Positive feedback processes, set in the context of all-industry locati
on factors, provide the theoretical background to the statistical anal
ysis of the shift-share residual component which is used to identify t
he localisation and delocalisation by county of high-technology manufa
cturing in Great Britain. Delocalisation from London appears to be wel
l established, but apart from the Welsh countries, where the stimulus
seems to be largely policy effects, none of the other shift-share resi
duals are significant enough to be distinguishable from random variati
on. However, this is after the elimination of an 'aerospace effect' wh
ich significantly disturbs the otherwise random spatial variation. The
techniques used in the paper separate signal from noise in the residu
al component of the shift-share method, which by itself offers no 'exp
lanation' of the residual.