Journal of Retailing and Consumer Services 9 (2002) 185199

Central place practice: shopping centre attractiveness measures,

hinterland boundaries and the UK retail hierarchy$
Charles Dennis*, David Marsland, Tony Cockett
School of Business & Management, Brunel University, Borough Road, Isleworth TW7 5DU, UK
Received 22 August 2000; received in revised form 21 February 2001; accepted 28 March 2001

Abstract
Christallers (Central Places in Southern Germany (translated by Baskin C (1966)), Prentice-Hall, Englewood Clis, NJ, 1933)
well-known and much criticised central place theory was based on classical, arguably unsustainable, economic assumptions such as
the uniformity of consumers and travel. Nevertheless, it has been claimed that the emergence of shopping areas in UK towns could
largely be explained in terms of central place principles (Retail Location: A Micro-Scale Perspective, Aldershot, Avebury, 1992).
Brown drew support from the example of the retail hierarchy of Cardi (UK, Store Location and Store Assessment Research,
Chichester, Wiley, 1984): a town centre core radiating progressively further out with greater numbers of district centres,
neighbourhood centres and nally local centres. Christallers theory was based on rigid laws of distribution of central places and
laws of settlement which often determine[d] with astonishing exactness, the location of central places in southern Germany. Guy
considered that for useful application to UK retail, a more exible interpretation was needed and that strict economic assumptions
could be relaxed in a more pragmatic approach. The classical approach fails to account for the positions and hinterland (or
catchment area) boundaries of modern out-of-town regional shopping centres. Except in dening the components of places at
various levels in the hierarchy, Christaller did not even consider the attractiveness of shopping areas in consumer choice. A number
of other authors have investigated various measures to dene positions in the retail hierarchy. In the Cardi example, Guy used
retail sales oor area as a surrogate measure. Systems have been proposed based on numbers and status of retail outlets (The New
Guide to Shopping Centres of Great Britain, Hillier Parker, London, 1991; Shopping Centres, Mintel, London, 1997; J. Property
Res. 9 (1992) 122160; J. Property Res. 9 (1985) 122160). This paper evaluates the authors empirically based measurement system
for attractiveness that can be applied to out-of-town as well as in-town shopping centres. The approach adapts previous simple
systems based on retailer counts. These have been combined in attractiveness measurements applied to denitions of position in the
hierarchy. Results support the prediction of central place hinterland boundaries based on the authors attractiveness measures and
adaptation of (The Law of Gravitation, Knickerbocker Press, New York, 1931) Law. The data t exemplar published empirical
data on shopping centre hinterlands more closely than do the commonly used drive-time isochrones. r 2002 Elsevier Science Ltd.
All rights reserved.
Keywords: Shopping centres; Central place; Hierarchy; Hinterlands; Catchment; Attractiveness

1. Introduction
Planned shopping centres comprise an essential part
of the UK economy, employing over three-quarters of a
million people and playing a key role in the investments
$
An earlier version of this paper was presented to the 7th
International Conference on Recent Advances in Retailing and
Services Science, Eindhoven, EIRASS. (Dennis, C.E., Marsland, D.,
Cockett, W., 2000d. Central place theory revisited: the use of
attractiveness measures in predicting shopping centre hinterland
(catchment area) boundaries).
*Corresponding author. Tel.: +44-(0)208-891-0121; fax: +44(0)208-891-8291.
E-mail address: charles.dennis@brunel.ac.uk (C. Dennis).

of pension funds (OXIRM, 1999; Davies et al., 1993).

In the context of this paper, a shopping centre is dened
as a planned retail development comprising at least
three shops, under one freehold, managed and marketed
as a unit (Guy, 1994a).
Modelling and predicting shopping centre hinterland
boundaries and positions in the retail hierarchy have
implications for developers and planners. Questions
might be explored, for example, as to whether there will
be sucient business to support a new shopping centre
development. Issues include the extent to which existing
facilities will be aected by a particular development
and the number of new customers who might be
attracted if an existing shopping centre was to be

0969-6989/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 6 9 - 6 9 8 9 ( 0 1 ) 0 0 0 2 1 - 2

186

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

extended. Despite research eort, this remains an

inexact science. Some centres have been reported only
50% let 12 months after opening (Kirkup and Raq,
1993, 1994a, b). Theoretical approaches to modelling
shopping centre hinterlands have produced considerable controversy and models in which the t of the
distance component must be adjusted by the Fiddle
Factor (Rogers, 1984, p. 320; drawing on Openshaw,
1973). Complex models are unpopular among
practitioners (Breheny, 1988).
This paper introduces and tests the authors empirically based measurement system for the attractiveness of
shopping centresFin- and out-of-townFwith a view to
modelling and predicting hinterlands and positions in
the hierarchy. The paper proceeds as follows. Firstly, the
implications of central place theory and the retail
hierarchy are briey outlined. The development of the
authors attractiveness model follows and the model is
used, in case study examples, to dene steps in the retail
hierarchy, based on shoppers choice behaviour. Scales
proposed by other researchers are examined and it is
demonstrated that such scales can be modied in the
light of the empirical research. A unied attractiveness
scale is proposed and demonstrated to have utility in
predicting hinterland boundaries, consistent with the
principles of central place theory. Finally, the authors
consider whether, when changes in attractiveness are
considered, central place theory modelling can become
dynamic. In discussing the results, indicators are drawn
for possible changes that might be anticipated in the UK
retail and population structure.

2. Central place theory

Central place theory is based on classical economic
assumptions such as uniformity of consumers and
travel, a theory described by Brown (1992, p. 40) as an
elegant andymuch maligned conceptualisation. Based
on the work of the German geographer, Christaller
(1933) and economist Losch (1940), the theory was,
according to OBrien and Harris (1991), widely
accepted by the planning profession as a model of retail
organisation. OBrien and Harris considered that
central place theory, for example, explains why London
[UK] contain[ed] the major fashion houses and top
stores, and why small places such as Durham [UK] [did]
not have department stores.
The theory has been criticised in recent decades and
dismissed as more elegant than practical by OBrien and
Harris, who drew support from Dawson (1979, p. 190):
Whilst the theory serves to describe and, in part,
explain locational patterns developed prior to the
1960s, it can no longer be used as anyexplanation of
present patterns or planning future.

According to Dawson, central place modelling was

awed because of the complex nature of retailing and
scrambled merchandise mixes. Despite acknowledged
drawbacks, in this paper, the authors revisit and adapt
central place ideas in modelling hinterland boundaries
and positions in the retail hierarchy.
In Christallers terminology, goods were described in
terms of threshold (the population needed to make
supply worthwhile) and range (distance from source of
supply beyond which demand falls to zero). According
to Brown:
Expensive and infrequently purchased [comparison] goodsyhave higher thresholds and ranges
than inexpensive, everyday [convenience] purchasesyThere will be a large number of purveyors of
[convenience] goods and relatively few sellers of
[comparison] merchandise.
By combining the concepts of threshold and range,
hinterlands are dened in which comparison goods are
only supplied from populous central places whilst
convenience goods are sold locally. Economic modelling
based on these assumptions produces maps of catchment hinterlands in terms of interlocking and overlapping areas (Fig. 1).
Brown points out that:
The spatial arrangement of shopping facilities in the
majority of British cities is explicable in terms of
central place principles. Typically, this comprises a
[comparison] retailing core, a series of [convenience]
shopping districts surrounding the core but closer to
it than the edge of the city, and a greater number of
shopping districts in the inner than the outer parts of
the urban area.
Eectively summing up researchers ambivalent approaches to central place theory, Tang and Ingene
(2000) drew attention to essential assumptions that
cannot be sustained in reality, for example the homogeneity of households and purchasing demands. Nevertheless, in calibrating their shopping model (using the
example of Shanghai), they observed that the retail
hierarchy predicted by the theory was actually observed
in practice.

3. The retail hierarchy

The distinction between higher order comparison and
lower order convenience goods has been observed in
empirical data, illustrated for example in Fig. 2.
Christaller (1933, p. 58) observed that:
We always nd great numbers of central places of a
lower order, i.e. lesser importance and smaller size.
Beside them, we nd a considerable number of

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

187

Fig. 1. Central place theoryFillustration of demand areas. (a) Hypothetical demand cones. (b) Hierarchy. Notes: (i) Dierent types and orders of
goods supplied from dierent levels of places will have dierent ranges and thresholds and thus sizes of market areas. A nested hierarchy will be
produced. (ii) This pattern is known as a K 3 hierarchy. The K value is determined by the number of lower order centres served by the next higher
level of place. The numbers of centres follows a geometrical progression, proportional to 1, 3, 9 and so on according to the number of levels. Sources:
(a) Brown (1992); originally sourced from Davies (1976). (b) OBrien and Harris (1991).

central places that have a somewhat greater importance, a still smaller number of places of a higher
order, and only very seldom, places of the highest
orderyThe greater a town is, the smaller is the
numberyin its respective category.
Analogous to Reillys (1929, 1931) approach to
measuring attractiveness, Christaller based his hierarchical system on population numbers. The (then) standard
German classication dened ve categories, from the

county town of 20005000 to the metropolis of

1,000,000+population. Despite having apparently utilised these classications, Christaller claimed to base his
denitions of the hierarchy on the classication of goods
on oer. For any particular good, the outer limit of its
range is dened by the distance that shoppers will travel
to obtain that good. Beyond a certain distance they may
purchase from another place, or may not purchase at all.
The inner limit is determined by the minimum salesyto
make the oering pay. Christaller claimed to classify

188

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

Fig. 2. Travel time decay curves, comparing convenience and shopping (comparison) goods. This (US) example indicated shopping frequencies for
three regional and three sub-regional shopping centres. Symbols represent the number of trips to those centres for customers at various travel times.
Sources: Jones and Simmons (1990, p. 40), based on Young (1975).

towns into categories based on levels of goods, but it is

not clear to the authors how the boundaries between the
categories were set. Nevertheless, in southern Germany,
especially for the plainsywhere there are no natural
barriers, the theory often determine[d] with astonishing
exactness, the locations of central places (pp. 190191).
An extract from Christallers mapping is reproduced in
Fig. 3, indicating striking regularity in the distribution
of places and hierarchy. Christaller, though, tested his
predictions against the distribution of population (and
telephone connections) but did not report doing so
against the availability of the specic goods that are
supposed to form the basis of the hierarchy. Other
authors have, however, reported conrmatory evidence.
Clark (1982) reported the applicability to market centres
in southwest Iowa (USA); once again, an area where the
map can be studied relatively free of the confusing eect
of natural barriers. On the other hand, the authors
question whether the (static basis of) central place
theory (i) explains why the biggest shopping centre in
the world is located in a sparsely populated province of
a sparsely populated country (West Edmonton Mall,
Alberta, CanadaFFinn, 2000); or (ii) accounts for the
shopping centre with the worlds highest number of
visitorsFover 40 millionFbeing located in a state of
only 5 million inhabitants (Mall of America, MOA,
Minnesota, USAFFeinberg et al., 2000). It is the
authors contention that the key to an understanding of
the locations of central places is dynamic. For example,
if entrepreneurs build shopping centres of sucient
attractiveness (which Finn and Feinberg have found to

underlie the successes of West Edmonton and MOA,

respectively), shoppers will come despite travelling long
distances.
Rather than dening hierarchies based on nominal
measures such as population or type of goods, a more
rigorous approach is based on attractiveness. Fig. 4
illustrates, for example, the retail hierarchy of Cardi,
UK, based on the use of m2 selling area as the proxy
measure. From the town centre core the hierarchy
radiates progressively further out with greater numbers
of district centres, neighbourhood centres and nally
local centres. An analogous pattern can be demonstrated to surround even a regional shopping centre such
as the Harlequin centre (Watford, UK), where local
centres survive and even thrive within a kilometre of the
shopping centre. Lakeside (Thurrock, UK) has another
shopping centre (in Romford) within the inner band of
its catchment area (20 km/15 min drive). In order to be
viable, those centres lower in the hierarchy need to
include a retailer mix appropriate to their position; in
Romford, satisfying the shoppers who spend less per
visit but come three times a week (the owners, quoted
in Retail Week, 6 June 2000).
Cardi comprised the basis of a long-term investigation into retail change. Guy (1999, p. 458, referring to
his 1996 study) observed that the development of a few
large, new food stores coincided with the closure of
several smalleryshops. Guy described these changes as
the outcome of a general process of concentration,
disproving any static nature of the retail hierarchy. This
point was further illustrated in a report of the impact of

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

189

Fig. 3. The distribution of towns as central places in southern Germany. Source: Adapted from Christaller, 1933, p. 225.

the Merry Hill regional shopping centre (UK) on the

adjacent town of Dudley. On completion of the
shopping centre in 1989, a number of major retailers
closed their stores and in eect moved them to Merry
Hill. Many other shops were closed and the premises
were reoccupied by low quality and discount stores
(Guy, 1999).
In the UK, retail areas have grown in a haphazard
manner, from origins as markets in the centres of towns
and in suburban areas through conversions of other
types of property (Guy, 1998). Patterns are not as
regular as in, for example, southern Germany or
southwest Iowa. Guy criticised central place theory as
relying on the notion that consumers would tend to buy
the goods required at the nearest available location, an
assumption:

Clearly incorrect for many shopping trips since other

determinants of shopping success are also importantyConsumer choice of shopping destination
reects several qualities, such as variety of goods,
price of goods, cleanliness, spaciousness and security
of the centre, and quality and quantity of car parking,
for example. Central place theory, which relies mainly
on distance as a choice criterion, becomes inadequate
in this situation.
Guy, though, goes on to review dimensions of retail
classication systems including goods sold, trip purpose,
size of store, store ownership, catchment area, physical
form, function, location, development history and type.
Guy pointed out that shoppers make their choices of
shopping centre destination from an evoked set with

190

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

boundaries, comparing with data not gathered by the

authors and not used in developing the model.

4. Propositions to be investigated
Based on the preceding discussion, two propositions
on the development of central places and on consumers
shopping centre choice behaviour can be derived.
(P1) Population and retail provision tend to cluster
around central places dened on a matrix.
(P2) Hinterlands and the retail hierarchy follow the
attractiveness of shopping and town centres.

Fig. 4. The urban retail hierarchy of Cardi (UK). Source: Brown

(1992); originally sourced from Guy (1984).

respect to a particular combination of location and trip

purpose, suggesting a hierarchical classication based
on trip purpose and size.
An example is the system used in the UK which sums
the numbers of non-food shops in a centre, ranking
centres in order of attractiveness (Reynolds and Schiller,
1992; Schiller and Jarrett, 1985). Reynolds and Schiller
examined the cumulative frequency of scores, tting the
divisions in the hierarchy to the gaps between size
clusters where possible. The hierarchy classications
illustrate the increases in numbers of centres at the lower
levels but cannot be observed to follow the strict
geometrical progression of central place theory. The
six levels were:

National
Metropolitan
Major regional
Minor regional
Major district
Minor district

Multiple branch
(Goad) score
186
93132
3574
2534
624
25

Number of
centres
1
6
99
60
290
370

In this paper, the authors have attempted a preliminary exploration of the use of retail attractiveness
measures in dening positions in the hierarchies and
hinterland boundaries for shopping centres and towns.
A description of the empirical measurement system for
shopping centre attractiveness developed by the authors
is as follows. Empirical measurements have been
compared with retailer count systems and a unied
scale has been proposed. In a further step, the
attractiveness scale has been utilised in dening positions in the retail hierarchy and predicting hinterland

P1 represents Christallers classical approach,

whereas P2 follows the more recent classication
systems of (e.g.) Guy; Reynolds and Schiller. P1 and
P2 are not necessarily mutually exclusive, but if P2 is
accepted, a trend towards the redistribution of population around attractive shopping destinations might be
expected. Accordingly, this study was designed to test,
on an exploratory basis, the extent to which P1 and P2
t available exemplar UK data.

5. Attractiveness
The use of attractiveness measurements in the context
of consumers choices of shopping centres has been
reported earlier (Dennis et al., 1999a). The study was
based on a questionnaire survey of 287 respondents at
six UK shopping centres varying in size from small intown sub-regional to large regional out-of-town centres.
A regional centre has a gross retail area of greater than
50,000 m2 and a sub-regional one 20,00050,000 m2
(based on Guy, 1994a, b; Marjanen, 1993; Reynolds,
1993). The study evaluated shoppers comparative
ratings of two shopping centres, one of them being the
centre where the interview took place. The alternative
centre was the one where they shopped most (or next
most after the interview centre) for non-food shopping.
The questionnaire instrument was based on the
attributes of image elements employed by McGoldrick
and Thompson (1992a, b) together with additional
constructs derived from analyses of preliminary unstructured interviews. Respondents stated their perceptions of the importance of each of 38 attributes
(including those identied by Guy as guring in
consumers choices of shopping destination, for example, quality of stores, cleanliness and availability of
toilets, following Hackett and Foxall, 1994). Each
attribute was also rated for both the centre studied
and the alternative centre. Respondents estimated
perceived travel distance and time to both centres and
supplied details such as age, location of residence and

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

occupation of the main earner in the household.

Examination of the characteristics of the sample
indicated the distribution of socio-economic groups,
age and sex reasonably representative of that anticipated
at UK shopping centres.
Further questions concerned typical perceived
monthly spend at each of the two centres. As
McGoldrick and Thompson pointed out, much of the
variation in shoppers expenditure relates to factors such
as income or socio-economic groups, rather than travel
distance or attributes of the shopping centre. Following
this approach, the main dependent variable used in this
study was the individual relative spend. A value of 100
indicated all expenditure at the centre studied none at
the alternative centre. A value of 50 indicated half of the
expenditure at each centre. The same approach was used
to scale perceived travel distance and time producing the
variables individual relative travel distance and time.
The view of attractiveness taken by the authors is
that any product (such as a shopping centre) can be
seen as a bundle of expected costs and rewards which
East (1997, p. 131) found was upheld by research. East
drew support from Westbrooks (1980) nding that an
overall measure of retail satisfaction correlated well with
a simple addition of the satisfactions and dissatisfactions
that customers experienced. In the authors methodology, the measures of satisfaction and dissatisfaction
have been taken from the respondents ratings of the
shopping centre compared to their main alternative
centre (on 5-point semantic dierential type scales).
These satisfactions for the individual attributes were
weighted, rstly by the importance of the attribute to
the respondent (also on a 5-point scale) and secondly by
the degree of association with the stated relative spend.
Once weighted, satisfactions were added, giving an
overall attractiveness measured value. The authors
combined the attractiveness measurements with the
relative travel time or distance variables, to derive
(statistically signicant) models of individuals relative
spend. More detailed derivations of the attributes and
models have been reported elsewhere (Dennis et al.,
1999a, b, 2000a).

6. Steps in the retail hierarchy

The positions of steps in the retail hierarchy of
shopping centres were investigated by rst dening what
constituted a substantial competitor in the empirical
survey. In this context, a substantial competitor was
dened as another shopping centre used by over 25% of
shoppers at the centre studied. To qualify as substantial,
a competitor must be attractive relative to the centre
studied. For example, considering centres featuring in
this empirical work, the Blue Rose Centre (not its real
name) has a competitor of 55% of the physical size

191

(m2 gross selling area) within its catchment. This

competitor was used as a main or alternative centre by
only 2% of the Blue Rose Centres customers and thus
cannot be considered as a substantial competitor.
Conversely, The White Water Centres substantial
competitor is 95% of the size of the White Water
Centre and used as a main or alternative centre by 40%
of the White Water Centres customers.
Using selling area as a proxy indicator of attractiveness, the dening point for a substantial competitor
(used by over 25% of a shopping centres customers) lies
somewhere in between above 55% and below 95% of
the size of the centre studied. A mid value of 75% has,
thus been adopted as the dening level of a substantial
competitor, based not on selling area but rather on the
more precise unied attractiveness scale reported below.
For a competitor to be counted as at the equivalent level
in the hierarchy to the centre studied, the competitor
must have an attractiveness of at least 75% of that of the
centre studied. It follows, therefore, that when two
shopping centres are compared, if the smaller one has an
attractiveness of less than 75% of the larger one, the
smaller is at a lower level in the hierarchy. Although
studies of many more centres are needed to conrm this
arbitrary-seeming denition, it is consistent with the
results of this study and those of other authors analysed
below. That is, a centre at a lower level can lie within the
catchment of a larger centre, whilst there will be a break
point in catchment boundary between two centres at
the same level.

7. Unied attractiveness scale

Mintel (1997) reported an attractiveness measurement
scale for shopping centres and towns, based on numbers
of retail outlets, scoring multiples higher than others and
certain specic named retailers higher still. In the light of
many considerations that are known to inform
shoppers choices of shopping destinations, measuring
shopping centre attractiveness, in such a manner,
initially appears trivial. Nevertheless, evidence in support can be inferred from at least two independent
sources. Firstly, Finn and Louviere (1996) investigated
all 17 regional and community shopping centres in
Edmonton, Canada. The t of their regression models
was highly signicant, with between 70% and 90% of
image item variance accounted for by the store tenant
variables such as the presence of particular major and
discount department stores. Other characteristics had
some signicant eects, but the additional eect on
image was generally rather small. It was concluded that
specic anchor stores had a substantial impact on
consumers images of shopping centres, accounting for
most of the variation in centre patronage. Secondly,
Feinberg and associates (2000) investigated the

192

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

prediction of (US) mall patronage from attraction scales

using stepwise logit regression. In every case, the rating
of specic stores that most attracted customers was
demonstrated to be the most signicant variable.
Findings such as these, are understandable when it is
considered that the most attractive shopping centres will
be expected to attract the most successful and popular
retailers.
In this study, the Mintel scale correlated well with
the empirical measurements from the questionnaire
surveys, R2 0:94 for Eq. (1):
A 0:978B;

where A is the Mintel score and B the empirical Brunel

Index for attractiveness (from the survey-measured
weighted attractiveness summations, rescaled to use an
equivalent numerical scale to the Mintel scoreFFig. 5).
Corrections must be made, though, for the dierences

Fig. 5. Mintel scores, A; of shopping centres vs. the Brunel

measured attractiveness index, B; of the centres on the Mintel
scaleFcorrected for towns (intercept set to 0).

between planned, covered shopping centre attractiveness

and the unplanned town centres, and in addition for the
eect that the surrounding town area has on shoppers
perceptions of the attractiveness of the centre. The
attractiveness measure was signicantly associated with
shopping centre sales turnover and rental income
(R2 =0.99 and 0.98, respectively, reported more fully
elsewhere: Dennis et al., 2000b).
To estimate and plot hinterland boundaries, maps
indicating the locations of towns and shopping centres
along with their Mintel scores would be useful. Such
maps can be produced to order by data owner(s). The
Hillier Parkers New Guide to shopping centres of
Great Britain (Goad Plans/OXIRM, 1991, the Goad
Plan) is useful in this connection. The Goad Plan is
based on the numbers of branches of 113 specic
multiple retailers (Goad scoresFbased on one of the
examples referred to by Guy and designed to reect the
presence of multiple comparison goods retailersFReynolds and Schiller, 1992). An extract from the Goad
Plan is reproduced in Fig. 6. The area selected deliberately typies a part of middle England relatively free
from natural barriers and having in consequence some
regularity in the distribution of towns. Below those
towns rated in the Goad hierarchy, there is a level of
smaller places, such that, in the Middle Ages, every
habitation was within a one-day return walk of a
market.
The authors have determined the Goad scores to be
valid measures of attractiveness for towns, correlating
with scales proposed by other research organisations
(such as Management Horizons, 1995: for 15 towns also
rated on the Goad Plan, R2 0:88; signicant as
p 0:01). The Mintel scores, though, are valid for both
towns and shopping centres. To facilitate data handling,
the Goad scores have been rescaled to the Mintel
equivalents.

Fig. 6. Towns and shopping centres in central England. Source: Extracted and adapted from Goad Plans/OXIRM, 1991.

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

Based on the best t of available data, it has been

estimated that, for plotting hinterland boundaries, the
attractiveness of towns must be scaled down by a xed
percentage compared to shopping centres. Validating
the combined scale as linear, estimates of sales values for
town and shopping centre combinations together with
out-of-town shopping centres correlated well with the
combined attractiveness scale (coecient of determination, R2 0:92; signicant at p 0:01). It must be
stressed that the authors new empirical attractiveness
measures should be used whenever practicable. Nevertheless, the scale based on Goad and Mintel values,
adjusted for town centres vs. shopping centres, can be
utilised as alternative where the empirical measures are
not available.

8. Distance exponent
The principle of the gravitational theory of retail is
that shoppers are more likely to shop in a more
attractive town or shopping centre, but the attractiveness decreases with distance. Spend can be considered as
being positively related to some measure(s) of attractiveness and negatively related to some measures(s) of
unattractiveness or deterrence, such as distance. Reilly
(1929, 1931) used analogy with gravitation as the basis
of his law which forms the bases of many approaches
to retail location. The frequency with which residents
trade with a town is postulated to be directly proportional to the attractiveness and inversely proportional to
some power of the distance that they travel. This inverse
eect of distance, though, need not be limited to its
square as in Reillys gravitational analogy. Rephrasing,
the distance exponent is not necessarily numerically
equal to 2. Based on the questionnaire survey outlined
above, the authors have previously reported on variations in the distance exponent (Dennis et al., 1999b,
2000a). The distance exponent was demonstrated to be a
variable related to attractiveness. The authors contend
that an estimation of the distance exponent, d; can be
predicted from the Brunel Index of attractiveness, B;
using a relationship derived from a linear regression
(SPSS):
d 0:0044B 3:19:

Fig. 7. Distance exponent, d; on visit rate per 1000 residents vs.

Brunel measured attractiveness index of shopping centre plus
correction for town centre (where relevant) on the Mintel scale, B:

to incorporate varying distance exponents (Dennis et al.,

2000c). Martin (1982) demonstrated that Reillys law
break points could be used to dene a notional
catchment area boundary, by calculating (and joining)
multiple break points with surrounding towns. Martin
compared the calculated break points around Northampton (England) and surrounding towns with a survey
data contour (from 1350 respondents) enclosing the area
from which 90% of Northamptons shoppers travelled.
Martins calculated break points for Northampton
produced a hinterland boundary in reasonable agreement with survey results, illustrated in Fig. 8. This
provided a quick quantitative method of estimating the
catchment area, but with two disadvantages. Firstly,
there was only a single measure of attractivenessFpopulation. Population must be an inappropriate measure
for the attractiveness of the new town shopping centre of
Milton Keynes, which acts in eect in a similar way to
an out-of-town shopping centre for a large hinterland.
Secondly, the distance exponent was treated as a
constant although it has been demonstrated to be a
variable. The authors have reworked the break point
calculations using unied attractiveness scores together
with estimated distance exponents (derived from Fig. 7).
An intuitive restatement of the Reilly/Converse break
point prediction of the position of the break point
between towns a and b would be

The empirical data comprising the basis of the

relationship are illustrated in Fig. 7.

9. The hinterland boundary

Reillys law has been restated (Converse, 1949) to
estimate the break point between two towns which
separates the areas of dominance of each. As reported
elsewhere, the break point calculation can be modied

193

distance from a

distance from a to b
p
p;
1 db  Ab =da  Aa

where da and db are the distance exponents for towns a

and b, respectively, Aa and Ab their attractiveness
scores. Eq. (3) is, though, only a rst approximation of a
relationship to which there is no straightforward
arithmetic solution. The full derivation has been
reported elsewhere (Dennis et al., 2000c) and space
limitations preclude inclusion here. In brief, the break
point derived from Eq. (2) must be multiplied by the

194

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

Fig. 8. Catchment area of Northampton, comparing the theoretical assessment from the Reilly/Converse law with the survey result. Source:
Adapted from Martin (1982, p. 71).

correction factor
r
Fa  Fb
;
d 1
2F
where the values of Fa and Fb ; respectively, are
Aa
Ddaa

and

Ab
Ddba

Da and Db are the distances of the calculated break

point (Eq. (3)) from a and b, respectively, d the higher of
the distance exponents and F the lower of the Fa and Fb
function values. In the authors experience, this simplied procedure has always estimated break points within
0.2 km of the theoretically optimum solution.
Martin dened the catchment area in what may seem
to be an arbitrary mannerFthe boundary which
enclosed 90% of the shoppers, pointing out that any
attempt to identify more than around 90% of catchment
area results in absurdities such as claiming that parts
ofyAmerica should be included. The authors have
investigated the use of values other than 90%. No other
value was more useful in modelling shopper spend. The
value of 90% has therefore been retained as a standard
(R2 value in predicting the eect of distance on
individual spend 0.85 when using the 90% value for

dening catchment, compared to 0.62 and 0.57 for 85%

and 95%, respectively).
By substituting attractiveness for population, and
using distance exponent values estimated from attractiveness, the break point method has been adapted to
help predict catchment area for an out-of-town shopping centre as well as for a town. The authors have
tested predictions against data not gathered by them and
not used in developing the models. There is a scarcity of
such data published in the public domain, but two
surveys mapping customer catchments have been used, a
town (Northampton, UKFMartin, 1982Fthe example
referred to above) and an out-of-town shopping centre
(Meadowhall, UKFHoward, 1993). Predicted hinterland boundaries have been compared with survey results
for these two areas, illustrated in Figs. 810.
For out-of-town regional shopping centres, break
point calculations in the main are less applicable than
for towns as such centres are usually not surrounded by
competitors attractive enough to gure at an equivalent
level in the retail hierarchy. Fig. 10 is an example of
catchment data for Meadowhall out-of-town shopping
centre, near Sheeld, England using data from Howards (1993) survey of 1244 respondents.
The authors break point calculations have been
based on estimated drive times from Automobile

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

195

Fig. 9. Catchment area of Northampton, comparing the theoretical assessment from the Reilly/Converse law vs. the survey result, with a plot based
on Brunel Index superimposed by the author (based on drive times and predicted distance exponents) Sources: The authors and adapted from Martin
(1982, p. 71).

Fig. 10. Catchment area of Meadowhall, comparing the theoretical assessment from predicted Brunel attractiveness, distance exponents and
population, plus the break points with Manchester (based on drive times) vs. the survey result. Sources: The author, adapted from Howard (1993,
p. 101); Census (1991) and Routemaster (B1998).

196

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

Association data (AA: Routemaster programme,B1998),

plotted via the major roads. For Northampton,
illustrated in Fig. 9, the radii of the predictions in the
directions of towns in the vicinity of the hinterland
boundary t the survey result well, with a Hoel and
Jesson index of t=0.82 (equivalent to the R2 valueF
Hoel and Jesson, 1982).
In directions other than Manchester, the eect
of competitors is not relevant and predictions
were made using standard calculations based on
population, drive time (or distance) and estimated
distance exponents only. Populations (residents in
households from Census, 1991) in radial segments in
the direction of the main population centres have been
analysed in conjunction with the (Routemaster, B1998)
drive times together with the estimated distance
exponent. Fig. 10 illustrates the catchment boundary
calculated on this basis (for directions other than
Manchester) compared with the boundary enclosing
90% of respondents (estimated by the authors from
Howards respondent spots).
The predicted catchment for Meadowhall is less
extensive than that observed in the directions of
Bradford, Leeds, Nottingham and Derby but at York,
Scunthorpe and Lincoln is close to the survey result. In
the direction of Manchester, the catchment area
boundary has been calculated using break point
calculationsFand the t is considerably closer to the
survey result than is the standard calculation.
Practitioners tend to favour the estimation of catchment area boundaries from drive time isochrones. For
example, the authors have calculated that the isochrone
that enclosed 90% of respondents in the Meadowhall
survey was at the 50 min drive time. This isochrone,
though, extended into the city of Manchester, considerably beyond the observed catchment. The break point
method predicts the catchment boundary in the direction of Manchester more closely than does either the
standard calculation or the isochrone. The radii of the
predictions using the break point method again tted
the survey result well. When combined with the standard
distance exponent approach for other radii directions,
catchment area was predicted with a Hoel and Jesson
index of t=0.77.

10. Discussion and conclusions

In this paper, the authors have eschewed the rigid
laws of settlement which, according to Christaller,
determined the location of central places. Rather, the
dynamic nature of shopper behaviour has been demonstrated with shoppers following the provision of
attractive shopping areas. Decisions of developers and
planners can lead to the building of large out-of-town
shopping centres that redene the retail hierarchy. As

briey alluded to above, Guy (1996, 1998, 1999) has

studied many examples, and concluded that a more
exible interpretation of central place theory is needed
for useful application to UK retail. The authors share
the view that strict economic assumptions can be relaxed
in a more pragmatic approach.
On the one hand, the distribution of market towns
in middle England, illustrated in Martins study of
shopping patterns around Northampton, supported
proposition P1. In that approach, population appeared
clustered around central places dened on a
matrix (classical central place theory), with hinterland
break points being dened by population (Reilly/
Converse). The authors, though, have demonstrated
an alternative scenario. In line with proposition P2,
hinterland catchment areas and the retail hierarchy
followed the attractiveness of shopping and town
centres, which in turn was closely related to the
provision of shops that shoppers considered attractive.
This proposition was demonstrated to hold not only for
traditional towns and shopping, but also for new out-oftown shopping centresFthus, extrapolating beyond the
scope of P1.
The dynamic nature of shopping behaviour following
the provision of shops can be further illustrated by
comparing the famous, traditional English University
cities of Oxford and Cambridge. These are of similar
population size (and socio-demographic prole), but
Cambridge supports 15 more of the specied multiple
comparison goods retailers included in the Goad score
(27% higher attractiveness score). Reynolds and Schiller
(1992) suggested this to be due to Cambridges expansion in catchment draw resulting from the opening of
the Grafton Centre on the edge of the central area.
The results from this study indicate a correlation
between attractiveness score and sales turnover
(R2 0:99; as mentioned above and reported elsewhere:
Dennis et al., 2000b). By modelling sales turnover on
this basis, it has been deduced that Cambridges
consumer spending is around a quarter more than that
of Oxford, directly resulting from the building of the
new shopping centre.
In the examples of Northampton and Meadowhall, a
t has been observed between the predicted break points
and the boundary containing 90% of shoppers. From
this, it can be concluded that there is an overlap of
catchment area boundariesFclose to 10%. Shoppers do
not necessarily shop at the nearest place that satises
their requirements for specic goods. Rather, they take
into account many aspects of shopping destinations and,
when considered on aggregate, distribute their (nonfood) shopping expenditure according to their assessments of the attractivenesses of the destinations. Despite
the many other considerations in shoppers decisions of
where to shop, there nevertheless appears to be a close
association between assessments of attractiveness and

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

the numbers of shops, particularly those shops especially

desired by the consumers. This is understandable on the
basis that the most attractive shopping centres will
attract the most successful retailers. Centres underperforming on attractiveness lack the big name and
designer stores.
The authors results have justied one aspect of
classical central place theoryFthe nesting hierarchy of
hinterlands. For example, from Fig. 9, the town of
Market Harborough can be seen to lie within the
hinterland of Northampton. On the other hand, there is
a break point boundary between Milton Keynes and
Northampton; although Milton Keynes is closer to
Northampton than is Market Harborough. Milton
Keynes (and the other towns used in the break point
assessments) is on the equivalent hierarchy level
to Northampton, based on the rule of at least 75% of
the attractiveness score of Northampton. Market
Harborough, though, in common with a number of
smaller towns not used in the break points assessment,
is below the 75% measure. The authors data (ongoing
study, to be reported later) indicate that the
typical resident of Market Harborough has some
comparison goods expenditure in that market town,
but substantially more in Northampton and/or Milton
Keynes. In the examples of both Northampton and
Meadowhall, the dening level of 75% of attractiveness
appears to have been eective in dening the level in the
hierarchy. Towns below 75% attractiveness do not
appear to aect the catchment area boundaries of the
larger centres.
The authors contend that classical central place
theory can be modied in the light of proposition P2,
taking as a basis for the provision of retail oerings,
which can be modied by planners and developers.
Thus, it can be forecast that movements of population
towards the new central places will follow developments of regional shopping centres. This eect has been
observed in the USA from as long ago as 1976 (Gosling
and Maitland, 1976; Lion, 1976). Young (1985) documented the US trend for residential and oce development around suburban growth poles centred on
regional, out-of-town shopping centres, sometimes
growing to minicities of 300,000500,000 residents
(for example, Brae, California). In these edge cities,
malls usually function as the village squares (Garreau,
1991). Des Rosiers and associates (1996, Canada)
reported a positive correlation between house prices
and proximity to shopping centres. In the UK, it is
understood that some of the best-known out-of-town
shopping centres (The Metro Centre and LakesideF
Mintel, 1997; BraeheadFLowe, 2000a) have attempted
to be reclassied for planning regulation purposes as intown, on the grounds that residential development has
been attracted to the area. According to Lowe (2000b),
UK shopping centres are becoming the new high

197

streets in a trend even more marked than in the US.

All of these indications support an essential tenet of
central place theory associating the number of population and provision of services with the availability of
shops. Retail has been stated to form the heart of
virtually all [UK] towns and cities (Retail Week, 9 June
2000). That article went on to claim that the rst step in
urban regeneration is renovating shopping centres and
retail, citing the example of the transformation of Wood
Green Shopping City [that] created a focus for the north
London community and improved the surrounding
area. The authors work has demonstrated the link
between shoppers and retail attractiveness to be part of a
dynamic process in which planners and developers might
take the initiative in providing shops, leading to changes
in population, expenditure, residence patterns and
indeed bringing new life into run-down areas.
If the ndings are conrmed by larger studies, there
may be a number of implications. Shopping centre
developers will be able to more accurately model the
likely catchment areas of new or extended centres, and
planners predict the impact on existing facilities with
more condence. By progressing to a further stage of
modelling, residential developers and institutional lenders can benet from improvements to the prediction of
house price changes. Planners will be able to model the
eects of regeneration projects in order to more
accurately assess required infrastructure improvements
and residential provision associated with retail and
shopping centre developments.
The work reported has been exploratory in nature.
Conrmatory studies of larger numbers of shopping
centres and respondents are recommended. There are
limitations to central place theory but the authors
contend that aspects of central place practice, when
suitably modied to place the main focus on the retail
provision, are worthy of further consideration by
planners, developers and managers. The indications
from the results to date are that, once the strict
economic assumptions are relaxed, the principles can
be applied to both in-town and out-of-town shopping
centres, improving the prediction of hinterland catchment area boundaries and dening positions of centres
in the retail hierarchy.

Acknowledgements
The authors thank Capital Shopping Centres PLC for
assistance with funding, Professor Peter McGoldrick of
UMIST for posing many searching questions on early
versions of the results, Professors Ross Davies of the
Oxford Institute of Retail Management, Cliord Guy of
the Cardi University of Wales and Heli Marjanen of
Turku School of Business Administration for providing
much extra useful information.

198

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

References
Breheny, M.J., 1988. Practical methods of retail location analysis: a
review. In: Wrigley (Ed.), Store Choice, Store Location and Market
Analysis. Routledge, London, pp. 3986.
Brown, S., 1992. Retail Location: a Micro-Scale Perspective. Aldershot, Avebury.
Census, 1991. Census of Population. HMSO, London.
Christaller, W., 1933. Central Places in Southern Germany (translated
by Baskin C (1966)). Prentice-Hall, Englewood Clis, NJ.
Clark, D., 1982. Urban Geography. Croom Helm, London.
Converse, P.D., 1949. New laws of retail gravitation. Journal of
Marketing 14, 379384.
Davies, R.L., 1976. Marketing Geography, with Special Reference to
Retailing. Retail Planning Associates, Corbridge.
Davies, R.L., Howard, E., Reynolds, J., 1993. The Shopping Industry
1993: Thirty Years of Growth. Oxford Institute of Retail
Management/British Council for Shopping Centres, Oxford.
Dawson, J.A., 1979. The Marketing Environment. Croom Helm,
London.
Dennis, C.E., Marsland, D., Cockett, W.A., 1999a. Why do people
shop where they do? Recent Advances in Retailing and Services
Science, Proceedings of the 6th International Conference, The
European Institute of Retailing and Services Studies, Eindhoven,
The Netherlands.
Dennis, C.E., Marsland, D., Cockett, W.A., 1999b. Objects of desire:
attraction and distance in shopping centre choice. Proceedings of
the 10th International Conference on Research in the Distributive
Trades, Institute for Retail Studies, Stirling.
Dennis, C.E., Marsland, D., Cockett, W.A., 2000a. Objects of desire:
attraction and distance in shopping centre choice. International
Journal of New Product Development and Innovation Management 2 (2), 4360.
Dennis, C.E., Marsland, D., Cockett, W.A., 2000b. Can a shopping
centre be a brand? Proceedings of the 11th International EARCD
Conference on Retail Innovation, ESADE, Barcelona.
Dennis, C.E., Marsland, D., Cockett, W.A., 2000c. The break point
revisited, Contemporary Issues in Retailing and Marketing Conference 2000, the Planning and Management of
Retail Locations, Manchester Metropolitan University,
Manchester.
Des Rosiers, F., Lagana, A., Thierault, M., Beaudoin, M., 1996.
Shopping centres and house values: an empirical investigation.
Journal of Property Valuation and Investment 14 (4), 411462.
East, R., 1997. Consumer Behaviour. Prentice-Hall, Hemel Hempstead.
Feinberg, R., Stanton, J., Keen, C., Kim, I.-S., Hokama, L., de
Ruyter, K., 2000. Attraction as a determinant of mall choice.
Proceedings of the 7th International Conference on Retailing and
Services Science, European Institute of Retailing and Services
Science, Eindhoven.
Finn, A., 2000. Retail entertainment: lessons from the worlds rst
mega-multi-mall. Proceedings of the 11th International EARCD
Conference on Retail Innovation, ESADE, Barcelona.
Finn, A., Louviere, J.J., 1996. Shopping centre image, consideration
and choice: anchor store contribution. Journal of Business
Research 35, 241251.
Garreau, J., 1991. Edge City: Life on the New Frontier. Anchor, New
York.
Goad Plans/OXIRM, 1991. The New Guide to Shopping Centres of
Great Britain. Hillier Parker, London.
Gosling, D., Maitland, B., 1976. Design and Planning of Retail
Systems. Architectural Press, London.
Guy, C.M., 1984. The urban pattern of retailing: within the UK. In:
Davies, R.L., Rogers, D.S. (Eds.), Store Location and Store
Assessment Research. Wiley, Chichester.

Guy, C.M., 1994a. The Retail Development Process: Location,

Property and Planning. Routledge, London.
Guy, C.M., 1994b. Whatever happened to regional shopping centres.
Geography 79 (4), 293312.
Guy, C.M., 1996. Corporate strategies in food retailing and their local
impacts: a case study of Cardi. Environment and Planning A 28,
15751602.
Guy, C.M., 1998. Classication of retail stores and shopping centres:
some methodological issues. Geojournal 45, 255264.
Guy, C.M., 1999. Retail location analysis. In: Pacione, M. (Ed.),
Applied Geography: Principles and Practice. Routledge, London.
Hackett, P.M.W., Foxall, G.R., 1994. A factor analytic study of
consumers location specic values: a traditional high street and a
modern shopping mall. Journal of Marketing Management 10,
163178.
Hoel, P.G., Jesson, R.J., 1982. Basic Statistics for Business and
Economics. Wiley, New York.
Howard, E., 1993. Assessing the impact of shopping centre development: the Meadowhall case. Journal of Property Research 10,
97119.
Jones, K., Simmons, J., 1990. The Retail Environment. Routledge,
London.
Kirkup, M., Raq, M., 1993. An examination of occupancy dynamics
in new shopping centres. Marketing Education Group, Proceedings
of Annual Conference 2, 559569.
Kirkup, M., Raq, M., 1994a. Tenancy development in new shopping
centres: implications for developers and retailers. The International
Review of Retail Distribution and Consumer Research 4 (1),
1732.
Kirkup, M., Raq, M., 1994b. Managing tenant mix in new shopping
centres. International Journal of Retail and Distribution Management 22 (6), 2937.
Lion, E., 1976. Shopping Centres Planning Development and
Administration. Wiley, New York.
Losch, A., 1940. The Economics of Location (translated by Woglam
W H, Stolper W F (1954)) Yale University, New Haven, CT.
Lowe, M.S., 2000a. Britains regional shopping centres: new urban
forms. Urban Studies 37 (2), 261274.
Lowe, M.S., 2000b. Britains regional shopping centres: new urban
forms. Proceedings of the 7th International Conference on
Retailing and Services Science, European Institute of retailing
and Services Science, Eindhoven.
Management Horizons, 1995. Index of Shopping Venues, Management Horizons, London.
Marjanen, H., 1993. Store location analysis and the mystery of
consumer spatial behaviour: competition between downtown
shopping areas and out-of-town shopping centres as a special
case, Turku School of Economics and Business Administration,
Turku, Finland.
Martin, P.G., 1982. Shopping Centre Management. London, Spon.
McGoldrick, P.J., Thompson, M.G., 1992a. The role of image
in the attraction of the out of town centre. The International
Review of Retail, Distribution and Consumer Research 2 (1),
8198.
McGoldrick, P.J., Thompson, M.G., 1992b. Regional Shopping
Centres: Out-of-Town versus In-Town. Aldershot, Avebury.
Mintel, 1997. Shopping Centres. London, Mintel.
OBrien, L., Harris, F., 1991. Retailing: Shopping, Society, Space.
David Fulton, London.
Openshaw, S., 1973. Insoluble problems in shopping model calibration when the trip pattern is not known. Regional Studies 7,
367371.
OXIRM, 1999. The Shopping Centre Industry: its Importance to the
UK Economy, Oxford Institute of Retail Management/MCB,
Oxford.

C. Dennis et al. / Journal of Retailing and Consumer Services 9 (2002) 185199

Reilly, W.J., 1929. Methods for the study of retail relationships.
University of Texas Bulletin, p. 2944.
Reilly, W.J., 1931. The Law of Gravitation. Knickerbocker Press, New
York.
Reynolds, J., 1993. The proliferation of the planned shopping centre.
In: Bromley, D.F., Thomas, C.J. (Eds.), Retail Change Contemporary Issues. UCL, London.
Reynolds, J., Schiller, R., 1992. A new classication of shopping
centres in Great Britain using multiple branch numbers. Journal of
Property Research 9, 122160.
Routemaster, 1998. SK 6801 St Albans, Pan Star.
Rogers, D.S., 1984. Modern methods of sales forecasting. In: Davies,
R.L., Rogers, D.S. (Eds.), Store Location and Store Assessment
Research. Wiley, Chichester, UK, pp. 319331.

199

Schiller, R., Jarrett, A., 1985. A ranking of shopping centres using

multiple branch numbers. Journal of Property Research 9,
122160.
Tang, E., Ingene, C., 2000. Determinants of consumer shopping
productivity. Proceedings of the 7th International Conference on
Retailing and Services Science, European Institute of Retailing and
Services Science, Eindhoven.
Westbrook, R.A., 1980. Intrapersonal aective inuences upon
consumer satisfaction. Journal of Consumer Research 7, 4954.
Young, R.A., 1985. Suburban growth poles. In: Dawson, J.A., Lord,
J.D. (Eds.), Shopping Centre Development: Policies and Prospects.
Croom Helm, London.
Young, W., 1975. Distance decay values and shopping centre size.
Professional Geographer 27, 304309.