Methodology of quantitative analysis and the projections

Contents - Previous - Next

Commodity and country detail in the analysis

The quantitative analysis and projections were carried out in considerable detail in order to provide a basis for making statements, generally policy related ones, about the future concerning: (a) individual commodities and groups of commodities as well as agriculture as a whole; and (b) any desired group of countries. These requirements dictated the need to carry out the analysis: (a) individually for as large a number of commodities as practicable and to account for a high share of total agricultural output; and (b) for individual countries for as many countries as practicable. Although many statements refer to regions, such statements are often required also for smaller geographic or functional country groups which partly overlap with regions as well as among themselves, e.g. grouping countries by income level, by degree of dependence on agriculture, by participation in economic cooperation schemes, etc.

The above-indicated degree of country and commodity detail makes it possible to use the results of the study to address issues at the most appropriate level of commodity/country interface. For example, issues of the food and agriculture futures of countries with high dependence on sugar (e.g. Cuba) can hardly be addressed without a view of the sugar sector, those of Ghana and Cote d'Ivoire of cocoa, those of Zimbabwe and Malawi of tobacco, those of Malaysia of rubber and palmoil, and so on. A study limited to, say, cereals and large country groups would not be very helpful in addressing these kinds of issues.

There are two more reasons why this degree of detail is necessary. The first is the focus of the study on problems of natural resources for agriculture, i.e. land and water use. 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 only about 50 percent of the total harvested area of the developing countries (excluding China). 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 second reason has to do with the interdisciplinary nature of the study and its heavy dependence on contributions provided by 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, no useful contribution can be obtained from the production, country, trade, etc., specialists for a commodity group (e.g. coarse grains or fruit) and a region. It is more productive to obtain contributions for specific countries and individual commodities, e.g. not coarse grains but separately for maize, barley, millet and sorghum; not fruit but bananas and citrus, and so on. Likewise, expression of views on production prospects requires further disaggregation in terms of agroecological conditions because, say, irrigated barley and rainfed semi-arid barley are practically different commodities for assessing yield growth prospects (this aspect is discussed below).

The commodities and countries covered individually are listed in Appendix 1. Concerning commodities, there are 26 crop and 6 livestock products. This is for the analysis leading to the derivation of the demand and supply balances. For this analysis, all oilcrops are recognized as one commodity (oil equivalent), and the same goes for sugar and milk. However, these are heterogeneous commodities at the production level. Therefore, for production analysis, the commodity "vegetable oil" is represented by 7 oilcrops (rising to 8, including the cottonseed), sugar by 2 sugar crops (cane, beet) and milk by cow and sheep/goat milk, raising the total to 33 crops and 7 livestock products. Although the criterion for selecting which commodities to analyse individually has been their importance for the developing countries as a whole, it is inevitable that some commodities of particular importance to some countries are not covered individually, e.g. rapeseed (important for South Asian countries) and safflower seed (India, Mexico, Ethiopia) which are lumped together into the group "other oilseeds".

Concerning countries, 127 are covered "individually", of which 93 developing (representing 98.5 percent of their total population) and 34 developed representing nearly all their population. The projections methodology for the developing countries is more detailed and more demanding in data and time than that for the developed countries, because for the latter the production projections are carried out summarily and without distinguishing agroecological zones and accounting for land constraints (see also Editor's Preface). Therefore, the workload involved for each developing country is much higher than for each developed country.

Data preparation

The variables projected in the study are: (a) the demand (different final and intermediate uses), production and net trade balances for each commodity and country; and (b) for the developing countries only, key agronomic variables, i.e. areas, yields and production by country, crop and agroecological zone (land class) for crops; and animal numbers (total stock, off-take rates) and yields per animal for the livestock products. 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 the 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) + Industrial non-food uses + Feed + Seed + Waste = Total domestic use = Production + (Imports-Exports) + (Opening stocks-Closing stocks)

The data base has one such SUA for each commodity entering the demand supply analysis, country and year (1961 to 1990 at the time the study was initiated). The data preparation work for the demand-supply analysis consists of conversion of the about 330 commodities for which the production, utilization and trade data are available into the 32 commodities mentioned above, while respecting the SUA identities. This is no simple matter as the accounting relationships between commodities range from the fairly simple (e.g. converting pasta products and wheat flour in the consumption and trade statistics to wheat equivalent, though also here complexities are not absent, e.g. converting imported flour into wheat at the extraction rates of the importing or of the exporting country), to the extremely complex (e.g. converting imported margarine into vegetable oil equivalent and interfacing it with the vegetable oil equivalent of domestic oilseeds; or converting orange juice into fresh fruit equivalent). Unavoidably there remain loose ends in this complex accounting framework. FAO has work under way to improve the system and a publication on this matter is in preparation.

The different commodities are aggregated into groups and into "total agriculture" using as weights world average producer prices of 1979/81 expressed in "international dollars" derived from the Geary-Khamis formula as explained in Rao (1993). These are the price weights used to construct the FAO production indices (Laspeyres formula). The growth rates for heterogeneous commodity groups or total agriculture shown in this study are computed from the thus obtained value aggregates. The measurement of changes in agricultural aggregates obtained from these price weights is subject to limitations for particular uses, e.g. for drawing inferences about the pressures on natural resources generated by a given increment in production (see Chapter, note 17). It is also noted that each commodity has the same price weight in all countries. This means that one unique set of relative price weights is used to aggregate the production in all countries. The resulting growth rates of production, consumption, etc., can, therefore, differ from those that would be obtained if country-specific relative price weights were used. But the use of unique price weights makes the resulting growth rates comparable among countries.

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 1988/90) into its constituent components of area, yield and production which are required for projecting production. For the 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 the agronomists to provide a good enough basis for the projections because of the widely differing agroecological conditions in which any single crop is grown, even within the same country. That is, judgements about future yields cannot be made without further information on whether, for example, barley is grown in irrigated or rainfed land and at what yield levels; and within the rainfed category, information is needed on whether it is grown on land with sufficient rainfall and good soils or in semi-arid conditions and poor soils. These considerations led to the decision to attempt to break down the base year production data from total area under a crop and an average yield into areas and yields for five rainfed and one irrigated categories. These categories (land classes) are described in detail in Chapter and are not repeated here. The problem is that such detailed data are not generally available in any standard data base. It became necessary to piece them together from fragmentary information: some of it contained in published documents giving areas and yields on irrigated and rainfed land or by administrative districts; some of it from unpublished documents. The results of this research were supplemented by guesstimates. The result of this operation is a matrix of size 33 x 15, with one row for each crop and one area and one yield column for each land class (2 x 6) and 3 columns of control totals (harvested area, yield, production at the national level, HA, Y, P respectively) subject to the following equalities for each country (i= 1..6 land classes, j= 1...33 crops):

(sum of harvested areas of crop j in each land class i equals total harvested area of crop j, the control total for harvested land in each crop)

(sum of harvested areas of each crop j equals total harvested area of the country HA)

(sum of production of crop j in each land class equals total production crop j, the control total for production of each crop)

In principle, there is no control total for HA in the standard data-set, but one is obtained by summing up the harvested areas reported for the different crops. There are, however, data for total arable land (AR) in agricultural use (physical area, not just harvested, called in the statistics "arable land and land in permanent crops"). It is not known whether the HA data (obtained by summation of the HAs given for each crop) and the AR data are compatible with each other. Compatibility of the two data-sets can be evaluated indirectly by computing the ratio of harvested area to arable land, i.e. the cropping intensity (CI). This is an important parameter which can signal defects in the land use data. Indeed, for several countries the implicit values of the CIs 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. The objective was to have a set of harvested and arable land use data which, to the belief of the specialists, were more credible, more compatible with each other and more representative of actual country situations than the data reported in the standard sources. This operation for adjusting the data for the countries with evidently inconsistent ones was facilitated by the fact that the fine-tuning of the CI parameter had to be performed for each of the six land classes, not for the country as a whole. For example, CI values of over 0.8 and up to well over 2.0 are acceptable for irrigated areas depending on country knowledge about climate, water shortages, double cropping, farming systems, etc. On the other hand, in rainfed semi-arid areas and in most rainfed land classes in countries with considerable shifting agriculture CI values are normally below 0.5. These problems are discussed in Chapter where also the adjusted and unadjusted data are given and compared.

Following this unfolding of the land use data, a final equality has to be satisfied:

(sum of arable land in each land class i equals total arable land in the country, the adjusted control total)


The bulk of the projections work concerns: (a) the drawing up of SUAs (by commodity and country) for the year 2010; (b) the unfolding of the projected SUA item "production" into area and yield combinations for up to six agroecological conditions (land classes); and (c) the drawing up of land use balances by land class, including irrigated land.

The projections for all SUA items and countries for the cereals, livestock and oilcrop commodities are derived, in the first place, from a formal multi commodity, multi-country, flex-price model which is used in the medium-term commodity projections work of FAO, the FAO World Food Model (WFM). A detailed description of the model together with its parameters is given in FAO (1993i) and is not repeated here. Suffice to say here that: (a) the model provides year-by-year world price equilibrium solutions for the commodities covered; (b) it has demand (for food, feed, other uses) and supply (area, yields, animal numbers, etc.) equations for each country; (b) each country's solution is influenced by those for every other country through the imports and exports which are equated at the world level by price changes; (c) the extent to which price changes are transmitted to each country is determined by wedges between the domestic and world prices. These wedges represent the policy variables and can be varied to generate alternative trade policy outcomes; and (d) the projections are subject to many rounds of adjustments following inspection by specialists on the basis of the criteria described below. The adjustments are "absorbed" by the model by fine-tuning its parameters and coefficients, usually the trend factors. The model does not, however, have natural resource (land, water) constraints, nor does it generate relevant balances and parameters, e.g. arable land, cropping intensities, etc. In conclusion, the results generated by the model are a major element, but only one among many, which enter the determination of the projections used in this study and only for the cereals, livestock and oilcrop commodities (the WFM commodities). For some other commodities (e.g. sugar, rubber, cotton, jute), single commodity models were used to generate the initial projections which were subsequently subjected to several rounds of inspection and adjustment.

For these same commodities as well as for all others, parallel projections are prepared for each SUA element, as follows.

The element food, as represented in the SUA (i.e. food availability for direct human consumption)', is projected in per caput terms using the base year data for this variable, a set of estimated food demand functions - Engel curves - for up to 52 separate commodities in each country and the assumptions of the growth of per caput incomes (GDP). The results are inspected by the commodity and nutrition specialists and adjusted taking into account any relevant knowledge and information, in particular the historical evolution of per caput demand and the nutritional patterns in the country examined. Subsequently total projected food demand is obtained by simple multiplication of the projected per caput levels with projected population. Projected food demand may be further revised in the process of projecting the other elements in the SUA, in particular production and net imports (see below).

Industrial demand for non-food uses is projected as a function of the GDP growth assumptions and/or the population projections and subsequently adjusted in the process of inspection of the results. This item is important for only a few countries and products, e.g. sugar in Brazil or maize in the USA, both for fuel. The historical data are particularly weak and they often represent the domestic disappearance not accounted for in the other SUA items.

Feed demand for cereals is derived simultaneously with the projections of livestock products from the relationships between these variables in the solution of the above-mentioned WFM and further checks are performed by multiplying projected production of each of the livestock products with country-specific input/output coefficients (feeding rates) in terms of metabolizable energy supplied by cereals and brans. The part that can be met by projected domestic production of brans is deducted and the balance represents cereals demand for feed. Feed demand for oilseed proteins (computed, in the first place, in terms of crude protein equivalent) is taken mostly from the relationships in the projections of the World Food Model. Feed use of other products with a feed-use component in the historical SUA data is obtained by ad hoc methods, mostly as a proportion of total production or total demand. It is noted that these feed-use projections do not provide a complete interface between animal production and feed supplies or resources in each country because of the lack of systematic data for complete feed balances, i.e. including non-concentrate feeds (cultivated fodder, natural grass, byproducts other than cereal brans, etc.).

Seed use is projected as a function of production using seeding rates per hectare. Waste (post-harvest to retail) is projected as a proportion of total supply (production plus imports).

This parallel method, unlike the WFM, does not project year 2010 stock changes. This does not mean that present stocks are assumed to remain constant but rather that adjustments in stocks between the base year level and any required level in year 2010 may occur in any year(s) between 1988/90 and 2010. In this case, the impact on production will appear only as temporary deviation(s) from the smooth growth path represented by a curve joining the base year production level to that of 2010, ignoring fluctuations in the intervening years. Whether or not year 2010 production includes a provision for "normal" stock changes (i.e. to maintain stocks at the desired percentage of consumption already achieved before 2010) makes little difference to the average growth rate of production for 1988/90-2010 if the deviations from the constant growth rate path in the intervening years are ignored.

Production and trade projections for each country involve a number of iterative computations and adjustments. The solution of the WFM provides the initial levels for cereals, livestock and oilcrops. For all commodities, the criteria for the projections and their iterative adjustments are as follows:

1. Commodities in deficit in the base year (developing countries only). A preliminary "target" level is set for 2010 taking into account the projected demand, production growth possibilities (evaluated in more detail in subsequent steps of the analysis, step below) and preliminary values for projected self-sufficiency ratios. The latter are used as a computational device to define preliminary levels of future production "targets" to be evaluated in subsequent steps. They do not reflect expression of any generalized preference for increased self-sufficiency, but they do take into account whatever is known about country preferences about self-sufficiency objectives which influence their policies. The more general issues of the pros and cons of the self-sufficiency objectives are discussed in Chapter 7.

2. Commodities exported in the historical period and the base year (developing countries only). It is assumed that they will continue to be exported in amounts which will depend on the country's possibilities to increase production, a preliminary assessment of import demand on the part of all the other countries which are deficit in that commodity and an assessment of the country's possibility to have a share in total world import demand resulting from whatever is known about policies and other factors influencing the country's competitive position (see, for example, the discussion on natural rubber in Chapter 3). Since for world balance total deficits of the importing countries must be equal to total surpluses of the exporting countries there is an element of simultaneity in the determination of the production levels of all commodities in all countries. This element of simultaneity is fully accounted for in the WFM solution, but only for the WFM commodities (cereals, livestock, oilcrops). For the other commodities, the problem is solved in a number of successive iterations rather than through a formal model, the key element being expert judgements on market shares in world exports and of somewhat more formal evaluations of the production possibilities, as explained below. Based on the above considerations, preliminary production "targets" are set for the export commodities of each developing country. They are equal to their own domestic demand plus the preliminary export levels. Once the preliminary production targets are set for all commodities, the missing elements of the demand side of the SUA which depend on the levels of production (feed, seed, waste) can be filled in.

At this stage complete preliminary SUAs are available for the year 2010 for all commodities and all the developing countries, showing for each commodity and country all the domestic demand elements and production. The differences between total demand (total domestic use) and production are the preliminary net trade positions (imports or exports). The next step is to construct the projected SUAs for the developed countries whose net trade balances must match those of the developing countries, with opposite signs.

3. The demand items of the SUAs for the developed countries are projected in the same manner as for the developing countries, subject to the differences in the criteria used to evaluate and adjust the projections. The great uncertainty surrounding, and the special factors applying to, the prospective evolution of the main food and agriculture variables in the ex-CPEs is underlined (see Chapter).

4. For the commodities not produced, or produced only in insignificant quantities in the developed countries (tea, coffee, cocoa, bananas, natural rubber, jute, cassava), nearly all their demand translates into import requirements. These, together with the import requirements of the developing countries in deficit, define the total market for these commodities available to the developing exporting countries. Their provisional production and export levels, set as described above, are then adjusted judgementally to equate them to the total import requirements.

5. A second set of non-WFM commodities comprises those produced in substantial quantities in both the developed and the developing countries but for which the latter have been traditionally substantial net exporters (mainly sugar, citrus, tobacco, cotton). The aspects of these commodities taken into account in the projections are discussed in Chapter 3. For example, they include the prospect that there will not be any significant trade liberalization effects on the sugar trade flows and that outcomes will be influenced by changes in the sugar trade relationships between the ex-CPEs and Cuba; that the raw cotton trade will be influenced by the ever-growing role of the developing countries in world exports of cotton manufactures; and that the tobacco production and trade prospects will be decisively influenced by trends for per caput consumption to decline in the OECD countries and to increase in the rest of the world.

6. The last group of commodities comprises those for which the developing countries and, at present, also the ex-CPEs are major importers and the other developed countries as a group are the major suppliers of these imports (mainly wheat, coarse grains, milk). For these commodities the projections from the WFM solution constitute the initial values. They are adjusted iteratively in a number of rounds, as follows. The possible production and net trade positions of the ex-CPEs are adjusted on the basis of the criteria discussed in Chapter 3, e.g. the prospect that their demand will be in the future lower than in the base year and that their production will recover, leading initially to import substitution and eventually to net exports. The resulting net trade positions of the ex-CPEs, together with those derived earlier for the developing countries define the total net exports required to be generated by all the other developed countries as a group. Together with the projected growth for their own domestic use they define future production levels. By and large, the required growth in their collective production is modest. For example, future production of cereals for the three main exporting regions (North America, Western Europe, Oceania) is required to be 680 million tonnes. This implies a growth rate to 2010 of 0.7 percent p.a., measured from their 1991/92 two-year average production of 595 million tonnes. This growth rate is considered to be well within the bounds of their collective potential to increase production and, therefore, no further work is done to evaluate this projected level from the standpoint of natural resources and yield growth feasibility (see also Editor's Preface).

However, the issue how this additional production and exports will be shared between these three main exporting regions is not so easily solved. At the time this analysis was done only some of the domestic policy reforms were known and elements of those that were subsequently agreed under the Uruguay Round were known only in the form of provisional negotiating positions. It is only after April 1994 that the concrete country proposals about measures to implement the Uruguay Round Agreement on Agriculture started to become known. In the absence of this information, the projections for production and net trade are shown in Chapter 3 for the main exporting regions as a group (Table 3.17 and Annex to Chapter 3). Some of the factors that may influence the production and net trade outcomes of the individual regions are discussed in Chapter 3, e.g. the prospect that commitments to reduce subsidized exports and to allow for minimum import access will translate into Western Europe's projected net exports of cereals being in the future no higher than in the base year. If this prospect materialized, all additional net exports would be supplied by North America and Oceania. However, these are tentative conclusions, subject to the many caveats of the models which generate them. A more concrete evaluation on the basis of the concrete Schedules of the different countries for the implementation of the Uruguay Round Agreement on Agriculture will be undertaken in FAO for the WFM commodities and selected other commodities.

7. At this stage the projections of demand, production and trade are complete: there is one projected SUA for each country and commodity (but only one aggregate complete SUA for the main developed exporting regions) and world imports equal world exports. These projections are, however, still provisional pending a more detailed evaluation of the feasibility of the production projections of the developing countries, from the standpoint of land and water use and the growth of yields. The basis for this evaluation is provided by: (a) the detailed data-set constructed for the base year in the above-described phase of data preparation (matrix of base-year areas and yields by crop and land class); and (b) whatever knowledge and judgements the specialists on countries, commodities and the different agronomic disciplines could contribute. The objective of this operation is essentially to "test the feasibility" of the preliminary crop production projections for each developing country by generating for 2010 the 33 x 15 matrix of areas and yields by land class which was constructed for the base year, as described earlier. China is not included in this exercise for the reasons discussed in Chapters 3 and 4.

The relevant work brings to bear on the evaluation of the crop production projections whatever knowledge the country, commodity and discipline specialists possess concerning the production conditions, national plans, etc., for individual crops and countries. Examples include: (a) knowledge of existing or impending shifts in the different countries towards hybrid rice varieties, or alternatively, back towards the more palatable traditional varieties, is an invaluable element for forming an informed judgement about the rice sector prospects; (b) country plans, in formulation or in execution, for example irrigation or establishment of tree crops such as oilpalm, rubber, etc.; (c) knowledge about preferential trading arrangements that influence future production, e.g. for bananas, sugar; and so on for other products, country by country.

The data on the land resources of each country and their suitability for crop production (see Chapter 4) constitute an important input in the derivation of the possible future land-yield combinations by land class. Assumptions are first made of what are feasible rates of harvested land expansion by agroecological class (through use of more land from the reserves and or through increased cropping intensities, including expansion of irrigation). Similar assumptions are made for yield increases and the land allocation to each crop. Since a multitude of detailed assumptions and different specialists are involved, continuous iterative computations of the whole system are made to ensure that constraints of land availability and notional upper yield levels for 2010 (both by country and land class) are respected. The end-result is that either the initial production target is fine-tuned and accepted or is revised downwards for some crops because land resources (of the required class, where applicable) are not sufficient or because the target requires yield increases considered by the specialists to be beyond achievement by 2010 even under reasonably improved policies.

8. Similar production analysis procedures are applied to the livestock production, except that the relevant parameters are animal numbers and yields (off-take rates, carcass weight, milk yields, eggs per laying hen) for the livestock species considered. Whatever knowledge is available about the feed resources of the different countries is utilized in evaluating the production prospects.

9. Subject to the results of the two preceding steps (7 and 8), final adjustments are made in the other SUA items for the commodities and countries for which the provisional production "targets" had to be changed during the feasibility tests. If the changes in projected production result in increased import requirements which are considered "excessive" for particular countries, their projected demand may be adjusted downwards, to make up for the shortfall in supplies, in whole or in part. Following this, a final iteration is made to adjust production and trade balances of other countries to make up the shortfall in production and cover the resulting increased import requirements in the developing countries whose initial provisional "targets" had to be lowered.

At this stage, the world demand, production and trade picture is completely quantified, backed by full quantifications for the developing countries of land and water use balances, their yield-harvested area combinations (by crop and land class) and the relevant livestock parameters.

10. The last step projects fertilizer use. This is obtained, in the first instance, by multiplying fertilizer input coefficients per hectare by projected harvested area. The coefficients are specific to each crop, land class and yield. These coefficients were generated by the agronomists using whatever data and other information were available (e.g. in FAO, 1989c and FAO/IFA/ IFDC, 1992) and supplemented by a good deal of judgement based on agronomic norms. Since they are yield specific, the method is equivalent to using yield-fertilizer response functions for each land class and crop. These coefficients are not country specific in the first instance, but they are made so by calibrating them on the basis of the implicit scalars required to reproduce for the base year the control data of total fertilizer use by country, which are available in the standard data sources. More details on the method are given in Bruinsma et al. (1983).

A summary evaluation of the methodology

The key characteristics of the methodology may be summarized as follows: (a) the analysis is conducted in great detail as regards commodities, countries, land classes, etc.; (b) behavioural relationships are used explicitly in the projections only for the WFM commodities and for all commodities in the projections of food demand; (c) prices play an explicit role in bringing about demand supply balance only for the WFM commodities and then only in the generation of the initial set of projections which is subsequently subjected to many inspections and adjustments; (d) links of agriculture with the rest of the economy are not accounted for, except for the influence of income growth on the food demand projections; (e) the method generates land use balances in great detail (by agroecological classes and with distinctions between harvested and arable land) and controls for land availability constraints; and (f) the method uses very diverse sources of data and all kinds of knowledge and judgements contributed by the different specialists on countries and disciplines.

This method has positive and negative aspects. On the positive side, the great detail of analysis means that the projections and related statements contained in this book for country groups, regions or the world as a whole as well as those for large commodity aggregates and the whole of agriculture are underpinned by detailed country/commodity quantifications. In practice, each global statement is derived from a summation of, and can be decomposed back into, a number of constituent single-country or commodity statements. This characteristic sets this study apart from most other global studies in which analyses are carried out at the level of major countries and/or regions and usually only for the major products grouped into a few commodity aggregates.

In many respects, the great detail of the analysis and the heavy dependence on multidisciplinary specialist input are two faces of the same coin. This is because such specialist input can be provided and utilized only if solicited at the level of detail in which the relevant expertise is available. The latter usually embodies knowledge of local conditions in widely differing country situations. This is the second major advantage of using this method.

This heavy dependence on great detail and specialist input is at the same time the major weakness of the method. The flow of such specialist input is, of course, controlled by a large data processing system which generates, after many iterations, an internally consistent set of projections. Such consistency is, however, only an accounting one. Projections based on specialist input suffer from the fact that the criteria and assumptions used and the implicit decision making mechanism cannot be formally described and they can vary from one person to another and over time. It follows that the projections cannot be strictly replicated at will, including for estimating alternative scenarios by varying certain assumptions only. This would have been possible if a formal model had been used for the projections.

There are advantages and disadvantages in the use of formal models for this type of work (Alexandratos, 1976). In the case of this study, the great amount of detail makes it impossible to conduct the analysis using one single formal model representing behaviour of the different actors (producers, consumers, governments) and with price-based market clearing mechanisms. Many of the data required for such an effort are just not available. For a formal modelling approach, the choice would have been between having: (a) a roughly estimated formal model with much less commodity, country and land use detail; or (b) a huge model with all the detail of this study but with the bulk of the parameters and coefficients being "guesstimates" rather than data. The former case is a clearly inferior option since it would make it impossible to evaluate the results using multidisciplinary specialist input. Moreover, the findings of the study would be of limited value for drawing conclusions at very diverse levels because of the much reduced commodity and country detail.

The second option (large model with a wide range of guesstimates for the parameters) is really a formal variant of the expert judgement-based approach used in this study, the difference being that expert judgements would be embodied in the guesstimates of the values of the model parameters and coefficients. Such an approach would be superior to the one of this study, since the utilization of the expert judgement input is subject to the discipline that the implied values of the parameters and coefficients must fall within a certain acceptable range. Iterations and dialogue would be greatly facilitated, alternative scenarios could be estimated and greater transparency would be assured. These advantages must be set against the greater resources and time required for model preparation, particularly for the development of the computing algorithms. The latter aspect can be daunting for a global model that would recognize 100-odd countries communicating through trade flows, over 35 commodities, up to 6 sets of production conditions per crop, and so on. It could easily absorb a disproportionate part of the resources of the study without assurance of a satisfactory end-product.

In conclusion, future improvements in the methodology should aim at introducing some of the advantages of formal models in the form of explicit statements of the assumed behavioural relationships and their empirical verification, replication of results and derivation of alternative scenarios in a consistent manner. It is, however, important that the strong points of the present methodology be preserved, viz. the detail of analysis as regards countries, commodities and production conditions as well as the associated possibility to use multidisciplinary input and to draw on very diverse sources of knowledge and expertise. Given "reasonable" resource and time limitations, it is unlikely that this could be achieved by attempts to build a formal model in all the detail of this study. Scarce resources could be used more productively if they are concentrated on improving selected components of the present methodology, as indeed was done in this study by constructing a formal model (the WFM) to capture the interdependencies between the cereals, oilseeds and livestock sub-sectors. Future work on agriculture and the environment would benefit from efforts to develop analytical methods to link geo-referenced data on agricultural resources to the assessment of production prospects.


1. The terms demand, consumption, availability and per caput food supplies are used interchangeably to denote the SUA element food.

2. Samples of elasticity estimates from household budget, or food consumption, surveys are given in FAO (1989d).

3. In these projections the team benefited from the cooperation of researchers from the Center for Agricultural and Rural Development, Iowa State University, USA (S. Johnson and W. Meyers) and the Institut für Agrarpolitik, University of Bonn, Germany (K. Frohberg).

Contents - Previous - Next