Alternative Methods of Regionalization
| Author | BO Hallin,Ulf Lindström,Lars Ricknell,Sten Berglund |
| Published date | 01 October 1980 |
| Date | 01 October 1980 |
| DOI | http://doi.org/10.1177/019251218000100408 |
| Subject Matter | Articles |
561
ALTERNATIVE
METHODS
OF
REGIONALIZATION
STEN
BERGLUND
BO
HALLIN
ULF
LINDSTRÖM
LARS
RICKNELL
Political
scientists
have
a
long-standing
interest
in
identifying
regions
which
are
defined
in
terms
of
political
stability.
That
is
also
what
we
set
out
to
do,
and
in
the
process
we
get
to
evaluate
two
alternative
methods
of
regionalization
which
are
based
on
different
theoretical
models.
The
two
models
are
geared
to
the
specific
requirements
of
longi-
tudinal
analyses
and
represent
recent
efforts
at
methodological
improvements.
In
both
cases,
the
regionalization
is
performed
on
the
basis
of
the
distances
between
the
geo-
graphical
units
with
respect
to
component
scores
obtained
from
principal
components
analyses
of
Swedish
community-level
data
over
time.
In
the
first
case,
the
principal
components
analysis
is
carried
out
on
the
correlation
matrix
generated
on
party-strength
data
cast
as
percentages
from
the
local
arena
at
different
points
in
time;
in
the
second
case,
the
principal
components
analysis
is
performed
on
the
residual
matrix,
which
is
obtained
by
correlating
the
election-year
specific
error
terms
or
residuals
generated
by
regressing
the
community-level
party-strength
data
on
a
series
of
socioeconomic
indi-
cators,
such
as
class
structure
and
degree
of
industrialization.
1.
SCOPE
There
are
as
many
subdivisions
of
an
area
as
there
are
purposes
that
the
regionalization
is
supposed
to
serve.
Airlines,
breweries,
and
polling
institutes
do
not
necessarily
set
up
regional
offices
in
the
same
locations.
They
operate
on
a
commercial
basis
and
go
wherever
the
market
requires.
Political
scientists,
on
the
other
hand,
have
a
long-standing
interest in
identifying
regions
which
are
defined
in
terms
of
political
stability
(Schwartz,
1974;
Odum,
1964;
Merritt
and
Rokkan,
1966;
Gustafsson,
1974).
That
is
also
what
we
set
out
to
do;
and
in
the
process
we
get
to
evaluate
some
alternative
methods
of
regionalization
which
are
based
on
different
theoretical
models.
562
2.
THE
MODELS
The
most
common
method
of
political
regionalization
requires
the
least
in
terms
of
statistical
computations.
The
partitioning
of
an
area
into
political
regions
is
simply
performed
on
the
basis
of
the
community-
or
constituency-wise
deviations
from
the
parties’
national
averages.
The
underlying
theoretical
models
are
as
straightforward
as
can
be.
Voting
in
general
and
the
con-
comitant
regional
variations
in
party
strength
in
particular
are
approached
in
a
way
reminiscent
of
the
black-box
technique
in
decisional
analyses
(Edwards,
1969).
We
do
not
know
for
sure
what
it
is
that
makes
the
system
work
the
way
it
does;
but
we
want
to
pin
down
the
behavioral
variations
because
they
are
interesting
in
their
own
right
and
because
they
enhance
our
long-
range
predictive
capacity.
It
is
a
matter
of
general
interest
how
the
parties
fare
in
different
parts
of
the
country,
and
familiarity
with
the
party
strongholds
should,
presumably,
be
an
asset
to
modern
survey
research
which
still
draws
on
random
national
samples,
even
though
voting
is
a
local
and
regional
as
well
as
a
national
phenomenon.
The
traditional
methods
of
political
regionalization-devia-
tions
from
the
national
means,
percentile
deviations,
and
so
on-
only
lend
themselves
to
diachronic
studies.
A
party
may
be
said
to
be
firmly
entrenched
in
a
region
to
the
extent
that
it
comes
out
strong
there
in
one
election
after
the
other.
And
a
party
may
be
said
to
fight
an
uphill
battle
in
regions
where
it
consistently
fails
to
reach
its
national
average.
It
is
the
regions
which
do
not
always
remain
within
the
control
of
the
same
party
that
introduce
an
element
of
arbitrariness
into
the
analysis.
Success
or
failure
in
all
elections
but
one,
two,
three,
or
more
may
be
deemed
suffi-
cient
to
count
the
area
among
the
party’s
stable
strongholds
or
weak
spots.
The
cutoff
to
be
applied
depends
to
no
small
degree
on
the
personal
whims
of
the
individual
analysts,
which
high-
lights
the
need
for
methodological
innovations
(Rydenfeldt,
1954).
Two
of
the
methods
to
be
tried
out
at
length
in
this
article
are
geared
to
the
specific
requirements
of
longitudinal
analyses
and
represent
the
recent
efforts
at
methodological
improvements.
To continue reading
Request your trial