GIS and Evidence-Based Policy Making - Chapter 9 pot

9
Residential Property Utilization: Monitoring
the Government Intensification Agenda
Peter Bibby
CONTENTS
9.1 Introduction 177
9.1.1 Policy, Evidence, and GIS 178
9.2 Patterns of New Construction: Accommodating
Housebuilding within Urban Areas 181
9.3 Accommodating Housebuilding: Urban Areas and Beyond 185
9.4 Using Grids to Characterize Dispersal of Housebuilding 191
9.5 Using Grids to Explore Structural Effects and Market Relations 194
9.6 Within the Urban Areas: Intensification of Units of Occupation
1998—Reconstructing a Grid Using PAF 214
9.7 Within the Urban Areas: Intensification of Utilization
of Existing Property 217
9.8 Constructing a Fine-Grained Settlement Geography
to Identify Development Contexts 220
9.9 Conclusions 229
9.9.1 Development Patterns and Policy Objectives 230
9.9.2 Methods and Representations 230
9.9.3 Relation between Policy, Evidence, and GIS 233
References 235
9.1 Introduction
The Government is committed to promoting more sustainable patterns
of development, by:
.
concentrating most additional housing development within urban areas;
.
making more efficient use of land by maximising thereuseofpreviously
developed land and the conversion and reuse of existing buildings;

ß 2007 by Taylor & Francis Group, LLC.
.
assessing the capacity of urban areas to accommodate more housing;
.
adopting a sequential approach to the allocation of land for housing
development;
.
managing the release of housing land; and
.
reviewing existing allocations of housing land in plans, and planning
permissions when they come up for renewal.
(Department of Environment, Transport and the Regions;
DETR, 2000c, para 21)
It seems peculiar to her suddenly that they should be living in this space:
a hundred years ago it would have been a garment factory, where
immigrants from eastern Europe stitched fabric into human shapes
and practised getting their tongues around the muted diphthongs of
English. This is what Lily loves about London, that every building,
street, common and square has had different uses, that everything was
once something else, that the present is only the past amended.
(Maggie O’Farrell, My Lover’s Lover, London, Review 2002, p. 41)
9.1.1 Policy, Evidence, and GIS
In the opening years of the twenty-first century, planning policy in England
and Wales was clearly directed to conserving un developed land and to
the intensification of use of urban areas. DETR’s Planning Policy Guidance
Note 3 of 2000 (PPG3) encapsulated this emphasis. The term intensification
denotes ‘‘a combination of changes in built form and activity’’ and focuses
attention on the capacity of urban areas both to accommodate extra dwell-
ings and to adapt to new economic roles. At the microscale, the term implies
development of previously undeveloped pores within cities; the redeve-

lopment of existing buildings and previously developed sites at higher
densities; and the subdivision, conversion, and extensio n of existing build-
ings. All contribute to the intensification of use of existing buildings or sites
and changes of use allowing increases in the numbers of people living in, or
working in an area (Williams, 1999, p. 168). Policy has focused on amending
the past in a manner which provides for more sustainable development and
which celebrates—perhaps in the manner of O’Farrells’s Lily—the values
of urban living. Over the same period, across government, there was a
reinvigorated interest in founding policy upon evidence. It therefore
seems plausible that there might be some potential role for GIS (and indeed
for Geographic Information Science (GISc)) in developing and monitoring
policy for reshaping of the physical environment.
This chapter explores some of that potential. Its focus is on monitoring
urban growth and the conservation of undeveloped land, on monitor-
ing the media ting influence of urban land recycling, and on the reuse of
existing buildings. It attempts to contribute to debate at three levels.
Most immediately, it attempts to use GIS to draw some inferences about
development patterns in England and Wales which might be pertinent to
ß 2007 by Taylor & Francis Group, LLC.
the assessm ent of policy. Second, it considers how particular techniques,
including the use of natural language processing (NLP) with GIS, can
contribute to the exploitation of data for policy purposes. Third and most
fundamentally, it is concerned with the overall relationship between policy,
evidence, and GIS and with the manner in which GIS use is and might be
embedded within policy pro cesses.
A prerequisite of addressing the first of these concerns is a broad under-
standing of aspects of relevant government policy in 2000 and immediately
afterwards, while engagement with the third concern demands some explicit
consideration of how the term policy itself is to be un derstood. The emphases
of the 2000 revision of PPG3 reflect a commitment to regeneration and

intensification, which suffuses popular planning thought and rests in turn
on underlying concerns about sustainable urban living and broader notions
of environmental sustainability. The 2000 revision of PPG3 must, therefore,
be understood alongside a welter of other documents (including, for
example, the urban and rural white papers of DETR, 2000a,b) and Prescott’s
(2003) statement on sustainable communities which depend upon the
broader discourse of sustainable development. It must be emphasized, how-
ever, that other discursive currents influence present policy set out in the
Communities and Local Government’s Planning Policy Statement 3 (PPS3;
CLG, 2006). CLG is the successor department to the Office of the Deputy
Prime Minister (ODPM), DETR, Dep artment of Transport, Local Government
and the Regions (DTLR), and the Department of Environment (DoE).
The concept of policy pertinent to this chapter should neither be reduced
to the text of PPG3 (or PPS3) nor bloated to include the sum of concerns
about sustainability. In the traditi on of Heclo, policy might be regarded as a
‘‘course of action or inaction’’ (Heclo, 1972, p. 85). The policy process might
thus be seen as centering on the articulatio n of commitments intended to
guide subsequent action. From this perspective, the prime significance of
texts such as PPG3 is that they potentially allow such commitments to bind
actors such as local authority planners who may be distant from central
government policy making both in space and time. The policy process
involves ensuring such attenuation, so that policy becomes a ‘‘stance
which once articulated, contributes to the context within which a succession
of future decisions will be made’’ (Hill, 1997, p. 7 ascribed to Friend et al.,
1974, p. 40). The context reproduced by the policy process is sometimes
referred to as the policy setting and includes an assumptive world of
values, metaphors, and core narratives reflected in bureaucratic practices,
operational definitions, and procedural rules.
Evidence is always used to support or supplant a story. Policy rests upon
particular understandings of the nature of the world. Given the nature of

policy, its relation to evidence is less straightforward than might first
appear. Context denies the possibility of transparent empiricism, thereby
complicating the role of GIS in monitoring its effectiveness. Sustainability,
moreover, should perhaps be seen as an ‘‘essentially contested ’’ concept in
the spirit of Gallie (1955–1956). Without elaboration of a particular narrative,
ß 2007 by Taylor & Francis Group, LLC.
and of particular definitions, GIS, however useful, cannot provide a tool for
distinguishing sustainabl e and unsustainable patterns of development. It,
therefore, cannot somehow ground policy in evidence in an unproblematic
manner. The evidence assembled using GIS is constrained by the data which
it has been deemed worthw hile collecting and framed by particular narra-
tives and images within the policy setting.
Understanding the potential of using GIS in policy monitoring involves
appreciating the character of the traditional narratives. One such narrative
provides an account of urbanization which focuses on the construction of
dwellings, leading from the idea of exogenous household growth to expan-
sion of the contiguous urban area and concomitant reduction in undeveloped
land. The number of dwellings in Great Britain has increased by 80% in the
last 50 years (Matheson and Babb, 2002, p. 163). The traditional narrative has
moved with images such as ‘‘a Bristol a year,’’ directly from increasing
numbers of households to the expansion of the contiguous urban area, and
this provides the imagery by which the press expresses the environmental
consequences of household growth [see, for example, the transmutation of
forecast changes in numbers of households into ‘‘twenty-seven huge new
towns’’ (Daily Telegraph, 1996) or the invocation of ‘‘an area the size of
Manchester’’ (Observer, 2003)]. They converge with images of urban growth,
urban sprawl, and urban spread, which liken cities to organisms, demanding
responses such as CPRE’s Sprawl Patrol. Such images are reflected and sup-
ported by famil iar cartographic devices, which record the expansion of
particular towns over time, which may be replicated within GIS.

More recent narratives, however, qualify this story. Growth in numbers of
households remains at the core. Although population growth has been modest
in recent years, household growth—and hence urban growth—has continued
(sustained by rising real incomes). This growth is to be understood in relation
to changing lifestyle choices that show themselves statistically as continuing
falls in average household size. Variants of the narrative typically question
how new households or dwellings are to be accommodated, but not the
sustainability of those social choices that allow household size to continue to
fall (DoE, 1996). Through the 1990s policy discussion became increasingly
concerned with the extent to which development might be concentrated on
brownfield sites and hence mitigate pressure for urban expansion. This in turn
prompted GIS development including both small-scale analytic work under-
pinning urbanization forecasts (Bibby and Shepherd, 1996) and development
of a National Land Use Database (NLUD)—an inventory of brownfield sites.
In the absence of strong popul ation growth, by 2000, household
growth had come to coexist alongside crude housing surplus at national
level (Matheson and Babb, 2002, p. 164). In particular cities and regions,
problems of low demand for housing had come to assume prominence (e.g.,
Bramley et al., 2000) and these issues had risen high up the policy agenda.
Narratives of urban growth thus came to interact with rather different narra-
tives of local housing market collapse. These emphasized the rapid, extreme,
and essentially arbitrary nature of local market adjustment as withdrawal of
key actors (such as particular social landlords), vandalism against empty
ß 2007 by Taylor & Francis Group, LLC.
property, and outbreaks of social disorder might undermine the possibility of
continued occupation. The specter of urban expansi on running apace along-
side the dereliction of redundant urban quarters had become evident.
Policy, moreover, must be concerned not only with substantive goals but
also to the manner in which they are to be pursued. In a climate where
evidence is used to legitimize policy, where there is a lack of confidence in

forecasts, and where there is uncertainty over the performance of local
housing markets, monitoring came and remains to the fore (in principle at
least). The 2000 revision of PPG3 introduced a ‘‘plan, monitor, and manage’’
approach to planning for housing in preference to the previous regime—
somewhat disparagingly dubbed ‘‘predict and provide’’ retrospectively
(Prescott, 2000). This provides the context in which this particular series of
GIS applications is set. It is very different to one in which housing
demand—driven by population growth—would inevitably be met by the
construction of family housing immediately recognizable by remote sensing
and easily represented on large-scale maps.
9.2 Patterns of New Construction: Accommodating
Housebuilding within Urban Areas
The introductory quotation from the 2000 revision of PPG3 (DETR, 2000c)
focuses on three objectives: concentrating housebuilding on sites within
urban areas, concentrating housebuilding on previously developed sites,
and accommodating new dwellings within existing buildings. The remain-
der of this chapter treats each of these objectives in turn, using GIS to
explore how far patterns of housebuilding in the 1990s proved consistent
with the intentions set out in 2000 and exploring some of the issues arising.
In so doing it must have regard to the closely linked intentions to
avoid developments which make inefficient use of land (those of less
than 30 dwellings per hectare net)
encourage housing development which makes more efficient use of land
(between 30 and 50 dwellings per hectare net)
and
seek greater intensity of development at places with good public trans-
port accessibility . . . such as city, town, district and local centres or
around major nodes along good quality public transport corridors.
(PPG3; DETR, 2000c, para 58)
The location of new development in relation to existing urban areas would

appear to be an issue where there is a clear role for GIS and where the analytic
issues are trivial. Effective monitoring might appear to depend simply on the
availability of information on the location of new housing sites on the one
hand and the boundaries of urban areas on the other recorded with sufficient
ß 2007 by Taylor & Francis Group, LLC.
pre cision and accura cy. Fortunately, Ordnance Survey (OS)—the national
mapping agency—generate both sets of data. Since 1985 they have collected
Land Use Change Statistics (LUCS) for what is now CLG as an adjunct to
updating the national map base (Sellwood, 1987). This constitutes a tractable
source of very fine-grained information about the location of new housebuild-
ing (among other things). OS have also produced for CLG and its predecessors
highly detailed boundaries of physical urban areas for use alongside Census
statistics for 1981, 1991, and 2001 (for a discussion of these boundaries and their
relation to other urban definitions, see Shepherd et al., 2002).
LUC S data re fer to the land par cels shown on ba sic-scale maps (1:125 0
in urban areas 1:2500 at the urban fring e and 1:1 0,000 in mo untain
and mo orland areas ). Whe re the use of any suc h parcel changes (on the
bas is of a 24-categ ory classific ation) a LUC S record is create d. It will
incl ude a 10-m gri d referen ce for a rep resentat ive poi nt with in the parcel ,
a one charact er code (e.g ., R for residen tial) indicati ng the use befo re and
anoth er ind icating the use after the change , an estim ate of the year of
change, an estimate of the area of the site, and (in the case of residential
development) an estimate of the number of dwellings demolished and the
numbe r of uni ts buil t. As shown in Tab le 9.1, these da ta ind icate that in the
years from 1990 to 2000 (inclusive), 1.45 million houses were built in Eng-
land on 586 square kilometers of land (i.e., at an average density of 24.7
units to the hectare). It is important to note at the outset that the implied
annual rate is historically low, although the scale of development is of the
same order of magnitude as that required to meet household projections (e.g.,
DoE, 1995) or that suggested by the Barker (2004) review.

Digital boundaries of physical urban areas are generated for CLG by OS
on the basis of a series of rules. The rules are used to aggregate parcels on
the basis of their use and the distance between them. Any parcel on a basic-
scale map is treated as being in either urban or rural use. The classification
used is the same as that in LUCS, the individual uses being arranged into
these two divisions. Parcels in urban use are then joined with their neigh-
bors or other such parcels within 50 m to form areas of urban land. Open land
totally surrounded by an area of urban land (such as Hampstead Heath or
Richmond Park in London, or Sutton Park in Birmingham) is also treated as
forming part of it. (Under the 1991 definition, a subset of these areas of urban
land are deemed to be urban areas.)
Simply overlaying LUCS point data on the OS 1991 urban area polygons
reveals that over the 1990s, in the order of 57% of new dwellings
were accommodated within those urban areas (Table 9.1).* Although
* In the case of the boundaries produced by OS for use with the 1991 census, a distinction was
made between areas of urban land and urban areas. An urban area for this purpose was defined
as an area of urban land that impinged on four or more enumeration districts (the smallest units
for which 1991 census data were released). This implied a variable lower limit to the population
of urban areas (between 1000 and 2000 persons). The boundaries produced for the 2001 census
encompassed a far larger group of settlements. For this study the term urban areas refers to
physical settlements treated as urban areas in 1991 and with a 1991 population of 2000 or more.
ß 2007 by Taylor & Francis Group, LLC.
TABLE 9.1
New Dwellings Built and Housing Land Developed, 1990–2000, England: Urban Areas (UAs) and Elsewhere
Outside UAs Inside UAs Totals % Inside UAs
Year Units Hectares Density Units Hectares Density Units Hectares Density Units Land
1990 51,516 3,337.9 15.4 110,046 3,841.0 28.7 161,562 7,178.9 22.5 68.1 53.5
1991 43,846 2,110.9 20.8 79,643 2,543.3 31.3 123,489 4,654.2 26.5 64.5 54.6
1992 52,126 2,582.8 20.2 78,714 2,626.8 30.0 130,840 5,209.6 25.1 60.2 50.4
1993 59,919 2,886.0 20.8 83,875 2,702.1 31.0 143,794 5,588.1 25.7 58.3 48.4

1994 70,735 3,500.0 20.2 81,493 2,772.4 29.4 152,228 6,272.4 24.3 53.5 44.2
1995 61,403 3,194.4 19.2 76,938 2,589.3 29.7 138,341 5,783.7 23.9 55.6 44.8
1996 61,967 3,029.5 20.5 66,095 2,101.5 31.5 128,062 5,130.9 25.0 51.6 41.0
1997 66,924 3,284.6 20.4 74,027 2,345.9 31.6 140,951 5,630.5 25.0 52.5 41.7
1998 65,839 3,118.9 21.1 69,252 2,239.3 30.9 135,091 5,358.2 25.2 51.3 41.8
1999 48,189 2,390.7 20.2 55,083 1,781.7 30.9 103,272 4,172.4 24.8 53.3 42.7
2000 45,100 2,124.3 21.2 45,703 1,533.9 29.8 90,803 3,658.2 24.8 50.3 41.9
Total(1) 627,564 H(RU1) 31,559.84 L(RU1) 19.9 820,869 H(UA1) 27,077.3 L(UA1) 30.3 1,448,433 58,637.1 24.7 56.7 46.2
Total(2) 570,784 H(RU2) 28,433.4 L(RU2) 20.1 877,401 H(UA2) 30,194.89 L(UA2) 29.1 1,448,185 58,628.3 24.7 60.6 51.5
Total(3) 564,516.3 H(RU3) 28,128.4 L(RU3) 20.1 883,773.7 H(UA3) 30,503.5 L(UA3) 29.0 1,448,290 58,631.9 24.7 61.0 52.0
2004 Definitions
Total(4) 698,284 H(RU4) 34,752 L(RU4) 20.1 750,005.9 H(UA4) 23,880.2 L(UA4) 31.4 1,448,290 58,631.9 24.7 51.8 40.7
Note: The year-by-year values and Total(1) values have been calculated by treating LUCS data as points and overlaying them on urban area polygons. The
values for Total(2) have been obtained by treating LUCS data as points and overlaying them on a 100-m grid derived from the urban area polygons. The
values for Total(3) have been obtained by spreading LUCS data across a 100-m grid as described in Section 9.4 and overlaying the derived values on a 100- m
grid derived from the urban area polygons.
ß 2007 by Taylor & Francis Group, LLC.
there are no quantitative targets for the proportion of housebuilding to be
accommodated within existing urban areas, it appears that these areas
were able to absorb well in excess of 800,000 new dwellings in the period.
The table appears to provide a substantial degree of comfort to those
anxious to realize the government’s goal of ensuring that by 2008, 60%
of new housebuilding is accommodated on previously developed sites.
(Note that this table says nothing about previously developed sites per se.)
Those practitioners and commentators who remain profoundly skeptical
of the realism of such targets might also find within Table 9.1 some justifi-
cation for their position. They might question how long this pattern of
development might be sustained, pointing out that while more than two-
thirds (68%) of new dwellings appear to have been accommodated in
urban areas in 1990, this proportion fell steadily through the decade, so

that only half of all new dwellings were being accommodated in this way
by 2000 (Figure 9.1). Moreover, it appears that less than half of all house-
building land was found within the confines of urban areas as they had
stood in 1991, and that this proportion too followed a distinct downward
trend. This is consistent with the familiar view that with the passage of
time it becomes progressively more difficult to identify sites within the
urban area.
Accommodating housing with urban areas LUCS 1990–2000
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
80.0
Percent
% Units
% Land
Housing output recorded in LUCS 1990–2000
0
20,000
40,000
60,000
80,000
100,000
120,000
140,000
160,000

180,000
1990 1998 2000199619941992
1990 1998 2000199619941992
Units
in UAs
Total
FIGURE 9.1
Housebuilding within urban areas in England between 1990 and 2000. (From LUCS. With
permission.)
ß 2007 by Taylor & Francis Group, LLC.
Unravel ing these mixed me ssages and draw ing out their im plication s
demands a mo re thorough examin ation of the evidenc e, questio ning the
usual nar ratives more closel y, and depl oying GIS more creativel y. Table 9.1
moves only a tiny step towards understanding how new dwellings have
been accommodated or the extent to which they might be accomm odated in
urban areas in the future. The rest of this chapter attempts to move succes-
sively closer to definitions that are substantively meaningful in policy terms.
This first definition of urban areas will be called UA1. The number of
dwellings accommodated within the 1991 urban areas will be referred to
as H(UA1), and the corresponding area of land developed L(UA1). Subse-
quent definitions of urban areas will be referred to as UA2 and so on, the
general case being termed UAi (and the corresponding rural residual RUi).
Development within UAi will be referred to here as urban consolidation
(accommodating additional households within existing urban areas
through either infilling of green pores or recycling of previously developed
sites). In the 1990s, debate counterposed such urban consolidation against
rural land conversion in the form of either urban extensions (UXi)orof
new settlements (NSi) (e.g., Breheny et al., 1993). It is, of course, usually
assumed that demand can be diverted between these different contexts
and so it is impossible to understand the volume of new dwellings being

accommodated in cities in isolation. As a next step we attempt to partition
the total number of dwellings built over the 1990s, H(TO), into these
components.
9.3 Accommodating Housebuilding: Urban
Areas and Beyond
As the very idea of urban extensions embodies the metaphor of the city as
polygon, elementary GIS operations should in principle allow for their
direct measurement and for examination of their contri bution to the housing
land supply. Urban extension polygons might be defined as a subset of the
difference polygons created by overlaying the urban area polygons defined
by OS for use with the 2001 census with those for 1991 (defining UX1). New
settlements (NS1) might be represented by urban area polygons not present
in 1991 but found in 2001. Urban consolidation would occur in the polygons
forming the intersection of the two sets (UA1). This simple geometric logic
demands the recognition of two further types of circumstance which are
more marginal to policy discourse. The first is represented by difference
polygons referring to land considered urban in 1991 but not in 2001. These
might be thought of as urban contraction polygons. The second comprises
an outside remaining rural throughout also represented by a polygon (or in
principle more than one). This last class of circumstance thus constitutes
what might be termed as an exurban context (specifically XC1). The number
ß 2007 by Taylor & Francis Group, LLC.
of new dwellings accommodated in eac h of these contexts might in principle
be assessed by overlaying LUCS point data on the polygons defined. Thus
H(TO) ¼ H(UAi) þ H(RUi)
or
H(TO) ¼ H(UAi) þ H(UXi) þ H(NSi) þ H(XCi)
and specifically
H(TO) ¼ H(UA1) þ H(UX1) þ H(NS1) þ H(XC1):
Alfreton

Sutton
Kirkby
Pinxton
Mansfield
Urban areas 1991
Urban areas 2001
Selston
c
c
c
x
x
Z
Z Inset
1 km
3 km
Ravenshead
c
d
e
FIGURE 9.2
Urban area polygons in 1991 and 2001. Detail from part of the Nottinghamshire Coalfield.
Note: Because of the procedural rules used to define urban area polygons (see text), they are
very convoluted. Comparison of polygons for 1991 with those for 2001 provides a clear
indication of urban expansion (see for example areas of expansion such as those marked ‘‘X’’
on the western fringe of Ravenshead. Although they are not consistent with the notion that
changes to urban use are fundamentally irreversible, areas of urban contraction are also found
(such as those marked ‘‘c’’ above). While some of these appear to reflect change on the ground,
other change appears to reflect differences of view. This seems particularly clear in the inset
which shows the south-western limit of Sutton-in-Ashfield. Here differences in the western

settlement margin appear to reflect a digitzing decision, and the minor contraction along the
southern limit an arbitrary decision that in 1991 the A38 dual carriageway should be included
within the urban area although it was excluded in 2001. The apparent contraction (c) seems to
reflect a change of view, whereas the expansion (d) seems consistent with change on the
ground, though the apparent contraction (e) seems to arise from another change of view.
ß 2007 by Taylor & Francis Group, LLC.
In practice, such partitioning proves troublesome, both computationally and
conceptually. Computational diffi culties a rise in overlaying two large sets of
highly convol uted bound aries (Figure 9.2). The conceptu al dif ficulti es
become apparent in examining the nature of change between the polygons
representing the urban extent in 1991 and 2001 urban area polygons.
Broadly speaking, such change s may arise
.
Where building operations or changes of use imply an extension
or contraction of the urban area on the basis of the definitional
rules used by OS
.
Where there is a difference in view of how the definitional rules
should be applied
.
Where there is a difference in view of the appropriate relationship
between polygons and intensional settlements
The first source is, of course, the focus of immediate interest. As CLG and
its predecessors have taken the view that rural to urban land conversion is
essentially irreversible, it is not clear from a policy perspective how urban
contraction polygons should be treated. In what follows, development that
occurs within them is regarded as being within the urban area UA1. Differ-
ences of view become apparent when changes in the urban area polygons are
examined in relation to (unchanging) detail of basic-scale map s. This second
source of change is not necessarily conceptually difficult, but although po-

tentially resoluble implies that measures of change based on the polygon
datasets are not simply attributable to change in the built environment.
Consistent with the expedient adopted with respect to urban contraction,
areas viewed as urban in 1991 are treated here as remaining urban thereafter.
The third source of change raises more fundamental issues deriving from
the general relationship between bounding and naming, which can only
receive the most cursory treatment here (but see Jubien, 1993; Bibby and
Shepherd, 2000; Bibby, 2005). The term intensional settlement is used here to
refer to the places that those involved in the policy process talk about, or
(more strictly) have in mind. (On intension, see Searle, 1995.) Despite the
rationale for defining OS urban area polygons set out above, there is not a
one-to-one correspondence between them and the named places assigned a
unique identifier in the datasets. Many intensional settlements (e.g., Lincoln
or Milton Keynes) are represented by more than one polygon. Although the
polygons discussed approximate physical urban areas, unstated intensional
definitions are in fact privileged. Conversely, a single contiguous area of
urban land may be partitio ned into several adjoining polygons. Hence
although the physical definition of London does not extend to its adminis-
trative boundary in some places (e.g., LB Bromley) and extends beyond it in
others (e.g ., LB Hillingdon), boundaries between the boroughs are imposed.
The relationship between naming and bounding gives rise to a range of
curiosa. Ch anges in view of the appropriate relation between places and
polygons may, for example, alter apparent population sizes. Critically, the
ß 2007 by Taylor & Francis Group, LLC.
add ition of a new urban are a polyg on cann ot signal develo pment of a new
settlem ent without regard to the nam e or urban-area seria l number attache d
to the place. Automated application of the city as polygon metaphor within
GIS might potentially lead to interpretations too literal to be valuable for policy
purposes (for example, placing too much emphasis on literal physical connec-
tion or disconnection and lacking real implications for likely travel patterns).

Recourse to the intensional definitions mitigates this, providing a first illustra-
tion of how policy monitoring entails the management of metaphor.
For pre sent purposes , it wi ll be suf ficient simply to flag each LUCS record
with an indicator showin g wh ether it fell within any urban area polygo n: (a)
in 1991 and (b) in 2001. Thi s allows dir ect as sessment of the ro le of urban
conso lidation H(UA1). Using lower cas e letters to express the role of par-
ticu lar contex ts (e .g., UX1) rela tive to Eng land as a wh ole (TO ), it also allows
ass essment of pro portion of new dwel lings (h(U X1)) and of hous ebuildin g
land (l(UX 1)) within implie d urb an extensio ns wh ile abs tracting from the
det ails of the geome try of these areas. The identi fication of dwel lings in new
settlem ents is necessaril y slight ly more compl ex, as it mu st take into account
not only of the matt ers just discussed , but also must recognize that polyg ons
appear ing for the first time in the 2001 dataset may dem arcate preexis ting
settlem ents. It is theref ore not possi ble to def ine in practice a new settlem ent
contex t on the polygo n logic (NS1 in pri nciple). The approach taken, there-
fore, involved first iden tifying candid ate new settlem ents (denote d by
urba n area polygons with cod es appear ing for the first time) and then
over laying these areas on concen tration s of new housebui lding evident in
LUCS. This was accomplished in fact by using a hectare grid representation
of both the OS urban areas and the LUCS data, producing a definition of
new settlement contexts (NS3) compatible with definitions of other contexts
(UA3, UX3, and XC3) introduce d below.
This implies an imperfect geometrical distinction between urban extension
and new settlement, and so in the summary entries in Table 9.2, building in
new settlement has been subsumed within UX1. It is immediately clear, how-
ever, that the contribution of new settlement to accommodating dwellings in
the 1990s was minimal (NS3). Only five urban areas identified in the 2001
urban areas dataset but not in that for 1991 had concentrations of residential
development in the 1990s. These were Cambourne in Cambridgeshire, Dickens
Heath within Solihull District in the West Midlands, Whitley in Hampshire,

Cotford St Luke near Taunton in Somerset (a new village on a former hospital
site), and Hatton near Warwick, a village of mediaeval foundation with a
former hospital site identified as a local growth point in the Warwick District
Local Plan (Warwick District Council, 2003).* Using the grid method, it is
* Other locations identified in applying the procedure that strictly fail to meet the criteria are
Millisons Wood (in the West Midlands), Dunkeswell near Honiton in Devon, Southfields in
Essex Thameside (Thurrock), and Tanfield (a village abutting the urban area of Cheshunt in the
Hertsmere District of Hertfordshire).
ß 2007 by Taylor & Francis Group, LLC.
TABLE 9.2
New Dwellings Built and Housing Land Developed, 1990–2000, England: by Mode of Accommodation
Units Built, 1991–2000 Land Developed, 1990–2000 Densities Achieved, 1990–2000
Outside
UAs 2001
Inside
UAs 2001 Total
Outside
UAs 2001
Inside
UAs 2001 Total
Outside
UAs 2001
Inside
UAs 2001 Total
Outside UAs 1991 205,505 422,059 627,564 14,248.67 17,311.17 31,559.84 14.4 24.4 19.9
Inside UAs 1991 5,256 815,613 820,869 261.16 26,816.13 27,077.29 20.1 30.4 30.3
Total 210,761 1,237,672 1,448,433 14,509.83 44,127.3 58,637.13 21.6 25.7 24.7
Percentages Descriptions
Outside UAs 1991 14.2 29.1 43.3 24.3 29.5 53.8 Exurban development Urban extension
Inside UAs 1991 0.4 56.3 56.7 0.4 45.7 46.2 Reclassification Urban intensification

Total 14.6 85.4 100.0 24.7 75.3 100.0
Units Built 1990–2000 Land Developed 1990–2000 Densities Achieved 1990–2000
Exurban development 205,505 14.2 14,248.67 24.3 14.4
Reclassification 5,256 0.4 261.16 0.4 20.1
Urban extension 422,059 29.1 17,311.17 29.5 24.4
Urban consolidation 815,613 56.3 26,816.13 45.7 30.4
1,448,433 100.0 100.0
Totals(1)
Exurban development 205,505 14.2 XC1 14,248.7 24.3 14.4
Urban extension 422,059 29.1 UX1 17,311.2 29.5 24.4
Urban consolidation 820,869 56.7 UA1 27,077.3 46.2 30.3
Overall 1,448,433 58,637.2 100.0
(continued )
ß 2007 by Taylor & Francis Group, LLC.
TABLE 9.2 (continued )
New Dwellings Built and Housing Land Developed, 1990–2000, England: by Mode of Accommodation
Units Built, 1991–2000 Land Developed, 1990–2000 Densities Achieved, 1990–2000
Outside
UAs 2001
Inside
UAs 2001 Total
Outside
UAs 2001
Inside
UAs 2001 Total
Outside
UAs 2001
Inside
UAs 2001 Total
Totals( 2)

Exurban development 170,910 11.8 XC2 11,789.8 20.1 14.5
Urban extension 399,874 27.6 UX2 16,643.6 28.4 24.0
Urban consolidation 877,401 60.6 UA2 30,194.8 51.5 29.1
Overall 1,448,185 100.0 58,628.2 100.0 24.7
Totals( 3)
Exurban development 185,628 12.8 XC3 12,597.9 21.5 14.7
Urban extension 378,888 26.2 UX3 15,530.5 26.5 24.4
Urban consolidation 883,774 61.0 UA3 30,503.5 52.0 29.0
Overall 1,448,290 100.0 58,631.9 100.0 24.7
2004 Rural Definition
Exurban development 398,165 27.5 XC4 22,949.6 39.1 17.3
Urban extension 300,120 20.7 UX4 11,802.1 20.1 25.4
Urban consolidation 750,006 51.8 UA4 23,880.2 40.7 31.4
Overall 1,448,291 100.0 58,631.9 100.0 24.7
Note: The crosstabulated values and Total(1) values have been calculated by treating LUCS data as points and overlaying them on urban area polygons. The
values for Total(2) have been obtained by treating LUCS data as points and overlaying them on a 100-m grid derived from the urban area polygons.
The values for Total(3) have been obtained by spreading LUCS data across a 100-m as described in Section 9.4 and overlaying the derived values on a 100-m
grid derived from the urban area polygons.
ß 2007 by Taylor & Francis Group, LLC.
estimated that new settlements accommodated barely more than 2000 units
over the period [or 0.2% of all dwellings; H(NS3) is 2207; h(NS3) is 0.2%].
Planning practitioners and analysts are unlikely to be surprised by the
nature of these results (whatever the precise numerical values). Whatever
definition might be adopted, enthusiasm for prospective new settlements
waned over the decade, given the difficulties of overcoming risk on the one
hand and public opposition on the other. It should also be clearly under-
stood, however, that in contrast to the presumptions of more journalistic
commentators, the 1990s witnessed the accommodation of more than 1.45
million new dwellings, without the construction of new settlem ents. Equa lly
important in considering appropriate policy responses to household projec-

tions or to the recommendations of the Barker (2004) review, it becomes
important to examine quite how such an apparently implausible volume of
development has in fact been accommodated historically.
While the minimal role of new settlement should occasion no surprise, it
is evident that new settlement and urban extension together accounted for
barely more than one house in four [h(UX1) is 29%]. Despite the traditional
narrative, the number of houses accommodated through urban consolida-
tion was almost double that in new settlements and urban extensions.
Besides urban consolidation, which accounts for 57% of new dwellings,
there remains, however, a further component of accommodation, termed
here exurban development. Although in GIS-based analyses it is not pos-
sible to ignore development in such contexts, it is relatively little discussed
by policy makers and practitioners. Exurban development appears to
account for some 200,000 dwellings in the 1990s [h(XC1) is 0.14]. Such
development seems, moreover, to have proceeded at particularly low dens-
ities (14.2 dwellings to the hectare on average) and thus to account for a
disproportionate share of all land developed for housing [l(XC1) is 0.243].
Examination of the potential for urban consolidation cannot be reduced
simply to assessing the capacity of urban areas (on the one hand) and
understanding (on the other) the competing attractiveness of sites within
the urban area and those which would extend it. The scale and character of
exurban development may imply a threat to urban consolidation and
demands more sustained analysis. It therefore seems important to take care
to see just what sort of development is involved—whether dispersed prop-
erties in sparsely populated areas, for example, or simply buildings very
close to urban areas but deemed outside by imposition of a particular bound-
ary. It will also be necessary to have regard to the overall level of demand.
9.4 Using Grids to Characterize Dispersal of Housebuilding
As a first step in visualizing the pattern of development, and in trying to
understand the reasons for place-to-place variation in the relative import-

ance of the modes of accommodation, it is convenient to transform the data
ß 2007 by Taylor & Francis Group, LLC.
to a fine (100 m) gri d (each ce ll rep resenting one hectar e). Mapping the raw
data points for LUCS is not hel pful as som e 223,538 observ ations und erlie
the summar ies in Ta bles 9.1 and 9.2. Grid repres entation form s the basis for
the remain der of the analys es in this chapte r. Conver ting the 1991 and 2001
urba n areas to hectar e gri d rep resentat ions forms the ba sis for a revised
defin ition of contex ts UA2, UX2, XC2 allowin g identi fication of urban
conso lidation, urb an extensio n, and exurban devel opment, respecti vely.
Tab les 9.1 and 9.2 therefo re also sho w along side estimates mad e on a
point -in-poly gon basis (UA1, UX1, etc.) varia nt estim ates suc h as H(UA 2),
L(U X2), etc. cal culate d havin g conve rted the phys ical ur ban areas to a 100-m
grid wh ile still treating the LUC S data as point s. The figures marke d UA3,
UX3, and so on are estim ates calculate d havin g co nverted bot h the physical
urba n area and the LUCS da ta to 100 m grids , with the LUCS measure s
spre ad as describ ed abov e.
As LUC S data point s refer to parce ls of very dif ferent size, som e pre pro-
ces sing is required . Whil e with in LUC S the me dian reco rded size of lan d
parce ls devel oped for residen tial use is 0.1 ha, 3.8% of point s refer to parcel s
grea ter than 1 ha in ex tent, wh ich accou nt for 23.1% of units. Obvio usly, if
a LUCS point re fers to an area gre ater than 1 ha, it mus t overflo w its cell.
To offset this, areas of develop ment in exces s of 1 ha must be locally spre ad
as the grid is created (imp lemente d here using Prolog) a lthough the conf ig-
urati on of the sites is unk nown. In the discuss ion of earlier sections , where
the relation betwee n LUCS records and urban areas has been reduced to the
point -in-poly gon trope, this prob lem has sim ply been igno red.
Exam inatio n of Ta bles 9.1 and 9.2 shows that spre ading out the LUCS
data has only a mo dest ef fect on interpret ation of how gro wth is accomm o-
date d [compar e H(XC 2) and H(XC 3), for exampl e]. Convers ion of the urban
area polygo ns does, however , have an impa ct on the overall pict ure. Table

9.1 sho ws that the propor tion of hou sebuil ding accomm odated with in the
urba n areas appear s to rise from 57% [h(UA1) ] to 61% [h(U A2)] as arou nd
60,000 dwel lings are reclassi fied. This stands as a war ning of the poten tial
sens itivity of any partitio ning of mod es of accommo dation to the precise
placem ent of the urban area bound ary.
Figure 9.3a attemp ts to show variatio ns in inte nsity of develop ment at
hectar e cell leve l, and dem onstra tes the im possibi lity of graspi ng the pattern
with out some gen eraliza tion. This can be achieved by usin g movi ng spatial
ave rages, thereb y commuti ng actual housing output in ce ll q [denoted here
H(TO( q,0)) ] as sho wn in Figure 9.3a, to a tendenc y to develo p over a
particular radius r around cell q. The tendency to develop within 2000 m
of cell q [denoted here H(TO(q,2000))] is shown in Figure 9.3b. When
represented as a 2-km moving average in this way, the pattern of develop-
ment in the 1990s becomes immediately obvious. Areas with limited or
highly dispersed housing development are shown by the lightest shades
(up to 1% of the land area being developed for housing over the period). In
tracts with the deepest grays more than 30% of the area was developed for
housing in the period.
ß 2007 by Taylor & Francis Group, LLC.
CH—Chafford Hundred IB—Ingleby Barwick Wo—Worcester
BS—Bradley Stoke MK—Milton Keynes
0–0.5
Wh
Wh
CH
CH
MK
MK
BS
BS

Wo
Wo
IB
IB
(b)(a)
0.5–2
>2
Wh—Whitley
FIGURE 9.3
Intensity of residential development in England (1990–2000) by hectare cell.
ß 2007 by Taylor & Francis Group, LLC.
Images suc h as Figure 9.3b provide the bas is for a more intuiti ve grasp of
the relati ve impor tance of differe nt mode s of accomm odation . Figu re 9.3 b
make s it a little easie r to visuali ze the co ntribution of diffuse exu rban
hou sebuil ding alon gside the rather smal l numbe r of majo r urban extensio ns
[inclu ding Milton Keyn es (MK), Bradle y Stoke (BS) on Bri stol’s north fringe,
Ch afford Hundre d (CH) in Kent Tham eside , Worceste r (Wo), or Ingleb y
Barw ick (IB) near Stockt on on Tees ]. The new settlem ent at Whitl ey (Wh ) in
Ham pshi re is also evide nt, as are the large st areas of ur ban consol idation ,
e.g., London Docklands .
App lying spat ial averagi ng at the 2-km scal e to the ind ividual accom-
mod ation mod es (u rban co nsolidat ion, urb an extension, and exurban
devel opment) allows visuali zation of geogr aphic variatio n in their
contr ibution— see Figure 9.4a–c fo r H(UA 3(2000)), H(UX3( 2000)), and H
(XC3( 2000)), respecti vely. The mi nimal contr ibution of urban extensio n
(Fig ure 9.4b ) arou nd Londo n (mor e clearly evident in Figure 9.5a), despi te
the volu me of uni ts constr ucted (Figure 9.2c) , must be unde rstood primaril y
in relation to Green Belt policy. Figure 9.4b illustrates the somewhat larger
contribution of urban extension to accommodation of new housebuilding in
the Midlands, highlighting in particular the continuing expansion of Telford

(a growth pole), and the expansion of Leicester (a city without a Green Belt)
while once again suggesting the influence of Green Belt policy in limiting
urban extension (around Birmingham, for example). More generally, the use
of moving spatial averages calculated over different radii allows patterns to
be analyzed and displayed at different scales, demonstrating the overall
dispersion of development.
9.5 Using Grids to Explore Structural Effect s
and Mar ket R elations
In order to understand these patterns, to assess whether they are in any way
remarkable and most critically to begin to assess the limits to urban con-
solidation, a more analytical approach is required. The balance between
urban consolidation and development elsewhere will depend in part on
the capacity of urban areas to intensify (an issue which has been brought to
the fore in policy terms), but also in part on the extent of development at
competing sites at the urban fringe, or beyond. Dep arting a little from the
dominant narrative, this section has regarded not only the geographic
structure of a locality and planning policy considerations, but also the
market for housing land. Actually representing markets within GIS is,
however, a sign ificant challenge. A market might (in the spirit of Cournot,
1838) be thought of as a conceptual space in wh ich free communication
ensures that identical goods command identical prices, but this begs the
question of what is to be treated as identical and what is to be considered
(geographically) unique. Moreover, the basic devices of market analysis are
ß 2007 by Taylor & Francis Group, LLC.
(a) Urban consolidation (b) Urban expansion (c) Exurban development
0–0.05
0.05–0.1
0.1–0.15
0.15–0.2
>

0.2
Additional
dwellings per
hectare
FIGURE 9.4
Components of accommodation, 1990–2000.
ß 2007 by Taylor & Francis Group, LLC.
London
Midlands
Harlow
CH
Birmingham
Bracknell
Bracknell
Bracknell
Thamesmead
Thamesmead
Thamesmead
Telford
Telford
Telford
Leicester
Leicester
Leicester
Derby
Derby
Derby
Nottingham
Nottingham
Nottingham

FIGURE 9.5
Dwelling units gained through urban expansion, 1990–2000 in London and the Midlands. (ß Crown Copyright=database right 2007. An Ordnance
Survey=EDINA supplied service.)
ß 2007 by Taylor & Francis Group, LLC.
concerned with the realm of intension—quantities which actors might wish
to provide or purchase at particular prices, whereas GIS are concerned
necessarily with extension; that is, with dateable, placeable parts of the
physical world (Bibby and Chowdry, 2001). So although the current frame-
work of government policy refers to spatially delimited housing markets
(ODPM, 2004), the structural metaphors of economics do not always easily
fit with those of GIS. Resolving these contradictions would constitute a
project of substantial practica l and theoretical significance. On the relation
between GISc and econometrics generally, see the work of Anselin and his
collaborators (Anselin, 2000, 2001; Anselin et al., 2004.). This is beyond the
scope of this chapter
To facilitate critical reflection on observed patterns, and to explore the
relation between these two styles of thought (and their corresponding struc-
tural metaphors), this section pursues a much more modest goal. It explores
what might be involved in assessing geographic variation in the degree of
urban consolidation that might realistically be expected on the basis both of
geographic structure and of market conditions. Founding expectations on
evidence is clearly more complex than merely providing evidence of out-
comes. The unit of analysis remains the hectare cell. Spatial averaging is used
to create variables that generalize various quanta (e.g., housing output and
house prices) over a 10-km radius around each cell, so that England is
effectively analyze d as 13 million overlapping circles. The analysis appeals
to a set of hypothetical circumstances, referred to as the 10-km radial model
(10 KRM). Units constructed are assumed to vary in both plot-size and
building footprint but to follow standard dwelling types and layouts, inci-
dentally implying the development of residential enclaves of fundamentally

similar character. They are assumed to occupy land of homogenous quality,
but located alternately within the urban area as at 1991 (UA3) or outside it
(RU3) [recall that H(RUi) ¼ H(UXi) þ H(XCi)]. Under these assumptions
housebuilding thus implies the construction of suburbs, albeit that they
may be discontinuous suburbs. Except in the presence of supply constraint,
prices and quantities for each cell, as averaged over 10 km, are assumed to
reflect market equilibrium. Rather than assuming geographically bounded
markets a priori, conditions of demand and supply, and hence the position of
the demand and supply curves, are assumed to change continuously from
cell to cell. Elasticities are assumed to reflect more fundamental behavioral
choices, conditioned by broader social values and hence to be constant across
all cells. From this perspective, it would be the failure of these relations that
would necessitate the identification of local markets.
The equilibria of the 10 KRM refer solely to two hypothetical homogenous
goods: housing land and housing space. Actual outcomes will differ because
of the varying relationship between the imaginary homogeneous good and
units actually traded, and will depend on the specific characteristics of
individual parcels and of their immediate environment. Nevertheless, 10
KRM should provide a benchmark identifying variability at the 10-km scale
to which shorter wavelength variation might be added.
ß 2007 by Taylor & Francis Group, LLC.
This model embodies a series of more specific assumptions:
(i) That markets for housing land reach equilibrium in such a manner
that within (overlapping) areas of 10-km radiu s around each hec-
tare cell demand is equal to supply at the reigning price, except in
the presence of supply constraints
(ii) That the demand for additional housing units in each cell stems
from the formation of new households, and thus is a function of
the distribution of existing households (no allowance being made
for the formation of additional households headed by people

living further afield)
(iii) That the demand for additional housing units is price and income
inelastic, but that the demand for housing space per unit decreases
with its price per square meter and increases with income
(iv) That the supply of urban land for housing is a function of the stock
of urban land and increases with the price of land
(v) That the supply of rural land is a function of the stock of land free
of planning constraint and also increases with the price of land
(vi) That the demand for housing land is derived by reference to the
demand for housing units and the demand for housing space
Following this logic, if supply is not constrained (e.g., by the planning
system), the number of housing units constructed within each overlapping
circle would be fixed by virtue of assumptions (ii) and (iii). Location within
these circles would not, however, and together with the area of land to be
developed for housing would depend on the intersection of a demand func-
tion based on assumption (vi) andasupplyfunctionbasedonassumptions (iv)
and (v). On the assumption of identical character (and price), the equilibrium
balance between development on urban and rural sites would be given by
their contribution to supply at the relevant point on the overall supply curve
(reflecting their different conditions of supply). The following paragraphs
work through the assumptions of 10 KRM, attempting ateach step to illustrate
the relationships that seem to hold and to consider their implications.
Assumption (i) is not directly testable. It provides the logical link between
extension and intension, allowing observable housing output across a circle
of 10-km radius to indicate the demand for housing units and housing space
at a given price—a hypothetical relation that is not directly observable. It
also allows housing output to indicate supply in those circles where the flow
of housebuilding land is not constrained.* Spatial averaging at the 10-km
* It is often assumed that the planning system constrains the residential land supply (or even
that the supply of residential land is fixed). While it is true that England’s area is roughly static,

it does not follow that the flow of land that owners will wish to make available for a given use at
a particular time is static. Neither is it self-evident that planning designations actually constrain
supply (in this sense) below its free market level. This is a matter to be determined empirically
(DoE, 1992).
ß 2007 by Taylor & Francis Group, LLC.
scale is intended to acknow ledge substitut ability of sites and hence the
possib ility of diver sion of dem and with in that radius .*
Assump tion (ii), that dem and follow s the existing distr ibution of house-
hold s and dwellin gs, appear s vind icated. The stock of hou seholds has been
estim ated by using the gri d referen ces on the postc ode addres s file (PAF; a
virtual ly comp rehensive reg ister of postal addres ses) to ass ign each address
to its corresp onding hectar e cell.
y

The re is a close relati onship betwe en the
stock of dwellin gs within 10 km of a ce ll [O(TO (q,100 00))]—t he deriv ed
struct ural variable —and the number of addition al units compl eted within
10 km of that cell [H(TO (q,100 00))]. This account s for 87.65% of the variabi -
lity of the volu me of new hou sebuil ding. It a ppears that the number of new
dwel lings built over the 1990s was typically equivale nt to 5.2% of the stock
of dwel lings in 2001 (imp lying a 5.47% increase rela tive to the start ing
stock).
Regr ession estimate s based simply on geogr aphic struct ure (ignorin g
price effects) highlight concen trat ion in and arou nd the majo r ci ties (Figure
9.6). The evident strengt h of this rela tionship sho uld temp er any tende ncy to
posit a contrast bet ween uni form decline in the No rth and rapi d urban
expansi on of the Southeast. No immed iately obvio us Nort h–South gradi ent
is appar ent in the reg ression residual s. It is clear, moreove r, that dema nd
around som e north ern ci ties (e.g ., Liverpool , Man chester , and Leeds) is
in fact higher than might be expecte d on the basis of the exi sting stock

of hou seholds, thoug h this is not true a round others (e.g. , Sheffield or
Birmingh am).
The findi ngs underscor e the nece ssity of havin g reg ard to both absolute
and relative change in plan ning poli cy analysi s (although this proves dif fi-
cult in practice) . The relati onship displ ayed in Figure 9.6 underpi ns the
absolute change in numbers of dwe lling units. It highli ghts localitie s with
substa ntial pop ulation s, and if the prime concer n is with accomm odating
growt h this shou ld be pr edominant bec ause the effect of place-to -place
variatio n in the stock of hou seholds dwar fs place- to-pl ace va riation in
rates of growt h (de parture from the regressi on parameter estim ated as
* This device is intended to capture a situation of local spatial competition, but without
imposing the housing market. The assumed spatial scale at which equilibrium is reached
might be compared with the modal journey to work distance from the 2001 census (10.79 km).
y
The generally close relationship between the distribution of households indicated by PAF [O
(TO(q,0))] and that evident from the 2001 census is clear from Figure 9.13. This high level of
matching confirms the potential value of PAF in monitoring change in land-use intensity.
Figure 9.13b was constructed by converting an ArcView shape file representing Census Output
Area boundaries to a 100-m grid with household density being calculated for each Output Area
on the basis of its original geometry. Figure 9.13a was produced 2 years before from the 2001
Quarter II PAF. From the grid-analytic perspective of this chapter, it seems appropriate to think
of statistical reporting units such as OAs, or units of administrative significance (such as
postcode sectors) as averaging underlying data over an irregular area of typical but varying
radius. The value of each cell q in Figure 9.13b might be thought of as O(TO(q,c)), c denoting the
average radius of a Census Output Area.
ß 2007 by Taylor & Francis Group, LLC.
(a) Predicted units (b) Residual units Dependent grid:
H(TO(q,10000))
Independent grid:
O(TO(q,10000))

r

2
= 0.8765
0 – 0.1
>
0.4
0.1 – 0.4
<
0.05
>
0.05
−0.05 to 0.05
y = 0.007796 + 0.05194x
Regression equation:
FIGURE 9.6
Additional residential units per hectare, 1990–2000.
ß 2007 by Taylor & Francis Group, LLC.
0.0078 in Figure 9.6). A focus on variation in rates of co nstructi on re lative to
the dwel ling sto ck, by co ntrast, highli ghts a tract of lan d stretchi ng from
Devon to Lincolnsh ire referr ed to a s Hall ’s Golden Belt (Figure 9.7; see also
Bibby and Sheph erd, 1991, 1996). It is impor tant not to be mi sled by hig h
rates of gro wth, howe ver intere sting they may be as an indicato r of the
leading edge of chang e. Cruc ially, areas such as the economi cally buoyan t
tract to the west of Londo n, where Hampsh ire, Su rrey, and Berks hire meet,
posses sed substa ntial stocks of exi sting hou seholds and exper ienced hi gh
growt h rates. This unde rlies the pattern of positi ve residuals (Fig ure 9.6).
Moreo ver, as 10 KRM takes no account of growt h originati ng beyon d
this dista nce, ot her positive residual s ine vitably highli ght the princip al
policy-dr iven gro wth centers (su ch as Milton Keyn es).

Assump tion (iii) posits that overal l demand for addition al hou sing units
does not vary with the price of uni ts, but that the demand for ho using space
per unit decreases with its price per square meter and increases with
income. This assump tion thus redresses the exclusive emphasis on
M(TO(q,10000))
0–350 m
2
350–500 m
2
> 500 m
2
FIGURE 9.7
Inferred plot size (m
2
).
ß 2007 by Taylor & Francis Group, LLC.