Distributions of Input-Output Coefficients and ProbabilisticInput-Output Models

随着更复杂的建模技术被用来评估更大、更复杂的数据集，对经济系统结构和动态的投入产出分析最近进入了一个新的发展阶段。一条特别有希望的调查路线是发展概率分布方法，为模型提供可操作的解决方案。该项目将继续开发这些新的概率解技术，以审查投入产出系数的区域差异，这些系数反映了特定地区不同部门在特定时间段内交易流量与总产出的比率。这项研究的主旨将是分析美国的全国投入产出数据，以改进解决方法，并估计不同地区和整个国家工业部门的概率密度派生系数。将分析系数的地理差异，以确定跨区域差异的模式和其他系统差异的来源，如工业设施的规模或类型。还将分析区域技术发展差异对经济活动的影响。通过探索对以前未开发的一组国家工业数据进行投入产出分析的新技术，该项目将从理论、方法和经验角度具有价值。与更传统形式的随机模型求解程序相比，基于概率分布的技术的改进将进一步加深对这些类型技术的实用性的一般理解。方法论的发展将为未来所有规模、地点和时间段的数据集工作提供基础，这些测试的结果应该有助于了解当代美国经济不断变化的地区构成。

Input-output analyses of the structure and dynamics of economic systems recently have entered into a new phase of development, as more sophisticated modelling techniques have been used to evaluate larger and more complex data sets. A particularly promising line of inquiry has been the development of probabilistic distribution approaches to provide operational solutions for models. This project will continue development of these new probabilistic solution techniques in order to examine the regional variation in input-output coefficients, which reflect the ratios of transaction flows to total outputs for different sectors in specific locales over given time periods. The main thrust of the research will be analyses of U.S. national input-output data in order to refine solution techniques and to estimate probability density-derived coefficients of industry sectors for different regions and for the nation as a whole. The geographic variation of the coefficients will be analyzed to determine patterns of cross-regional variation and other sources of systematic variation, such as the size or type of industrial establishment. Analyses also will be made of the impacts of differences in regional technological development on economic activity. By exploring new techniques of input-output analyses on a previously untapped set of national industrial data, this project will be valuable from theoretical, methodological, and empirical perspectives. The enhancement of techniques based on probabilistic distributions will further general understandings of the utility of these types of techniques in comparison with more conventional forms of stochastic model solution procedures. Methodological developments will provide a foundation for future work on data sets for all scales, locations, and time periods, and the outcome of these tests should contribute new knowledge about the changing regional composition of the contemporary American economy.