The repeated application of shift-share: a structural explanation of regional growth?

The views put forward by Fothergill and Gudgin that structural factors account for variation in the growth rates of UK regions, and that such structural factors are identified by the repeated application of shift-share are examined. First, the conditions under which the growth rate of a region can be said to be structurally determined are considered. These conditions are not only restrictive, but are also at odds with the expressed views of those who support such a theory. Second, it is shown that, even if the identified conditions hold, the repeated application of the shift-share technique will lead, in general, to an incorrect measure of the importance of these structural factors. An accurate simple procedure is given as an alternative to shift-share in these circumstances. Last, analysis of variance is considered and it is shown that the assumptions required for this technique to be appropriate are essentially the same as those required for the regional growth rate to be structurally determined. Although the use of analysis of variance is not advocated, it is suggested that it is incorrect for shift-share to be portrayed as a superior alternative.