Lessons from a decade of whole-farm modelling in Western Australia
David J. PANNELL
Agricultural and Resource Economics, University of Western Australia, Nedlands 6907, Australia
This article outlines important lessons and experiences in successful development and use of the MIDAS whole-farm linear programming models in Western Australia. The article includes discussions of whole-farm planning and whole-farm modelling in general and various aspects of the MIDAS experience, including the model's history, strengths and weaknesses, uses, positive outcomes and negative aspects. Uses of MIDAS have included research prioritization, extension, policy analysis, education and provision of a database for other uses. Extension has focused on general messages for groups of farmers, rather than provision of a service for individual farmers. However, the area in which MIDAS has had the biggest impact has been in influencing the biological research activities of the Western Australian Department of Agriculture. The model has (a) brought together researchers (of various disciplines) and extension agents who otherwise would interact little; (b) allowed scientists and extension agents to assess the economic significance of particular biological or physical information; (c) influenced the thinking of researchers and extension agents about the whole-farm system; and (d) highlighted a large number of data deficiencies and allowed prioritization of research to overcome them. Experiences with MIDAS point to a number of strategies and practices which may be beneficial to others undertaking similar modelling efforts. Initiate collaboration with scientists and advisors at the earliest stages, and maintain it throughout. Cultivate support from the organization's leadership. Model at the greatest level of detail within which you can deliver useful results in a reasonable time. Emphasise representation of biological and technical details as well as financial factors. Make model assumptions public and open to criticism. Do not emphasise individual model solutions. Use sensitivity analyses to promote understanding of results, to put issues in perspective and to deal with data limitations and uncertainty.
Many management decisions made by farmers depend on factors external to the particular crop or livestock enterprise directly affected. Issues which arise at the level of the whole-farm, such as the allocation of limited resources between alternative enterprises, can be especially difficult to analyze. Whole-farm models of various types provide a vehicle for better understanding these issues and decisions. In Western Australia a particular whole-farm model, MIDAS (Model of an Integrated Dryland Agricultural System), has been an influential tool for research prioritization, extension, education and other uses for over a decade. The purpose of this article is to outline experiences in the development and use of MIDAS with a view to providing guidance to others commencing or contemplating development of similar models. The discussion starts with points on whole-farm planning and whole-farm modelling in general and then considers a wide range of issues arising out of the several versions of the MIDAS model.
Managing a farm can be blindingly complex. In deciding on the best mix of farm enterprises and management practices, the diversity and extent of relevant information is enormous. The choice of farm strategy may be influenced by the farmer's knowledge of: scientific issues (biological and/or physical), machinery, economic/commercial factors, political events, legal constraints, historical trends, climate/weather, environmental issues, personal circumstances and any number of practical considerations. Even if a farmer had a complete grasp of all relevant information, the problem of combining it and appropriately evaluating its significance for decision making would be very substantial indeed. A thorough and detailed analysis would certainly be beyond the capability of any single human mind.
Despite the apparent difficulties, farmers in general seem to cope well with their planning and day-to-day management decisions. Many are enthusiastic collectors of information and ideas. They do agonize over some decisions, but they are not defeated by the intractability of the problem. They appreciate that their decision making processes are not rigorous and formal, but they must be careful not to spend too much time on formal analyses because there are so many decisions to make and most of their time is required for the many physical tasks which are needed to run the farm. Very important decisions may have to be made (or at least finalized) during times when the time available for contemplation and analysis is at a minimum; i.e. during crop seeding. They manage by using judgment, guesses, hunches, outside advice and some limited numerical analysis and they can do so because of their intimate knowledge of their farms (Malcolm).
In my judgment, farmers' decisions made in this ad hoc way are usually very good. They are not perfect (farmers are human!) but they are usually near enough to the theoretical ideal for their particular circumstances to obtain most of the potential benefits. They are helped in this by the forgiving nature of many agricultural decisions; often there is a range of strategies around the optimum which give near-optimal levels of profits (Anderson).
People who are not themselves farmers usually have at best a dim appreciation of the complexity and difficulty of decision making for farm management and just how much is at stake. If farmers are not doing something which an outsider (such as a scientist) thinks is obviously the right thing to do, it is probably for very good reasons which have not been recognized by the outsider. The reasons may be remote from the issue at hand, such as the availability of labor or an issue in the farmer's private life. This observation about non-farmers applies even to many professionals serving agriculture. Scientists, even those with a close relationship with farmers, are protected from the messiness of farm decision making by their own specialization. Farmers specialize to some extent too, but they can only afford to do so to a small degree. In general they have to be reasonably well in control of all aspects of their farm to stay in business.
Although whole-farm decision making is very complex, it is important for outsiders attempting to serve or influence farmers (e.g. through research or extension) to have some way of analyzing and understanding whole-farm issues. Without this, any advice given may be incompatible with the local farming practices, or may lead to lower, rather than higher, economic returns to farmers(1). One way in which non-farmers can assess the whole-farm implications of any change to the farming system is by the use of whole-farm computer models. These fall into one of two broad categories: simulation and optimization.
Simulation approaches to whole-farm modelling range from very simple to extremely complex. Simple simulation models are common and widely used. Most farm advisors and many farmers build simple whole-farm budgets, which are in essence simple simulation models, with almost all of the detail simplified away. They can be very valuable and revealing, especially in the hands of an experienced farmer or advisor (Malcolm). At the other end of the complexity spectrum, is an integrated system of bio-physical dynamic simulation models (one for each species of plant or animal on the farm), perhaps feeding directly into an economic model.
The second category of whole-farm models is optimization. These have a long but often disappointing history in the area of whole-farm planning for farmers. Optimization modellers working in agriculture seem to have had a dream of whole-farm optimization models serving farmers in a region. Beginning in the 1960s, there have been a number of attempts to implement such systems, but, with very few exceptions (e.g. McCarl et al.), benefits for farmers from these attempts have been very limited. For example Bradford (p. 362) observed that "after models are developed and partially validated, they are rarely applied beyond a few farm firms for a limited timespan". In my view there are several reasons for this:
These problems would also affect any large integrated whole-farm system of simulation models offered as a service to farmers. Nelson notes that, "Comprehensive planning programs are attractive to only a few of the most sophisticated farmers who have academic training in a closely related area and to a few who are temporarily infatuated by reading accounts that gloss over the data requirements of an effective model." (Nelson, p. 368).
Nevertheless there is an audience for which such models are potentially of very great value: outsiders attempting to serve or influence farmers. In Western Australia we have had a notable degree of success in targeting a whole-farm optimization model to this audience. The MIDAS model was originally developed at the Western Australian Department of Agriculture (WADA) and has been used for a wide range of applications (e.g. Abadi Ghadim and Pannell; Abadi Ghadim, Kingwell, and Pannell; Bathgate, Schmidt, and Pannell; Ewing and Pannell; Ewing, Pannell, and James; Morrison and Young; Pannell 1987, 1995; Pannell and Falconer; Schmidt and Pannell). This model is now in use in several institutions in Western Australia and is the inspiration for model developments in several Australian states and in other countries (e.g. Nordblom et al.). An indication of the model's success within Western Australia is provided by the following quote, from a Western Australian scientist in a 1994 meeting.
"What has MIDAS ever really achieved apart from changing the thinking about whole-farm systems within the Department of Agriculture?"
Although intended as a criticism, I take this as praise of the highest order. The fact that such an important achievement could be belittled in this way reflects the extent to which MIDAS and the understanding it generates have become so familiar that they are taken for granted.
The remainder of this article focuses on the MIDAS models and their derivatives. After brief descriptions of the model and its history, the article proceeds through the various uses to which the models have been put, the mode in which they have been used, positive and negative experiences with MIDAS and an outline of current and possible future developments. The objective is to highlight aspects of the MIDAS experience which may be of interest and of help to other modelling groups attempting to develop a whole-farm model of this type.
BRIEF DESCRIPTION OF MIDAS
MIDAS is a whole-farm linear programming model with a joint emphasis on biology and economics. There are several versions for representative farms in different regions of Western Australia, but all include components for crops (cereal and legume), pastures, sheep, feeds (crop residues, grain, pastures), machinery and finance. They are detailed in their representation of soil types and potential enterprise rotations, with different production figures for each phase of each rotation on each soil type.
The models are structured as an equilibrium, meaning that production coefficients represent a time when production has stabilized, possibly after more than one rotational cycle. Dynamics are represented in the sense that production depends on land use and agronomic practices in the previous year, but not in the sense of tracking the move from one equilibrium to another. It is assumed that climatic conditions are "average" every year.
The objective function represents profit maximization. There is no attempt to routinely trade off profit with non-profit goals which farmers may have. The linear programming matrix varies in size between versions. As an example, the eastern wheatbelt version includes approximately 300 constraints, 450 activities and 5000 non-zero coefficients. MIDAS has been described in detail elsewhere (e.g. Morrison et al., Kingwell and Pannell, Pannell and Bathgate).
BRIEF HISTORY OF MIDAS
A brief overview of the history of MIDAS is given in Table 1. The initiative for the model's initial development came from David Morrison (an economist in the head office of the Western Australian Department of Agriculture) and Mike Ewing (a pasture researcher in the Merredin regional office). Early model development was by Ross Kingwell, David Morrison and David Pannell. Throughout its history, primary responsibility for development and management of the MIDAS models has rested with a group of agricultural economists. At the time when MIDAS was established, economists had a very low profile within the Department of Agriculture, with active hostility from some quarters. This affected the MIDAS project in various ways. For example, the Department provided its own priority ratings of all applications for external research funds. The initial application for funding for MIDAS was given the lowest possible rating by the Department, but nevertheless received a high rating from the funding body, the Wheat Research Committee of Western Australia. In the main, however, there was little active antagonism. Most who were not active supporters left us to work, presumably expecting the work to fade with time. There has been a complete transformation in the attitude to economics and economists within most (but certainly not all) of the agricultural research and extension community in Western Australia. Economists are actively involved throughout the agricultural research and extension process to an extent which is remarkable and rare internationally.
Table 1. Milestones in the History of MIDAS Models
|1982||First meetings to discuss possibility of model.||0.5|
|1983||Initial model development.||1.2|
|1984||Model results extended/publicized for the first time.||1.2|
|1985||Develop microcomputer spreadsheets for inputs.||1.5|
|1986||First refereed publication.||2|
|1988||Microcomputers used for model solution.||2|
|1989||MUDAS (version with risk and uncertainty) developed.||3|
|1990||MARG (MP Automatic Run Generator) software developed.||3|
|1991||Serious interest from other states and overseas.||3|
|1992||Started giving general courses on MIDAS/MARG.||3.5|
|1993||Commencement of model development in other states.||4|
|1994||MID (MIDAS Interactive Database) approach developed.||5|
*Estimate of number of staff (in full time equivalents) working on model development and application in Western Australia.
The first version of MIDAS was representative of the Merredin region in the eastern part of mixed crop-livestock producing region of Western Australia (the "wheatbelt"). Due to the regional specificity of model results a number of other versions have been developed. Currently six regional versions are actively maintained, covering most of the wheatbelt of Western Australia. Several other regional versions have been developed but not maintained. This reflects a high demand for model development, but a difficulty in procuring sufficient resources for the time-consuming task of model maintenance.
Risk and uncertainty
From its inception, the most prominent and persistent criticism of MIDAS was that it did not allow for risk. The standard versions are based on the concept of an "expected season" (expected in the statistical sense). This eventually prompted development of MUDAS (Model of an Uncertain Dryland Agricultural System) (Kingwell, Pannell, and Robinson; Kingwell) in the late 1980s.
As well as providing a deeper insight into the farming system, MUDAS has allowed us to test the biases resulting from the simplifications inherent in MIDAS. As a result of this testing we now have more confidence using MIDAS for many analyses in which only a broad picture of the farming system is needed. The gross elements of the farming strategy (i.e. the types of rotations practiced, the levels of different enterprises) are reasonably consistent between MIDAS and MUDAS. Of course the MUDAS model specifies a different farming strategy for each year depending of climatic and economic conditions, but, for example, the expected value of optimal crop area selected by MUDAS is similar to the value selected by MIDAS. Also, evaluations of particular farm management practices are rarely changed by using MUDAS rather than MIDAS (e.g. Kingwell and Schilizzi). This is important because of the greater difficulty of using MUDAS and the existence of only one regional version of MUDAS. There is an ongoing need for careful judgment about whether any given issue can be adequately analyzed using MIDAS or whether MUDAS is needed.
Table 1 includes reference to three developments in the area of software. All three have increased the accessibility and usability of MIDAS. The first was the development of software capable of reliably solving large linear programming models on microcomputers. We use AESOP, a linear version of MINOS but there are now many packages available. The second was development of MARG (Pannell, 1990), a menu driven system for generating series of model solutions and summarizing results into tables. This has made the model much more accessible to non-specialist users and has allowed specialists to much more easily generate large series of runs, saving time and making practical much more detailed analyses of any given issue. The third development, MID (MIDAS Interactive Database), is a system for allowing users to interrogate a large database of MIDAS results addressing a particular issue. This allows rapid response to "what-if" questions from users and has allowed MIDAS results to be easily accessed by users with limited training and limited computer skills.
STRENGTHS AND WEAKNESSES OF MIDAS
The major strengths of MIDAS are its joint emphasis on biology and economics and its ability to address a range of whole-farm issues in a profit-maximizing framework. The issues include:
Additional strengths of MUDAS are:
On the other hand there are some clear weaknesses of MIDAS:
The major weakness of MUDAS is its limited geographic coverage. There is currently only one region with a MUDAS model.
USES OF MIDAS
The area in which MIDAS has had the biggest impact has been in influencing the research activities of the Western Australian Department of Agriculture. At one time in the mid 1980s, virtually every research project being conducted in the Merredin region was directly linked to the MIDAS model, either by filling a data deficiency highlighted during model development or pursuing a direction of research identified by MIDAS as likely to be of benefit to farmers. MIDAS greatly facilitates the most difficult part of economic evaluations of research: the estimation of economic benefits from a given change in the farming system.
As well as influencing which research is conducted, MIDAS has had an impact on the details of how the research is conducted. Scientists who have interacted closely with MIDAS have a better appreciation of the context and relevance of their work and are more likely to measure the variables which matter for the farm-management problem they are addressing. In a sense the models act as a broker for the farming community, asking the scientist the questions which matter most to farmers and asking them in a way which focuses on the most important details for those questions. Influencing the details of the research in this way may be even more important than influencing which areas of research are pursued.
MIDAS has contributed directly to the Department's extension program through provision of quantitative information on profitability of alternative enterprises or management strategies. Unlike some other whole-farm modelling groups (e.g. McCarl et al.) we have never offered a one-to-one service to farmers, but have instead focused on identifying robust messages which can be extended generally to groups of farmers or in the mass media. These tend most often to relate to novel technologies or practices which farmers may be considering or trialing, rather than to the standard year-to-year decisions with which farmers are most adept. Occasionally the models are adapted to represent individual farms, but this is more for the purposes of model testing than for extension.
The range of extension issues which have been analyzed using MIDAS is large. The issues include optimal crop-pasture rotations for particular soil types, profitability of new crop species (such as lupins, peas, chick peas, lentils) the optimal allocation of land between crops and pastures, and the economic value of "farming to soil type" (i.e. managing different parts of the farm differently to suit different soil characteristics).
As well as direct provision of information, MIDAS has also influenced extension indirectly, by improving the understanding of advisors about whole-farm management generally or about the whole-farm implications of a particular issue. Often its role has been in providing advisors with confidence that a particular strategy is or is not worth extending.
Policy measures with complex implications at the farm level have been investigated in order to support design or assessment of particular policies. For example payment of extra premiums for high-protein wheat, and provision of a guaranteed minimum price for wheat in Western Australia have been examined. Although the models have great potential to contribute in this area, we have focused our energies more in the previous two categories due to the greater receptiveness of the relevant decision makers.
Education and Training
The models play a valuable role in educating new extension staff on the nature of the farming system and how its components interact. Understanding which might otherwise take years to gain through accumulated experience can be achieved in weeks. Originally, training was mainly "on-the-job", but more recently we have been conducting formal courses of several days duration. Participants at these courses include researchers, extension staff and economists. MIDAS is also used at various tertiary education institutions, especially the University of Western Australia.
Each model represents a convenient summary of the key features of a farming system. Data from MIDAS are regularly used as indicative regional values in other economic analyses, both by researchers and by regional extension economists.
POSITIVE OUTCOMES OF MIDAS
Especially during the model development phase, MIDAS has provided a focus for bringing together people who otherwise have little interaction. Researchers from different disciplines are brought together, and researchers are brought together with extension agents and farmers. Most people would agree that this sort of thing is good, but without the focus and structure of a modelling effort, it is difficult for meetings between such disparate groups to be productive.
Putting biological information in an economic context
Farmers are not only interested in profit, but if they are going to continue to be farmers, they must give it a high priority. The capacity to estimate the economic significance of any new piece of biological or technical information is clearly of great value to an organization attempting to benefit farmers. MIDAS and/or MUDAS provide this capacity for many types of information.
The biggest single benefit of MIDAS is the way it influences the thinking of researchers and extension agents about their role in influencing farm profit. Sometimes this is a painful process which challenges cherished beliefs and paradigms. Some scientists resist the process so strongly that they remain unconverted, but in most cases, the result of close interaction with MIDAS is positive.
Indirect benefits to farmers from this influence on scientists and advisers are likely to be greater than the direct benefits which would arise if the resources involved were redirected towards direct extension of MIDAS results to farmers. This is because most farmers already have a good appreciation of whole-farm issues, whereas most scientists and many extension agents do not.
Highlighting data deficiencies
Especially during the model construction phase, many areas with little or no useful data are identified. While frustrating for the modellers, this serves a valuable role in directing the attention of researchers to these data gaps. Once a working model has been constructed, the relative importance of the various data gaps can be investigated to help further prioritize them and target attempts to fill the gaps.
NEGATIVE ASPECTS OF THE MIDAS EXPERIENCE
Maintenance and routine use of a MIDAS model is resource hungry. To do the job properly there is a need to dedicate half a person per model per year. This half a person is engaged in updating coefficients, debugging errors and problems as they arise, reviewing data, improving the model structure, and documenting the model.
Unfortunately it is much easier to attract resources for construction of new models than for maintenance of an existing model. As a consequence of this, there may have been too many versions of MIDAS constructed for different regions. It may have been preferable to do fewer things but to do them better.
Dependence on hierarchy
Success in a region is dependent on support from the Officer in Charge. If scientists and advisors are not given a clear message that the model is an important activity for the office, it can be difficult to involve them to the extent needed. There are too many day-to-day priorities to compete with. Unfortunately we have experienced indifference from leaders in a couple of regions. Of course this is better than outright antagonism, but it still means that the modelling effort falls well short of its potential.
Inevitably, some groups remain resistant to the charms of MIDAS. In Western Australia, some private agricultural consultants and some groups of scientists remain aloof. In some cases this can be seen to result from a MIDAS result which is unfavorable to their own interests, but in others, it is due to a more general attitude which is anti-economist, anti-modelling or, in the case of some private consultants, anti-government.
A model like MIDAS, with its reliance on subjective data in a number of important areas, is rather susceptible to malicious criticisms. The modellers must give a high priority to maintaining good relations with client groups. Often this means developing a very thick hide. In situations where one is receiving unfair criticism from someone with whom we would rather be on good terms, one is sometimes forced to swallow one's pride and concede issues which would not need to be conceded if they were to be settled on the merits of the arguments. Being of a temperament which can tolerate the psychological conflict which this entails is valuable as a whole-farm modeler.
Availability of suitable modellers
MIDAS makes extreme demands on its developers. They must have sufficient knowledge of biological and physical science to be able to win the confidence of relevant scientists. They also need computer skills and knowledge of finance/economics and mathematical optimization. A feel for practical farming and farmers is valuable, although this can be improved over time. Finally, interpersonal skills and communication (both verbal and written) are at least as important as any other skills. People with all of these attributes are rare indeed. Our experience to date is that we do need top quality people for the model to succeed. Their limited availability is probably the factor which most constrains development of the models.
Some outsiders to the MIDAS team are concerned about the danger of MIDAS results being misused. Given the degree of uncertainty about MIDAS coefficients and the limited applicability of MIDAS to any specific farm, it certainly would be possible for a too literal interpretation and implementation of the model's results to make a farmer worse off rather than better off. Similarly there is a risk that research managers might use MIDAS results to make far-reaching changes to the priorities for research without an adequate appreciation of the limitations of the model and its data.
However for a number of reasons, the risk of either of these outcomes is low. Primarily, both farm managers and research managers are too sensible to uncritically apply MIDAS results in a prescriptive way. This is reinforced by the natural (and justified) general scepticism about the relevance and reliability of results from a computer model. Furthermore, we mitigate against the risk by the way we use MIDAS and communicate results. As emphasized below, our approach is not to focus on individual results but rather to provide understanding and to put issues in perspective by examining a wide range of plausible scenarios. In this context the risk of misuse of MIDAS is low. In 11 years I have only witnessed two cases of MIDAS results being badly misinterpreted. These were both by advisers, and they appeared to have no negative consequences.
FACTORS CONTRIBUTING TO THE SUCCESS OF MIDAS
Strategies of MIDAS use
There are a number of distinguishing features of the way MIDAS has been used in Western Australia. Part of its impact is attributable to the way in which analyses have been conducted collaboratively, with interaction between different disciplines and between researchers and extension agents. For this reason, the MIDAS team relies just as much on communication and interpersonal skills as on computer skills. Maintaining the good-will and the communication channels is time consuming, but essential. It is surprising how quickly a group's confidence in a particular MIDAS model can diminish if the level of contact is not high enough.
The model is primarily used to put issues in perspective, rather than to provide definitive or precise numerical results. Our philosophy is that our role is to provide inputs to the judgments of decision makers (whether they be farm managers or research managers) rather than to be prescriptive.
Consistent with this view, little emphasis is placed on detailed interpretation of any single model solution. The approach is to conduct extensive sensitivity analyses to investigate how the optimal farm strategy and profitability varies in different plausible scenarios. This approach is partly necessitated by the uncertainty about data and partly by a desire to improve understanding rather than provide a prescription. For any given issue, the number of model solutions generated may be very large (up to 10,000). These can then be interpreted and summarized prior to dissemination or alternatively entered into a "MID" database to allow quick and easy interrogation of the model's behavior in particular scenarios.
Another similar approach is to conduct break-even analysis. For example, we might ask how high does the yield of a new crop species have to be before it would be as profitable as the best existing enterprise on a particular soil type in a given region? This provides scientists with a target and a perspective on the probability that the new species might ever be adopted.
Other elements of MIDAS strategy
Early in the life of MIDAS, a high emphasis was placed on regular production of up-to-date model documentation. Our approach is to make assumptions as public and open to criticism as possible. This strategy creates the risk of providing ammunition for unconstructive criticism, but given the subjective nature of many of the assumptions in MIDAS, there is no sensible (or ethical) alternative to acknowledging the many weaknesses in data in a model of this scope. If somebody does not agree with one or more assumptions, the model is re-run with their preferred values included.
The production of paper documentation has received less emphasis since the development of spreadsheet templates for displaying, storing and changing model assumptions. However, paper documentation is still needed as well to better present an overview of the models' assumptions and record background details about the reasons for assumptions. Some client groups also prefer to work with paper than with a computer.
Development of a reputation and profile has helped the MIDAS project in a number of ways: attraction of funds, support from local scientists, attraction of international and interstate visitors, etc. A key element in developing this profile has been an emphasis on publications of all types. The aggregate number of publications of all types in now over 120.
Appropriate level of detail
MIDAS is unusual in its joint emphasis on economics and biology, and especially on the interactions between enterprises. The level of biological detail is low by the standards of biological models, but high for an economic model. The choice of level of detail has been crucial to the success of MIDAS. There is enough detail to validly analyze a wide range of whole- farm issues, but not so much that we are overwhelmed by data collection and validation.
In part the degree of detail is dictated by the optimizing framework used. This limitation on the level of detail which can practically be represented in an optimization model is a great advantage. It provides a constraint and a discipline which ensures that the model does get finished and applied. It also forces us to identify the key parameters to which the farming system is most sensitive, rather than falling into the "black hole" of detail which can consume some simulation modelling projects.
One of the details which is excluded from the standard MIDAS models is risk. Some consider this to be a critical omission. However there are many issues for which MIDAS provides useful information and understanding without formally representing risk. Partly this is possible because of the way risk is addressed informally using sensitivity analysis. In solving the model for a wide range of plausible scenarios and examining how the farm system changes, it is possible to develop a feel for the importance of uncertainty about particular parameters. This is not as good as formally and explicitly representing parameters as random variables with defined probability distributions, but it achieves many of the same things.
One of the less-obvious advantages of the optimizing approach is discussed above. The more obvious advantage is the ability to quickly and automatically identify optimal responses to changes in the farming system. This is crucial in many cases. For example the introduction of lupins into the farming system in Western Australia led to changes in optimal rotations (including on soils where lupins are not grown), machinery use, livestock feeding strategies, stocking rates, and fertilizer use. In order to correctly estimate the impact of lupins on farm profit, it is essential to identify and quantify all of these changes. This is extremely difficult in a simulation framework.
A supportive institutional environment
WADA is an excellent environment for interdisciplinary work like MIDAS. It is unusual in the extent to which people from different disciplines talk to each-other. We have also been very fortunate in the level of real support given by the hierarchy of WADA. We have seen how important it is to have support from the local officer in charge, as reflected in the different success levels of MIDAS in different regions. In the Merredin office, where support has been active and long-term, the model has had a big impact on the whole office. We also benefited early on from a change in the Directorship of the Department, to one with a greater sympathy to our approach.
After 11 years, it seems that the impact and influence of MIDAS is still increasing. There will be an ongoing role for MIDAS in tackling the issues of the day in agricultural research and extension in Australia. Our level of resourcing has never been higher, but demand for our services has risen rapidly so that we are able to satisfy only a fraction of the requests received. On the national scene, we are increasingly called upon to help other states with model development. Prioritizing our own activities among the competing demands is increasingly difficult.
It is notable that MIDAS has been so successful despite our use of a very old and unfashionable modelling technique. This reflects the power and flexibility of linear programming for this type of modelling, as well as our particular strategies for building and using the models.
Although our approach to the development of MIDAS cannot be used as a prescription for similar models elsewhere, there were a number of principles which should be of general relevance:
I am grateful to Steven Schilizzi, Ross Kingwell and three anonymous referees for helpful comments.
Abadi Ghadim, A.K., and D.J. Pannell. "The Economic Trade-Off Between Pasture Production and Crop Weed Control." Agricultural Systems 36(1989): 1-15.
Abadi Ghadim, A.K., R.S. Kingwell and D.J. Pannell. "An Economic Evaluation of Deep Tillage to Reduce Soil Compaction on Crop-Livestock Farms in Western Australia." Agricultural Systems 37(1991): 291-307.
Anderson, J.R. "One more or less cheer for optimality." Journal of the Australian Institute of Agricultural Science 41(1975): 195-97.
Bathgate, A.D., C. Schmidt, and D.J. Pannell. "Economics of Changing Rotation to Combat Herbicide Resistance." Proceedings, 10th Australian and 14th Asian-Pacific Weed Conference, Brisbane, September 1993, pp. 480-4, 1993.
Bradford, G.L. "Optimizing Models: Some Observations and Opinions on the State of Progress." Modeling Farm Decisions for Policy Analysis. K.H. Baum and L.P. Schertz (eds). Boulder: Westview, 1983, pp. 358-63.
Ewing, M.A., and D.J. Pannell. "Development of Regional Pasture Research Priorities Using Mathematical Programming." In: J.L. Wheeler, C.J. Pearson and G.E. Robards (Eds) Temperate Pastures: Their Production, Use and Management, Australian Wool Corporation/CSIRO, Melbourne, pp. 583-5, 1987.
Ewing, M.A., D.J. Pannell, and P.K. James. "The Profitability of Lupin:Cereal Rotations." In: R.S. Kingwell and D.J. Pannell (Eds). MIDAS, A Bioeconomic Model of a Dryland Farm System, Pudoc, Wageningen, pp. 82-90, 1987.
Kingwell, R.S. "Risk Attitude and Dryland Farm Management." Agricultural Systems 45(1994): 191-202.
Kingwell, R.S., and D.J. Pannell (Eds). MIDAS, A Bioeconomic Model of a Dryland Farm System, Pudoc, Wageningen, 207pp, 1987.
Kingwell, R.S., D.J. Pannell, and S.D. Robinson. "Tactical Responses to Seasonal Conditions in Whole-Farm Planning in Western Australia." Agricultural Economics 8(1993): 211-26.
Kingwell, R.S., and S. Schilizzi. "Dryland Pasture Improvement Given Climatic Risk." Agricultural Systems 45(1994): 175-90.
Malcolm, L.R. "Fifty Years of Farm Management in Australia: Survey and Review." Review of Marketing and Agricultural Economics 58(1990): 24-55.
McCarl, B.A., W.V. Candler, D.H. Doster, and P.R. Robbins. "Experiences with Farmer Oriented Linear Programming for Crop Planning." Canadian Journal of Agricultural Economics 25(1977): 17-30.
Morrison, D.A., R.S. Kingwell, D.J. Pannell, and M.A. Ewing. "A Mathematical Programming Model of a Crop-Livestock Farm System." Agricultural Systems 20(1986): 243-68.
Morrison, D.A., and J. Young. "The Value of Increasing Lambing Percentages." Australian Journal of Agricultural Research 42(1991): 227-41.
Nelson, T.R. "Specification and Use of Farm Growth and Planning Models in Farm Management Extension Activities." Modeling Farm Decisions for Policy Analysis. K.H. Baum and L.P. Schertz (eds). Boulder: Westview, 1983, pp. 367-74.
Nordblom, T., D.J. Pannell, S. Christiansen, N. Nersoyan, and F. Bahhady. "From Weed to Wealth? Prospects for Medic Pastures in Mediterranean Farming Systems of North-West Syria." Agricultural Economics 11(1994): 29-42.
Pannell, D.J. "Crop-Livestock Interactions and Rotation Selection." In: R.S. Kingwell and D.J. Pannell (Eds). MIDAS, A Bioeconomic Model of a Dryland Farm System, Pudoc, Wageningen, 64-73, 1987.
Pannell, D.J. MARG, MP Automatic Run Generator, User Manual, Western Australian Department of Agriculture, Perth, 128 pp., 1990.
Pannell, D.J. (1995). Economic aspects of legume management and legume research in dryland farming systems of southern Australia. Agricultural Systems 49: 217-236.
Pannell, D.J., and A. Bathgate. MIDAS, Model of an Integrated Dryland Agricultural System, Manual and Documentation for the Eastern Wheatbelt Model Version EWM94-1, Department of Agriculture, Perth, Western Australian, 166 pp., 1994.
Pannell, D.J., and D.A. Falconer. "The Relative Contributions to Profit of Fixed and Applied Nitrogen in a Crop-Livestock Farm System." Agricultural Systems 26(1988): 1-17.
Schmidt, C. and Pannell, D.J. "Economic issues in management of herbicide resistant weeds." Review of Marketing and Agricultural Economics 64(1996): 301-308.
1. For example, in the early 1980s, various agricultural consultants were advising farmers in Western Australia to plant all their farm area to crops. This advice was wrong for most farmers since it was based on simple analyses which did not consider whole-farm factors such as resource availability, resource heterogeneity, risk and positive interactions between enterprises (i.e. jointness in production).
2. The MID approach, described above, is a way of allowing a limited interrogation of MIDAS in a simple and usable framework without the need to learn the model in detail.
Citation: Pannell, D.J. (1996). Lessons from a decade of whole-farm modelling in Western Australia. Review of Agricultural Economics 18: 373-383.
David Pannell's Web Page