estimating the value of a range of local environmental...
TRANSCRIPT
1
Estimating the Value of a Range of Local Environmental Impacts
Mark Wardman1, Abigail Bristow2, Jeremy Shires1, Phani Chintakayala1
and John Nellthorp1
1Institute for Transport Studies, University of Leeds
2Transport Studies Group, Department of Civil and Building
Engineering, Loughborough University
April 2011
Prepared for:
Department for Environment, Food and Rural Affairs
2
INSTITUTE FOR TRANSPORT STUDIES DOCUMENT CONTROL INFORMATION
Title Estimating the Value of a Range of Local Environmental Impacts
Authors Mark Wardman, Abigail Bristow, Jeremy Shires, Phani Chintakayala and John Nellthorp
Editors Mark Wardman, Abigail Bristow, Jeremy Shires
Version Number V5
Date 26th April 2011
Distribution Internal, Confidential
Availability ITS and Defra
File defra_LEF_final_report.docx
Signature
3
Contents
EXECUTIVE SUMMARY ....................................................................................................................... 5
ACKNOWLEDGEMENTS ...................................................................................................................... 9
GLOSSARY OF TERMS....................................................................................................................... 10
1. INTRODUCTION ....................................................................................................................... 11
1.1 Aim and Objectives .......................................................................................................... 11
1.2 Approach and Structure of the Report. ............................................................................. 12
2. EXPERIMENTAL DESIGN ........................................................................................................... 14
2.1 Introduction ..................................................................................................................... 14
2.2 Representation of the Local Environmental Factors .......................................................... 14
2.3 Focus Groups ................................................................................................................... 17
2.4 Survey and Experimental Design ...................................................................................... 19
2.4.1 Introduction ............................................................................................................. 19
2.4.2 Stated Preference Experimental Design .................................................................... 19
2.4.3 The First ‘Quality of Life’ SP Experiment (SP1) ........................................................... 21
2.4.4 The Second ‘Local Environmental Factors’ SP Experiment (SP2) ................................ 23
2.4.5 Background and Attitudinal Questions ...................................................................... 25
2.5 Pilot Surveys .................................................................................................................... 26
2.6 Final Survey Design .......................................................................................................... 28
2.6.1 Section 1: Quality of Life ........................................................................................... 28
2.6.2 Section 2: SP1 ‘Quality of Life’ Experiment ................................................................ 28
2.6.3 Section 3: Ratings ..................................................................................................... 29
2.6.4 Section 4: SP2 ‘Local Environmental Factors’ SP Experiment ..................................... 30
2.6.5 Section 5: About You and Your Household ................................................................ 31
2.7 Conclusions ...................................................................................................................... 31
3. SURVEY IMPLEMENTATION AND SAMPLE CHARACTERISTICS ................................................... 32
3.1 Introduction ..................................................................................................................... 32
4
3.2 Area Selection .................................................................................................................. 32
3.3 Survey Implementation .................................................................................................... 33
3.4 Sample Characteristics ..................................................................................................... 37
3.5 Ratings of SP Attribute ..................................................................................................... 41
3.6 Rating of the Images and Descriptions used in SP2 ........................................................... 44
3.7 Respondents “As Now” Situations .................................................................................... 46
3.8 Conclusions ...................................................................................................................... 49
4. EMPIRICAL FINDINGS ............................................................................................................... 51
4.1 Modelling Approach ......................................................................................................... 51
4.2 SP1 Quality of Life Experiment Results ............................................................................. 56
4.3 SP2 Local Environment Factors Experiment Results: Dummy Variables ............................. 59
4.4 SP2 Local Environment Factors Experiment Results: Rating Scale Models ......................... 64
4.5 SP2 Rating Scale Models: Socio-Economic and Attitudinal Segmentations ........................ 66
4.5.1 Modelling Approach ................................................................................................. 66
4.5.2 Interaction Effects .................................................................................................... 67
4.5.3 Separate Models ...................................................................................................... 68
4.5.4 Preferred Segmented Model .................................................................................... 69
5. CASE STUDIES ...................................................................................................................... 72
6. CONCLUSIONS ......................................................................................................................... 76
REFERENCES .................................................................................................................................... 80
SEPARATE APPENDICES
APPENDIX A The Focus Groups
APPENDIX B Images Used in the Focus Group
APPENDIX C Focus Group Importance and Satisfaction Ratings
APPENDIX D Local Environmental Factors SP
APPENDIX E Ratings Photographs
APPENDIX F Final Questionnaire
5
EXECUTIVE SUMMARY
The local environment influences people‟s perceptions of their quality of life and their overall
well-being in many different ways. Whilst there are a wide range of local environmental
factors that can impact on individuals‟ well-being, there is relatively little empirical evidence
on this subject.
In particular, there is a dearth of knowledge on their economic valuation, commonly
expressed in terms of how much money individuals are prepared to pay for improved
conditions. Such evidence would be of value to policy makers and those involved in the
economic appraisal of public spending priorities since willingness to pay valuations are
commonly taken to be a measure of the benefits that individuals experience after changes in
their circumstances.
The aim of this study was to estimate how much individuals would be prepared to pay, in
terms of council tax, to obtain improvements in a wide range of local environmental factors.
The focus is very much upon individuals‟ willingness to pay for improvements or to avoid
deteriorations. Detailed examination of the dynamics of household decision making and
household well-being were beyond the scope of this study, and indeed this is commonly the
case in comparable studies.
The environmental factors that were the subject of this investigation were:
urban quiet areas
fly-tipping
litter
fly-posting
graffiti
dog-fouling
discarded chewing gum
trees
light pollution (obscuring the stars)
light intrusion (into the home)
odour
The emphasis here was on local or neighbourhood effects, by which we mean individuals‟
willingness to pay for improved conditions as experienced in their locality. The study does
not cover the benefits of improved environmental factors for those who are visitors to an
area or indeed the respondents‟ experiences of these environmental factors in places other
than their locality.
The key method used to estimate willingness to pay values was Stated Preference. This is a
widely accepted and used method of valuing non-market goods and services and has been
used on countless occasions to support decision making both by government bodies and
commercial organisations. This method was supplemented with questions about how
important different quality of life and environmental factors were to individuals, as well as
with information on a range of socio-economic, attitudinal and location characteristics.
6
A particular challenge is how to represent the wide range of environmental factors listed
above in a realistic manner in a survey setting. The limited literature on the subject, two
focus groups, consultation with Keep Britain Tidy, Defra and colleagues, and a pilot survey,
all contributed in a short timescale to a mix of photographic and verbal presentations.
Respondents were asked to rate on a 10 point bad to good scale each of the different types
of local environmental factor offered to them, whether it be presented in verbal or
photographic form, and also to indicate their current situation for each factor. This provides
background information and was critical for the analysis of the Stated Preference data.
The methodological approach adopted was two-fold. First of all, a Quality of Life Stated
Preference exercise was offered, where dog-fouling and access to quiet areas were chosen
from the above list of local environmental factors to be included alongside what might be
regarded to be more substantive quality of life attributes of local crime levels, local school
quality, road traffic levels, traffic noise at home, the general condition of pavements and
council tax levels.
The aim of this Stated Preference exercise was to mask the exact purpose of the study, by
offering local environmental factors in a broader quality of life dimension which avoids
placing undue emphasis on local environmental factors, and therefore to be less suspect to
people providing answers that are aimed at influencing policy makers. It provides valuations
of dog-fouling and access to quiet areas that can be used, if necessary, to adjust the
valuations obtained for these and other valuations from the second Stated Preference
experiment which is explicitly and transparently focussed on local environmental factors.
It turned out that there was no convincing case for amending the valuations obtained from
the specific Stated Preference exercise relating to local environmental factors according to
the valuations obtained from the quality of life Stated Preference exercise.
Surveys were conducted in late January and early February 2011 in Manchester, Coventry
and London, providing a representative mix of respondents across England. Within each
city, surveys were conducted in three specific locations that covered inner-city, suburban
and rural/semi-rural areas, because both current environmental conditions and preferences
towards them might be expected to vary across each. The surveys took the form of „Group
Hall Tests‟ where up to 25 people at a time were invited to a venue and the questionnaire
was administered by skilled personnel.
A large sample of 561 respondents was obtained which was broadly representative of the
population of England in terms of gender, age and socio-economic make-up.
In terms of the importance of different quality of life factors, rated on a scale from 1
(extremely important) through to 5 (Not at all important), the mean ratings and associated
95% confidence intervals in parentheses were:
7
Importance of Quality of Life Factors (1=Extremely Important, 5 = Not at all Important)
Level of Local Crime 1.45 (1.38 - 1.52)
Condition of Roads and Pavements 1.88 (1.80 - 1.96)
Amount of Local Council Tax 1.92 (1.84 – 2.00)
Level of Dog Fouling 1.98 (1.90 – 2.06)
Quality of Local Schools 1.98 (1.87 – 2.09)
The Amount of Road Traffic in Your Area 2.00 (1.92 – 2.08)
Access to Quiet Areas 2.00 (1.92 – 2.08)
Neighbourhood Air Quality 2.05 (1.97 – 2.13)
Road Traffic Noise Experienced at Home 2.26 (2.16 – 2.36)
Crime is clearly the most important factor, but dog-fouling is, amongst the general
population, fourth equal in importance and on a par with the quality of local schools. Access
to quiet areas, the other local environmental factor incorporated within this broader quality of
life scenario, is only slightly less important and is deemed to be more of an issue than road
traffic noise and local air quality.
Turning to the local environmental factors, it is illuminating to establish where respondents
currently perceived themselves to be, with a value of 1 denoting the worst condition offered
and either 3, 4 or 5 denoting the best condition depending on the number of levels offered.
As might be expected, rural residents almost always rate their conditions as superior to the
ratings of suburban residents who in turn almost always rate their conditions more highly
than inner-city residents. The extent to which the rural conditions exceed the suburban
conditions is similar, on average, to the extent that the suburban conditions exceed the
inner-city conditions. We can discern some entirely expected patterns, such as rural areas
performing very well in terms of access to quiet areas, light intrusion and pollution differing
little between inner-city and suburban areas, and litter and chewing gum being particularly
prevalent in inner-city areas.
Respondents’ Current Situation for Each Local Environmental Factor
Attribute Scale Inner Suburban Rural Total
Light Pollution 1-3 2.20 2.23 2.72 2.33
Discarded Chewing Gum 1-3 1.95 2.20 2.64 2.21
Litter 1-4 2.31 2.86 3.18 2.74
Light Intrusion at Night 1-4 3.02 3.01 3.25 3.07
Trees 1-4 2.29 2.54 3.06 2.57
Fly Tipping 1-4 3.16 3.54 3.59 3.42
Access to Quiet Areas 1-5 2.92 3.53 4.32 3.49
Graffiti 1-5 2.76 3.53 4.14 3.40
Odour 1-5 3.76 4.17 3.76 3.93
Fly-posting 1-5 3.42 3.86 4.16 3.77
Dog Fouling Occurs 1-5 2.78 3.60 3.86 3.37
8
In terms of the willingness to pay valuations for improvements in local environmental factors,
all expressed on a common 0-10 scale from bad to good, we obtained the following
monetary values in £s per person per month for a unit change in a rating and for the
movement from the worst to the best position. 95% confidence intervals are given in
parentheses.
Willingness to Pay Valuations (£s per person per month) and Ranking of Importance
Value of a Unit
Rating Change
Value of a
Move from
Worst to Best
Stated
Preference
Rank
Importance
Rating
Rank
Chewing Gum 2.17 (1.96 – 2.38) 21.7 4 7
Dog Fouling 1.89 (1.69 – 2.09) 18.9 6 3
Fly Posting - - - 11
Fly Tipping 3.71 (3.39 – 4.03) 37.1 2 2
Graffiti 0.56 (0.42 – 0.71) 5.6 9 8
Light Intrusion 0.34 (0.02 – 0.65) 3.4 10 9
Litter 3.95 (3.59 – 4.31) 39.5 1 1
Light Pollution 0.63 (0.29 – 0.98) 6.3 8 10
Odour 1.91 (1.72 – 2.10) 19.1 5 6
Quiet 1.37 (1.20 – 1.53) 13.7 7 4
Trees 2.33 (2.07 – 2.59) 23.3 3 5
There are some clear priorities here; the valuations exhibit considerable variation across
attributes and it would have been disconcerting to have obtained similar values.
The largest valuations are quite clearly for litter and fly-tipping. Then there are a series of
attributes with similar „medium‟ valuations. These are trees, odour, chewing gum, dog fouling
and quiet areas. Light pollution, graffiti and light intrusion have relatively minor valuations.
This pattern of valuations seems plausible.
It can be seen that the ranking of the local environment factors according to the Stated
Preference valuations corresponds well with the ordering obtained from the importance
ratings1. This is an encouraging finding.
We were able to detect a number of variations in the valuations of local environmental
factors according to socio-economic, attitudinal and location factors. We have used the
results to demonstrate how willingness to pay valuations (in £s per person per month) vary
across circumstances.
1 Technically, a Spearman correlation coefficient of 0.77 indicates a high degree of association
9
Valuations (£s per person per month) for Improvements by Area Type
Inner-City Suburban Rural
One Level To Best One Level To Best One Level To Best
GUM 1.99 2.10 0.83 0.78 0.08 0.08
LITTER 9.75 15.81 12.85 16.20 11.33 12.54
TREES 0.61 1.82 3.11 4.46 2.15 2.95
FLY-TIP 8.43 8.70 5.84 6.18 5.02 5.02
GRAFFITI 1.12 2.78 0.83 1.55 0.21 0.29
FLY-POST - - - - - -
QUIET 0.27 0.58 1.03 1.91 0.53 0.60
DOG FOUL 4.16 8.87 5.12 7.79 1.20 2.72
ODOUR 0.87 1.69 2.25 2.70 2.45 4.05
INTRUSION 0.02 0.03 1.58 2.25 0.55 0.57
POLLUTION -0.23 -0.26 2.37 2.40 0.07 0.07
We have not been able to detect any plausible effect from fly-posting, although we note that
this was returned as the least important factor of all those considered. The negative value
for light pollution, and some other low values for light pollution and intrusion, may be due to
confounding effects with security and safe navigation.
The results above seem plausible. However, they will depend upon the particular
improvements under consideration and the situations from which the improvements are
made, as well as the background socio-economic and location characteristics. These imply
much greater variation in values across the sample than if we simply rely on different ratings
of environmental factors across the sample.
The valuation models here reported can be used to provide willingness to pay measures of
the economic benefits of a wide range of improvements to local environmental factors.
These environmental improvements are often very context specific. By conducting a simple
survey of residents, and asking them to rate on a 0-10 scale existing situations and
proposed ones, the models reported here can be used to provide estimated monetary
valuations of the benefits in those circumstances.
ACKNOWLEDGEMENTS
The authors are grateful for the assistance and advice provided by the Defra Project Board
of Adam Geleff, Roald Dickens, Yvette Hood, Tony Poole and James White. The views
expressed in this document are those of the study team and do not necessarily reflect those
of Defra.
10
GLOSSARY OF TERMS
Attribute: Used synonymously with factor and variable although more commonly associated
with the design features of Stated Preference experiments.
Contingent Valuation (CV): A technique commonly used in environmental valuation which
asks individuals how much they are willing to pay achieve some improvement or to avoid
some deterioration. It can also be expressed in terms of willingness to accept compensation.
Factor: Used synonymously with attribute and variable.
Level: These are the particular values taken on by variables. They can be discrete in nature
reflecting, for example, different categories of litter, or continuous in nature, an example
being the different amounts of time taken to access a quiet area.
Priority Ranking (PR): A particular form of SP exercise where, having determined an
individual‟s current situation across a range of attributes, a number of improvements (and
then deteriorations) are offered and the respondent makes choices that effectively rank the
improvements (deteriorations) in order of preference.
Stated Choice (SC): A form of Stated Preference exercise where respondents evaluate
hypothetical scenarios composed of a number of attributes at different levels and choose
between them.
Stated Preference (SP): A generic term covering methods asking for willingness to pay or
accept compensation either directly in the case of CV or through the choice, ranking or rating
of multi-attribute and level hypothetical scenarios.
Variable: Used synonymously with attribute and factor, although most commonly associated
with the modelling process (eg, as in independent variable).
11
1. INTRODUCTION
1.1 Aim and Objectives
The quality of the local environment clearly has an effect on perceived quality of life and
hence well-being. The local environment itself is made up of and influenced by a range of
diverse factors. Although efforts have been made to value different aspects of the local
environment, at present there are no consistent, comprehensive and therefore usable
monetary values that cover a broad range of environmental factors. This makes decision
making on resource allocation in this domain challenging. Defra has therefore funded this
research to apply a consistent approach between factors, to fill gaps and to reflect public
preferences.
The aims and objectives are to:
“estimate marginal monetary amenity values across local environmental factors, which will
feed into the building of a comprehensive evidence base. These estimates will then be used
to:
Inform appraisal and evaluation of national government policies in relation to each of
the factors;
Inform the consideration of local environmental factors for national prioritisation and
forward planning; and
Provide Local Authorities with an additional resource that could be used in local
decision making.”
The project aimed to establish marginal values for eleven diverse local environmental quality
factors. These are:
urban quiet areas
fly-tipping
litter
detritus
fly-posting
graffiti
dog-fouling
chewing gum
trees
light pollution
odour.
The values obtained will be indicative and provide an input into policy making with respect to
prioritisation in resource allocation decisions in England and Wales.
12
1.2 Approach and Structure of the Report.
This study uses a Stated Preference (SP) methodology. SP methods ask respondents to
consider hypothetical scenarios and express a choice between them (stated choice) or place
a value directly on them (contingent valuation method). Here we use SP experiments that
seek to obtain respondents‟ choices amongst different hypothetical scenarios. These
scenarios contain the variables of interest (in this case local environmental factors) at
different levels (for example, a quiet area could be within five minutes of home or within 10
minutes of home). The choices made allow models to be estimated that provide willingness
to pay values for changes in the levels of the factors (since one of the factors is a money
cost).
A two stage SP approach was adopted here, with both SP experiments using the priority
ranking approach (Wardman and Bristow 2008) whereby a range of attribute improvements
(and deteriorations) on the currently experienced situation are ranked in order of preference.
This method is explained in detail in chapter 2.
The first SP experiment (SP1) contains a broad range of factors influencing local quality of
life to provide a „top level‟ value. Its purpose is to mask the purpose of the study, of valuing
local environmental factors, and thereby offer a reduced incentive to strategic bias2 which
can be present in SP responses.
The second SP experiment (SP2) focuses on the eleven local environment factors of interest
to provide detailed willingness to pay values for different factors and levels. It is more
transparent that this experiment is concerned with valuing local environmental factors. These
valuations may then be scaled to the values obtained in SP1 if deemed necessary.
Our approach addresses a number of key challenges in the valuation of local environmental
factors:
The representation of each factor at clearly distinct levels and relating these to
current experience.
Presenting attributes in a fashion that can be understood by respondents.
Reducing the scope for biased responses, especially responses designed to
influence the policy outcome rather than to express genuine preferences.
Adopting a method that allows the valuations obtained to be transferred across
circumstances.
2 Strategic response bias represents the deliberate biasing of responses in order to influence policy when there
are no consequences from so doing. For example, respondents might overstate their preferences for reduced
dog fouling if they perceive that to be the purpose of the study and especially if they feel they would not have
to pay for it.
13
The first key challenge is to be able to identify the different levels of the relevant attributes.
They must be perceived as different and capable of representation in a survey. The factors
and their different levels must also be clearly understandable and in some way relate to an
objective measure. In order to address these challenges, we draw upon existing knowledge,
and also conducted two focus groups prior to the experimental design and a pilot survey in
somewhat challenging timescales. These elements all helped to shape the final design and
are reported in Chapter 2.
The means of reducing the incentives to strategic bias was to conduct a first SP exercise
that focuses on quality of life in a broader sense, although also covering two local
environmental factors to provide a means of linking to the second SP exercise and adjusting
the values obtained if necessary. By requiring respondents to rate each level of each
attribute on a scale between 0-10 provides a metric in which valuations can be expressed
and transferred to other circumstances conditional upon obtaining ratings for the situation of
interest. The survey implementation is reported in Chapter 3.
Chapter 4 contains the results of the analysis of the SP experiments. Overall models are
estimated, followed by models which identify the key socio-economic and attitudinal factors
that drive variations in monetary valuations of local environmental factors across the
population.
The monetary valuations are applied to evaluate various „case studies‟. These demonstrate
how valuations differ according to the precise local environmental factor concerned and the
differences specified in that factor. The results of this procedure are reported in Chapter 5.
Concluding remarks are contained in Chapter 6.
14
2. EXPERIMENTAL DESIGN
2.1 Introduction
This chapter outlines the design of the SP experiments and the overall survey instrument.
Section 2.2 addresses the key challenge of representing different levels of local
environmental quality factors in a clear and understandable manner. The focus groups that
informed the design are discussed in section 2.3. Section 2.4 integrates these indicators
into a particular form of SP experiment, termed the Priority Ranking (PR), and explains the
design of the survey instrument. Section 2.5 reports on the experiences with the pilot survey
and how this further informed the final survey design which is described in section 2.6. Brief
conclusions on the design process appear in section 2.7.
2.2 Representation of the Local Environmental Factors
The aims and scope of the project are very clear as to the environmental factors to be
included in the valuation exercise. It is also clear that some of these aspects are more easily
defined and represented to respondents than are others. The factors are:
urban quiet areas
fly-tipping
litter
detritus
fly-posting
graffiti
dog-fouling
chewing gum
trees
light pollution
odour.
The challenge here is to find methods of representation that can effectively communicate
different levels of each factor. As there are eleven factors each must also be as concise as
possible. A range of options for presentation exist including:
objective measure
categorical scales
verbal or visual description
simulation
proxy
experienced variation.
15
With respect to these factors there are no obvious, objective and clearly understood
measures. Simulation is beyond the resources of this study and anyway impractical given
the number of factors whilst categorical scales, such as very good, good, neither good nor
bad, bad and very bad, can be vague and difficult to interpret.
The use of proxy measures, whereby a measurable and clearly understood factor is used as
a substitute for the factor of interest, such as the use of aircraft movements as a proxy for
aircraft noise, is not appropriate here, again in part due to the number of factors to be
examined, but also because for these factors there are no obvious proxy measures that
would ease understanding.
Similarly experienced variation, either temporal or spatial, is an attractive approach but it
would not be feasible for respondents to have experienced all the levels of all the factors of
interest.
In this study we therefore use verbal or visual presentation of different levels. The next
steps are:
to identify any existing representations of different levels of environmental factors that
could be applied here
to develop new scales where none exist
to test these representations within focus groups and a pilot survey
For four of the factors photographs exist that depict different standards as part of the
National Indicator 195 framework (Defra, undated) and are applied at Local Authority level. It
therefore seems appropriate to adopt these in the first instance and to seek appropriate
images for other factors, where possible, and where this is not possible or doesn‟t work to
use verbal descriptors. The use of similar approaches for each factor should avoid giving
undue emphasis to any particular factor in presentation.
Here we examine each of these with respect to our initial development of presentations of
different levels for testing in focus groups.
For four factors (litter; detritus; fly-posting and graffiti) a four level national indicator has
been developed and illustrative photographs and descriptions are available (Defra, undated).
Litter and detritus both have four images as they include an image of “no” litter or detritus,
whereas for fly-posting and graffiti only three images are available as they do not contain an
image showing the absence of graffiti and fly-posting. These images at four levels for litter
and detritus and three levels for graffiti and fly posting were taken forward for testing in the
focus groups.
Two factors are fairly homogenous in nature and may be treated as discrete or multiple
events, these are dog fouling and discarded chewing gum. It was decided at an early stage
that a verbal descriptor was appropriate for dog fouling rather than the use of images. It is
16
nonetheless challenging to find an appropriate scale that reflects how people perceive the
issue. In the end four different potential scales were tested:
a five point scale ranging from “not a problem” to “extremely bad problem”
three four point scales examining frequency of occurrence in different ways all
commencing with “never” as the best level:
every X minutes when walking
every X metres when walking
once a month/week/year.
For discarded chewing gum three images were sourced from the web
(www.powercleanservicesuk.com/Chewing-Gum-Removal-02 and
www.guardian.co.uk/business/2008/may/07/3) showing no, some and a great deal of gum.
More complex factors include fly tipping and trees where the individual events may have
very different characteristics; mature trees against recent plantings; mattresses against
hazardous chemicals.
Urban trees perform a number of practical valued roles including: shading and cooling; air
pollutant absorption; reducing water run-off and supporting biodiversity alongside amenity
benefits to residents. With respect to the amenity benefits of trees a number of dimensions
could be examined:
quantity
size and age
variety (including greenness throughout the year)
care and maintenance, for example pollarded v non-pollarded.
Clearly these cannot all be addressed in one experiment with eleven attributes. It was
decided to focus on quantity. For trees we trialled two means of presentation:
aerial images
street level images, using photographs taken for this project, showing different
numbers of mature trees in similar types of residential street.
Fly-tipping as with trees poses a challenge through the diverse nature and quantities of
waste that might constitute fly-tipping. Levels could cover either type of incident and hence
severity of nuisance or number of events. It might be possible to cover both but if a standard
common incident type can be defined, this should probably be household waste/black bags
as this constitutes the most common form of fly-tipping in England and Wales (Defra 2010;
Welsh Assembly Government 2010).
Varying the frequency of event might also be a way of presenting this impact from: no fly-
tipping; through occasional fly-tipping (once a month; through to more regular or frequent
17
occurrence). In the focus groups we used images of different severity, using household
items dumped in a residential street (pictures were taken for this study).
The remaining three factors of quiet areas, light pollution and odour are, arguably, the most
difficult both to measure and represent. Odour may be best treated by verbal description as
bad smells of whatever type will be familiar to most people. However, clearly different
people will have experienced different types of odour and may differ in their response to
specific smells. Therefore, a generic five point scale based on frequency of occurrence was
tested ranging from “no bad smells” to “bad smells all the time”.
Light pollution is a very difficult area. As the brief points out artificial lighting offers both a
benefit in terms of safe navigation and influences perceptions of safety and security – yet it
can also disturb sleep and obscure the night sky (Powe et al 2006). In order to avoid any
confounding effects on safety and security, the descriptions used did not include explicit or
implicit reductions in levels of street lighting. The literature on artificial light and street
lighting in particular (RCEP, 2009; Powe et al, 2006) suggests a need to split this attribute.
We therefore distinguish between light intrusion into the home and the obscuring of the
night sky as these are very different aspects of light pollution. In the focus groups images
were used in an attempt to represent these two factors.
Quiet areas also pose challenges as there is no definitive agreement as to definition or
measurement of quiet areas. The Environmental Noise Directive (European Parliament and
Council, 2002) leaves the definition in terms of noise indicator in agglomerations to the
member states and defines quiet areas in open country as “undisturbed by noise from traffic,
industrial or recreational activities”. Guidance on definition in the UK is available from Defra
and suggests a 55dB Lday filter (Defra 2006). The key dimensions are the relative quiet
offered or the degree of respite from prevailing noise levels and access. Two approaches
are tested:
A five point verbal scale ranging from “no where quiet around here”, through varying
levels of relative quietness to “It‟s really quiet around here”.
A five point scale based on access, ranging from “No quiet areas” to “Quiet area
within 1 minute walk of home”
2.3 Focus Groups
Two focus groups were conducted to obtain to explore key aspects of the representation of
the environmental factors. Firstly, are the images/words clearly understood to be relating to
the factor of interest. Secondly, are the levels portrayed clearly distinguishable and finally
are the levels seen to cover an appropriate range.
The focus groups were recruited from staff and students at the University of Leeds; the
extremely short time frame of the project made a convenience sample the only feasible
option. Nevertheless, particular care was taken to ensure that a wide range of socio-
economic groups were presented and this was achieved (See Appendix A for the
18
composition of the focus groups). Both focus groups, each with 8 participants, took place on
13th January at the University of Leeds.
Each group consisted of three stages: firstly an ice breaker discussion of the local
environment; secondly the main component being a discussion of the images and
descriptions that might be used to depict different levels of the various factors and finally a
rating exercise. Appendix A contains a detailed account of the discussions and findings and
Appendix B the images and descriptions presented. Here we present the key findings from
the focus groups and their implications for the survey design factor by factor.
Litter – whilst the N195 pictures do illustrate different levels of littering, they also introduce
confounding factors by showing very different types of area, the main difference being city
centre and residential streets. This is understandable as they are designed to illustrate
measurable standards to those with responsibility for such measurement, not to present
different levels to the general public. To avoid any confusion a decision was made to obtain
pictures of similar residential areas with differing degrees of litter.
Detritus - There was a consensus within and between groups that the NI195 pictures drew
attention to the differing state of the road and yellow lines rather than the detritus. There
was also a view that detritus was not a problem at these levels. Again the decision was to
take further photos, perhaps at only two or three levels given the difficulty in showing
variation and the relative lack of importance attributed to this attribute.
Graffiti - The image perceived to be of “artistic” graffiti was removed from the set due to the
feeling within the groups that this was a good thing in the environment shown. The picture of
the shop front was effective and again there was a need to seek further images.
Fly-posting - Again there were confounding effects in the images and the response is not
clear with the most severe image being viewed favourably and the second worst seen as
more about rubbish. There was some ambiguity in response to this variable with some
valuing such posters for their information content and a view that fly-posting can brighten up
a run down area. It is clear that different images are needed or a move to a verbal scale.
Again it seems that three levels may be adequate.
Dog fouling – Several verbal scales were tested with a consensus that words were better
than pictures for this factor. The most successful scale was frequency of occurrence when
walking and this was adopted.
Discarded chewing gum - The pictures do show the levels clearly but images should be
obtained to show the different levels on the same surface. Three levels do though seem
appropriate.
Fly-tipping - The groups clearly felt that fly-tipping does not occur in standard residential
contexts. Therefore images were needed that show the “typical” levels of such waste – bin
bags and/or furniture in a more appropriate context.
Trees – Aerial and street view images were tested, with consensus that the street view
images were best and also had the advantage of depicting similar residential streets.
19
Quiet areas – Two scales were tested, the first being distance from your home in minutes
and second describing the relative level of quietness. The second scale was seen as too
wordy and was not easily understood. The distance in time from your home was clear and
was adopted at 5 levels. A 30 minute walk was deleted as it was perceived to be beyond the
local area and a “best” level of “It is very quiet immediately outside” was added between the
focus group and the pilot survey. This was subsequently amended as a result of the pilot
survey.
Odour - The verbal scale showing frequency of occurrence worked well and a four point
scale was adopted.
Light intrusion – The images did not really illustrate light intruding into the home. A picture
indicating a lamp shining into a house was sought. This attribute may be best presented as
two levels – on or off. If necessary this scale could be verbal.
Light pollution - The general view was that the pictures got across the general gist of the
grading but not perfectly with some people noting that the context changed across the
pictures, such as with city centre sky lines and non city centre sky lines. Again we sought to
improve the images within a two or three point scale.
Essentially some scales worked well in the focus groups and others were substantially
revised prior to the pilot survey. The NI195 images do not work well in this context having
been devised to guide grading.
Images shown in PowerPoint lost some precision and we therefore decided to provide
“photo packs” for each respondent. It was also found to be necessary to be very clear about
the area respondents should think of as their local environment and this should be the area
where they live.
2.4 Survey and Experimental Design
2.4.1 Introduction
The survey design will be examined in stages, mostly working through the survey from start
to finish. The survey was designed to be pen and paper implemented in a hall test
environment in sections. This context enables staff to help respondents who have
questions. Section 2.4.2 presents the overall approach with respect to the SP experiments,
section 2.4.3 outlines the design of the first SP experiment, section 2.4.4 presents the
second SP experiment and section 2.4.5 sets out the other questions in the survey.
2.4.2 Stated Preference Experimental Design
There are a number of challenges to be addressed in the development of SP experiments to
value local environmental factors. Key issues include:
20
The representation of each factor at clearly distinct levels and relating these to
current experience.
Presenting attributes in a fashion that can be understood by respondents.
Developing an approach that can cope with the number of attributes
Reducing the scope for biased responses, especially responses designed to
influence the policy outcome rather than to express genuine preferences.
Adopting a method that allows the valuations obtained to be transferred across
circumstances.
An additional challenge specific to this project was the very short time frame within which all
elements of the work had to be completed.
Previous experience and the focus groups have enabled the production of clearly defined
levels for each factor. A key feature within the SP experiments is the linking of the levels for
each attribute to respondents‟ experiences by identifying the “as now” level which is that
currently experienced in their local environment. Respondents are also asked to rate every
level of each attribute which this will allow the derivation of values for moving along a rating
scale which may be highly policy relevant. These features contribute to the first two and fifth
points above.
The Priority Ranking (PR) approach was designed to deal with a large number of attributes
at one time, thereby contributing to the third point, and partly as a result of this and partly as
a result of masking the purpose of the exercise, it offers a lesser invitation to strategic bias
than in more conventional SP exercises by presenting local environmental factors in a
broader quality of life dimension which places less emphasis on them, thereby contributing
to the fourth point. Further detail on this method and these issues is provided in section
2.4.3, but we note here that reducing incentives to bias is an important feature, clearly most
important where issues are contentious. In our work on aircraft noise, the values derived
from the PR exercise were lower and more believable than those derived from a standard
and transparent Stated Choice exercise.
A common feature of SP experiments and most especially contingent valuation is the
difference between willingness to pay (WTP) and willingness to accept (WTA) which in this
context of environmental public goods may be tenfold (Horowitz and McConnell, 2002). Such
a difference is not compatible with conventional economic theory. In a previous application
to aircraft noise, the results for Manchester and Lyon found a relatively small discrepancy
between WTP and WTA, suggesting that the PR approach, by offering a different form of
trade off from the norm, may reduce incentives to bias in the form of free-riding or
overstatement (Wardman and Bristow 2008). Here losses are traded against losses and
gains against gains, essentially a WTA measure to forgo a gain (equivalent gain) and a
WTP measure to avoid a loss (equivalent loss), in contrast to the more familiar
compensatory format of WTA to suffer a loss (compensating loss) and WTP to achieve a
gain (compensating gain). We note that Bateman et al. (2000) suggest that equivalent gain
and equivalent loss may be the most appropriate measures in precisely the tax versus
services trade-off envisaged here as they could both avoid falsely identifying gain and loss
21
effects that are really based on caution and more accurately reflect the context within which
policy choices are made.
2.4.3 The First „Quality of Life‟ SP Experiment (SP1)
SP1 contains a broad range of factors affecting perceptions of local environmental quality
including a sub-set of the local environmental factors of interest here. SP1 has the capacity
to examine preference over a large range of factors and levels. The approach took
inspiration from the priority evaluator developed by Hoinville (1971) to identify public
preferences in decisions affecting the quality of life. It has been used when there has been a
need to evaluate a large number of variables, such as the many different types of rail rolling
stock and station facility attributes and diverse quality of life issues. However, the
conventional priority evaluator has problems in that the process of allocating a fixed points
budget across attributes variations with different „points prices‟ induces linear-dependency3.
The approach used here is similar in offering a wide range of factors. But instead of using a
budget to purchase improvements from the current situation we ask respondents to identify
their most preferred improvement from a set. The preferred improvement is then eliminated
from the set and the respondent is asked for their preferred improvement from this revised
set and so on until all improvements are ranked in order of preference.
Accommodating such a large number of factors in a conventional choice experiment is
feasible, but the demands placed upon individuals in trying to evaluate two options
characterised by for example the thirteen attributes used in the EUROCONTROL study
(Bristow and Wardman 2006; Wardman and Bristow 2008) or the twelve attributes under
consideration here would be considerable. There is evidence to indicate that task complexity
can influence valuations, largely through the use of simplifying but inappropriate choice rules
or ignoring attributes.
The challenge therefore is to be able to cover a wide range of factors in a single exercise yet
ensure that the task is manageable. To do this an approach has been developed which
involves the evaluation of factor variations one at a time rather than the conventional
procedure of multiple trade-offs. It is thought that if offered a whole series of improvements
(or deteriorations) to specific factors, respondents can more readily state which (one-
dimensional) factor variation they would most like to achieve than they can weigh up the net
benefit of (multi-dimensional) differences in a whole range of factors between two
alternatives.
The ten variables used in this experiment are local crime rates, school pass rates, traffic
congestion, traffic noise at home, air quality, dog fouling, the general condition of roads and
pavements, access to quiet areas, personal security and council tax.
3 This is a statistical property whereby there is an exact linear relationship between the attributes and hence it
would not be possible to estimate parameters indicating the importance of each attribute. Such a property is
built-in with this form of priority evaluator.
22
Here each variable has five levels, except local tax which has seven levels in order to
introduce more variation into this key factor and to allow for uncertainty as to individuals‟
monetary valuations. Council tax was used as the most appropriate payment or
compensation mechanism for local quality of life factors. As the key purpose of this
experiment is to provide a top level value for a sub-set of the local environmental factors in
which we are interested, two of the latter, namely dog fouling and quiet areas, were selected
for inclusion. Table 2.1 illustrates this form of PR exercise as implemented in the pilot
survey. The starting point is to identify the respondent‟s current situation. Respondents were
asked to state their perceived current position where this was reasonable to do so. This is
the case for street traffic noise at home, air quality, dog fouling, general condition of local
roads and pavements, access to quiet areas and personal security.
Where an objective measurable baseline existed, but would not be known with precision by
respondents, this was pre-coded (shaded) on the response sheet. This was done for local
crime rates and school pass rates. This was specified to be the central level around which
there were two improvements and two deteriorations. For the remaining factors of traffic
congestion and council tax, the levels of the factors are centred on the current situation,
marked “as now”.
The respondents were then asked to consider improvements to the established current
situation. These are all the levels of the factors to the right of the current levels. The
respondent was asked to state which improvement would be most preferred from this set.
Initially, this should logically be a factor level in the right hand column. They were then asked
to disregard this improvement, treating it as if it were no longer available, and asked to state
which was now the preferred improvement. This process continued until all the possible
improvements had been ranked in order of preference.
An alternative approach would have been to ask individuals to start at the current situation
and gradually improve it by moving to the right one attribute at a time. In contrast to the
approach that was actually adopted, which essentially involves evaluating the improvements
from the right to the as now situation, this would mean that the reference situation is
continually updated to contain the improvements that have been selected. Whilst this might
be viewed as a simpler procedure for respondents to undertake, there are two problems
facing it. The first, and critical problem, is that this process would build in linear dependency
between the attribute levels, whereupon it would not be possible to analyse the data.
Secondly, it might be that an individual selected an improvement that is not actually
preferred simply to subsequently achieve an improved level of that attribute which was
preferred overall.
Having completed the ranking of improvements, the respondent then proceeded to evaluate
the deteriorations. The deterioration which was regarded to be worst was first identified. As
with the improvements respondents were then asked to disregard this level and identify the
worst from the remaining set and so on until all the deteriorations had been evaluated.
23
Table 2.1: SP1 Pilot Survey: Leeds
Local Crime - Burglaries per 1000 households
42 35 28 21 14
Local Schools - % gaining 5 GCSEs grades A to C
30% 40% 50% 60% 70%
Road Traffic 10% More Traffic
5% More Traffic As Now 5% Less Traffic 10% Less Traffic
Traffic Noise at Home
Extremely Noisy Very Noisy Moderately Noisy
Slightly Noisy Not at all Noisy
Air Quality Very Poor Poor Neither Good nor Poor
Good Very Good
Dog Fouling occurs
Always dog mess in view
Every minute when walking
Every 5 minutes or so when walking
Every 15 minutes or so when
walking
Never or very rarely
General condition of local Roads and Pavements
Very Poor Poor Neither Good nor Poor
Good Very Good
Access to quiet areas
No quiet areas around here
Quiet area within 15
minutes walk of home
Quiet area within 5
minutes walk of home
Quiet area within 1 minutes walk of
home
Quiet area immediately outside the
door.
Personal security Feel very unsafe Feel unsafe Neither safe nor unsafe
Feel safe Feel very safe
Council Tax you would pay
£10 more each
month
£4 more each month
£1 more each
month
As Now £1 less each month
£4 less each
month
£10 less each
month
2.4.4 The Second „Local Environmental Factors‟ SP Experiment (SP2)
SP2 allows a very focused approach to be taken on disaggregated elements of the factors of
interest. We now have 12 factors of interest as light pollution has been divided into two
factors, the first relating to light intrusion in the home and the second to light pollution
obscuring the stars. With the cost factor this becomes 13 in total. The final SP2 design may
be found in Appendix D. The key changes for the pilot from the focus groups were:
24
Visual
Litter was retained at four levels but the images were replaced with photographs
taken for this survey in residential streets in Leeds.
Graffiti increased from three to five levels using photographs taken for this survey in
residential local shopping areas in and around Leeds.
Discarded chewing gum was retained at three levels but using photographs taken for
this survey such that all images showed a similar paved surface.
Grit and dirt moved from four to three levels and using photographs taken for this
survey on similar looking streets in and around Leeds.
Four levels of trees were selected from the six street level images presented to the
focus groups.
Fly-tipping was reduced from five to four levels, using new images of typical fly-
tipping at different magnitudes.
Fly-posting increased from three to five levels using some new and some old but
cropped images.
Verbal
Five levels of quiet area were used based on minutes walking, including no quiet
areas and quiet area immediately outside as the extremes.
The dog fouling scale was based on frequency of occurrence when walking which
seemed to work best and was adopted at five levels.
Odour was specified at five levels based on frequency of occurrence as in the focus
groups.
Light intrusion moved away from images to a verbal scale reflecting the level of
nuisance caused at four levels.
Light pollution moved away from images to a verbal scale based on ability to see
stars at three levels.
Other changes identified in section 2.3 were also made to improve the clarity of presentation
of SP2. A book of photographs was given to each respondent with the images for each
factor on a separate page. This gave a reasonably sized picture for each image.
Respondents rated each level of each attribute on a 0-10 scale, which provides a metric that
is useful for modelling purposes as well as familiarising them with the detail of the attributes
prior to undertaking the SP exercise. The photographs are reproduced in Appendix E.
25
2.4.5 Background and Attitudinal Questions
The first section of the survey contains general questions about the quality of life in the
respondents‟ neighbourhood and includes specific questions on:
The importance of different factors in choosing to live here
Rating the importance of a range of quality of life factors that would later appear in
SP1.
Rating satisfaction with the same range of quality of life factors
Agree/disagree questions on the environment to enable segmentation by
environmental attitudes.
This is immediately followed by SP1 in section 2. Section 3 asks respondents to rate on a 0-
10 scale from bad to good each of the attribute levels that will later appear in SP2. This
allows us to check firstly that the levels we have are discriminable by respondents, secondly
to identify inconsistencies and thirdly to scale monetary values onto a rating scale.
Respondents are then directly asked about the clarity of the presentation and asked to rate
the importance of each factor. SP2 follows in section 4. Section 5 asks some questions
about council performance with respect to these key environmental indicators and the
likelihood that this might change. Section 6 contains questions about the respondent and
their household largely for segmentation purposes. The full questionnaire is set out in
Appendix F.
In terms of the environmental segmentation questions this was based upon an adapted
version of Defra‟s 7 segment model which was developed for Defra by AD Research &
Analysis (2008) using a Combined Block Method (CBM) that involves asking respondents to
answer a set of attitudinal questions and then using an algorithm to allocate respondents‟ to
a unique segment:
1. Positive Greens
2. Waste Watchers
3. Concerned Consumers
4. Sideline Supporters
5. Stalled Starters
6. Honestly Disengaged.
The method was simplified by ITS, University of Leeds and approved by AD Research &
Analysis (2008a) and was asked as part of the questionnaire for this current piece of
research. The list of questions used is outlined in Table 2.2.
26
Table 2.2 Environmental Segmentation Questions
Q5 Please indicated how much you agree or disagree
with the following statements using the scale against
each statement. Please only circle one option for each statement
Strongly
Disagree Disagree Neutral Agree
Strongly
Agree
The effects of climate change are too far in the future to really worry me 1 2 3 4 5
I don‟t pay much attention to the amount of water I use at home 1 2 3 4 5
It's not worth me doing things to help the environment if others don't do the
same 1 2 3 4 5
If things continue on their current course, we will soon experience a major
environmental disaster 1 2 3 4 5
It's only worth doing environmentally-friendly things if they save you money 1 2 3 4 5
People who fly should bear the cost of the environmental damage that air
travel causes 1 2 3 4 5
It's not worth Britain trying to combat climate change because other countries
will just cancel out what we do 1 2 3 4 5
I don't really give much thought to saving energy in my home 1 2 3 4 5
For the sake of the environment, car users should pay higher taxes 1 2 3 4 5
The environment is a low priority for me compared with a lot of other things in
my life 1 2 3 4 5
It takes too much effort to do things that are environmentally friendly 1 2 3 4 5
We are close to the limit of the number of people the earth can support 1 2 3 4 5
I would be prepared to pay more for environmentally-friendly products 1 2 3 4 5
Which of these best describes how you feel about your current lifestyle and
the environment? 1 2 3 4 5
2.5 Pilot Surveys
The pilot surveys took place in Leeds on 22nd January with three groups commencing at
11am, 1pm and 3pm respectively. A total of 21 people participated, with group sizes of 5, 7
and 9 respectively. Recruitment was from different areas of Leeds and was a dry run for the
method to be used in the main survey. Four survey staff participated, each group was lead
by a different member of the survey team as this was in part a training exercise. Abigail
Bristow was present throughout as an observer and to carry out a short de-briefing session
with each group. Jeremy Shires observed two of the groups.
Overall the pilot surveys fulfilled their role, allowing us to identify difficult and/or redundant
questions and areas where greater clarity is required. The key insights were as follows:
27
1. As suspected one hour was insufficient to complete the survey. The first session
was very slow, in the others it was possible to get through everything except the
deteriorations in the first choice experiment. Changes to reduce the time taken and recruiters
should specify one and a half hours for the survey.
2. The question on why you moved to the area took a surprisingly long time to get
through and generated a lot of questions. This question was replaced with an open question
on the 3 best/worst things about where the respondent lived. This would allow us to identify
any of the factors of interest that are mentioned spontaneously (as previously suggested by
Defra).
3. The SP experiment needs a more careful and precise explanation and consistency
between survey staff. A tighter script was provided.
4. It is necessary to avoid any ambiguity as to whether we are asking for the
preferences of the individual or the household. There are arguments both ways:
A household value is useful in this context, especially as the council tax is a
household tax, which some respondents will not be responsible for.
An individual value may be easier for the respondent.
We opted for an individual value on the grounds of simplicity but did ask who paid the
council tax.
5. We found that the first SP caused the most difficulty but the second was much easier
as people had grasped the principle. As the first SP is primarily for masking and establishing
top level values we reduced the number of attributes to simplify it.
6. There was a general feeling that council tax reductions were not plausible and that
reductions in tax could mean cuts in services. This was revealed in the debriefing and
occasionally during the sessions.
Nevertheless, observation of the responses reveals that there was trading in the sample, in
the sense that respondents did not simply rank one attribute change then turn to the ranking
of another but there is a spread across attribute changes. People were considering their
responses and looking at different factors. However, it appeared that a lower priority was
given to council tax reductions and some did not trade on the cost variable. Some who do
trade do so only at the £10 level and hence the monetary scale needed to be stretched.
A better explanation was required, along the lines of “we know that council tax is more likely
to go up than down and that reductions may seem unlikely but for today please consider that
tax reductions are as likely to happen as increases”.
28
A question was added at the end of SP2 exercise about whether the respondent found the
tax reductions believable. This replaced the questions on council performance and how likely
things are to change.
7. A number of issues surrounded the presentation of levels in SP2.
The presentations was fairly clear and unambiguous with the exception of grit and
dirt.
It was difficult to see each photograph when the SP exercise was reproduced on a
single size of A4. Hence the layout was increased to A3.
More space between attributes would make presentation a little clearer
Having a clear space in which to write the number would be useful
In section 3 we numbered the pictures consecutively rather than as 1,2,3 for each factor, in
response to respondents‟ requests.
In the debrief it was apparent that grit and dirt was the least clear factor, with grit and dirt,
light intrusion and light pollution deemed least important.
9. The detailed questions on job type and income at the end were often left incomplete.
This was subsequently simplified.
10. There were clearly problems with recruitment. We had groups of people who knew
each other in every session. For example, three women were walking together and recruited
on street together. Recruitment must be on a more random basis. The letter participants are
given by the recruitment agency must request that they arrive at a specific time and that they
bring reading glasses if these are needed.
2.6 Final Survey Design
2.6.1 Section 1: Quality of Life
The questions on why you moved to the area and the satisfaction ratings of the quality of life
attributes were dropped due to time constraints and a lack of centrality to the objective of the
survey. This section has basic facts about where the respondent lives, then asks for the
three best and worst things about living there. This is followed by some agree/disagree
environmental statements for segmentation purposes and finally importance ratings for all
the factors that will appear in the SP1 quality of life experiment.
2.6.2 Section 2: SP1 „Quality of Life‟ Experiment
As a result of the pilot survey, the personal security and air quality attributes were dropped
from the quality of life SP to reduce effort. The explanation was improved and clarified and a
statement about treating tax increases and decreases as just as likely added.
29
The example in Table 2.3 is for Southwark, one of the areas on the main survey, and
contains the mid-range of cost levels. Three sets were finally used to represent a broad
range of possibilities: these were £1, £5 and £12; £2, £8 and £15 and £3, £10 and £20.
Table 2.3 Quality of life SP (SP1) Southwark
Worse Situation Better Situation
Local Crime -
Burglaries per
1000
households
17 13 9 5 2
Local Schools -
% gaining 5
GCSEs grades
A to C
40% 50% 60% 70% 80%
Traffic Noise at
Home
Extremely
Noisy
Very Noisy Moderately
Noisy
Slightly Noisy Not at all Noisy
Road Traffic 10% More
Traffic
5% More
Traffic
As Now 5% Less
Traffic
10% Less
Traffic
Dog Fouling
occurs
Always dog
mess in view
Every
minute
when
walking
Every 5
minutes or so
when walking
Every 15
minutes or so
when walking
Never or very
rarely
General
condition of
Pavements
Very Poor Poor Neither Good
nor Poor
Good Very Good
Access to quiet
areas
No quiet areas
around here
Quiet area
within 15
minutes
walk of
home
Quiet area
within 10
minutes walk
of home
Quiet area
within 5
minutes walk
of home
Quiet area
within a 1
minute walk of
home
Council Tax
you would pay
£15 more
each
month
£8 more
each
month
£2
more
each
month
As Now £2 less
each
month
£8
less each
month
£15
less each
month
2.6.3 Section 3: Ratings
The verbal scales or photographs were revised given the changes in the specification of the
attributes and the number of levels of each. In addition to rating each photograph or verbal
30
description on a 0-10 scale, respondents are also asked about clarity of presentation and to
rate the importance of each attribute.
The main change here was the decision to omit detritus (or grit and dirt) from further
consideration. In both the focus groups and the pilot it was seen as the least important factor
and the most difficult to interpret. Moreover, it is more of a technical indicator for
measurement issues than a perceptual indicator, and Keep Britain Tidy confirmed that this
factor was less appropriate for this type of exercise.
2.6.4 Section 4: SP2 „Local Environmental Factors‟ SP Experiment
The experiment is reproduced in Appendix D and is similar to SP1 but instead contains more
attributes and uses a mix of photographs and text to describe them. The full list of attributes
is now outlined, along with the description offered where a verbal presentation is used:
Verbal
1. Light Pollution (on a clear night)
I can’t see the stars
I can see some stars
I can see many stars
2. Light Intrusion (at night)
Light intrusion that affects my sleep or that of someone else in my household
Light intrusion that I can’t block out with heavy curtains but doesn’t affect my sleep
Light intrusion into my home that I can block out with heavy curtains
No light intrusion from any source
3. Access to Quiet Areas
No quiet areas around here
Quiet area within 15 minutes walk of home
Quiet area within 10 minutes walk of home
Quiet area within 5 minutes walk of home
Quiet areas within a 1 minute walk of home
4. Odour
Bad smells all the time
Bad smells occur weekly
Bad smells occur every month or so
Bad smells occur once or twice a year
No bad smells
5. Dog Fouling (occurs)
Always dog mess in view
Every minute when walking
Every 5 minutes or so when walking
Every 15 minutes or so when walking
Never or rarely
31
Photographs
6. Discarded Chewing Gum (3 photographs)
7. Litter (4 photographs)
8. Trees (4 photographs)
9. Fly-tipping (4 photographs)
10. Graffiti (5 photographs)
11. Fly-posting (5 photographs)
The photographs used for each level are reproduced in Appendix D.
Two debriefing questions were asked after the SP experiment. The first on what the
respondents thought of the changes in council tax was aimed at identifying any who did not
believe these would happen and so did not consider them. The second was aimed at those
who did not complete the SP and asked why. These replaced questions about the
performance of the local council which were seen as less relevant.
2.6.5 Section 5: About You and Your Household
This section was broadly the same as in the pilot but with some simplification of the
occupation classes and a reduction in the number of income groups.
2.7 Conclusions
The survey design process, including focus groups and piloting, was completed in less than
one month of elapsed time. This was a very intense process and, as has been illustrated,
contributed a great deal to our understanding of the attributes and the final design.
32
3. SURVEY IMPLEMENTATION AND SAMPLE CHARACTERISTICS 3.1 Introduction
This chapter outlines the process by which the survey locations were selected, as set out in
section 3.2, before moving on to the implementation of the surveys in section 3.3. Section
3.4 presents key socio-economic characteristics of the sample. The next three sections
present figures relating to respondents‟ perceptions of local environmental quality. Section
3.5 reports the relative importance of different factors to respondents. Section 3.6 indicates
the average score for each level of each factor that was presented as either a photograph or
on a verbal scale and was rated on a 0-10 bad to good scale. Section 3.7 summarises the
reported as now situations for each of the local environmental factors.
3.2 Area Selection
Before recruitment could commence three general locations and nine sub-locations had to
be identified and agreed upon. In the proposal, ITS had suggested that one of the general
survey locations would be a city located in the north of England, the second would be based
in or around London, whilst the third would probably be located in the Midlands. After
discussion with the client, it was agreed that the major locations would be:
Manchester
London
Coventry
The ITS proposal had suggested that for each general location three sub-locations should be
identified which reflected different area types, namely, inner-city, suburban and rural/semi-
rural. The client wished ITS to pursue this sampling strategy on the understanding that the
aggregated sample for each general location reflected national population statistics.
In order to achieve this, a number of potential sub-locations were identified for each general
location along with a number of key socio-economic and demographic characteristics (as
taken from the latest census data) for each sub-location. These were:
a) Gender
b) Age
18-29 yrs
30-44 yrs
45-59 yrs
60+
c) Socio-economic classification
Large employers (1), higher managerial occupations (2) and higher
professional occupations (3) = A
Lower managerial and professional occupations (4) with intermediate
occupations (5) and small employers and own account workers (6) = B
33
Lower supervisory and technical occupations (7) with semi routine
occupations (8) and routine occupations (9) = C
Long term unemployed = D
OAPs = E
Never worked = F
The national (English) averages for the above characteristics are presented in Table 3.1.
Table 3.1: Socio-economic Characteristics of England
% of 16-74 % of Total Population Aged
Males Females 18-29 30-44 45-59 60+
49% 51% 15% 23% 19% 21%
Socio Economic Categories
A B C D E F
9% 35% 28% 1% 18% 10%
After consultation with the client the following sets of sub locations were agreed upon:
Manchester
Wythenshawe (inner city sub location)
Heald Green (suburban location)
Chelford (rural/semi rural location)
London
Peckham (inner city location)
Nunhead (suburban location)
East Dulwich (suburban location)
Coventry:
Holbrooks (inner city location)
Kenilworth (suburban location)
Bulkington (rural/semi rural location)
The client was satisfied that these locations were representative of the national average. It
is worth noting that no rural/semi-rural location was identified for London. This was done in
agreement with the client who felt that two rural/semi rural locations across the whole survey
was sufficient to reflect the current distribution of the population in England across the three
area types identified.
3.3 Survey Implementation
The main survey took just over three week with initial recruitment for the first location taking
place on 22nd January and the final hall tests taking place on 13th February. Three general
locations were surveyed: Manchester, Coventry and London. Within each location three
34
sub-locations were chosen to be surveyed to reflect three different types of area: (a) inner-
city; (b) suburban; and (c) rural/semi rural.
The methodology required the recruitment of participants to take part in supervised „hall
tests‟. The hall test would take participants around 1 hour and 15 minutes to complete, in
return for which they would receive a cash incentive of £25.
Responsibility for the recruitment was tendered out to four market research firms. The
successful bidder was UKFS, a market research firm based in Leamington Spa. Their
methodology was based upon house to house recruitment within a pre-defined location as
specified by ITS. Responsibility for conducting the hall test was ITS‟s solely.
Once the nine sub locations had been identified UKFS could be instructed to carry out the
recruitment. UKFS were provided with boundaries of the area to recruit from along with the
following quotas to be met:
Age Profile
18-30 - around 30% - so no more than 35%
40-59 - around 45% - so no more than 50%
60+ - around 25% - so no more than 30%
Employment Status
Employed (full or part) and Self Employed - No less than 70%
Retired/home maker with no children in household – No more than 20%
Unemployed/homemaker with children at home or other – No more than 10%.
Gender – equal split
In order to ensure that participants reflected the national average it was important that the
participants were recruited randomly across each sub-location. As such UKFS were issued
with the following guidelines:
The sample should be drawn from across the area within each location and not
focused in particular streets or clusters of streets to ensure a random sample.
Only one person per household should be recruited.
Only the person that the recruiter has spoken to can be recruited.
To help ensure this, UKFS produced a series of „blown up‟ maps that covered small sections
of each sub location. Recruitment was then undertaken in each of these small sections to
ensure that the entire area was covered. In addition, UKFS undertook recruitment at times
when participants were likely to be at their homes, such as weekends and the late
afternoon/evenings. In order to allow for „no shows‟ at the hall tests UKFS were asked to
over recruit by 25%, i.e. 750 participants.
In order to minimise the number of „no shows‟ UKFS recruited participants the week prior to
the hall tests at each location so that the date of the hall test was „fresh‟ in people‟s mind and
the likelihood of other arrangements being made was minimised. As such, recruitment took
35
place on the dates for each locations as set out in Table 3.2, with final participant lists
provided to ITS the evening before the survey to allow ITS hall test staff to control
attendance. The number of participants included in the final analysis (561) was lower than
the target of 600, mainly due to low turn outs at Bulkington (atrocious weather) and
particularly Nunhead (hall test unavoidably located on the outskirts of the sub location).
Table 3.2: Hall Test Dates and Locations
General Location
- Sub Location
Recruitment Dates Hall Test
Dates
Numbers
Recruited
Completed
Questionnaires
Manchester –
Heald Green
22nd January till
27th January
29th January 84 77
Manchester –
Chelford
22nd January till
27th January
30th January 87 73
Manchester –
Wythenshawe
29th January till 3rd
February
5th February 84 73
London – Peckham 5th February till 10th
February
12th February 79 59
London - Nunhead 5th February till 10th
February
12th and 13th
February
79 37
London – East
Dulwich
5th February till 10th
February
12th February 83 61
Coventry –
Holbrooks
5th February till 9th
February
10th February 97 67
Coventry –
Bulkington
5th February till 10th
February
12th February 86 54
Coventry -
Kenilworth
5th February till 10th
February
13th February 84 60
The methodology for the hall tests was based upon one which had been used to good effect
for a previous ITS study in the summer of 2010. That itself was an exact repeat of a 2002
study. As such, ITS survey staff were very familiar with the processes to be employed and
the pitfalls to be avoided. In particular, we had improved the guidance on the SP exercises
and were aware of problems to look out for. As a result, we experienced much fewer
incomplete rankings and the improved completion rate continued in this survey.
The hall tests involved taking a group of participants (between 20 and 25) through each
section of the questionnaire. Three hall tests were carried out at each sub location with
venues being chosen that had enough space to accommodate up to 30 participants and
which were well known by the local communities in order to reduce non attendance. A list of
the venues used is outlined in Table 3.3.
36
Table 3.3: Hall Test Venues
Sub Location Venue
Heald Green The Unity Café
Chelford Chelford Village Hall
Wythenshawe The Forum Library
Peckham Peckham Library
Nunhead Peckham Library
East Dulwich The Barnabas Centre
Holbrooks Holbrooks Community Centre
Bulkington Bulkington Club for Young People
Kenilworth Kenilworth Community Centre
Each hall test followed the same set of procedures. After signing in, each participant was
given an information letter and asked to sign a consent form whilst waiting for the session to
begin. Once all the participants were present the survey team leader would welcome
everyone to the hall test, run through health and safety procedures and explain how the
session would be run. During the introduction and throughout the session participants were
told that the study was concerned with quality of life issues in their local neighbourhoods.
The participants were then taken through the questionnaire as five separate stages, with
each stage corresponding to one of the sections of the questionnaire. Each stage was
introduced and, where appropriate (for example section 2 and 4), explained in detail by the
survey team leader. Participants where then asked to fully complete the section in front of
them and were made aware that survey staff (three in total) would be circulating around the
room in case any further explanation was required. The survey team leader would have an
indicative time allowance for each section and would update participants as to how long they
had left to complete the section. Once the allowed time was finished the survey team leader
would ask respondents to finish and to put the completed section in a brown envelope in
front of them. The next section would then be distributed to participants and the same
process followed.
Sections 2 and 4 required detailed explanation since they involved the most complex
exercises, namely SP1 and SP2. In order to facilitate this, the survey team leader took
participants through a similar version of the SP but for an area in Leeds (Lawnswood) in 3
stages. The first stage asked participants to highlight what their „as now‟ boxes were for
each row were no shaded boxes existed. The second stage saw respondents rank all the
possible improvements to their „as now‟ situation (i.e. all the boxes to the right of the „as
nows‟) in order of preferred priority. Finally the third stage asked respondents to rank all the
possible deteriorations (i.e. all the boxes to the left of the „as nows‟) in order of the worst
possible deteriorations that could take place. The survey team leaders were given extensive
training prior to conducting the hall tests to ensure that they explained each stage clearly
and without undue influence on participants‟ responses.
37
Particular attention was paid to ensuring that the participants realised that
They were answering for themselves and not their households.
That any changes in tax would be for themselves to receive or pay personally.
That the local authority would be able to make the changes described and that the
changes in council tax would not impact on any other local services, just those in the
table.
A problem with some SP studies is that there is an ambiguity as to whether the valuation
relates to the individual or the household and clearly such an undesirable feature has to be
avoided.
When the final section had been completed, and if time permitted, the participants were
offered the chance to complete any sections still remaining. An incentive form was then
distributed to the participants to sign and take, along with the brown envelope containing
their responses, to the survey team leader who paid them their incentive and asked them to
sign out from the session. Each envelope was then sealed and allocated an ID number so
that for data coding purposes it was clearly identifiable.
3.4 Sample Characteristics
The following tables outline the characteristics of our sample of 561 respondents. It can be
seen in Table 3.4 that the sample is balanced in favour of females although little different
from the census, whilst Table 3.5 reflects an age profile broadly in line with the census.
Table 3.4: Gender
Gender % of sample England 16-74
Male 45.1 49
Female 53.5 51
Missing 1.4
Table 3.5: Age
Age
Category
% of sample England
18-24 14.4 19
25-29 10.7
30-44 28.0 30
45-59 29.2 24
60-64 8.0
27 65-74 6.2
75+ 2.7
Missing 0.7
38
Similarly, annual household income is well profiled (Table 3.6) and would appear to reflect
the national picture with the exception that we have under-represented the wealthiest in
society. In terms of who pays the council tax, Table 3.7 indicates that around 70% of the
sample either paid or made some contribution to it.
Table 3.6: Annual Household Income
HH Income
Categories
% of sample England
2009
<£10,000 11.8 10
£10-19,999 19.1 27
£20-29,999 15.0 18
£30-39,999 11.2 14
£40-49,999 6.8 9
£50,000+ 9.8 20
Don‟t know 13.0
Missing 13.4
Source: Family Resources Survey May 2011. Department of Works and Pensions
Table 3.7: Who Pays the Council Tax?
Person % of sample
Me 40.1
Me jointly with others 29.9
Someone else 27.6
Missing 2.4
Table 3.8 reveals that around 9.5% of the participants belonged to an environmental
organisation whilst in Table 3.9 the majority of respondents lived in semi-detached
accommodation (35%). The distribution of type of home in the sample is similar to the
national distribution with the exception that those in detached homes form too small a
proportion of our sample.
Table 3.8: Do You Belong to an Environmental Organisation?
Yes/No % of sample
Yes 9.4
No 84.7
Missing 5.9
39
Table 3.9: Type of Home
Type of Home % of sample England
Flat/Apartment 16.2 17
Terraced House 30.7 28
Semi-detached House 35.1 32
Detached House 13.9 22
Other 2.9
Missing 1.2
Source: Communities and Local Government (2011) Housing Table 1.17
In terms of the monthly household council tax payment, there is quite a large spread, with
the largest category being those paying between £100 and £125 (24%). 7.5% paid less than
£80 per month. Our sample reflects the national situation well except for the low proportion
in the lowest council tax bracket.
Table 3.10: Monthly Household Council Tax
Type of Home % of sample England
<£80 7.5 £79 25
£80-99.99 15.3 £93 20
£100-124.99 24.1 £106 22
£125-149.99 13.9 £119 15
£150-174.99 4.5 £146 9
£175-199.99 4.5 £173 5
£200+ 5.0 £199 4
Don‟t Know/Paid for me 20.9 £239 1
Missing 4.5
Source: Communities and Local Government Statistical Release. Council Tax Levels Set By
Local Authorities in England 2011-12
Finally, with regards to employment status and occupation classification a substantial
percentage of participants did not provide information on occupation (40%), making it difficult
to draw many conclusions from the data, shown in Table 3.11.
40
Table 3.11: Employment Status and Occupation Classification (% of respondents)
Employment Status
Occupation Class
Missing Managerial Professional Supervisory & Technical
Semi-skilled or Skilled Manual
Clerical Other
Missing 5.0 0.4 2.1 0.4 0.5 0.5 0.4
Run Own Company
2.3 0.0 0.5 0.0 0.0 0.0 0.0
Self Employed
0.7 0.9 3.9 0.5 1.2 1.2 0.0
Employed Full-time
1.4 2.0 8.6 2.9 4.6 4.6 4.3
Employed Part-time
0.9 0.9 4.1 1.1 3.4 3.4 3.2
In part/full time Education
2.0 0.0 0.2 0.0 0.0 0.0 0.0
Retired 13.4 0.0 0.0 0.2 0.0 0.0 0.2
Home Maker 5.2 0.0 0.0 0.2 0.0 0.0 0.0
Not Working at Present
7.8 0.0 0.0 0.0 0.2 0.2 0.0
Other 1.2 0.0 0.0 0.0 0.2 0.2 0.0 Total 39.9 4.1 19.4 5.2 10.2 10.2 8.0
In terms of employment status alone it is possible to see how closely the sample represents
the quotas set for the recruitment company. The quotas are outlined below with the actual
figures (discounted for missing) in brackets:
a. Employed (full or part) and Self Employed - No less than 70% (63%)
b. Retired/home maker with no children in household – No more than 20%
(22%)
c. Unemployed/homemaker with children at home or other – No more than 10%
(15%)
It would appear that the recruitment under sampled with regards to employment/self
employed and over sampled with regards the last two categories. The overall sample size
however is sufficiently large to enable segmentation of values by each of the relevant
cohorts and the disparities are not major.
To summarise the representativeness of the sample it is useful to reproduce a version of
Table 3.1 that outlined the national average characteristics in terms of gender, age and
socio- economic categories. In this new Table (Table 3.12) the figures not in brackets reflect
the national average and figures in brackets the sample.
41
For gender it can be seen that males are slightly under represented in the sample but not
sufficiently so to cause concern. With regards to age, because the survey excluded those
over 18 years of age, the national population figures have been adjusted to take this into
account. The age characteristics of the sample over represent the 18-44 and 45-59 year
age categories and under represent quite significantly those aged 60+years. Thus it will be
important to segment by age group to be able to allow for this mis-representation of the
population in drawing up any average valuations.
Unfortunately it is difficult to derive corresponding tables for socio-economic class due to the
difficulties apparent in Table 3.11. However, if we take the broad categories of employed,
OAPs and other from Table 3.11 and apply these to the socio-economic categories it is clear
that the sample under represents the employed population and over represents OAPs and
Other (Table 3.12).
Table 3.12: Socio-economic Characteristics of England and the Sample
% of 16-74 % of Total Population Aged
Males Females 18-44 45-59 60+
49% (45.1%) 51% (53.5%) 48% (53.1%) 24% (29.2%) 27% (16.9%)
Socio Economic Categories
A B C D E F
9% 35% 28% 1% 18% 10%
Employed OAPs Other
72% (63%) 18% (22%) 11% (15%)
3.5 Ratings of SP Attribute
Table 3.13 shows the importance ratings for the range of factors in the first quality of life SP
(SP1). The lower the rating then the more important the factor. Dog fouling is ranked fourth
equal and quiet areas sixth equal out of the ten factors although there is actually very little
difference in the rating of the two. Local crime is by far the most important factor. The
proportion of individuals who did not provide a rating is very similar across attributes at
around 5%.
Table 3.14 shows respondents‟ ratings of the importance of the different environmental
factors in the second SP (SP2). Clearly litter, dog fouling and fly-tipping are the most
important and equally clearly fly-posting, light pollution and light intrusion are the least
important. The proportion of individuals not providing ratings is higher, at around 11%, than
for the first set of quality of life factors. The difference between dog fouling and access to
quiet areas is now larger than for the first SP exercise.
42
Table 3.13: Importance Ratings for SP1
Extremely Important
Very Important
Moderately Important
Slightly Important
Not at all Important Missing Average
Standard Deviation
Code 1 2 3 4 5
Quality of Local Schools
Number 257 151 52 36 40 26 1.98 1.23
% 45.7 26.9 9.3 6.4 7.1 4.6
The Amount of Road Traffic in Your Area
Number 172 232 104 20 9 25 2.00 0.90
% 30.6 41.3 18.5 3.6 1.6 4.4
Road Traffic Noise Experienced at Home
Number 148 194 123 37 30 30 2.26 1.11
% 5.3 26.3 34.5 21.9 6.6 5.3
Neighbourhood Air Quality
Number 181 200 107 34 10 30 2.05 0.98
% 32.2 35.6 19.0 6.0 1.8 5.3
Condition of Roads and Pavements
Number 215 191 102 19 4 31 1.88 0.89
% 38.3 34.0 18.1 3.4 0.7 5.5 Level of Local Crime Number 361 119 39 11 3 29 1.45 0.77
% 64.2 21.2 6.9 2.0 0.5 5.2 Access to Quiet Areas Number 197 190 107 29 10 29 2.00 0.98
% 35.1 33.8 19.0 5.2 1.8 5.2 Amount of Local
Council Tax Number 225 166 107 24 10 30 1.92 0.99
% 40.0 29.5 19.0 4.3 1.8 5.3 Level of Dog Fouling Number 231 160 90 33 21 27 1.98 1.10
% 41.1 28.5 16.0 5.9 3.7 4.8
43
Table 3.14: Importance Ratings for SP2
Extremely Important
Very Important
Moderately Important
Slightly Important
Not at all Important Missing Average
Standard Deviation
Code 1 2 3 4 5
Litter Number 267 158 47 7 4 61 1.60 0.79
% 49.08 29.04 8.64 1.29 0.74 11.21
Graffiti Number 182 150 95 34 23 60 2.10 1.13
% 33.46 27.57 17.46 6.25 4.23 11.03
Fly-posting Number 148 117 127 54 35 63 2.40 1.23
% 27.21 21.51 23.35 9.93 6.43 11.58
Dog fouling Number 281 124 51 15 12 61 1.66 0.96
% 51.65 22.79 9.38 2.76 2.21 11.21
Chewing Gum Number 180 149 94 35 18 68 2.08 1.10
% 33.09 27.39 17.28 6.43 3.31 12.50
Fly-Tipping Number 280 124 43 18 12 67 1.65 0.97
% 51.47 22.79 7.90 3.31 2.21 12.32
Presence of Trees Number 229 109 94 37 14 61 1.96 1.11
% 42.10 20.04 17.28 6.80 2.57 11.21
Access to Quiet Areas
Number 221 141 88 20 10 64 1.87 0.99
% 40.63 25.92 16.18 3.68 1.84 11.76
Odour Number 177 158 92 42 11 64 2.07 1.06
% 32.54 29.04 16.91 7.72 2.02 11.76
Light Intrusion Number 141 137 130 51 26 59 2.35 1.16
% 25.92 25.18 23.90 9.38 4.78 10.85
Light Pollution Number 128 144 132 55 24 61 2.39 1.14
% 23.53 26.47 24.26 10.11 4.41 11.21
44
3.6 Rating of the Images and Descriptions used in SP2
Table 3.15 reports the mean ratings and associated standard deviations for each level of
each local environmental factor used in the second SP exercise. This is split by the specific
location within an area. Each attribute was rated on a scale of 0 to 10, with a higher number
denoting a better perception.
Higher levels of attributes are better and hence should be associated with higher ratings. In
almost all cases, the relationship between the ratings and the levels takes the expected
form, and the pattern does not vary greatly across locations. It is not uncommon to note
relatively minor variation in the mean rating across levels with a relatively large increase for
the movement to the best level.
The exceptions to the expected relationship all occur in the rural setting; for dog-fouling, light
intrusion and light pollution. In the latter two cases, there is a monotonic effect apparent but
of the reverse form to that which might be expected. It may be that in a rural setting, light
provides benefits in terms of safe navigation and security and hence, at least by some, it is
welcomed.
45
Table 3.15: Ratings for Each Level of the Local Environmental Factors
Inner Suburban Rural
Mean SD Mean SD Mean SD
GUM Level 1 0.81 1.419 0.87 1.532 0.70 1.246
GUM Level 2 4.31 2.038 4.25 1.997 3.86 1.993
GUM Level 3 9.07 2.016 9.00 1.897 9.20 1.658
LITTER Level 1 0.73 1.405 0.90 1.418 0.63 1.329
LITTER Level 2 2.31 1.695 2.34 1.698 2.23 1.728
LITTER Level 3 3.88 2.083 3.57 2.096 3.26 2.126
LITTER Level 4 9.23 1.656 9.27 1.647 9.63 0.958
TREES Level 1 4.09 3.173 3.11 2.951 3.35 2.838
TREES Level 2 5.52 2.257 5.66 2.034 5.46 2.225
TREES Level 3 6.24 2.298 6.64 1.875 6.36 2.131
TREES Level 4 7.60 2.918 8.25 2.556 8.08 2.692
FLY TIP Level 1 0.35 1.186 0.29 0.932 0.23 0.939
FLY TIP Level 2 0.58 1.352 0.56 1.125 0.35 1.014
FLY TIP Level 3 2.04 2.210 1.87 1.856 1.82 1.875
FLY TIP Level 4 9.07 2.067 9.21 1.672 9.43 1.831
GRAFFITI Level 1 1.73 1.750 1.88 1.947 1.40 1.540
GRAFFITI Level 2 2.44 1.823 2.69 1.947 2.02 1.849
GRAFFITI Level 3 3.08 2.133 3.51 2.053 3.09 2.308
GRAFFITI Level 4 5.12 2.406 5.19 2.489 4.32 2.514
GRAFFITI Level 5 8.11 2.603 8.70 1.917 9.00 1.753
FLY POST Level 1 2.01 2.320 2.10 2.301 1.91 2.193
FLY POST Level 2 2.70 2.389 2.50 2.070 2.29 1.938
FLY POST Level 3 3.50 2.282 3.67 2.053 3.50 2.062
FLY POST Level 4 4.73 2.356 5.04 2.122 4.82 2.305
FLY POST Level 5 8.21 2.449 8.50 2.045 8.87 1.809
QUIET Level 1 2.49 3.022 1.62 2.612 2.60 3.564
QUIET Level 2 4.80 3.055 4.57 2.791 4.93 3.398
QUIET Level 3 5.36 2.690 5.71 2.597 5.63 3.004
QUIET Level 4 6.01 3.000 7.12 2.592 7.08 2.659
QUIET Level 5 6.69 3.760 7.90 3.138 8.39 2.781
DOG FOUL Level 1 1.71 2.839 1.29 2.725 2.09 4.283
DOG FOUL Level 2 2.23 2.647 1.86 2.568 4.24 3.180
DOG FOUL Level 3 2.76 2.397 2.72 2.407 3.43 2.616
DOG FOUL Level 4 4.13 2.734 4.15 2.764 3.10 2.960
DOG FOUL Level 5 7.50 3.344 8.08 2.991 3.74 4.332
ODOUR Level 1 2.55 3.261 1.92 2.960 2.54 3.565
ODOUR Level 2 3.26 2.865 2.98 2.584 3.31 2.830
ODOUR Level 3 4.20 2.800 4.08 2.531 4.54 2.579
ODOUR Level 4 5.89 2.723 5.99 2.686 5.74 2.749
ODOUR Level 5 8.14 2.917 8.76 2.315 8.39 2.692
INTRUSION Level 1 3.24 3.499 2.68 3.466 6.84 3.812
INTRUSION Level 2 4.11 3.294 3.82 2.931 5.57 3.025
INTRUSION Level 3 5.55 2.793 5.39 2.466 4.86 3.047
INTRUSION Level 4 7.82 2.853 7.96 2.761 4.62 4.179
POLLUTION Level 1 3.26 3.437 2.61 3.209 6.85 4.055
POLLUTION Level 2 5.40 2.784 5.53 2.575 6.60 2.868
POLLUTION Level 3 8.06 2.749 8.62 2.178 5.03 4.336
46
3.7 Respondents “As Now” Situations
The final series of tables outlines the distribution of the „as now‟ situation with regards to
each of the factors considered within SP2. It should be noted that the worst possible „as
now‟ equates to level 1 and improves the higher the level. So taking light pollution as an
example, level 1 equates to „I can‟t see the stars‟ whilst level 3 equates to „I can see many
stars‟.
The results for each of the factors have been disaggregated by area type and for most of the
factors it is expected that the „as nows‟ for rural will tend to have the highest levels whilst the
inner-city areas will tend to have lowest levels. The tables show the percentage of the
sample in each location that have an as now situation in each category.
Table 3.16: Light Pollution - As Now (%)
Level Inner Suburban Rural Total
1 6.5 5.1 2.4 5.0
2 67.3 67.2 22.8 57.2
3 26.1 27.7 74.8 37.8
Total 100.0 100.0 100.0 100.0
For light pollution there is little variation between inner-city and suburban areas with the
majority in each case able to see some stars. In rural areas, nearly three quarters of the
sample can see many stars as would be expected in a less built up area.
Table 3.17: Discarded Chewing Gum - As Now (%)
Level Inner Suburban Rural Total
1 22.1 7.7 0.8 11.2
2 60.8 65.1 34.6 56.7
3 17.1 27.2 64.6 32.1
Total 100.0 100.0 100.0 100.0
About a fifth of inner-city dwellers categorise themselves as being in the worst category for
discarded chewing gum, falling to around 8% in suburban areas. In rural areas, as might be
expected, the majority experience no discarded chewing gum and very few experience the
worst level.
47
Table 3.18: Litter - As Now (%)
Level Inner Suburban Rural Total
1 21.1 6.8 2.4 10.9
2 32.7 19.1 11.0 22.1
3 40.7 54.9 52.8 49.4
4 5.5 19.1 33.9 17.6
Total 100.0 100.0 100.0 100.0
The pattern for litter reflects closely the pattern for discarded chewing gum, with the worst
experiences occurring in inner-city areas.
Table 3.19: Light Intrusion at Night - As Now (%)
Level Inner Suburban Rural Total
1 5.0 3.8 3.9 4.3
2 14.6 12.8 7.1 12.1
3 54.3 61.7 48.8 56.1
4 26.1 21.7 40.2 27.5
Total 100.0 100.0 100.0 100.0
Light intrusion, like light pollution, is a similar issue for both inner-city and suburban areas.
The problem is much less for rural areas but is still experienced to some degree by nearly
60% of the respondents from that area type.
Table 3.20: Trees - As Now (%)
Level Inner Suburban Rural Total
1 28.1 17.4 7.9 19.1
2 25.1 26.8 17.3 24.1
3 36.7 40.0 36.2 38.0
4 10.1 15.7 38.6 18.9
Total 100.0 100.0 100.0 100.0
In terms of trees, those in suburban areas are more likely to have trees around them at
some level than in inner-city areas. Again this is to be expected given the, typically, lower
density of the built environment in suburban areas. As expected, those in rural areas
experience trees more frequently.
48
Table 3.21: Fly Tipping - As Now (%)
Level Inner Suburban Rural Total
1 5.5 1.3 1.6 2.9
2 8.0 4.7 4.7 5.9
3 51.3 32.3 26.8 37.8
4 35.2 61.7 66.9 53.5
Total 100.0 100.0 100.0 100.0
The „as now‟ results for fly-tipping are somewhat counter to existing thinking and contrary to
the messages that came out of the focus groups, which saw fly tipping as very much a rural
problem, occurring in isolated locations where the chance of being caught was minimal. The
results here suggest that the problem is much more prevalent in inner-city areas, whilst
suburban and rural areas experience the problem to a lesser extent.
Table 3.22: Access to Quiet Areas - As Now (%)
Level Inner Suburban Rural Total
1 11.6 3.8 0.8 5.9
2 29.6 12.3 3.9 16.6
3 24.6 25.5 11.8 22.1
4 24.1 43.4 29.1 33.3
5 10.1 14.9 54.3 22.1
Total 100.0 100.0 100.0 100.0
As expected access to quiet areas is more readily found in rural areas and less available in
inner-city areas. However, even here around 60% of the sample report a quiet area within a
ten minute walk.
Similar expected patterns emerged for graffiti, odour, fly-posting and dog fouling, with the
latter emerging as a particular problem for inner-city areas.
Table 3.23: Graffiti - As Now (%)
Level Inner Suburban Rural Total
1 17.1 5.5 2.4 8.9
2 30.7 17.0 8.7 20.0
3 21.6 19.1 8.7 17.6
4 20.1 35.7 33.1 29.6
5 10.6 22.6 47.2 23.9
Total 100.0 100.0 100.0 100.0
49
Table 3.24: Odour - As Now (%)
Level Inner Suburban Rural Total
1 6.0 2.6 0.0 3.2
2 12.6 5.5 16.5 10.5
3 19.6 9.8 16.5 14.8
4 23.1 37.0 41.7 33.2
5 38.7 45.1 25.2 38.3
Total 100.0 100.0 100.0 100.0
Table 3.25: Fly-Posting - As Now (%)
Level Inner Suburban Rural Total
1 7.0 2.1 0.0 3.4
2 10.6 6.4 0.8 6.6
3 31.7 20.4 10.2 22.1
4 35.2 46.0 61.4 45.6
5 15.6 25.1 27.6 22.3
Total 100.0 100.0 100.0 100.0
Table 3.26: Dog Fouling Occurs - As Now (%)
Level Inner Suburban Rural Total
1 21.1 7.2 4.7 11.6
2 16.6 6.0 6.3 9.8
3 34.2 25.5 19.7 27.3
4 19.1 42.6 37.0 33.0
5 9.0 18.7 32.3 18.4
Total 100.0 100.0 100.0 100.0
3.8 Conclusions
The surveys were successfully implemented in nine areas within three cities in January 2011
and a large sample was achieved. The ratings show a high level of consistency between
inner city, suburban and rural respondents on most levels and factors. Exceptions are the
higher scores for the highest levels of quiet as the area becomes more rural and lower
scores for the best levels of dog fouling, light pollution and light intrusion in rural areas.
There is a good spread in the reported “as now” situations for all factors with differences
between areas as expected. The information on the relative importance of factors, the
scores for different levels and the “as now” information may all be useful both in interpreting
the model results and as logic checks. An overall summary table is reported in Table 3.27
which outlines the mean „as now‟ levels for each factor categorised by area type.
50
Table 3.27: Average As Now Levels
Attribute Inner Suburban Rural Total
Light Pollution 2.20 2.23 2.72 2.33
Discarded Chewing Gum 1.95 2.20 2.64 2.21
Litter 2.31 2.86 3.18 2.74
Light Intrusion at Night 3.02 3.01 3.25 3.07
Trees 2.29 2.54 3.06 2.57
Fly Tipping 3.16 3.54 3.59 3.42
Access to Quiet Areas 2.92 3.53 4.32 3.49
Graffiti 2.76 3.53 4.14 3.40
Odour 3.76 4.17 3.76 3.93
Fly-posting 3.42 3.86 4.16 3.77
Dog Fouling Occurs 2.78 3.60 3.86 3.37
51
4. EMPIRICAL FINDINGS
4.1 Modelling Approach
As was pointed out in section 2.5, and for both the SP1 quality of life exercise and the SP2
local environment indicators SP exercise, respondents ranked in order of preference the
various improvements offered on their current situation. They then turned their attention to
various deteriorations on the current situation, identifying which was worst and then ranking
all the deteriorations through to that least disliked.
The way that the data have been analysed is to convert them into pairwise comparisons,
comparing each improvement (deterioration) with the other possible improvements
(deteriorations).
Suppose that, with reference to the SP1 exercise set out in Table 2.3, a respondent‟s current
positions for those variables which are not predetermined are very noise for traffic at home,
always dog mess in view, good pavement conditions, and quiet areas within 15 minutes.
We start by comparing the local crime improvement to a level of 5 burglaries per 1000
households with all other improvements (although not with the improvement to 2 burglaries
per 1000 households since the latter is logically superior). Given the above current
situations, and the pre-specified current situations for crime, school quality, road traffic and
tax, then we would observe the following two options (A and B) for the comparison of the first
improvement in crime with the first improvement in school quality.
Crime School Traffic Noise Dog Pave Quiet Tax
Option A 5 60% As Now Very Always Good 15 As Now
Option B 9 70% As Now Very Always Good 15 As Now
If the crime improvement to 5 burglaries per 1000 households is ranked higher, then option
A which offers the crime improvement is chosen over option B which offers the improvement
in school quality, all else constant.
We would next compare the improvement in crime to 5 burglaries with the improvement to
80% school pass rates, with the options then being:
Crime School Traffic Noise Dog Pave Quiet Tax
Option A 5 60% As Now Very Always Good 15 mins As Now
Option B 9 80% As Now Very Always Good 15 mins As Now
This would be followed by a comparison of the crime improvement to 5 burglaries with the
improvement of 5% less traffic represented as:
Crime School Traffic Noise Dog Pave Quiet Tax
Option A 5 60% As Now Very Always Good 15 mins As Now
Option B 9 60% -5% Very Always Good 15 mins As Now
52
We proceed through all the improvements with the final comparison for the improvement to 5
burglaries per household being against a £15 tax reduction, all other things equal, and
represented as:
Crime School Traffic Noise Dog Pave Quiet Tax
Option A 5 60% As Now Very Always Good 15 mins As Now
Option B 9 60% As Now Very Always Good 15 mins -£15
The process is then repeated for the improvement in burglaries to 2 per household,
comparing it first with 5% less traffic and then all other improvements through to the £15
council tax reduction.
Comparison of the improvement to 70% school pass rates would then be compared with all
subsequent improvement and we continue through all remaining variables whose
improvements are compared against all other improvements presented below in the table.
The final comparison in this process would be to compare the improvement in access to
quiet areas from 15 minutes to 1 minute with the £15 saving in council tax, whereupon the
two options would be represented as:
Crime School Traffic Noise Dog Pave Quiet Tax
Option A 9 60% As Now Very Always Good 15 mins As Now
Option B 9 60% As Now Very Always Good 1 min -£15
If a respondent currently experiences the best level of a variable, there can be no
improvements considered and all of an individual‟s pairwise comparisons will specify that
variable to be the same for option A and option B.
An entirely analogous process is followed in creating the pairwise comparisons from the
rankings of deteriorations.
The total sample includes 561 respondents who had provided rankings of the improvements
and deteriorations, although note that some individuals do not rank the full set of
improvements or deteriorations available to them.
By far the most common method used to explain choices between discrete options, as is the
format of the data here, is some form of logit model. The logit model which is used to
analyse choices at the disaggregate (individual) level is based on the assumption that each
individual chooses that alternative from the n on offer which yields maximum utility (U) or
satisfaction. Thus individual i chooses alternative 1 if:
1,1 nnallforUU ini (1)
53
In turn, the overall utility for each alternative is made up of the part-worth utilities associated
with a range of explanatory variables. However, the analyst cannot possibly observe all the
influences on each individual‟s choices, whilst others are difficult to measure or too minor to
merit inclusion. An error term (i) is therefore introduced to represent the net effect of the
unobserved influences on an individual‟s choices. Hence as far as we are concerned,
individual i bases decision making on what might be termed random utility which for
alternative k (Uik) is made up as:
ikikik VU (2)
Vik is the observable part of utility, termed deterministic utility. In the case of the choice
between n options with different levels of what for now we will term a generic local
environmental variable (E) and council tax (T), the deterministic utility associated with option
1 for individual i could be represented as:
111 iii TEV (3)
The utility for other options are specified in an entirely analogous fashion. As analysts, by
definition we can proceed only by observation of Vik, yet this ignores the influence of what is
to us unobservable. We cannot be sure that alternative 1 is preferred if V i1 is the highest, yet
the analysis must proceed on the basis of this observable component of utility alone.
The way forward is to specify the problem as one of explaining the probability of an
individual choosing a particular alternative. We would expect the likelihood of choosing
alternative 1 to increase as its overall random utility increases. The probability that an
individual chooses alternative 1 (Pi1) from the n on offer can be represented as:
1,Pr 111 nnallforVVP ininiii (4)
By assuming some probability distribution for the in, the probability of choosing alternative 1
can be specified solely as a function of the observable component of utility. In the case of
the choice between just two alternatives, as we are here dealing with, assuming that the
errors associated with each alternative have a type I extreme value distribution yields the
familiar binomial logit model:
)(1121
1
ii VVie
P
(5)
The coefficients in the disaggregate logit model‟s utility function (equation 3) are estimated
by the technique of maximum likelihood to provide the best explanation of individuals‟
discrete choices.
54
More sophisticated estimation allows the parameters in the utility function to have a
distribution across the sample rather than assuming them to be fixed across all individuals.
Similarly, more flexible forms of utility function that are not linear in parameters can be
directly estimated.
The estimated coefficient weights ( and of equation 3) denote the relative importance of
the variables. We will have expectations as to the sign of the coefficient estimates. A
variable which as it becomes larger is disliked more, such as council tax, will have a
negative coefficient weight. Similarly, attributes relating to improved environmental
conditions will have positive coefficients.
The logit model produces standard errors for each of its coefficient estimates, allowing t
ratios and confidence intervals to be derived. These are interpreted in the same manner as
for the more familiar multiple regression analysis and indicate the degree of confidence that
can be placed in the coefficient estimates. The t ratio is derived as the ratio of the coefficient
estimate and its standard error. The critical value is commonly taken to be two, given that
then a coefficient value of zero lies outside the 95% confidence interval and we can be 95%
confident that the true parameter is not zero. However, we ought to be prepared to retain
variables whose coefficients have t ratios of less than two if the estimates are expected to
influence choice and are plausible, even though not precisely estimated.
The 2 statistic is a measure of goodness of fit, analogous to the more familiar R2 measure of
regression analysis. However, the interpretation of what is a reasonable figure is somewhat
different. Louviere et al. (2000) state that, “Values of 2 between 0.2 and 0.4 are considered
to be indicative of extremely good model fits. Simulations by Domencich and McFadden
(1975) equivalenced this range to 0.7 to 0.9 for a linear function”. 2‟s of around 0.1 are
typical of the goodness of fit obtained in standard SP choice models.
The valuation of an attribute is a relative concept. It is the change in an attribute that has the
same utility as a change in some other attribute. We are here interested in monetary
valuations. If, say, a worse level of environmental quality incurs the same disutility as an
extra £3 of council tax, then the monetary valuation of the environmental deterioration is £3.
More formally, the marginal monetary value of a variable is defined as the ratio of the
marginal utility of that variable and the marginal utility of money. In the case of linear-additive
utility functions of the form of equation 3, the marginal value of environmental variable E is
simply the ratio of the coefficient estimates for the environmental variable and the council tax
(/). In this case, the monetary value is constant, and the average and marginal values are
the same. However, this is not so where the utility function is not linear-in-parameters and
the marginal value will depend on the levels of the attributes and can no longer be taken as
a simple ratio of coefficients.
Many of the variables covered in the two SP experiments are not continuous, like council
tax, but instead have a discrete number of levels and we wish to estimate the utility
associated with each level. The levels of these categorical variables can be represented by
55
dummy variables. If there are five levels of a particular variable, we can specify four dummy
levels to represent any arbitrarily selected four categories. We would specify these in the
utility function alongside council tax as follows:
iiiiii ddddTV 55443322 (6)
The d2i through to d5i are dummy variables for the second through to the fifth categories.
Since the arbitrarily omitted category is here the first one, the coefficient estimates represent
the utility of moving from the first category to the category in question. Thus λ3 is the utility
effect of moving from the base level of some environmental factor to the level represented by
dummy variable 3. Other environmental factors can be entered alongside this particular
categorical variable in an analogous fashion.
One final point is the estimation of how coefficients and values vary across the sample
according to socio-economic and trip characteristics. One way is to estimate separate
models for each category of interest, say by income group. A more parsimonious approach,
and one that is here adopted, is to allow specific coefficients to vary according to socio-
economic variables in line with theoretical expectations. This approach specifies a base level
of a coefficient and incremental departures from it according to specific socio-economic and
situational characteristics.
So, for example, if we believe that the sensitivity to council tax (T) varies with income group,
and there are four income groups, we could specify the utility function as:
.....443322 iiiii TdTdTdTV (7)
where d2, d3 and d4 are dummy variables denoting income groups 2, 3 and 4 respectively
The arbitrarily omitted income category is the lowest one (1). Thus α2 denotes the extent to
which those with incomes in group 2 have a council tax coefficient (β+α2) that differs from
the base category of those in the lowest income group whose council tax coefficient is β.
A variant upon this occurs where the variable influencing the marginal utility is continuous,
such as income or household size. If we believe that the marginal utility of money falls as
income increases, we could instead of specifying income in categorical form as in equation 7
enter it as a continuous interaction. This would take the form:
.....
Y
TV i
i
(8)
Y is household income and is taken as the mid-point of the income category. The λ term
allows the relationship between the marginal utility of money and income to be other than
proportional. The marginal utility of income (MUY) is in this case:
56
YMUY
(9)
and λ is the income elasticity denoting the proportionate change in a valuation after a
proportionate change in income.
4.2 SP1 Quality of Life Experiment Results
The models for the entire SP1 data, for both the improvements on the current situation and
deteriorations to the current situation, are reported in Table 4.1. The t ratios4 associated with
the coefficient estimates are provided (in parentheses) and indicate that the vast majority of
coefficients are estimated with an extremely high degree of precision. The overall goodness
of fit (ρ2) is not, in our experience, atypical of SP choice models, particularly when we bear in
mind the task is not based around everyday decision making.
Council tax is specified in pounds per month. Local crime, local school quality and traffic are
specified in the units presented. Traffic noise and pavement conditions are specified in
dummy variable form, with four dummy variables specified to cover the five levels offered.
Dog fouling can be specified in terms of whether or not there is dog fouling and then the
frequency of occurrence. Thus a dummy variable denotes whether it is present and then,
conditional upon there being dog fouling, we specify the minutes it is observed when
walking. The latter can be every 15 minutes, every 5 minutes, every minute or always which
is taken to be every quarter of a minute.
Similarly, access to quiet areas is represented by a dummy variable denoting whether one is
present and then the number of minutes (15, 10, 5 or 1) it is away from the respondent‟s
home.
The council tax has the expected negative sign in both the improvements and deteriorations
models, as do the local crime and road traffic variables. On the other hand, improving school
quality is a good and hence its coefficient has the expected positive sign.
The relationship between the different traffic noise and pavement conditions in the
improvements model is what would be expected. However, this is not so with improvements
upon the base for traffic noise and pavement condition in the deteriorations model. This
might be because few currently experience very good levels of these two variables from
which the deteriorations increment.
4 The t ratio is an indication of the level of confidence that can be placed in an estimated coefficient. A
t ratio of 2 indicates that we can be 95% confident that true coefficient estimate is not zero.
57
Of greater interest here, since they are local environmental indicators which is the main
focus of this study, are dog fouling and access to quiet areas. The presence of dog fouling
should induce disutility, and indeed the dummy variable denoting its presence is negative
and significant in both models. We then observe in both models that, as expected, disutility
falls as the frequency of observing dog mess reduces.
Table 4.1 SP1 Models for Whole Sample
Improvements Deteriorations
Council Tax Pounds per Month -0.101 (47.7) -0.0783 (41.1)
Local Crime Burglaries per 1000 homes -0.239 (69.4) -0.2770 (71.0)
Dog Fouling Dog Fouling Minutes 0.0711 (24.6) 0.0660 (32.3)
Dog Fouling Present -2.25 (49.3) -0.529 (14.7)
Traffic Noise
Extremely Noisy -0.564 (6.7) -1.93 (51.9)
Very Noisy -0.164 (3.3) -0.986 (29.2)
Moderately Noisy Base Base
Slightly Noisy 0.656 (15.7) -0.0627 (2.0)
Not at all Noisy 1.85 (39.9) -0.0009 (0.1)
Pavement
Condition
Very Poor -0.568 (11.0) -1.1500 (32.4)
Poor -0.218 (6.2) -0.319 (9.3)
Neither Good nor Poor Base Base
Good 0.625 (17.7) -0.365 (10.3)
Very Good 1.58 (41.4) -0.480 (5.3)
Access to
Quiet Areas
Access Minutes -0.0644 (20.4) -0.0082 (3.0)
Quiet Area Present 1.17 (19.0) 1.09 (27.7)
Local School % gaining 5 GCSEs A to C 0.117 (66.7) 0.0831 (54.7)
Road Traffic % change in traffic -0.195 (59.4) -0.187 (61.3)
Observations
48903 64635
ρ2
0.12 0.11
The coefficients for access to quiet areas also possess correct signs. The presence of a
quiet area is regarded to be beneficial, and hence its coefficient is positive in both models.
Increasing the access time to quiet areas reduces their attractiveness.
There are instances where respondents ranked a better improvement in a specific variable
(farther to the right) less highly than a lesser improvement, and similarly cases where a
worse deterioration (farther to the left) was ranked higher than a lesser deterioration. Such
illogical responses can be removed and the re-estimated model are reported in Table 4.25.
We note only small reductions in sample size and improvements in fit. Otherwise the pattern
of results is very similar, which is not surprising given that the samples are little different.
5 This assumes that the illogical responses are in fact errors. However, it could be envisaged in the SP2 exercise
that these are in fact rational answers; for example, no light intrusion might not be most preferred since it
implies disbenefits in terms of safe navigation and security.
58
Table 4.2: SP1 Models for Sample Excluding Illogical Rankings
Improvements Deteriorations
Council Tax Pounds per Month -0.106 (48.0) -0.0828 (42.1)
Local Crime Burglaries per 1000 homes -0.245 (68.0) -0.286 (70.6)
Dog Fouling Dog Fouling Minutes 0.0739 (24.7) 0.0697 (33.1)
Dog Fouling Present -2.32 (49.0) -0.589 (16.0)
Traffic Noise
Extremely Noisy -0.536 (6.1) -2.01 (52.3)
Very Noisy -0.171 (3.2) -0.999 (28.3)
Moderately Noisy Base Base
Slightly Noisy 0.613 (13.8) -0.0773 (2.4)
Not at all Noisy 1.9 (39.4) 0.0109 (0.3)
Pavement
Condition
Very Poor -0.597 (11.1) -1.21 (33.1)
Poor -0.192 (5.2) -0.307 (8.8)
Neither Good nor Poor Base Base
Good 0.620 (16.7) -0.358 (9.8)
Very Good 1.640 (41.4) -0.647 (6.6)
Access to
Quiet Areas
Access Minutes -0.0685 (20.8) -0.00981 (3.5)
Quiet Area Present 1.24 (19.1) 1.17 (29.0)
Local School % gaining 5 GCSEs A to C 0.12 (65.7) 0.0848 (54.3)
Road Traffic % change in traffic -0.2 (58.7) -0.193 (61.4)
Observations
45569 61389
ρ2
0.13 0.12
We take the models in Table 4.2 that remove illogical preferences as preferred, although
recognising that there is no material difference to the results for the whole sample. The key
parameters of interest to us relate to access to dog fouling and quiet areas. The implied
monthly values are reported in Tables 4.3.
We note little difference in valuations of dog fouling and of access to quiet areas according
to whether illogical rankings are removed. However, there are some large differences
between improvements and deteriorations valuations, that is a difference between:
Willingness to accept compensation to forgo an improvement; and
Willingness to pay to avoid a deterioration.
For dog fouling, the improvements offered are very much more highly valued than are the
deteriorations. In contrast, for quiet areas the values are somewhat larger for those
presented with deteriorations. It is not readily apparent why this is so, although the
correlation between the fixed effect (presence) and the variable effect (minutes) could be a
contributory factor. Whilst we might expect some differences in valuations according to
current positions, such as lower valuations for improvements given that those with lower
values are more likely to live in places with poorer levels and hence more likely to face
improvements in our SP exercises, this cannot explain the pattern of results observed since
59
the marginal effects (minutes) are less for the deteriorations model and these valuations will
tend to relate to movements from a relatively good position.
An artefact of the modelling process is that the valuations can become wrong sign, since the
positive minutes on dog fouling for large numbers of minutes offsets the negative effect of
the presence of dog fouling and the negative effect of longer access times to quiet areas
offsets the positive effect of the presence of quiet areas. This should be interpreted as there
being some level of accessibility at which quiet areas have no value and some infrequency
of dog fouling such that it is not regarded to be a problem.
Table 4.3: Values (£ per person per month) Dog Fouling and Quiet Areas (Relative to
Never/None)
Dog Fouling 20 mins 10 mins 5 mins 1 min Always
All Data: Improvements 8.2 15.2 18.8 21.6 22.3
Omit Illogical: Improvements 7.9 14.9 18.4 21.2 21.9
All Data: Deteriorations -10.1 -1.7 2.5 5.9 6.8
Omit Illogical: Deteriorations -9.7 -1.3 2.9 6.3 7.1
Quiet Areas 20 mins 15 mins 10 mins 5 mins 1 min
All Data: Improvements -1.2 2.0 5.2 8.4 10.9
Omit Illogical: Improvements -1.2 2.0 5.3 8.5 11.1
All Data: Deteriorations 11.8 12.3 12.9 13.4 13.8
Omit Illogical: Deteriorations 11.8 12.4 12.9 13.5 14.0
Note that the purpose of the SP1 exercise was to mask the real purpose of the study to
value local environmental indicators and hence provides a means of deflating possibly
inflated valuations from the SP2 exercises. We return to this issue in section 4.3.
4.3 SP2 Local Environment Factors Experiment Results: Dummy Variables
The SP2 data is modelled in the same fashion as for the SP1 data, with the rankings
expanded into a large number of pairwise comparisons. We use dummy variables to
represent discrete levels and where possible, as with odour, dog fouling and quiet areas,
specify continuous variables to represent the frequency of occurrence alongside a dummy
variable denoting the presence of odour, dog fouling or quiet areas.
The preferences amongst the environmental factors relate to the conditions experienced in
the respondent‟s neighbourhood. They do not cover the preferences of visitors to a
neighbourhood, nor our respondents‟ valuations of these factors in other locations.
Table 4.4 presents the SP2 results for the entire sample of respondents, with separate
models for improvements and deteriorations. To ease comparison, both models have
common base categories which are clearly specified. The council tax has the expected
negative sign, and is not greatly different between the two models.
60
Table 4.4: SP2 Models for Whole Sample
Improvements Deteriorations
CGUM1 (Chewing Gum Worst Situation) Base Base
CGUM2 0.563 (10.7) 1.25 (50.7)
CGUM3 (Chewing Gum Best Situation) 1.47 (26.3) 1.27 (42.5)
Council Tax -0.071 (35.6) -0.0657 (45.4)
Dog Fouling Minutes 0.0795 (29.3) 0.0807 (49.4)
Dog Fouling Present -2.07 (46.8) -1.06 (37.4)
FLYP1 (Fly-Posting Worst Situation) Base Base
FLYP2 0.0405 (0.5) 0.6 (20.3)
FLYP3 -0.0824 (1.0) 0.97 (34.7)
FLYP4 -0.0808 (1.0) 0.845 (33.1)
FLYP5 (Fly-Posting Best Situation) 0.321 (3.9) 0.574 (18.2)
FLYT1 (Fly-Tipping Worst Situation) Base Base
FLYT2 0.373 (3.2) 1.1 (31.8)
FLYT3 0.882 (7.8) 2.27 (71.2)
FLYT4 (Fly-Tipping Best Situation) 2.11 (18.0) 2.63 (77.7)
GRAF1 (Graffiti Worst Situation) Base Base
GRAF2 0.0575 (1.2) 0.731 (25.7)
GRAF3 0.091 (1.8) 1.26 (41.5)
GRAF4 0.316 (6.3) 1.08 (39.0)
GRAF5 (Graffiti Best Situation) 0.973 (18.5) 0.976 (31.8)
LINT1 (Light Intrusion Worst Situation) Base Base
LINT2 0.707 (9.1) 0.827 (29.1)
LINT3 0.931 (12.7) 0.988 (40.1)
LINT4 (Light Intrusion Best Situation) 1.55 (20.4) 0.864 (28.5)
LITT1 (Litter Worst Situation) Base Base
LITT2 0.189 (3.7) 1.47 (49.5)
LITT3 0.626 (12.6) 2.02 (68.7)
LITT4 (Litter Best Situation) 1.88 (33.6) 2.28 (58.9)
LPOL1 (Light Pollution Worst Situation) Base Base
LPOL2 0.489 (5.9) 0.581 (23.9)
LPOL3 (Light Pollution Best Situation) 1.11 (13.0) 0.514 (18.1)
ODOUR Present -0.785 (22.5) -0.138 (6.6)
ODOUR Days -0.0028 (13.7) -0.0042 (58.9)
QUIET Present 1.21 (20.1) 1.16 (40.9)
QUIET Minutes -0.0651 (21.9) -0.0228 (11.8)
TREE1 (Trees Worst Situation) Base Base
TREE2 0.368 (8.5) 0.705 (24.5)
TREE3 0.657 (15.3) 0.833 (30.2)
TREE4 (Trees Best Situation) 1.4 (30.4) 1.09 (32.1)
OBSERVATIONS 50826 134285
ρ2 0.07 0.12
61
Table 4.5: SP2 Models for Sample Excluding Illogical Rankings
Improvements Deteriorations
CGUM1 (Chewing Gum Worst Situation) Base Base
CGUM2 0.62 (11.1) 1.32 (51.0)
CGUM3 (Chewing Gum Best Situation) 1.68 (28.1) 1.33 (43.4)
Council Tax -0.0843 (39.4) -0.0742 (48.6)
Dog Fouling Minutes 0.0852 (29.2) 0.0926 (53.2)
Dog Fouling Present -2.28 (48.8) -1.18 (39.7)
FLYP1 (Fly-Posting Worst Situation) 0 Base
FLYP2 0.0353 (0.4) 0.632 (20.3)
FLYP3 -0.0914 (1.0) 1.04 (35.4)
FLYP4 -0.0573 (0.7) 0.909 (34.2)
FLYP5 (Fly-Posting Best Situation) 0.484 (5.5) 0.728 (21.6)
FLYT1 (Fly-Tipping Worst Situation) Base Base
FLYT2 -0.0433 (0.3) 1.16 (32.2)
FLYT3 0.502 (4.1) 2.37 (71.4)
FLYT4 (Fly-Tipping Best Situation) 1.91 (15.4) 2.74 (77.5)
GRAF1 (Graffiti Worst Situation) Base Base
GRAF2 0.106 (2.0) 0.803 (27.0)
GRAF3 0.169 (3.1) 1.33 (41.8)
GRAF4 0.357 (6.6) 1.17 (40.3)
GRAF5 (Graffiti Best Situation) 1.19 (21.0) 1.1 (34.2)
LINT1 (Light Intrusion Worst Situation) Base Base
LINT2 0.732 (9.0) 0.915 (30.6)
LINT3 0.982 (12.9) 1.09 (42.5)
LINT4 (Light Intrusion Best Situation) 1.76 (22.3) 0.885 (27.9)
LITT1 (Litter Worst Situation) Base Base
LITT2 0.24 (4.3) 1.56 (50.5)
LITT3 0.657 (12.1) 2.13 (69.5)
LITT4 (Litter Best Situation) 2.09 (34.7) 2.38 (58.9)
LPOL1 (Light Pollution Worst Situation) Base Base
LPOL2 0.447 (5.0) 0.643 (25.4)
LPOL3 (Light Pollution Best Situation) 1.25 (13.6) 0.565 (19.0)
ODOUR Present -0.924 (25.4) -0.117 (5.3)
ODOUR Days -0.003 (13.5) -0.0045 (60.8)
QUIET Present 1.45 (22.4) 1.23 (41.4)
QUIET Minutes -0.081 (25.8) -0.019 (9.3)
TREE1 (Trees Worst Situation) Base Base
TREE2 0.428 (9.1) 0.781 (25.9)
TREE3 0.693 (14.9) 0.92 (31.8)
TREE4 (Trees Best Situation) 1.59 (32.4) 1.15 (32.1)
OBSERVATIONS 46171 122857
ρ2 0.07 0.12
62
The „fixed‟ (presence) and „variable‟ (frequency) elements of the odour, dog fouling and quiet
area valuations are all of the correct sign and highly statistically significant in each model. Of
the other eight attributes, the correct relationship is observed in the improvements model for
all attributes other than for fly-posting. For the latter, only the best situation is significantly
different from the base. We note that in Table 3.14 fly-posting was revealed as the least
important of the local environmental factors. Moreover, it may be that respondents had
difficulty discerning differences between the different levels of fly-posting.
As for the deteriorations model, the expected relationships are apparent for chewing gum,
fly-tipping, litter and trees. There are significant effects for fly-posting and graffiti but no clear
relationship. Light intrusion and pollution have some of the expected effects but are not
entirely consistent, which could be because deteriorations in these attributes, which lead to
more light, are of benefit for safe navigation and security.
Table 4.5 reports the SP2 models after removing the relatively small proportion of illogical
responses. This does not make a great deal of difference, although some increases in the t
ratios associated with the coefficient estimates denote that the data is now of better quality.
The valuations for dog fouling, access to quiet and odour, where we have used a mix of
discrete and continuous variables, and where for two attributes the results can be compared
with those obtained from the SP1 model, are presented in Table 4.6 for the model that
removes illogical responses.
The figures in parentheses denote the relationship between the values recovered by the SP2
model and the corresponding valuations of the SP1 model where correct sign values are
implied. Although the figures do vary across the different valuations, there is a clear pattern
in the sense that the valuations from the SP2 experiment are almost always higher than from
the SP1 experiment, as expected, but the divergence is not particularly large in general. We
feel that there is not a convincing case to scale the values in SP2 to those in SP1.
Table 4.6: SP2 Values (£ per person per month) of Dog Fouling, Access to Quiet and
Odour (Relative to Never/None)
Dog Fouling 20 mins 10 mins 5 mins 1 min Always
Omit Illogical: Improvements 6.8
(86%)
16.9
(113%)
22.0
(119%)
26.6
(125%)
27.0
(125%)
Omit Illogical: Deteriorations -9.1 3.4
9.7
(234%)
14.7
(133%)
15.9
(124%)
Access to Quiet Areas 20 mins 15 mins 10 mins 5 mins 1 min
Omit Illogical: Improvements -2.0 2.8
(140%)
7.6
(143%)
12.4
(146%)
16.2
(146%)
Omit Illogical: Deteriorations 11.4
(97%)
12.7
(102%)
14.0
(109%)
15.3
(113%)
16.3
(116%)
Odour Annual Monthly Weekly Daily
Omit Illogical: Improvements 11.0 11.4 12.8 24.0
Omit Illogical: Deteriorations 1.6 2.3 4.7 23.8
63
The remaining valuations implied by the SP2 models of Table 4.5 are reported in Table 4.7,
along with 95% confidence intervals in parentheses. For trees, the pattern of results in the
improvements and deteriorations models is broadly similar whilst for chewing gum and
graffiti the valuations of the movement from the worst to the best level are similar. The
improvements model provides larger valuations in the case of light pollution and light
intrusion for the movement from the worst to the best level. For litter, fly-tipping and fly-
posting, the deteriorations model generally yields somewhat larger valuations.
Table 4.7 SP2 Valuations £ per person per month (Illogical Responses Removed)
Improvements Deteriorations
CGUM1 (Chewing Gum Worst Situation) Base Base
CGUM2 7.35 (-5.94 – 8.77) 17.79 (16.73 – 18.85)
CGUM3 (Chewing Gum Best Situation) 19.92 (18.03 – 21.83) 17.92 (16.76 – 19.09)
FLYP1 (Fly-Posting Worst Situation) Base Base
FLYP2 0.42 (-1.68 – 2.51) 8.52 (7.58 – 9.46)
FLYP3 -1.08 (-3.24 – 1.08) 14.01 (12.99 – 15.04)
FLYP4 -0.68 (-2.62 – 1.26) 12.25 (11.33 – 13.17)
FLYP5 (Fly-Posting Best Situation) 5.74 (3.61 – 7.87) 9.81 (8.78 – 10.84)
FLYT1 (Fly-Tipping Worst Situation) Base Base
FLYT2 -0.51 (-3.93 – 2.91) 15.63 (14.42 – 16.85)
FLYT3 5.95 (3.00 – 8.91) 31.94 (30.28 – 33.60)
FLYT4 (Fly-Tipping Best Situation) 22.66 (19.39 – 25.92) 36.92 (35.17 – 38.68)
GRAF1 (Graffiti Worst Situation) Base Base
GRAF2 1.26 (-0.01 – 2.52) 10.82 (9.87 – 11.78)
GRAF3 2.00 (0.71 – 3.00) 17.92 (16.74 – 19.11)
GRAF4 4.23 (2.90 – 5.57) 15.77 (14.70 – 16.83)
GRAF5 (Graffiti Best Situation) 14.11 (12.56 – 15.68) 14.82 (13.72 – 15.93)
LINT1 (Light Intrusion Worst Situation) Base Base
LINT2 8.68 (6.68 – 10.69) 12.33 (11.34 – 13.33)
LINT3 11.65 (9.78 – 13.52) 14.69 (13.73 – 15.65)
LINT4 (Light Intrusion Best Situation) 20.88 (18.64 – 23.12) 11.92 (10.90 – 12.95)
LITT1 (Litter Worst Situation) Base Base
LITT2 2.85 (1.50 – 4.19) 21.02 (19.76 – 22.28)
LITT3 7.79 (6.39 – 9.20) 28.70 (27.17 – 30.24)
LITT4 (Litter Best Situation) 24.79 (22.80 – 26.79) 32.07 (30.41 – 33.74)
LPOL1 (Light Pollution Worst Situation) Base Base
LPOL2 5.30 (3.14 – 7.47) 8.67 (7.86 – 9.47)
LPOL3 (Light Pollution Best Situation) 14.82 (12.45 – 17.20) 7.61 (6.72 – 8.51)
TREE1 (Trees Worst Situation) Base Base
TREE2 5.08 (3.91 – 6.25) 10.52 (9.57 – 11.48)
TREE3 8.22 (7.08 – 9.36) 12.40 (11.39 – 13.41)
TREE4 (Trees Best Situation) 18.86 (17.28 – 20.44) 15.50 (14.29 – 16.71)
64
4.4 SP2 Local Environment Factors Experiment Results: Rating Scale Models
In addition, and as described in chapter 2, each respondent rated either the photographic or
verbal representation of each level of each variable. In part this was as a familiarisation
exercise, so they could consider each level before the more challenging task of ranking the
improvements and deteriorations on the current situation. However, the rating of each level
of each variable, on a 0-10 scale, allows each generic attribute to be represented by its
rating rather than as a dummy variable or a mix of dummy and continuous variables. We
have also reported models based on this approach.
The rating scale simply replaces the dummy variable that was used to represent levels and a
single coefficient is reported for each attribute. Given that 10 indicates the best level, the
estimated coefficients should be positive. Where a respondent did not provide a rating of a
level of an attribute, the mean value across all respondents was used.
Table 4.8 reports the estimated rating scale models. With the exception of fly-posting, where
problems were experienced in the previous models, all the coefficients are of the correct sign
and significant in the improvements model. The deteriorations model has wrong sign
coefficients for light intrusion and light pollution, in line with the problems encountered in the
equivalent model based on dummy variables. For the remaining variables, the improvements
and deteriorations models yield coefficients which are generally not greatly different.
We therefore estimated a combined model, pooling improvements and deteriorations data,
allowing for scale differences between the two data sets and for variations in parameters that
were somewhat different between the two separate models. The scale for the SP2 data in
the combined model (0.92) is not significantly different from one indicating that the two data
sets have essentially the same scale. Separate coefficients for improvements and
deteriorations were specified for light intrusion, light pollution and fly-posting.
We then accounted for those who stated that, in some form or other, they did not fully
account for council tax in making their decisions. Some people stated that did not believe
council tax reductions would occur. An incremental effect on the tax coefficient for these
respondents, using the formulation set out in equation 7, was significant and positive,
denoting they have a lesser sensitivity to the tax reductions offered as expected. Others
stated that they focussed primarily on the environment factors. These were also found to
have a positive and significant incremental effect on the tax coefficient, as expected, to the
extent that it implies they took no account of tax whatsoever. Finally, there were some who
stated that they paid more attention to increases, who had a very strong incremental effect
denoting a tax sensitivity almost twice as high as the base. This increased sensitivity does
not seem plausible and we take this to represent protest responses towards council tax
increases.
65
Table 4.8: SP2 Ratings Models (Whole Sample)
Improvements Deteriorations Combined Combined
Chewing Gum 0.0931 (20.4) 0.0667 (23.5) 0.0769 (24.5) 0.0725 (29.9)
Dog Fouling 0.142 (30.4) 0.0324 (11.4) 0.0689 (18.7) 0.0631 (27.0)
Fly-Posting -0.0003 (0.1) -0.0124 (4.2)
Fly-Posting Imp -0.0083 (1.6) -0.0125 (2.5)
Fly-Posting Det -0.0121 (3.7) -0.0092 (3.2)
Fly-Tipping 0.109 (23.6) 0.126 (61.5) 0.132 (36.5) 0.124 (67.3)
Graffiti 0.052 (11.7) 0.00875 (3.2) 0.0203 (7.8) 0.0188 (8.1)
Light Intrusion 0.0209 (3.8) -0.0067 (2.2)
Light Intrusion Imp 0.0141 (2.6) 0.0112 (2.1)
Light Intrusion Det -0.0054 (1.7) -0.0045 (1.5)
Litter 0.129 (30.5) 0.138 (41.7) 0.138 (38.7) 0.132 (51.8)
Light Pollution 0.0329 (5.4) -0.0274 (8.5)
Light Pollution Imp 0.0253 (4.2) 0.0211 (3.6)
Light Pollution Det -0.0279 (7.8) -0.0255 (7.9)
Odour 0.0704 (13.6) 0.0621 (27.8) 0.0676 (25.1) 0.0638 (30.1)
Quiet 0.0281 (5.2) 0.0486 (19.7) 0.0477 (18.8) 0.0456 (20.2)
Trees 0.109 (19.7) 0.0663 (17.8) 0.0827 (21.7) 0.0779 (25.3)
Council Tax + not believe + focus environment + attention increases
-0.0333 (20.3) -0.0241 (20.4) -0.0278 (23.4) -0.0334 (23.3) 0.0110 (3.3)
0.0325 (15.3) -0.0315 (11.0)
Observations 50826 134285 185111
ρ2 0.07 0.12 0.040 0.041
The monetary valuations implied by the combined ratings model are reported in Table 4.9.
These use the council tax numeraire (-0.0334) that is free of the effects of not believing tax
reductions, placing attention on environmental factors and protest response.
For light intrusion and light pollution, the values for the improvements model are used. We
are unable to obtain sensible values for fly posting. Values are provided for a unit change in
the rating of a factor and also for the maximum possible improvement from a rating of zero to
ten.
The valuations express considerable variation across attributes; it would have been
disconcerting to have obtained similar values across variables. Since we have expressed
each attribute in common units, that of a 0-10 rating scale, we can readily identify the
importance of different attributes. The largest valuations are quite clearly for litter and fly-
tipping. Then there are a series of attributes with similar „medium‟ valuations. These are
trees, odour, chewing gum, dog fouling and quiet areas. Light pollution, graffiti and light
intrusion have relatively minor valuations. This pattern of valuations seems plausible.
66
Table 4.9: Monthly Values Implied by SP2 Ratings Model (£ per person per month)
Value of a Unit
Rating Change
Value of a
Move from
Worst to
Best6
Stated
Preference
Rank
Importance
Rating
Rank
Chewing Gum 2.17 (1.96 – 2.38) 21.7 4 7
Dog Fouling 1.89 (1.69 – 2.09) 18.9 6 3
Fly Posting - - - 11
Fly Tipping 3.71 (3.39 – 4.03) 37.1 2 2
Graffiti 0.56 (0.42 – 0.71) 5.6 9 8
Light Intrusion 0.34 (0.02 – 0.65) 3.4 10 9
Litter 3.95 (3.59 – 4.31) 39.5 1 1
Light Pollution 0.63 (0.29 – 0.98) 6.3 8 10
Odour 1.91 (1.72 – 2.10) 19.1 5 6
Quiet 1.37 (1.20 – 1.53) 13.7 7 4
Trees 2.33 (2.07 – 2.59) 23.3 3 5
We have also reported in Table 4.9 the ranking of each factor in terms of the reported
importance ratings which can be compared with the ranking of each factor in terms of
implied SP valuation. A Spearman correlation coefficient of 0.77 indicates a high degree of
correspondence between the SP valuations and the importance ratings. This is an
encouraging finding. Moreover, our inability to recover a significant correct sign fly-posting
valuation may be because, as indicated by the importance ratings, this factor is the least
important of all those here considered.
4.5 SP2 Rating Scale Models: Socio-Economic and Attitudinal Segmentations
4.5.1 Modelling Approach
A wide range of socio-economic, location and attitudinal factors can be examined to test
whether there is an impact from them on the valuations of local environmental factors.
Two approaches were used in identifying candidate segmenting variables for final models.
Where there is a clear theoretical relationship, we used the interactions approaches set out
in equations 7 and 8. This might be with the expected relationship between the sensitivity to
cost and the level of income, with those with higher incomes being less sensitive to cost.
Where there is no expected relationship or where the socio-economic, location or attitudinal
variable might be expected to impact on the valuation of a number of environmental factors,
we simply estimate separate models for each category of interest. Inspection of the results
6 Not everyone will rate the worst level we offered as zero and the best level we offered as 10. Hence this
valuation will overstate the benefit of moving from the worst to best level. Table 5.1, based on the actual
ratings supplied, provides a more accurate and lower indication of willingness to pay, often around half the
amount here.
67
then indicates which segmentations are worthy of further attention using the more
parsimonious interaction approach.
The model selected to explore these segmentations is the combined model of Table 4.8.
This avoids us having a separate set of segmentation effects for improvements and
deteriorations and halves the number of model estimations. The segmenting variables,
whose categories are evident from the questionnaire set out in Appendix F, were:
Age Group
Gender
Number of Children in Household
Household Size
Member of Environmental Organisation?
Environmental Attitude
House Type
Presence of Garden
Pets in Household
Current Council Tax Level
Who Pays Council Tax?
Employment Status
Occupation
Household Income
Perceptions of Council Tax changes
Location
The incremental interaction approach of equation 7 was followed at the outset in exploring
the impact of who pays the council tax, household income and perception of council tax
changes. For all other variables, the initial segmentation analysis was based around a
separate model for each category.
4.5.2 Interaction Effects As far as perceptions of council tax were concerned, an incremental effect was specified for
those who did not believe that reductions would happen and so did not consider them. This
interaction was specific to cost improvements and was found to be significant and to indicate
that these respondents essentially did not consider council tax improvement.
For those who stated that they focussed on environmental factors and did not take the
council tax seriously, an incremental effect was specified on cost increases and reductions.
Its coefficient was positive and significant, indicating that this group do indeed take less
account of council tax.
Finally with regard to the perceptions of council tax changes, an incremental effect was
specified on the council tax increases for those who stated that they paid more attention to
increases than reductions. This coefficient was negative, as expected, and significant. Its
68
magnitude is such that it cannot plausible be taken to represent a true difference in response
to deteriorations and improvements in council tax, especially when perceived unlikely
improvements have been accounted for. We therefore take this to be discerning a protest
towards council tax increases.
Given these three effects were significant and correct sign, relative to a base that the
respondent considered both increases and reductions to be equally likely, they are retained
in the final, preferred segmented model.
With regard to who pays the council tax, the base was that the tax was paid by the
respondent. Incremental effects were specified for those who stated that someone else paid
or else the council tax was paid jointly Both these incremental effects were insignificant.
Separate incremental effects for household income were specified for £10-20k, £20-30k,
£30-40k, £40-50k, more than £50k and not given, compared to a base of less than £10k.
The results were not particularly promising. The only significant and correct sign coeff icients
were for the two highest income groups where there was a lesser sensitivity to tax. We
therefore explored whether there was any relationship between the sensitivity to council tax
and income using the continuous interaction specified in equation 8.
We specified the income variable both as the midpoint of the income category and also as
that amount deflated by the number of people in the household. The specification of
household income alone, without any adjustment for household size, provided a better fit.
The income elasticity was significant and took the value of 0.28. This is a relatively low
income elasticity but this is by no means uncommon in environmental valuation.
4.5.3 Separate Models For the remaining segmenting socio-economic, location and attitudinal variables, separate
models were estimated for the different categories.
Age was throwing up some variations in valuations across the groups of under 30, 30-59 and
60 and over. The patterns were not always entirely clear and consistent but the results
indicated that more detailed specification using incremental terms was warranted.
Some gender specific effects were apparent in the segmented models. Females had a
greater sensitivity to graffiti, trees and dog fouling, whilst there were also some less clear
effects apparent for light pollution, light intrusion and fly posting. These were all taken
forward for further consideration as interactions in an incremental model.
We specified four categories of household size, of single person household, two people,
three or four people, and five or more. Positive relationships on the marginal utility were
apparent for dog fouling, chewing gum, trees and graffiti. These were taken forward for
further analysis as continuous interactions related to the number of people in the household.
Simply whether there were children in the household lowered the values and this was
because the council tax coefficient was higher. We attributed this to an income effect,
whereby those with children had a lesser income per person and as a result of this could be
69
more sensitive to council tax. However, we have already identified that household income
rather than household income per person provides a better account of the SP responses.
There were no variations in values apparent for whether the respondent was a member of an
environmental organisation, whether there were pets in the household, housing type and
whether the home had a garden, although the vast majority of homes did have a garden.
For employment and occupation, we split the employed respondents into managerial and
professional, supervisory and technical with semi-skilled or skilled manual, and clerical, with
the non-employed as a separate category. Surprisingly there were no clear patterns in the
coefficient estimates or valuations across these four categories.
With regard to location, we split by both city and the specific area (inner-city, suburban and
rural) within the city. Across the separate models, there were indications that a number of
effects might be present and would warrant further analysis with the interaction approach.
These were for specific areas:
Chewing Gum – Inner
Dog fouling – Inner and Suburban
Graffiti - Suburban and Rural
Fly Tipping - Suburban and Rural
Litter - Suburban and Rural
Light Pollution - Suburban and Rural
Light intrusion - Suburban and Rural
Odour - Suburban and Rural
Quiet - Suburban and Rural
Trees - Suburban and Rural
whilst for cities they were:
Chewing Gum and Graffiti - Manchester
Light Intrusion - Coventry
Light Pollution, Odour, Quiet and Trees - London
4.5.4 Preferred Segmented Model The preferred segmented model is based on the interactions modelling approach, since this
is more parsimonious and much more usable than separate models for different socio-
economic segments. Nonetheless, the final model has been developed using the insights
from the separate model segmentations
The final model is reported in Table 4.10, based around the combined ratings model. The
incremental effects, which can be positive or negative, are denoted by + and – after the main
effect of a variable.
The council tax was entered in the form of equation 8, that is, deflated by household income
raised to some power (λ) which is the income elasticity. The income elasticity was estimated
70
to be 0.4, which is not atypical of environmental valuation studies. The council tax coefficient
of -2.67 is isolated of effects where respondents did not account for it appropriately.
For chewing gum, older age groups, surprisingly, are less sensitive to it and its valuation
also falls with household size, the latter entered as a continuous variable so that the
incremental effect for a household of three is three times that of a single person household.
Those in Manchester particularly dislike chewing gum as do those in inner-city areas. This
may be because it is here more prevalent.
Table 4.10: SP2 Segmented Ratings Model (Whole Sample)
Chewing Gum
+ Age30to60
+ AgeOver60
+ HHsize
+ Manchester
+ Inner
0.0523 (5.9)
-0.0449 (7.2)
-0.0200 (2.4)
-0.0120 (5.8)
+0.0305 (6.0)
+0.0314 (5.5)
Graffiti
+ Age30to60
+ AgeOver60
+ Manchester
+ Suburban
+ Rural
n.s.
+0.0307 (6.4)
+0.0210 (3.1)
+0.0443 (8.9)
-0.0177 (3.3)
-0.0317 (5.3)
Dog Fouling
+ AgeOver60
+ Male
+ HHsize
+ Suburban
+ Rural
0.0750 (8.6)
+0.0617 (7.9)
-0.0393 (7.9)
+0.0166 (7.9)
-0.0113 (1.8)
-0.0173 (25.3)
Light Intrusion Imp
+ Age30to60
+ AgeOver60
+ Male
+Coventry
+ Suburban
+ Rural
0.0195 (2.3)
-0.0197 (3.2)
+0.0249 (2.7)
-0.0289 (5.3)
+0.0387 (6.6)
+0.0352 (5.3)
-0.0314 (4.3) Fly-Posting Imp -0.0142 (2.5)
Fly-Posting Det -0.0008 (0.3) Odour
+ Age30to60
+ AgeOver60
+ London
+Suburban
+ Rural
0.0202 (3.8)
+0.0271 (5.5)
+0.0236 (3.2)
+0.0249 (5.2)
+0.0311 (6.3)
+0.0335 (5.5)
Fly-Tipping
+ Age30to60
+ AgeOver60
+ Rural
0.1250 (52.8)
+0.0258 (3.7)
+0.0204 (2.5)
+0.0300(7.3)
Light Intrusion Det 0.0025 (0.3)
Litter
+ AgeOver60
+ Suburban
0.148 (27.0)
-0.0104 (2.0)
+0.0165 (3.0)
Trees
+ Age30to60
+ AgeOver60
+ Male
+ HHsize
+ London
+ Rural
n.s.
-0.0224 (3.7)
-0.0293 (2.9)
-0.0223 (3.6)
+0.0249 (7.8)
+0.0591 (7.9)
+0.0556 (7.3)
Light Pollution Imp
+ Male
+ London
+ Suburban
n.s.
-0.0329 (5.9)
+0.0213 (2.8)
+0.0430 (6.8)
Light Pollution Det -0.0357 (8.3)
Council Tax
λ (income elasticity)
+ not believe
+ focus environment
+ attention increases
-2.67 (2.8)
0.40 (11.6)
1.00 (2.3)
2.55 (2.8)
-3.43 (2.4)
Quiet
+ AgeOver60
+ London
+ Suburban
+ Rural
n.s.
-0.0187 (2.9)
+0.0371 (6.7)
+0.0470 (10.4)
+0.0775 (19.6)
ρ2 0.048
71
Dog fouling is of greater concern to the elderly and increases in importance with household
size. Males are less concerned by dog fouling. Surprisingly those in suburban and rural
areas place less emphasis on it, compared to those in inner-city areas. Presumably this is
because it is more of a problem in the latter.
The value of flying tipping increases with age and, not surprisingly, it is highest for rural
dwellers.
Graffiti is more of an issue as age increases, which might be expected. It is particularly
highly disliked in Manchester and is less important in suburban and rural locations, perhaps
because it is there less widely experienced.
In rural areas, light intrusion is valued somewhat lower than elsewhere, and this may reflect
a preference for it for security and safety reasons. Otherwise, those in suburban areas have
a stronger preference for improvements to it. Males are less concerned about light intrusion
although the effect from age group is not clear-cut.
The base level for light pollution was not significant (n.s.) and the negative incremental effect
for males indicates that they in fact like it! Those living in London and particularly suburban
residents do value improvements to light pollution.
There were only two significant incremental effects on sensitivity to litter. Those aged over
60 are, surprisingly, less bothered about it whilst suburban dwellers are more concerned.
The impact of odour increases with age group and is also somewhat larger for suburban and
rural residents and additionally for those in London.
Quiet areas had an insignificant effect (n.s.) as a base term applying across the entire
sample. With the exception of Londoners, those in inner-city areas had somewhat lower
values than others. The negative coefficient for those over 60 would imply that quiet areas
are actually disliked. This could be because of association with anti-social behaviour.
Finally, whilst the base level of trees was not significant, its value is dependent upon
household size. Residents of London and those in rural areas more appreciate additional
trees but males have lower values and the value falls with age group.
72
5. CASE STUDIES
By case study, we mean applying the willingness to pay values that have been obtained. We
do not have the resources to take a case study area, establish the current situation and then
evaluate improvements to it, ideally with some estimate of the costs of making those
improvements in practice.
However, given that our sample is broadly representative, we can apply the results obtained
to our sample of respondents and consider a series of changes to individuals‟ current
circumstances. This is termed a sample enumeration approach. We can also use our overall
model to come up with representative valuations across the whole sample.
First of all, we use the overall ratings model of Table 4.8, based on the combined model, to
derive valuations for movements between each level used of each local environmental
factor. We use the final model in Table 4.8 that isolates the incremental effects on the
council tax.
We then use the segmented model reported in Table 4.10, applying the relevant coefficient
estimates to each individual and then taking an average across individuals for a series of
improvements.
Table 5.1 reproduces the mean ratings for each level of each attribute split by location that
were reported in Table 3.15. We then take the unit valuations from Table 4.9, which are
based on the combined SP2 ratings model, and calculate the valuation of an improvement
from the base of level 1 to level 2, and for successive improvements from level 2 to level 3,
level 3 to level 4 and level 4 to level 5 as appropriate.
Given that the ratings are broadly similar across locations, the valuations follow a similar
pattern. What is apparent is that the valuation for the improvement to the best level is
generally the largest. In some cases, notably litter, fly-tipping and to some extent dog fouling
and chewing gum, it is by far the highest valuation. It is not uncommon in the results
presented in Table 5.1 that the valuations increase with successive improvements.
It is not clear why the ratings, which yield negative valuations, are as they are for dog fouling
and rural dwellers. We also note negative values for improvements in light intrusion and light
pollution for rural dwellers. Presumably this is because this is correlated with the security
and safety aspects of light pollution.
73
Table 5.1: Valuations of Successive Improvements (£ per person per month)
Inner Suburban Rural
Rating Value Rating Value Rating Value
GUM Level 1 0.81 Base 0.87 Base 0.70 Base
GUM Level 2 4.31 7.60 4.25 7.34 3.86 6.86
GUM Level 3 9.07 10.33 9.00 10.31 9.20 11.59
LITTER Level 1 0.73 Base 0.90 Base 0.63 Base
LITTER Level 2 2.31 6.24 2.34 5.69 2.23 6.32
LITTER Level 3 3.88 6.20 3.57 4.86 3.26 4.07
LITTER Level 4 9.23 21.14 9.27 22.53 9.63 25.17
TREES Level 1 4.09 Base 3.11 Base 3.35 Base
TREES Level 2 5.52 3.34 5.66 5.95 5.46 4.92
TREES Level 3 6.24 1.68 6.64 2.29 6.36 2.10
TREES Level 4 7.60 3.17 8.25 3.76 8.08 4.01
FLY-TIP Level 1 0.35 Base 0.29 Base 0.23 Base
FLY-TIP Level 2 0.58 0.85 0.56 1.00 0.35 0.45
FLY-TIP Level 3 2.04 5.42 1.87 4.75 1.82 5.46
FLY-TIP Level 4 9.07 26.10 9.21 27.36 9.43 28.25
GRAFFITI Level 1 1.73 Base 1.88 Base 1.40 Base
GRAFFITI Level 2 2.44 0.40 2.69 0.46 2.02 0.35
GRAFFITI Level 3 3.08 0.36 3.51 0.46 3.09 0.60
GRAFFITI Level 4 5.12 1.15 5.19 0.95 4.32 0.69
GRAFFITI Level 5 8.11 1.68 8.70 1.98 9.00 2.63
FLY-POST Level 1 2.01 Base 2.10 Base 1.91 Base
FLY-POST Level 2 2.70 - 2.50 - 2.29 -
FLY-POST Level 3 3.50 - 3.67 - 3.50 -
FLY-POST Level 4 4.73 - 5.04 - 4.82 -
FLY-POST Level 5 8.21 - 8.50 - 8.87 -
QUIET Level 1 2.49 Base 1.62 Base 2.60 Base
QUIET Level 2 4.80 3.15 4.57 4.03 4.93 3.18
QUIET Level 3 5.36 0.76 5.71 1.56 5.63 0.96
QUIET Level 4 6.01 0.88 7.12 1.93 7.08 1.98
QUIET Level 5 6.69 0.93 7.90 1.06 8.39 1.79
DOG FOUL Level 1 1.71 Base 1.29 Base 2.09 Base
DOG FOUL Level 2 2.23 0.98 1.86 1.08 4.24 4.06
DOG FOUL Level 3 2.76 1.00 2.72 1.62 3.43 -1.53
DOG FOUL Level 4 4.13 2.59 4.15 2.70 3.10 -0.62
DOG FOUL Level 5 7.50 6.37 8.08 7.42 3.74 1.21
ODOUR Level 1 2.55 Base 1.92 Base 2.54 Base
ODOUR Level 2 3.26 1.36 2.98 2.02 3.31 1.47
ODOUR Level 3 4.20 1.80 4.08 2.10 4.54 2.35
ODOUR Level 4 5.89 3.23 5.99 3.65 5.74 2.29
ODOUR Level 5 8.14 4.30 8.76 5.29 8.39 5.06
INTRUSION Level 1 3.24 Base 2.68 Base 6.84 Base
INTRUSION Level 2 4.11 0.29 3.82 0.38 5.57 -0.43
INTRUSION Level 3 5.55 0.48 5.39 0.53 4.86 -0.24
INTRUSION Level 4 7.82 0.76 7.96 0.86 4.62 -0.08
POLLUTION Level 1 3.26 Base 2.61 Base 6.85 Base
POLLUTION Level 2 5.40 1.35 5.53 1.84 6.60 -0.16
POLLUTION Level 3 8.06 1.68 8.62 1.95 5.03 -0.99
74
The figures in Table 5.1 do not make use of our segmented model, but instead use single
parameters across all individuals. The figures are also based on the mean ratings for each
factor level. Nor do they distinguish improvements from a current situation. We therefore
finally report the use of the segmented model‟s results. This uses a sample enumeration
approach based on each individual‟s valuations implied by the segmented model of Table
4.10 and each individual‟s ratings of the factor levels. The resulting mean valuation of an
improvement is an average across individuals‟ valuations in the sample.
One set of improvements is to move the respondent to the next improved level of those
contained in our SP exercise, with those already at the best level facing no improvement and
hence a valuation of zero. The other set of improvements is to move the respondents from
their current level to the best possible.
The results in Table 5.2 are the mean and standard deviation (in parentheses) of the
valuations of the various improvements specified. They are not directly comparable with
those in Table 5.1 since different improvements are being evaluated. However, what is clear
is that when we allow for variations in parameter estimates across areas, rather than just
relying on differences in ratings across areas as in Table 5.1, we observe much more
variation in values across the inner-city, suburban and rural sites. This is to be expected
given the findings of the segmented model in Table 4.10 which indicates large variations in
sensitivity not only across areas but across factors, such as income, that will vary across
areas.
Table 5.2: Valuations of a One Level Improvement and Improvement to Best (£ per person per month)
Inner Suburban Rural
One Level To Best One Level To Best One Level To Best
GUM 1.99 (3.36) 2.10 (3.51) 0.83 (3.6) 0.78 (3.7) 0.08 (2.65) 0.08 (2.65)
LITTER 9.75 (9.22) 15.81 (9.89) 12.85 (12.3) 16.20(12.88) 11.33 (11.12) 12.54 (11.51)
TREES 0.61 (4.56) 1.82 (5.60) 3.11 (5.93) 4.46 (7.94) 2.15 (4.92) 2.95 (6.43)
FLY TIP 8.43 (11.06) 8.70 (11.17) 5.84 (10.78) 6.18 (11.01) 5.02 (10.67) 5.02 (10.67)
GRAFFITI 1.12 (1.94) 2.78 (3.11) 0.83 (2.10) 1.55 (2.83) 0.21 (1.90) 0.29 (1.90)
FLY-POST - - - - - -
QUIET 0.27 (0.91) 0.58 (1.44) 1.03 (3.52) 1.91 (4.07) 0.53 (2.87) 0.60 (3.64)
DOG FOUL 4.16 (6.32) 8.87 (8.53) 5.12 (7.37) 7.79 (9.08) 1.20 (3.18) 2.72 (5.98)
ODOUR 0.87 (1.55) 1.69 (3.08) 2.25 (3.83) 2.70 (4.60) 2.45 (4.10) 4.05 (5.67)
INTRUSION 0.02 (2.04) 0.03 (2.32) 1.58 (3.00) 2.25 (3.76) 0.55 (1.71) 0.57 (1.74)
POLLUTION -0.23 (1.47) -0.26 (1.50) 2.37 (3.70) 2.40 (3.72) 0.07 (0.72) 0.07 (0.72)
75
The relatively large valuation of chewing gum in the inner-city is not only a function of their
higher valuation of a specific improvement but also because there is more scope for
improvement. Similar reasoning applies in the case of graffiti. In contrast the high value of
fly-tipping for inner-city residents, despite the dampening effect of their lower incomes on
valuations, will be almost entirely due to the greater scope for improvement. Compare this
with trees where despite being green places, the suburban and rural areas have higher
valuations in part due to higher incomes but also because of stronger preferences in rural
areas and London where two of the three areas were suburban.
In general, the valuations of improvements in local environmental factors obtained seem
reasonable.
76
6. CONCLUSIONS
This study has conducted innovative research into the valuation that residents place upon
local environmental quality. The valuations are specific to individuals, as opposed to
households, and relate to their experiences of a range of environmental factors in their
locality. The analysis of visitor values or of the valuations our respondents place upon these
factors in other localities was beyond the scope of this study, but could of course be pursued
in further research in the area.
A wide range of variables have been covered, with improvements or deteriorations in them
set alongside each other and reductions or increases in council tax as the logical monetary
instrument in the context of local quality of life. The environmental factors that were the
subject of this investigation were:
urban quiet areas
fly-tipping
litter
fly-posting
graffiti
dog-fouling
discarded chewing gum
trees
light pollution (obscuring the stars)
light intrusion (into the home)
odour
A large sample of 561 respondents covering three cities, and a blend of inner-city, suburban
and rural settings, was obtained. From observing respondents at the group hall test
interviews through to inspection of the data, it is clear that respondents have been able to
engage well with the SP exercises and the data is of good quality.
Two SP exercises were offered, similar in nature and what we have termed the priority
ranking. From the various levels of each variable offered, respondents identify their current
situation. From all those improvements to the current situation, they are asked to select the
most preferred. This is then removed from consideration and they are asked to identify the
preferred improvement from the remaining on offer. The process is repeated until all the
improvements have effectively been ranked in order of preference. The process is then
repeated, in an entirely analogous manner, until all the deteriorations in the attributes on the
current situation are ranked in order.
The first SP exercise included two local environmental variables, namely access to quiet
areas and the contentious subject of dog fouling, alongside what might be regarded to be
more substantive quality of life issues. The latter were local crime levels, local school quality,
road traffic conditions, traffic noise and the general condition of pavements.
The reason for this was to have willingness to pay results for local environmental indicators
without having placed undue emphasis on them. This would be expected to reduce
77
incentives to strategic bias and provide a means of scaling values obtained from the second
SP exercise where there might be a greater incentive to bias responses since the purpose of
the exercise was more transparent. It turned out that we did not feel that there was a
convincing case for any re-scaling.
In terms of the importance of different quality of life factors, rated on a scale from 1
(extremely important) through to 5 (Not at all important), the mean ratings along with 95%
confidence intervals in parentheses were:
Importance of Quality of Life Factors (1=Extremely Important, 5 = Not at all Important)
Level of Local Crime 1.45 (1.38 - 1.52)
Condition of Roads and Pavements 1.88 (1.80 - 1.96)
Amount of Local Council Tax 1.92 (1.84 – 2.00)
Level of Dog Fouling 1.98 (1.90 – 2.06)
Quality of Local Schools 1.98 (1.87 – 2.09)
The Amount of Road Traffic in Your Area 2.00 (1.92 – 2.08)
Access to Quiet Areas 2.00 (1.92 – 2.08)
Neighbourhood Air Quality 2.05 (1.97 – 2.13)
Road Traffic Noise Experienced at Home 2.26 (2.16 – 2.36)
Crime is clearly the most important quality of life factor, but dog-fouling is, amongst the
general population, fourth equal in importance and on a par with the quality of local schools.
Access to quiet areas, our other local environmental factor incorporated within this broader
quality of life scenario, is only slightly less important and is deemed to be more of an issue
than road traffic noise and local air quality which are both of significant concern to policy
makers.
The second SP exercise covered the 11 local environmental factors listed above in addition
to council tax. In six cases, of chewing gum, litter, trees, fly-tipping, graffiti and fly-posting,
photographic representation of the different levels was used, with a verbal description in the
remainder.
Each of the photographic or verbal descriptions of each level of each attribute was rated on
a 0-10 scale. This provides useful information for modelling but also serves as familiarisation
with the levels prior to conducting the SP exercise.
The selection of attributes and their levels, and how best to present them, was informed by
two focus groups and a pilot survey. Extensive changes were made on the basis of these
exploratory stages of the study.
The preferred model is based on respondents‟ ratings of the attributes and their levels. We
therefore obtain valuations of a unit change in respondents‟ ratings. These can be related
back to the categorical levels of each attribute, thereby providing valuations of each level, or
else surveys can be done of residents to determine their rating of different levels of
environmental indicator, which can include levels not covered here, to allow the evaluation of
78
practical schemes in a more flexible manner than is the case by restricting ourselves to
estimating valuations to categorical variables.
In terms of the willingness to pay valuations for improvements in local environmental factors,
all expressed on a common 0-10 scale from bad to good, we obtained the following
monetary values in £s per person per month for a unit change in a rating and for the
movement from the worst to the best position. 95% confidence intervals are given in
parentheses.
Willingness to Pay Valuations (£s per person per month) and Ranking of Importance
Value of a Unit
Change
Value of a
Move from
Worst to Best
Stated
Preference
Rank
Importance
Rating
Rank
Chewing Gum 2.17 (1.96 – 2.38) 21.7 4 7
Dog Fouling 1.89 (1.69 – 2.09) 18.9 6 3
Fly Posting - - - 11
Fly Tipping 3.71 (3.39 – 4.03) 37.1 2 2
Graffiti 0.56 (0.42 – 0.71) 5.6 9 8
Light Intrusion 0.34 (0.02 – 0.65) 3.4 10 9
Litter 3.95 (3.59 – 4.31) 39.5 1 1
Light Pollution 0.63 (0.29 – 0.98) 6.3 8 10
Odour 1.91 (1.72 – 2.10) 19.1 5 6
Quiet 1.37 (1.20 – 1.53) 13.7 7 4
Trees 2.33 (2.07 – 2.59) 23.3 3 5
There are some clear priorities here; the valuations exhibit considerable variation across
attributes and it would have been disconcerting to have obtained similar values.
The largest valuations are quite clearly for litter and fly-tipping. Then there are a series of
attributes with similar „medium‟ valuations. These are trees, odour, chewing gum, dog fouling
and quiet areas. Light pollution, graffiti and light intrusion have relatively minor valuations.
This pattern of valuations seems plausible.
It can be seen that the ranking of the local environment factors according to the Stated
Preference valuations corresponds well with the ordering obtained from the importance
ratings, with a Spearman rank correlation coefficient of 0.77. This is an encouraging finding.
We were able to detect a number of variations in the valuations of local environmental
factors according to socio-economic, attitudinal and location factors. We have used the
results to demonstrate how willingness to pay valuations (in £s per person per month) vary
across circumstances.
79
Valuations (£s per person per month) for Improvements by Area Type
Inner-City Suburban Rural
One Level To Best One Level To Best One Level To Best
GUM 1.99 2.10 0.83 0.78 0.08 0.08
LITTER 9.75 15.81 12.85 16.20 11.33 12.54
TREES 0.61 1.82 3.11 4.46 2.15 2.95
FLY-TIP 8.43 8.70 5.84 6.18 5.02 5.02
GRAFFITI 1.12 2.78 0.83 1.55 0.21 0.29
FLY-POST - - - - - -
QUIET 0.27 0.58 1.03 1.91 0.53 0.60
DOG FOUL 4.16 8.87 5.12 7.79 1.20 2.72
ODOUR 0.87 1.69 2.25 2.70 2.45 4.05
INTRUSION 0.02 0.03 1.58 2.25 0.55 0.57
POLLUTION -0.23 -0.26 2.37 2.40 0.07 0.07
We have not been able to detect any plausible effect from fly-posting, although we note that
this was returned as the least important factor of all those considered. The negative value
for light pollution, and some other low values for light pollution and intrusion, may be due to
confounding effects with security and safe navigation.
The results above seem plausible. However, they will depend upon the particular
improvements under consideration and the situations from which the improvements are
made, as well as the background socio-economic and location characteristics. These imply
much greater variation in values across the sample than if we simply rely on different ratings
of environmental factors across the sample.
The valuation models here reported can be used to provide willingness to pay measures of
the economic benefits of a wide range of improvements to local environmental factors.
These environmental improvements are often very context specific. By conducting a simple
survey of residents, and asking them to rate on a 0-10 scale existing situations and
proposed ones, the models reported here can be used to provide estimated monetary
valuations of the benefits in those circumstances.
80
REFERENCES
A.D. Research Analysis (2008) Defra Environmental Segmentation Mode Guidance for
Qualitative Recruitment: The Combined Block Method, and Quota Questions.
A.D. Research Analysis (2008a) Defra Segmentation: Evaluation of Qualitative Recruitment
Methods:The Combined Block Method.Modified by ITS Leeds
Bateman I.J., Langford I.H., Munro A., Starmer C. and Sugden R. (2000), Estimating Four
Hicksian Welfare Measures for a Public Good: A Contingent Valuation Investigation. Land
Economics, 76(3) 355-373.
Bristow AL and Wardman M (2006), What influences the value of aircraft noise?, Paper
presented at the European Transport Conference.
Defra (2006), Research into quiet areas: recommendations for identification.
Defra (undated), Cleanliness National Indicator (NI195) Manual
http://cleanliness-indicator.defra.gov.uk/manual.aspx
Defra (2010) Flycapture 2009-2010 data
http://www.defra.gov.uk/environment/quality/local/flytipping/flycapture-data.htm
Domencich T.A. and McFadden D. (1975) Urban Travel Demand a behavioural analysis.
North-Holland, Amsterdam.
European Parliament and Council (2002), Relating to the assessment and management of
environmental noise, Directive 2002/49/EC.
Hoinville G. (1971) Evaluating Community Preferences, Environment and Planning A 3 33-
50.
Horowitz JK, McConnell K.E (2002) A review of WTA/WTP studies. Journal of Environmental
Economics and Management 44: 426–447
Louiviere J.J., Hensher D.A. and Swait J.D. (2000) Stated Choice Methods: Analysis and
Application. Cambridge University Press, Cambridge.
Powe N.A., Willis K.G. and Garrod G. (2006), Difficulties in valuing street light improvement:
trust, surprise and bound effects, Applied Economics 38(4) 371-381.
Royal Commission on Environmental Pollution (2009), Artificial Light in the Environment,
Wardman MR and Bristow AL (2008), Valuations of aircraft noise: experiments in stated
preference, Environmental and Resource Economics, 39 (4), 459-480.
81
Welsh Assembly Government (2010) Fly-tipping in Wales 2009-10
http://wales.gov.uk/docs/statistics/2010/100909sdr1422010en.pdf