SHIFT‐SHARE ANALYSIS OF REGIONAL GROWTH AND POLICY: A CRITIQUE*
| Author | J. K. Swales,Alasdair G. M. Nairn,Darryl R. Holden |
| Published date | 01 February 1989 |
| DOI | http://doi.org/10.1111/j.1468-0084.1989.mp51001002.x |
| Date | 01 February 1989 |
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 51, 1(1989)
0305-90'9 $3.00
SHIFT-SHARE ANALYSIS OF REGIONAL
GROWTH AND POLICY: A CIUTIQUE*
Darryl R. Holden, Alasdair G. M. Nairn and J. K. Swales
Shift-share has been extensively used in regional analysis and the evaluation
of the effectiveness of regional policy (Fothergill and Gudgin, 1982; Mackay
and Thompson, 1979; Moore and Rhodes, 1973 and 1974; Moore, Rhodes
and Tyler, 1986; Perloff, Dunn, Lampard and Muth, 1960; Thirlwall, 1967;
Tyler, 1980 ).1 However, its use has always been the cause of some concern.2
In this paper, we argue that shift-share might have some value as an account-
ing procedure, but that its ability to act as a useful standardisation technique
is questionable. Section I briefly describes the shift-share technique. Section
II examines the application of shift-share in the analysis of inter-regional
growth differences. Section III considers the use of shift-share in the
evaluation of regional policy. Section W is a discussion of possible ways
forward if the arguments of the preceding sections were accepted.
I. THE SHIFT-SHARE TECHNIQUE
What follows is a brief description of the most basic, straightforward and
popular method of applying the shift-share technique to the analysis of
regional employment growth. Other variants exist, but the general criticisms
which will be made in this paper apply equally to these other methods
(Bishop and Simpson, 1972; Danson, Lever and Malcolm, 1980; Estriban-
Marquillas, 1972; Stiiweil, 1969).
Imagine a region with an industrial structure represented by the row vector
eÇ with element i (e) the proportion of the region's total employment (Er) jj
industry i. Therefore:
eÇ=E/E' (1)
so that:
(2)
*Holden and Swales are at the University of Strathclyde, Nairn is at Murray Johnstone Ltd.
West Nile Street, Glasgow. They would like to acknowledge the help of two anonymous referees
who made useful comments on a previous draft of this paper.
This list is not exhaustive.
2See particularly the acidic critique of shift-share in Richardson (1978, pp. 202-6) and the
spirited defence by Fothergill and Gudgin (1979).
15
BULLETIN
Employment growth can be represented by a colunm vector gÇ with
element j (g) the regional growth rate of employment in industry j. The
regional employment growth rate, GÇ is the sum of the regional industrial
growth rates weighted by the region's industrial structure:
G=er.gr (3)
Similarly, the national employment growth rate (G") can be expressed as the
sum of the national industry growth rates (g"), weighted by the nation's
industrial structure (e"):
G" "e"g" (4)
Conventional shift-share analysis divides the regional growth rate into
national (N), structural (S) and differential (D) components:
Gr=N+S+D (5)
The national component is simply the national employment growth rate, G"
N=G" (6)
The structural component is measured by first constructing an 'expected'
regional employment growth rate (ar), which is the rate of growth of employ-
ment in the region which would occur if individual industries in the region
grew at the same rate as their national counterparts. This 'expected' growth
rate is the sum of the national industrial growth rates, weighted by the
regional industrial structure:
r(7)
The structural component is then calculated as the difference between the
'expected' regional employment growth rate and the national employment
growth rate. Using (4) and (7):
S = - G" =e'g" e"g" =(er e")g" (8)
The differential component is generally calculated as a residual and is
conventionally interpreted as that part of the regional growth performance
which is attributable to regional specific factors, e.g., regional entrepreneur-
ship and location. As a residual, the differential component can be calculated
using(S), (6) and (8): D=Gr_N_S=GrÔr (9)
The region's differential employment growth rate is, therefore, the difference
between the region's actual growth rate and the 'expected' growth rate, given
its industrial structure. Substituting (3) and (7) into (9) gives:
(10)
where d=g'g"
16
To continue reading
Request your trial