Improving regional I-O models: Evidence against uniform regional purchase coefficients across rows

Most non-survey input-output methods, either implicitly or explicitly, attempt to identify the regional component of the direct requirements coefficient. The major non-survey methods, LQ, supply-demand pool, Stevens' RPC and RAS, define the regional component to be the same across a row: that is, all industries requiring input i will buy the same proportion of it within the region as other industries purchasing input i. In this paper, primary data is used to estimate the regional component of manufactured input purchases for several different industries in each row. This allows the hypothesis of uniformity for several sectors' regional purchases of an input to be evaluated.Specifically primary data about the location of specific input purchases, i.e., regional purchase coefficients, are introduced into an existing Delaware input-output model which was estimated using the supply-demand pool method. The changes are significant and, therefore, the implications of these changes for the understanding and use of models prepared using non-survey methods are discussed.