EMPLOYMENT AND HOUSING STREAMS IN BRITISH INTER‐REGIONAL MIGRATION
| Published date | 01 June 1975 |
| Date | 01 June 1975 |
| DOI | http://doi.org/10.1111/j.1467-9485.1975.tb00056.x |
| Author | Ian Gordon |
Scottish
Journal
of
Politicat
Economy
Vol.
XXII,
No.
2,
June
1975
EMPLOYMENT AND HOUSING STREAMS
MIGRATION
IN BRITISH INTER-REGIONAL
IAN
GORDON
I
Recent studies of regional migration in Great Britain have progressed from
analyses of net or gross rates
of
flow for particular regions to attempts to
explain complete migration flow matrices.
This
approach offers
a
number
of advantages for the tasks of understanding or predicting levels of regional
mobility. Not least
is
the increase in the number of observations available
to the analyst, although it must be recognised that the observations dis-
tinguished (i.e. the cdls of the matrix) will not normally be independent
of
each other. Perhaps the main difference in analysing complete flow matrices,
rather than the marginals alone, is that accessibility must be introduced as
an explanatory variable. The treatment
of
the distance factor has generally
been cursory, however, and the results unsatisfactory in at least one respect.
Quite consistently, models developed to explain flows between standard
regions have tended to under-predict flows between adjacent regions, and
over-predict other flows.
Weeden
(1973)
provides the clearest demonstration of this problem
by
introducing into his regression analyses a dummy variable to represent
contiguity.
This
variable proved highly significant in accounting for varia-
tions in flow levels, reducing the conventional distance measure to
a
very
low level of significance, but given the arbitrary nature of regional boundaries
cannot
be
regarded in its
own
right
as
offering an explanation
of
the pattern.
Weeden’s results might have been explicable in a number of ways, some of
which are peculiar to his rather arbitrary specification of the interaction
model. For example, he makes use of a linear distance function rather than
the multiplicative forms conventionally employed, despite the obvious argu-
ments against the former
on
a
priuri
grounds.
A
contiguity measure would
‘correct’ for some
of
the effects of this mis-specification.
A
more general
problem for analysts
of
inter-regional flows stems
from
the use of distances
between population centroids
of
large regions to approximate the actual
distance travelled by migrants between pairs of regions.
On
average, these
movement lengths will almost always be less than the distance between
centroids and, depending
on
the distribution
of
population, may be very
much shorter.
A
better approximation can
be
obtained on the basis of
simulated
flows
between smaller units,
if
an estimate of the ‘true’ distance
decay function for migrants is available. Re-analysis of Weeden’s data
showed that better results could have been obtained with
a
power function
and a weighted distance measure, with some weakening
of
the contiguity
161
162
IAN
GORDON
effect, but the latter still remained very signif3cant.l Since the tendency to
under-predict flows to adjacent regions was evident in a number of studies
prior to Weeden’s, this result was probably to have been expected.
Another possible explanation, of more substantive interest, is touched
on by both Weeden and Brown
(1972)
in discussing the former’s results.
They note that movements across a single regional boundary may
be
under-
taken for housing reasons, without necessarily involving a change of work-
place, while longer distance moves will entail such a change. The implication
is that, even within the working age group considered by Weeden, there are
at least two distinct classes
of
migrants with quite different patterns
of
move-
ment conflated in the inter-regional figures. The break between the
two
sorts of move is quite sharpindividuals either change their workplace or
not-implying that the two classes of migrants ought to
be
handled separ-
ately in any migration model.
The practical difficulty, however, is the lack
of
any comparable informa-
tion about workplace movements by inter-regional migrants,
so
that the
model itself must be capable of separating out the two streams of migrants.
The remainder of this paper outlines a simple two-stream model and applies
it to the data on migration of working age males between
1961
and
1966
used
by Weeden. The results suggest that there are important inter-dependences
between the
two
streams. The contiguity effect is satisfactorily accounted for
within this framework, but analysis
of
the residuals from the model, in the
last section, points to further aspects
of
inter-regional ‘distance’
still
to be
incorporated.
I1
A
TWO-STREAM
MIGRATION MODEL
In general, gross migration flows between pairs
of
regions may be
represented as functions of three sets of variables relating to the size of
the regions concerned, particular attributes serving as ‘push’ or ‘pull’ factors
and the ‘distance’ separating the regions. We may assume that these
sets
of
variables combine together multiplicatively and that, within the fist set,
the indices of size for the two regions should also be multiplied together.
Within the second set, it
is
conventional to assume again that
a
generalised
‘push’ and ‘pull’ factor respectively, for the two regions should combine
multiplicatively, although
a
prim
an additive relation seems equally plaus-
ible. For the present at least, we shall assume further that population is
the relevant size measure for both regions and that physical distance is an
appropriate measure of separation. With the conventional treatment
of
push and pull factors, we obtain a gravity-type model of inter-regional
migration
:
or
with the additive version
:
Mi,
=
PiPjAiBjf(Dij)
(1:
Mi,
=
lJ,Pj(A,+Bj)f(D,j)
(2:
1
Further details
of
the statistical analysis are reported in a note obtainable from
the
author.
To continue reading
Request your trial