Shift-Share, Simpson’s Paradox and Multifactor Partitioning

One of the most widely adopted methodologies to study regional growth and disparities is shift-share analysis. This paper contrasts traditional shift-share analysis with Multifactor Partitioning (MFP) by comparative analyses and theoretical arguments. It is demonstrated that MFP, which is based on standardized rather than crude rates, overcomes some flaws associated with the use of traditional shift-share analysis by providing an accurate measure of the region effect and by identifying the region-industry interaction effects. Further, MFP allows the introduction of additional factors into the decomposition. These characterising features of MFP could be extended to other modifications of shift-share analysis. Applications to first the UK motor vehicle industry and second the Italian labor force disaggregated by region, industry, gender and age are presented to compare the two methods and to show how the MFP approach can be extended to include more than two factors.