections, the importers are estimated to need net imports of 35.2 million tonnes in 2015, an increase of 23 percent. If the discrepancy in the base year were to be ignored, the export surplus of the exporters should also be 35.2 million tonnes, i.e. only 9 percent above the 32.5 million they exported in 1997/99, thus greatly distorting the analysis of their export prospects. By necessity, the unsatisfactory solution of assuming that a discrepancy of roughly equal magnitude to that of the base year will also prevail in the future had to be adopted (see Table 3.23, Chapter 3).
There are good reasons why discrepancies arise in the trade statistics, e.g. differences in the timing of recording of movement of goods in the exporting and importing country, although this can hardly explain some very large discrepancies, e.g. world exports of refined sugar are 20 percent higher than world imports, while for concentrated orange juice world exports (85 percent of them from Brazil) are double world imports. At the same time, world imports of single strength orange juice exceed world exports by an almost equal amount.
Some problems with the exogenous assumptions. As an example, the impossibility of foreseeing which countries may face extraordinary events leading to their being worse off in the future than at present is mentioned. In Chapter 2 it was noted that several countries suffered declining levels of food consumption, some of them in the form of collapses within the span of a few years, e.g. the Democratic People’s Republic of Korea, Iraq, Cuba, Afghanistan, the Democratic Republic of the Congo and many transition economies. In most cases such collapses result from the occurrence of difficult to predict systemic changes or crises, or from outright unpredictable events, such as war or civil strife. It is impossible to predict which countries may be in that class in the future. Therefore, in the projections each and every country is shown with a higher food consumption per person than at present, some significantly better, others less so and several remaining with critically low levels. This is the result, in the first place, of the exogenous income growth assumptions that allow only rarely for the eventuality that per capita income of individual countries might in 30 years be lower than at present.
The prospect that only few countries may suffer income declines is, of course, at variance with the empirical evidence that shows quite a few countries having lower incomes today than three decades ago. The World Bank has data for 80 of the developing countries covered individually in this study (World Bank, 2001b; Table 1.4). No fewer than 28 of them have had negative growth rates in per capita GDP in the period 1965-99 (the number is larger if the transition economies are included). They include many of the countries devastated by war or civil strife at some period from 1965 to 1999. As noted, it would be foolhardy to predict or assume which countries may have similar experiences in the future. For example, in the 1988 edition of this study (Alexandratos, 1988) with projections to 2000, the collapse of food and agriculture in the formerly centrally planned economies of Europe and their virtual disappearance as large net importers of cereals had not been predicted.
Why only one scenario. In this study, only one possible outcome for the future based on a positive, rather than normative, assessment is presented. Alternative scenarios have not been explored for a number of reasons, some conceptual, some practical, and usually a mix of both. Producing an alternative scenario is essentially a remake of the projections with a different set of assumptions. On the practical side, the major constraint is the time-consuming nature of estimating alternative scenarios with the methodology of expert-based inspection, evaluation and iterative adjustments of the projections. On the conceptual side, defining an alternative set of exogenous assumptions that are internally consistent represents a challenge of no easy resolution. For example, among the major exogenous variables are the projections of population and income (GDP). As discussed, for population the medium variant demographic projections of the United Nations were used. There are also high and low variants. In estimating an alternative scenario with, say, the high variant, it would not be known how the exogenous GDP projections should be modified so as to be internally consistent with the high population variant. If the GDP growth rates were retained unchanged, projected per capita incomes would be lower, and this would mean implicitly accepting that population growth is detrimental to economic welfare. If the GDP growth rates were raised to keep projected per capita incomes unchanged, it would mean accepting that population growth made no difference. Neither of the two views can be correct for all countries. In actual life, some countries would be better off with higher population growth and some worse off (see more discussion in Box 2.3, Chapter 2). It would be impossible to define in an empirically valid manner what the relationships could be for each of the more than one hundred countries analysed individually in this study.
The one alternative scenario that it would be highly desirable to have is one that would introduce feedbacks from agriculture to the overall economy, at least for the countries in which agriculture is a substantial component of the economy. The methodology used in this study is of the partial equilibrium type, that is, interdependence is accounted for among, and balance is brought about in, the demand and supply of the individual agricultural commodities, at the country and world levels. Other aspects of interdependence and balance in the wider economy are ignored, e.g. how a more robust agricultural performance would eventually contribute to a higher GDP growth rate than originally assumed and how the latter would in turn stimulate demand for food and agriculture itself. To introduce such general equilibrium elements in the analysis, rather sophisticated economy-wide models would have to be built and validated for the individual countries. This is a quasi impossible task, partly because of the time and resources required and partly because the data available for many of the countries for which such analyses would be most appropriate (low-income ones with high dependence on agriculture) are generally not adequate to support such an undertaking. Circumventing the problem by assuming arbitrarily the existence of linkages between agriculture and the rest of the economy (e.g. that a 1 percent increase in agricultural GDP causes the rest of the economy to grow by x percent) would not do. As noted above, in a number of countries robust agricultural growth was associated with meagre or declining growth for the rest of the economy, implying a negative link if this simplistic approach were followed. Obviously, rather sophisticated economy-wide analysis would reveal the reasons why such “perverse” relationships exist in the data, and could even lead to the conclusion that some of the data are outright wrong.
In conclusion, alternative scenarios would be certainly useful for exploring the future in the face of uncertainties about how key variables of the system may evolve. In this study, an attempt was made to contain this uncertainty by bringing to bear the expert judgement of the discipline and country specialists on the future values of the relevant variables (e.g. rates of growth of yields, land, irrigation, etc.). Running a scenario with alternative values for one or more of these variables would mean repeating this process. Much of the work for a new scenario would be devoted to the definition of plausible alternative values. It is just not a question of assuming that, for example, irrigation would expand at a higher rate than in the baseline projection in each and every country. This would be impossible for some countries because of physical water constraints. The same holds for higher yield growth rates: the potential exists in some countries and crops but not in others. If realistic alternative paths for such values cannot be defined, the results of estimating alternative scenarios with blanket assumptions about uniform changes in the values of some variables in each and every country would certainly be misleading rather than illuminating.
