Search Results

Econometrics, as a total subject, is older than macroeconometric model building and deserves a separate historical inquiry. Early investigations of demand–supply functions, income distributions, family budgets, cost, and production functions have been the subject of historical study. The joint founding of the Econometric Society and the Cowles Commission for Research in Economics represented the beginnings of major steps forward from the 1930s. In addition to some of the papers presented at the Cowles Foundation anniversary party, there have been some separate studies by scholars from the group. More than ten years ago, a seminar was organized to compare models to appreciate their differences and to look for commonalities. In a first phase, the Model Comparison Seminar looked at distributions of multipliers across macroeconomic models, and the outcome of common applications of control theory. This chapter reviews briefly the focal points of interest during the historical period of macroeconometric model development.

This chapter focuses on the response of the various econometric models to a number of alternatively specified external shocks. Altogether, 11 model groups supplied simulation results for the model comparisons. These groups represent a cross section of the currently active model operators and forecasters. They run from the now traditional Keynesian to the monetarist and the rational expectations approaches. The participating model groups and the special characteristics of their models are listed in the chapter. Early in the meetings of the model comparison seminar, the philosophy of model comparisons was a topic of some extensive discussion. The approach was to compare alternative “disturbed” model solutions to a base solution. Each model operator was instructed to prepare a so-called “tracking solution” that would approximately reproduce history over the period 1975 to 1984.

Comparisons of the performance of econometric models serve a number of purposes. These goals motivate the comparisons of econometric models of less developed countries (LDCs) as they have previous model comparison projects dealing with the United States and other industrial countries. The most important goal of the round of LDC model performance comparisons presented in Taiwan in May of 1987 was to get a first idea of the general characteristics of these models and a preliminary diagnostic of their principal strengths and weaknesses. A request to produce a standard set of simulations was sent to a dozen model groups in different countries. The simulations were defined carefully, restricting as much as possible the freedom of individual model builders so as to attain maximum comparability of results. Even so, since model structures and procedures differ, it is not certain that all the simulations are fully comparable.

Economic forecasts differ because forecasters use different macroeconomic models. However, even if everyone used the same model, all forecasts would not be identical. Most forecasts reflect a complex interaction among three elements. Unfortunately, little is known about the relative importance of these elements. This chapter addresses three kinds of question. The initial stage of the Model Comparison Seminar's project, starting in early 1986, has been the collection of relevant data, a laborious and time-consuming part of the project. The following results are a preliminary report on an ongoing effort. The conclusions, based on the limited experience so far, must be regarded as highly tentative. Any success that has been achieved should be largely credited to the modelers who participated in this exercise. This chapter compares model solutions based on different sets of conditioning information. In general, a model can be thought of as a conditional statement about the relationship between inputs (Xs) and outputs (Ys), or Y = f(X).

The Model Comparison Seminar has produced some striking findings. The differences in the properties of the major macroeconometric models are much bigger than one might have expected given their emphasized theoretical foundations, or given the similarity of the published economic forecasts. F. Gerard Adams and Lawrence Klein conclude that there is “considerable, one might even say astonishing,” variation in the behavior of models “largely intended for the same purposes.” Roger Brinner and Alert Hirsch refer to these sharp differences as “disturbing, particularly insofar as models share a common theoretical basis.” Stephen McNees shows that the models themselves agree much less on forecasts than do the forecasters: the forecasts of the exogenous policy variables and the ad hoc adjustments serve to make the different models' forecasts much more similar than they would be if the models alone accounted for the differences in forecasts.

Social learning is fundamental to human cultural evolution and an important aspect of social rationality. This chapter examines how advice influences decision making and learning. A brief review of the advice-taking literature shows that people seldom have full insight into the reasons for the usefulness of advice; nevertheless, they often successfully use advice to improve judgments. To investigate the effect of advice on learning from experience, participants of two experiments performed a four-armed bandit task in which they tried to find the best of four choice options. Before the task, they received trustworthy advice about which option is most beneficial. The results revealed a sustained effect of advice, so that the recommended option was preferred over the nonrecommended options, even if the nonrecommended option led to the same average reward. Surprisingly, this effect of advice lasted for more than 100 learning trials. The comparison of social learning models, incorporating different assumptions about the influence of advice on learning, showed that social learning was best explained by the outcome-bonus model. This model assumes that rewards from recommended options are evaluated more favorably than those from nonrecommended options. An additional simulation study revealed the social rationality of this outcome-bonus model, because it accumulated more rewards in the learning task than alternative models. In sum, these results suggest that people combine advice with individual learning in an adaptive manner.

This chapter analyzes stochastic models, demonstrating a powerful technique for model comparison and model calibration. The method uses the mathematical construct known as the Wasserstein pseudometric and is general enough to address many modeling problems of interest, such as model comparison, parameter estimation, and model invalidation. This chapter concentrates on stochastic reaction networks. It shows that the Wasserstein pseudometric method proposed has been developed with an eye toward maximum generality. In principle, it can be employed with any class of stochastic model given enough time and enough sample data.

In this chapter I concentrate on continuous inferential problems: problems where the dependent variable is continuous, such as classical regression problems. As in the previous chapter, using duality theory, I show that the info-metrics framework is general enough to include the class of information-theoretic methods as a special case. The formulation is developed for the classical regression problem, but the results apply to many other problems. A detailed discussion of the benefits and costs of using the info-metrics framework is provided and contrasted with other approaches. I use theoretical examples and policy-relevant applications to demonstrate the method. The common problem of misspecification is also discussed and studied within the info-metrics framework. I show that a misspecified model and a correctly specified one can yield similar answers. The appendices provide detailed discussions of the generalized method of moments and the Bayesian method of moments. Both are connected to info-metrics.

Because habitat is so crucial to the survival and reproduction of individual organisms and persistence of populations, it has long been studied by wildlife ecologists. However, the modern concept of habitat originated with ecologists before the field and practice of wildlife ecology arose. The fields of ecology and wildlife ecology have developed along separate historical paths, but, given that research in each field continues to involve the study of species–habitat relationships, there is common ground for practitioners and students in both fields to better engage with one another. Such collaboration could involve a shared recognition that habitat largely determines a species spatial distribution in nature. Through a behavioral process of dispersal, settlement, and establishment, an individual organism finds appropriate habitat by searching and responding to environmental cues. These cues may primarily be characteristics of the habitat such as vegetation structure. Characterization or statistical analysis of habitat is an obvious and important component of studying the habitat requirements of a species. It is recommended that multiple logistic regression will often be the most appropriate method for characterizing habitat. Of most importance, a habitat analysis should recognize that the habitat of a species involves an integrated set of environmental variables that synergistically influence the survival and reproduction of the individual and existence of the species. The study of habitat can help us learn more about the autecology of the focal species, its role in ecological communities, and proper strategies for its preservation.

This chapter moves beyond viewing nodes as homogeneous dots set on a plane. To introduce more complicated underlying space, multiplex networks (which are defined with layers of interaction on the same underlying node set) and temporal (time-dependent) networks are discussed. It shown that despite the much more complicated underlying space, many of the techniques developed in earlier chapters can be applied. Heterogeneous nodes are introduced as an extension of the stochastic block model for community structure, then extended using methods developed in earlier chapters to more general (continuous) node attributes such as fitness. The chapter closes with a discussion of the intersections and similarities between the many alternative models for capturing topological features that have been presented in the book.