Modeling for insights,not numbers: the experiences of the energy modeling forum

It has often been contended that the primary goal of policy modeling should be the insights quantitative models can provide, not the precise-looking projections—i.e. numbers—they can produce for any given scenario. Students of the energy policy process, in particular, have noted that preoccupation with the plethora of detailed quantitative results produced by large-scale computer models has substantially impeded their influence on key policy decisions. The creation of the Energy Modeling Forum (EMF) at Stanford University in 1976 represents one potential remedy for that situation. The EMF was formed to foster better communication between the builders and users of energy models in energy planning and policy analysis. The EMF operates through ad hoc working groups, composed of national and, more recently, international energy modeling and policy experts. These working groups conduct studies concentrating on a single energy topic. The diversity of backgrounds of the working group members ensures that the language of the EMF studies is English, not computer. Each working group identifies existing models relevant to the study's focus. A series of tests is then designed by the group to illuminate the models' basic structure and behavior. A comparison of results is published in a widely distributed report that identifies the models' strengths and weaknesses in the context of the study's topic. Seven EMF studies have been initiated to date: (1) Energy and the economy, (2) Coal in transition, (3) Electric load forecasting, (4) Aggregate elasticity of energy demand, (5) US oil and gas supply, (6) World oil and (7) Macroeconomic impacts of energy shocks. Each EMF study has broadened the understanding of the nature of the relevant policy issues and the models that have been, are, or could be used to address them. The present paper describes how each study's key insights were developed in the context of a simplified analytical framework that provided the proper perspective for understanding the model results.     

Limit theorems for a general weighted process under random censoring

Necessary and sufficient conditions for weak and strong convergence are derived for the weighted version of a general process under random censoring. To be more explicit, this means that for this process complete analogues are obtained of the Chibisov-O'Reilly theorem, the Lai-Wellner Glivenko-Cantelli theorem, and the James law of the iterated logarithm for the empirical process. The process contains as special cases the so-called basic martingale, the empirical cumulative hazard process, and the product-limit process. As a tool we derive a Kiefer-process-type approximation of our process, which may be of independent interest.     

Value of a Life: What Difference Does It Make?1

A comparison of the costs and benefits of 57 lifesaving programs reveals striking disparities across agencies and programs in cost/life saved and even greater disparities in cost/life-year saved. Within a broad range the monetary value assigned to the benefits of averting a death usually does not alter the policy implications of the analyses. The findings suggest that despite the substantial disagreements and uncertainties in the theory and practice of valuing lives, careful quantitative analysis can be helpful in setting health, safety, and environmental priorities.     

Elementary models for population growth and distribution analysis

A version of the Lotka-Volterra interaction model is adapted to describe population growth and migration processes in a two-region system. The regions are identified as a metropolis and its non-metropolitan hinterland. Several conditions on growth and migration regimes are imposed. The time behavior of the systems are analyzed, noting especially situations where total depopulation or population explosion eventually occur in one or both populations. Neither growth control nor migration control alone results in a condition of long-run stability in both regions. If at least a momentary condition of zero growth is achieved in both regions, it is possible to maintain finite populations if each population follows a logistic natural growth process and migration flow is proportional to the volume of interaction. It is necessary also that the natural increase limitation is strong relative to migration rates. This result holds even if one population has a net migration advantage over the other.     

Action and institution,network and function: The cybernetic concept of social structure

A program of research on the formal representation and analysis of institutional structures is taken a step further by integrating it with recent developments in the formal representation of hierarchical levels of inclusion or part-whole relations. We begin by reviewing a cybernetic conception of action and show how this relates to the construction of production system models of institutional structures. Thereafter, we treat the inclusion hierarchy to show how the production rule constitutes the conceptual unit integrating social knowledge and social action upon which are built two hierarchies, involving institutional entities and social networks, respectively. We indicate some of the detailed forms of control involved in these hierarchies and then show how a form of functional analysis can be undertaken on this basis. Finally, we provide a lengthy discussion of the promise and problems of this mode of structural analysis.     

Automatic selection of the proper family for simultaneous confidence intervals

Statisticians often employ simultaneous confidence intervals to reduce the likelihood of their drawing false conclusions when they must make a number of comparisons. To do this properly, it is necessary to consider the family of comparisons over which simultaneous confidence must be assured. Sometimes it is not clear what family of comparisons is appropriate. We describe how computer software can monitor the types of contrasts a user examines, and select the smallest family of contrasts that is likely to be of interest. We also describe how to calculate simultaneous confidence intervals for these families using a hybrid of the Bonferroni and Scheffé methods. Our method is especially suitable for problems with discrete and continuous predictors.