This paper presents a conceptual framework for understanding the influence of alternative paradigms on policy conclusions. Two types of assumptions are associated with mathematical models--meta-assumptions or methodological priors and specification assumptions. Because two different paradigms must assume two different sets of methodological priors, the possibility exists that different problem definitions and hence policy conclusions may emerge from two parallel studies of the same area. In each of two cases presented here, a single problem area has been analyzed with two different methodologies. In each case, different policy conclusions have been reached as a result of the different methodological priors of the two paradigms. The first case involves two models used to analyze changes in retirement policies within the military enlisted system of the United States Armed Services. The second case involves two models used to analyze the determinants of equal educational opportunity in the United States. The dependence of the policy conclusions upon the analytic paradigm employed in a given study has important practical implications for the use of quantitative models in the analysis of social policy situations.
Views of knowledge contain methodological theories--theories of how knowledge progresses-- and epistemological theories-- theories about the nature of knowledge. The former serve four particularly important functions: providing formulas for the generation of knowledge, criteria for the legitimation of knowledge, reasons to suspect other ideas, and rules for the propogation of ideas.
Even the experienced practitioner of system dynamics can encounter serious conceptual problems in getting started on a model, and is tempted to add more and more to his model. A technique â âlist extensionâ â is described which, from the purpose of the project and the importance of feedback loops, guides the evolution of the simplest adequate model. This model is expressed as an influence, or causal loop, diagram. The influence diagram should be tested to ensure that its structure contains the necessary elements of a dynamic model. If it fails the test attention is directed to the area of the system where further elucidation is needed. The techniques have been applied in many practical cases and have been found to give useful results and to increase the efficiency of the modelling process.
The paper describes a system dynamics model of the consumer durables manufacturing industry in the United Kingdom. The model purpose is to analyse the causes and effects of cyclical fluctuations in the industry with a view to encouraging government or operational policies that might improve industry stability. The paper extensively examines the consumer durables industry and explains the model in detail, each equation being accompanied by an account of its construction. The results of the simulation experiments conducted on the model using various test inputs are described. The paper appraises the technique of spectral analysis, which has served as one means of assessing model validity. The model, once validated, should form part of a larger model which will also represent the steel stockholding and steel manufacturing industries. Work on the larger model is in progress.
The purpose of this paper is to convey the techniques and considerations normally involved in formulating and estimating parameters in system dynamics models. Ideally, model equations should be formulated so that the associated parameters each describe some unique observable characteristic of the real system. Thereby, translating observations and measurements below the level of aggregation of model structure (estimation from disaggregate data) into specific parameter values becomes very straightforward. Fewer assumptions about the structure of the system are needed than if the parameters were set by equation estimation or model estimation from data at the level of aggregation of model structure. Making additional assumptions provides more opportunities for systematic errors to creep into the parameter-setting process. Rather than using data at or above the level of aggregation of model structure to set parameters, such information might better be reserved for validity testing. When such data are not already used to set parameter values, the validity tests become simpler and depend upon fewer assumptions.Parameters need only be set accurately enough to allow the model to fulfill its purpose. One time-saving research strategy is to determine, by using only roughly-set parameters at first, how accurately the parameters must be set before investing time and effort in setting them accurately. Then, sensitivity testing can identify the relatively small number of parameters whose values significantly alter the model behavior or response to policy changes. The model can then be reformulated, the policies redesigned, or the sensitive parameters reset by more elaborate and hopefully more accurate techniques.
Delays are a ubiquitous feature of dynamic systems; they are present at every stage of an action. An understanding of delays is necessary if policy makers are to foresee the consequences of their actions. It is often not sufficient to rely on âexpertâ opinion to tell how long it will take for the repercussions of an action to be complete, because even the âexpertsâ can seriously underestimated delay times. It is, therefore, important to have systematic methods of estimating the length of delays in system dynamics models. The time structure of delays is also important.Whether a delay is destabilizing or stabilizing will depend on whether the repercussions are concentrated or dispersed, as well as whether the time lag is long or short. Systematic methods of estimating the orders of delays are, therefore, also useful. This paper presents five statistical methods that can be used to estimate lengths and orders of delays in system dynamics models. The presentation contains a discussion of when each method is applicable and what problems may be encountered in using it. Empirical results from applying two of the methods are discussed. The empirical studies respectively involve the problem of estimating the delay between changes in export prices and changes in export market shares and the problem of estimating the delay between capital appropriations and capital expenditures.The paper also offers guidelines for choosing an estimation technique and discusses validation of the estimates obtained.
Often system dynamics, and particularly the DYNAMO- language, is attacked for not integrating other modelling approaches into the field. This investigation offers alternatives that will hopefully stand up against the critics. The first part of this paper concerns the integration of external functions into system dynamics models. Modifications of the DYNAMO simulation language and of the DYNAMO compiler are explained, and conceptional questions about the integration are discussed. By means of examples of LP programs and statistical methods, the paper shows the philosophical improvements entailed by the system dynamics method.The same criteria are applied in the second part of this paper to the model-method integration of a system dynamics model with an input-output method, considered to be representative of a complete economic structure.The last part of the paper explains the integration of system dynamics model into the higher program structure of an optimizing feedback loop. The best combination of input vector parameters is calculated in the feedback loop at any time so that the output vector follows a predetermined objective function. The overall paper contents demonstrate the flexibility of the system dynamics method.
The opening address at the 1976 International Conference on System Dynamics points out that today's social ills are diffuse difficulties rather than clear-cut problems. Remedial action must start with attempts to clarify the problem, and develop alternative comprehensive strategies that consider a wide segment of society and also the long-term future in an open minded fashion. System Dynamics may serve as a tool for broad policy analysis of this kind.
The principle of conservation states that physical quantities are confined to their own identifiable channels and can enter, circulate within, or depart from a system only by explicit processes. This paper applies the conservation principle to an analysis of the multiplier-accelerator theory of business cycles. Section I describes and critiques a well-known model of the multiplier-accelerator interaction. By ignoring accumulations of inventory and fixed capital investment, the model fails to observe the conservation of important physical flows. Section II proposes a system dynamics model that incorporates the multiplier and accelerator processes within a closed, conserved-flow framework. Section III uses computer simulation to portray the impact of conservation on the multiplier-accelerator interaction. Simulations of the system dynamics model reveal plausible long-term cycles, rather than the short-term fluctuation associated with traditional multiplier-accelerator models. At the end of Section III, the model is modified to account explicitly for labor, as well as capital, in the production process. This revised model produces both short-term and long-term oscillation when submitted to a noise input. The short-term oscillations, averaging about 5 years, reflect the attempt to adjust inventories by varying the labor input to production. The longer fluctuations in capital stock, averaging 19 years, reflect the management of investment in fixed capital. In all of the tests, the incorporation of conserved flows considerably reduces the sensitivity of system behavior to changes in parameter values. The simulations provide theoretical evidence for divorcing short-term business cycles from the interaction of the multiplier and accelerator.
This paper contrast two viewpoints for analyzing the concepts of supply and demand. The first viewpoint, which dominates most economic thinking, treats supply and demand as rates of flow. For example, supply in economic models tends to be measured by a rate of production, while demand is measured by a flow of consumption or purchases. The second viewpoint sees supply and demand primarily as stock variables or integrations. According to this viewpoint, for example, supply would be measure by the available inventory of a commodity while demand would be measured by a backlog of unfilled orders. The central point of the paper is that stock-variable concepts of supply and demand must be incorporated explicitly in economic models in order to capture the full range of disequilibrium behaviour characteristics of real socio-economic systems. More specifically, the paper shows that consideration of stock-variable measures of supply and demand is necessary to describe the price- and quantity-adjustment mechanisms linking supply and demand; to analyze properly the stability characteristics of an economic systems; to analyze short-run and long-run disequilibrium behaviour; and to assess the desirability of economic policies intended to influence such disequilibrium modes behaviour as economic growth and fluctuation.
This paper contrasts two approaches to testing the importance of model variables: single-equation statistical tests and model-behavior tests. The paper demonstrates that, both theoretically and operationally, tests which analyze the impact of individual variables on model behavior are better suited to the task of selecting model variables. Conversely, the statistical tests should not be viewed as tests of model specification per se, but as tests of a particular type of data usefulness. When viewed as tests of data usefulness, the statistical tests have a clear, albeit quite narrow, role in model validation: they warn the modeler when available data do not permit accurate estimation of model parameter. However, as a detailed example illustrates, a model relationship may be difficult to estimate yet extremely important for overall model behavior.
This paper is a summary of the major assumptions underlying the field of computer modeling and the specific assumptions that differentiate four modeling methods used to represent social systems: system dynamics, econometrics, input-output analysis, and optimization. The primary conclusions are: 1. Each modeling method is based on a set of techniques and priors that suit it well to some sorts of policy problems and poorly to others. 2. Misunderstandings between different kinds of modelers and between modelers and clients often arise from failures to recognize these implicit priors and the various strengths and weaknesses of the various modeling schools. 3. Some modeling schools, especially system dynamics and econometrics, are based on such different basic world views and assumptions about the nature of human knowledge that communication from one school to another is almost impossible.
In this report we discuss our possibilities to attain insight about social phenomena. In the first part of the report we argue the nature of social phenomena is different from natural phenomena. Therefore there is a danger connected with the fact that social science for so long time has been dominated by techniques and goals which were successfully developed for the purpose of natural science. In the second part of the report we identify and discuss four essential problems in the study of social phenomena. The problems are: (a) the definition problem, (b) the issue of limitation, (c) the problem of causality and (d) the problem of stability. In the last part of the report we discuss in what way social phenomena can be understood. Six conditions for a successful paradigm in social science are presented and we can conclude that used in a proper way System Dynamics can be one paradigm that fulfill these conditions.
Conclusions derived from world models have little value if they do no include an estimate of the uncertainty associated with the outputs. This paper describes the System Analysis Research Unit World Model and gives an account of the application of Monte Carlo techniques to testing the model. Samples of uncertain data encoded in probability densities are used as input for model runs. The model output is analysed statistically and the contribution to total uncertainty by the variance of the inputs is determined. The output is also to be additive over a limited range. Due to the strong negative feedback loops in the model, the model usually attenuates any variation in inputs. The cost of Monte Carlo methods is justified by the quality of the results obtained.
This paper focuses on the aggregation that is implicit in the use of distributed delays in dynamic models. The aggregation process relates the continuous time-dependent response of a delay structure to the underlying distribution of delay times of the disaggregated events which constitute the delay. The discussion covers in particular the special case of exponential delays used in system dynamics models.
Industrialized societies are presently characterized by rapid change, strong interactions and an abundance of new phenomena. To increase the likelihood of policies having the intended effects, there is a need for policy analysis with a broader perspective and longer time horizon. The main task in such broad policy analysis should be to integrate the vast amount of available information into a useful conceptual structure of the problem area. System Dynamics (SD) –relying heavily on descriptive information for a data base, on a theory of the structure of social systems for theory formation, and on computer simulation for relating structure to behaviour—offers one method of attaining such broad policy analysis. This paper reviews the historical development of the field and examines the major system dynamics literature. The impatient questions of “what is?”, “why does one do?”, “when should one do?”, and “how does one do SD?” are all answered in summary fashion. Within the system dynamics profession, intense conflicts abound as to what constitutes “proper procedure” for the policy analysis process, particularly concerning model conceptualization and testing. Much disagreement arises from implicit differences in modeling objectives. Explicit recognition of objectives and procedures could reduce much of the conflict.
The Scandinavian countries are approaching full utilization of the regrowth in domestic forests, and the forest industry is facing a period of much slower expansion in volume than in the past. Slower growth implies problems for the industry, forestry, and society at large. The “transition” from ample to scarce wood resources could take several forms, depending on actions taken both inside and outside the forest sector. A system dynamics simulation model has been constructed to describe different possible transition paths, and to highlight potential problems. The model purpose is not to predict what will actually happen in the future, but to describe possible futures in an internally consistent way. Such insights about the consequences of various management strategies are useful to interest groups as a basis for discussing how to reach their goals. Within the industry, there is a tendency toward temporary overexpansion of capacity. The forest sector's ability to survive under slow growth conditions could be enhanced by technological and organizational remedies. The necessary remedies will be less traumatic the earlier one accepts and acts upon the problems of finite wood supply.
The process of attaining a useful model embraces the conceptualization, formulation, and testing stages. This paper argues that effective conceptualization can be achieved through a dynamic hypothesis (that is, a chosen time development of interest and hypotheses about the underlying mechanisms). The resulting rough, conceptual model should then be improved gradually through a recursive procedure where the model is tested, redesigned and tested again, in as many ways as possible and as long as is feasible. The paper attempts to structure the hazy topic of model construction by defining a number of terms and presents lists of dysfunctional tendencies in and guidelines for model construction.