Regional Purchase Coefficients Estimates from Value-Added Tax Data

Abstract

Regional input-output models are generally developed with non-survey methods that adjust the national table to reflect a region. These adjustments require data on the regional output levels, data on how value-added in a particular industry differs between the region and the nation, and the degree of trade between the region and the rest of the world. While data are readily available on the first item, typically they are not on the other two. Traditional methods of handling the imports are to estimate regional purchase coefficients (RPCs) based on location quotients (LQ), supply-demand pooling (SDP) or econometric models. A number of regional modelers have suggested that RPCs are the weakest link in non-survey I-0 methods. This paper describes a Value-Added Tax (VAT) method of estimating RPCs that is theoretically superior to traditional methods. Then empirical estimates of RPCs are developed using the VAT, LQ, and SDP methods for the North-Trondelag County in Norway. The SDP multiplier impact estimates were often many times that of the VAT estimates. The impacts on specific sectors are overestimated even more. These results suggest the VAT should be used when available for developing RPCs for non-survey models. In the United States and Australia, the only two major countries without VAT, the results suggest that primary data methods should be used for estimating the RPCs in principal sectors.