A Review of the Literature Supporting the Hybrid Approach to Constructing Regional Input–Output Models

Abstract

Fueled by the high costs of survey-based models and the inherent inaccuracies of non-survey models, a literature has emerged on hybrid regional input–output models. Although this literature has been instructive on the general approach that should be taken in the construction of such models, authors have provided little information about the precise methods they have used to idenfy. the portions of the models for which superior data are worth the effort to obtain. In addition, the role of non-survey techniques used to produce survey-based tables has not been well articulated in the literature. This paper reviews approaches that have been used in input—output analysis that might be adopted for use in constructing hybrid input–output models. It is argued that hybrid model constructors should pursue the most accurate non-survey model of their region as possible (use accurate regional purchase coefficients and minimize aggregation), always seek superior data for households and establishments in resource-based and ‘miscellaneous’ sectors, and sequentially identify. other sectors that should receive superior data. This last task should be performed by examining the sensitivity of sectors to changes in the proportion of their inputs that are imported. Other than intermediate inputs and outputs, labor income and intrasectoral shipments for any specified sector, it is suggested that, for each targeted sector, modelers seek superior data only for a limited number of cells associated with the sector in the direct-coefficients matrix. These cells should be those that are critical to the overall accuracy of the Leontief inverse. Finally, it is suggested that a procedure ought to be used to reconcile the various data sources that will inevitably be employed.