Optimal Aggregation of Input-Output Coeffictents - A Decision Theoretic Approach

We present a criterion for the optimal aggregation of industries in input-output analysis, when the aggregation is made prior to data collection. This criterion is based on a model assuming that the input-output coefficients are used in some decision process, and that they are unknown (or random) prior to data collection. We show that our criterion, based on decision theoretic considerations, differs considerably from traditional criteria for good aggregation. Our model is also applied, as an example, to the Netherlands economy.