An introduction to decision analysis
by
Hector McNeill1
The purpose of this brief
This brief is an introduction to decision analysis and is part of a series in support of the ISEE Workshop 2007-2008 concerning Feedstocks, Fibres & Food. Its purpose is to explain the general methodology of decision analysis used to ensure that all relevant information, required to take a decision, has been obtained, assessed for quality and analysed in an objective and logical fashion.
The state of decision analysis in agriculture
Decision analysis is applied in agriculture in four principal areas:
- genetic resources development in the form of animal & plant breeding & selection strategies
- technical resources development through the application of specific experimental designs
- production operations to optimise short term resource allocation for production and supply chain logistics
- design and optimization of proposed investments
In all of these areas decision analysis is a mature methodology. However, we have to admit that there is a varying record of success in all areas when outcomes are compared with expectations; what was expected and determined on the basis of careful analysis, does not always occur.
Investment
The area of agricultural investment projects, in support of economic development, is where the most striking failures occur. Reviews by multilateral development Banks, such as the World Bank, indicate that well over 50% of investments do not perform. The number of projects where expectations meet outcomes achieved is very small. In some cases large projects have such poor performance that new loans are knowingly advanced to help countries pay back previous non-performant loans.
A review of some of the principal reasons for project failures on technical grounds is included in the second brief in this series entitled, "Fundamental relationships in the feedstocks, fibres & food domain". The less technical areas of failure will be covered in the related briefs 3, "Private & public money", 4, "Ethics & professional standards" and 5,"Law & evidence").
What is a decision?
In the context of this brief a decision is an act which identifies the actions to be taken using resources to achieve some objective. Thus resources are committed as a result of a decision. These include human, physical, technical, economic and financial resources.
In contrast to this definition of a decision (the commitment of resources) an agreement by a group, or a mental resolve of an individual, to "do something" does not become a decision until committment of resources take place. Such a mental resolve remains no more than an intent and for the purposes of this discussion is not considered to be a decision.
The implications of resource commitment
 As soon as resources are committed, some are used to organise and manage the implementation of the decision. This involves costs. If there were a subsequent re-evaluation leading to another decision to cancel the former decision, this would commit further resources to bring about the necessary actions to bring to a halt the consumption of resources justified by the former decision. The point is that the outcome of a decision involves resource costs and time. Normally the quantity of resources consumed on implementing decisions involving complex projects increase significantly during the implementation phase. This is why it is so important to invest time, and a lower quantity of resources, in pre-decision activities which can help raise the likelihood that the desired outcome will be achieved. This is the fundamental purpose of decision analysis.
Decision maker preferences - a note
Most decision analysis responds to a defined set of decision maker preferences where the decision maker is a person or agency which has the power to commit resources and is in possession of the legal basis to do so. There are many cases in agriculture where the decision maker and those affected by the decision are completely separate entities. Sometimes where there is a legal basis for the decision maker to commit resources the legal basis for taking a decision which affects people, who have not been consulted on the decision, places the resulting project on tenuous grounds. These are matters for other briefs.
There are usually two basic levels of decision analysis applied to any review of investment options in agriculture:
- a review of the production options in a specific area or an identification of the potential areas where it is feasible to produce a specific set of produce
- a more detailed project appraisal or evaluation is completed setting out the details as to the expectations of outcome for the best crops and/or locations
Resource commitment in agriculture
The success of crop production determined by the genotypes of plants and husbandry techniques applied is subject to the direct influence of predominant bioclimatic and natural resource factors including:
- edaphic regimes (E)- soil conditions including texture, structure and nutrient content
- water regimes (W) - availability to roots and humidity of air
- temperature regimes (T) - temperatures in the soil and ambient
This predominant influence will be referred to in this brief as the EWT complex.
Husbandry techniques involve the management, by farmers, of several economic factors of production which include:
- information
- manpower
- tools & equipment
- energy &/or traction
- variable inputs
- land area
Production Functions (PF)
In any geographic location, there is a predominant influence of the EWT complex over crop potential and the quest is to identify the production system which offers the highest likelihood of profitable sustainable production. Conventional production systems need to be assessed from the standpoint of their capacity to improve the standards of living of farmers. Where the decision maker's preference is to introduce systems of production which are both real income enhancing as well as economically, financially and environmentally sustainble in the long run, this presents a complex challenge for decision analysts.
The current relationships between output achieved and the inputs of labour, seed, pesticides and fertilisers provide the basis for describing and quantifying production relationships. These can be described as input-output relationship or a production function. Where the objective of a project is to improve production in some way then the decision analyst needs to build a rational production function of what can be expected under the proposed (new) circumstances. To achieve this, the results of breeding experiments and multiplication programmes provide some guidance for estimating likely improved production functions. On the other hand, the observed practice of farmers who are achieving better economic performance within areas with similar EWT complexes is another basis for determining what can be achieved.
In the end the decision analyst is left with the challenge of building a realistic production function or "model" of what can be expected in terms of output from production options defined in terms of different ranges of inputs and sometimes different husbandry methods (technique).
Project evaluation
In the area of investment analysis in agriculture there are well-established procedures of decision analysis which can be summarised as project evaluation (the word appraisal can be used interchangably with evaluation) whose purpose is to project the likely physical output achieved, quantify employment generation, economic margins, financial results in terms of a cash flow and other aspects such as import substitution and foreign exchange generation associated with exports. This decision analysis model has been applied since the mid-1950s and the organization which pioneered many of the more refined analytical techniques associated with project evaluation/appraisal was the International Bank for Reconstruction and Development, today known as the World Bank. Indeed, examples of World Bank projects were used as teaching reference materials in post-graduate university economics courses in the 1960s (e.g. School of Political Economy at Cambridge). Since then, extensions to project evaluation have occurred in terms of additional emphasis being given to environmental impacts and social issues which has been a tacit admission that decision maker preferences were not comprehensive enough in the past. This does not mean that there was an over-emphasis on economics, rather it means the production functions being used were deficient in that they did not make allowances for these important areas of concern.
Don't blame the economists - at least, not all the time
This evolution in project evaluation emphasises that unless the production functions devised for a project reflect reality then the whole project is compromised. No amount of sophisticated operations research models and optimization routines can compensate for a poor production function. A poor production function is one which does not represent reality. In professional terms this has an important message. The onus to establish relevant production functions, in physical input-output terms rests with those who work in this sphere. The relevance of the work of economists and subsequent financial work rests fundamentally upon the relevance of the production functions selected as the foundation of any project. The information coherence for the whole project is dependent upon the production function/s selected; indeed, the main reason for many project failures is lack of information on what constitutes a realistic production function. This constitutes a significant constraint on decision analysis effectiveness in agriculture and it is also an area where least advances have been made during the last 50 years.
Don't blame the agronomists either - it is often a matter of decision maker confusion
A common circumstance in the agricultural sector is that governments as decision makers will willingly promote large international loans to support agricultural investment projects but will not invest adequate funds in agricultural research. Such research as field experiments such as crop and animal performance trials, are fundamental sources of information in helping define relevant production functions and thereby increasing the likelihood of agricultural project success. Governments often see a false distinction between agricultual research which they consider to be long term and short term gains from promoting large agricultural projects which non-the-less, because of inadequate information, end up as long term failures.
Structural Production Functions (SPF)
One of the most apparent gaps in production function research is the lack of coherence between the EWT complex and production functions. Production functions will vary according to EWT and this in turn with geographic location (longitude and latitude) as well as altitude and time (within the crop season). Brief 2 in this series will describe how production functions can integrate EWT into Structural Production Functions and help move project evaluation work onto a firmer foundation.
The decision analysis cycle
An idealised summary of what decision analysts are doing in the project evaluation process, albeit not applied at that time to agricultural projects, was produced by Ronald A. Howard of Stanford University in 1968. This can be summarised as the decision analysis cycle and in diagramatic form is presented below:
In summary a deterministic phase is undertaken where the production functions to be applied are defined. The confidence with which the production function can be relied upon established the probabalistic phase of the assessment. Where there are doubts as to the model's representation of reality, then by identifying the areas where doubts exist one identifies where further information is required; this is the informational phase. It is possible that once further information is gathered and assessed that doubts as to other aspects of the model arise. As a result the decision analysis process cycles again with careful application of new information to the deterministic and probabalistic phases.
When the decision analysts are satisfied that the marginal gains to collecting more information are low and that their representation of solution options is adequate then the selection of options on the basis of decision maker preferences (criteria) can be made and a reccommendation for a decision made subject to any provisos which have arisen during the information phase. These might relate to assumption on market conditions, for example.
Opening up the field
So far, discussion has revolved around conventional agricultural project evaluation but decision analysis can be applied to any decision relevant to any level in the supply chain from the selection of where to grow a feedstock, the best locations for processing plants and, for example, the best location for a biofuels conversion complex. Logistics (transport, storage & distribution) are of fundamental importance in agriculture and decisions in this area can be handled adequately applying decision analysis. Systems for traceability, hazards analysis, packaging and wholesale and retail distribution are also areas where decision analysis has an important role to play. The specific methodologies and techniques applied in calculating and optimizing different aspects of a problem which can be applied during decision analysis will be covered in other briefs in this series.
All briefs are subject to update. Current status of this brief. Updated 9th March 2007, Version 1.04.
1 Hector McNeill is the Systems Coordinator at the Systems Engineering Economics Lab.
References:
An introduction to Decision Analysis, Matheson J.E. & Howard R.A., Decision Analysis Group, Stanford Research Institute, 1968.
Towards Structural Production Functions, McNeill H.W., CRESTFILM project, Systems Engineering Economics Lab, 1983.
| | ISEE Copyright© Statement
In the case of all briefs prepared for the Workshop Series 2007-2008
the copyright of content rests with the named author/s |
|