Impacts of Aggregation on Relative Performances of Nonsurvey Updating Techniques And Intertemporal Stability of Input–Output Coefficients

In many instances, and for variety of reasons, input–output researchers are compelled to both employ mechanical techniques to update older survey-based tables as well as using more aggregated ones. This combination, however, gives rise to several concerns. The present paper is an attempt to investigate two such questions. First, the effects of aggregation on the accuracy ranking of selected updating methods, and second, the effects of aggregation on intertemporal stability of the input–output coefficients. To probe these issues, three updating methods were selected. These methods are NAÏVE or constant coefficient hypothesis, RAS or biproportional method, and LaGrangian optimization technique. Two survey-based tables from the former Soviet Union along with the selected updating techniques are used to generate updated target year’s direct and inverse transaction matrices at four aggregation levels. Comparison of the resultant estimates at these four levels of aggregation with their counterparts in the actual benchmark table reveals that a higher level of aggregation neither affects the rankings of the updating methods nor does it universally and unequivocally leads to a higher degree of intertemporal stability of input–output coefficients. Copyright Springer Science+Business Media, Inc. 2005