Straightening up shift-share analysis

Abstract

Introduced in 1960, shift-share analysis has been widely applied in regional science with frequent dispute over its efficacy and accuracy. During the 1970s and 1980s, inherent shortcomings of the traditional formulation were criticised by a number of workers while others attempted to circumvent the problems by extending the model. To address links between regional economies and their global counterpart, Markusen et al. (1991) applied the traditional model to incorporate trade via a disaggregation of the national growth and industry mix components. A development of the technique by Noponen et al. (1997) into an import/export disaggregated dynamic shift share model was found deficient in a number of aspects (Dinc and Haynes 1998). Noponen et al. (1998) re-evaluated and corrected their 1997 work. They contrasted their approach with that of Dinc and Haynes, who concluded the debate positively with a rejoinder. Stilwell (1969) provided an apparently definitive traditional account complete with a working tableau and relevant equations. Chalmers (1971) pointed to shortcomings in Stilwell's proportionality modification shift and argued a different angle on industrial composition. Later, Edwards et al. (1978) claimed to correct the 1969 error by reworking Stilwell's nomenclature, re-specifying equations and offering graphical explanations. An appraisal of Edwards et al. reveals remaining flaws, correction of which is the aim of this research note.