16. Selection of a Suitable Forecasting Method

The selection of a suitable method is naturally governed by the purpose to be served by the forecast. The objectives can range from the immediate need of arranging transport and storage fertilizer distribution in the next season to long-term decisions concerning production capacity for research, extension, infrastructure, policy and legislation. In the short run, there is not enough time for the effect of new strategies to be felt and the momentum of the past has a dominant influence on demand. This is why the time series approach is more suited for short and medium-term needs. Even so, it is necessary to correct the results of trend projections by allowing for factors such as price reductions, changes in credit availability, increased supplies, improved distribution or a new irrigation project. Time series analysis and trend extension modified by expert opinion can therefore meet short-range forecasting needs.

For long-term forecasting, causal models seem better suited. The effect of extension is felt more in the long run, eg. through extending fertilizer adoption to new users, increasing fertilizer application rates or extending adoption of high-yielding varieties. If expertise and data are available, the effect of each of the factors can be quantified on the basis of past demand behaviour and applied to future points of time and sets of circumstances for which we need the forecast. In view of the complexity of the causal method, however, it is normally necessary to rely on the trend extension method, duly corrected for new developments affecting demand. Long-range demand projections based on time series analysis can be upgraded through expert assessment of the effects of new developments.

Long-term demand cannot be considered in isolation from total potential demand which, as discussed, can be assessed using either the output approach or the input approach. The results of the output approach can be made more realistic by taking expert opinion on the likely state of environmental factors, such as the effect of non-fertilizer factors on agricultural production, increase in irrigation or speed of adoption of crop management practices. Similarly, the accuracy of the input method can be enhanced by specific sample surveys on current fertilizer application, extent of fertilizer adoption or crop acreage.

For short-term sales forecasts (as distinct from demand forecasts) the closer we get to the ultimate buyer, i.e. the farmer, to obtain information the less inaccurate the estimate will be. From this angle, the buyers' intention method (retailer survey) seems well suited. The second best would be the composite of sales force opinions. Sales forecasts must be viewed in the context of the overall demand growth in the country or region. Overlooking this aspect may result in the company losing its market share.

Trend extension and causative methods require extensive data on a variety of factors influencing fertilizer demand, such as crop area, fertilizer usage by crop, shifts in cropping patterns, application levels, etc. which are very difficult to collect and not up-to-date in most countries. Consequently, these methods are not of practical significance to most countries.

Marketing organisations just entering the fertilizer business in countries that have recently privatised fertilizer distribution have a problem in estimating their sales. With no background of fertilizer distribution, they have to rely on two methods, (i) survey of buyers' intentions, i.e. ascertaining from their newly appointed retailers the quantity they are likely to buy in the next season or year and extrapolating and amending the aggregated estimate using the judgement of their senior marketing staff or (ii) the market share approach, i.e. assessing their organisational capability, strength of their sales force and extensiveness of the distribution network and thereby estimating the likely initial market share.

Expert opinion and sample surveys play an important role as a substitute for past data when it is not available and are, therefore, very useful as supports for both short and long-term forecasts. When use of sophisticated tools is hampered by want of data and or expertise, simpler informal methods can be used, with suitable innovations to suit local limitations.

The following table suggests practical forecasting methods to meet different needs of governments and marketing organisations. Annex IV contains a selection chart summarising different methods and their characteristics. Flow charts and specimen workings for methods that are of practical significance appear in Annexes V to XV.