You are on page 1of 12



P.S Nelson
Booz Allen & Hamilton
J.G Towriss
Cranfield School of Management Centre for Logistics and Transportation



Transport conveys many benefits to mankind but also has negative side effects such as
road casualties, and environmental impacts such as noise, vibration, air pollution, etc.
"Transport is necessary for people and firms, it is important for the economy as a
whole, but it produces a certain number of undesirable effects such as pollution,
congestion, accidents and consumption of scarce resources",(Alexandre, 1995).
Economists argue that at present many of the environmental costs of transport are being
externalised (i.e. not included in the market). Many goods and services have prices that
can be observed in the market place, for example the costs of road construction and
maintenance. Environmental goods and services are not invariably bought and sold in
the market place (e.g. air quality), and are therefore external to it. It is argued that the
effect of these external costs is an imbalance in the market, so that the environment is
under valued or not valued at all. The environment is therefore seen as a free
commodity that can be over utilised. "Excessive depletion of environmental resources
occurs when their utilisation is external to the cost functions of those supplying or
using transport services" (Button, 1990).
In an attempt to internalise these costs, monetary values are increasingly called for, so
that the true cost of man's impact on the environment is considered. Roads have a long
lifetime, therefore in terms of long-term environmental protection the correct decisions
need to be made and a 111 life cost-benefit assessment is required. Environmental
factors are often via a series of quite sophisticated and ingeniousprocedures, examined
in a quantitative manner, but are rarely reduced to the common denominator ofmoney
which is the currency of the standard cost-benefit jamework used for hansport
injastructure appraisal" (Bannister and Button, 1993).

In the United Kingdom the evaluation b e w o r k for the appraisal of trunk road
infrastructure considers both the environmental and economic impacts of a trunk road

scheme with the economic and environmental impacts being separated. The results are
presented in tabular form showing both quantified and unquantified impacts on a
number of different population groups. The decision-maker is therefore presented with
a balance sheet of a large number of disparate elements, from which he/she has to base
a decision.

SACTRA, the Standing Advisory Committee on Trunk Road Assessment, in the UK,
published its report on the "Assessment of Environmental Impacts of Road Schemes" in
1992. The report suggests that a common unit of measurement of economic and
environmental factors would make the decision process simpler and more consistent.


That is that the environmental factors should be incorporated into a cost-benefit

appraisal process, where monetary values are used as relative weights on the different
impacts of transport infrastructure. Support for the use of economic tools to value
environmental impacts is not solely from SACTRA. The following quotation typifies
the support for economic appraisal of the impacts of transport infrastructure on the

whendeciding upon transportpolicy initiatives it is important that environmental

considerations are placed on the same footing as other potential impacts (such as those
relating to mobiliiy, vehicle operating costs, regional development, industrial growth,
etc.). This can be achieved by adopting a common unit of evaluation (OECD, cited in
Alexandre, 1995).
This paper reports upon reasearch undertaken by the authors on the use of Stated
Preference techniques to place monetary values on the environmental impacts of new
road schemes.

Monetary Valuation Techniques

In order to include environmental factors in a cost-benefit framework, monetary values

need to be assigned. Therefore, some method of placing monetary values on

environmental impacts is required. Rendel Planning (1992) suggest that such methods
must be:


robust and technically acceptable;

have institutional acceptability;
have public acceptability;
be user-friendly and cost-effective.

Several methods of monetary valuation have been used in a variety of different fields,
generally on individual projects, e.g. power stations, airports, recreation and amenity
sites. The techniques can be divided into two types, indirect and direct.
i) Indirect Techniques
The methods use either a surrogate market or data from other sources to generate
valuations. The indirect methods are shadow pricing, dose-response, expert opinion and
surrogate market techniques: travel cost method and hedonic pricing. There are a
number of difficulties with theses methods. For example expert opinion has the
weakness that it is assumed that people like district valuers have sufficient knowledge
of factors that influence house prices to provide surrogate market prices as a function of
environmental impacts which are systematically varied. Hedonic pricing suffers from
the weakness that the explanatory variables are often inter-related resulting in problems
of multi-collinearity in the regression analysis. Kroes and Sheldon (in Button, 1993)
state that the limitations of surrogate market methods are as follows:

difficulty in obtaining information on the full range of options open to the

cannot easily handle new forms of environmental impacts or impacts of existing
it is assumed in Revealed Preference that those involved have full information



the options available;

can handle qualitative information, but only in a restricted fashion.

ii) Direct Techniques

These use market research survey techniques to gather valuations directly from
individuals. The direct technique currently most commonly used is the contingent
valuation method (CVM), which asks people directly for their willingness to pay for
environmental improvement in a hypothetical scenario. One of the main weaknesses of
CVM is bias which can arise from variety of sources. Johansson (1990) identified
these as follows:

Strategic bias respondents try to bias their answers to encourage a strategy they
prefer. This deliberate bias or overstatement is described by SACTRA (1992),
"respondentsmay state sums of money which are much greater than those which would
in truth satisfi them in the knowledge that no real charge is being threatened and their
answers may influence policy in a desired direction".
Learning bias can also affect strategic bias in that the simple nature of the CVM
bidding games make them relatively simple to learn. This learning process can allow
strategic bidding and can therefore affect the final bid.
Starting Point bias- the starting point for the bidding game affects the final bid.
Johansson (1990) suggests that the higher the starting biarange the higher the
maximum willingness to pay.

Information bias the type and amount of information provided can affect the final
bid. "Thereis evidence in the CVMresearch literature, too, that the amount and kind of
information provided to the respondents in a CVMsuwey may aflect their response to
valuation questions" (Tunstall and Coker, in Coker and Richards (Ed), 1992). Baughan
and Savill(l994) describe one of the major problems in CVM to be the description of
not only the physical conditions but also their consequences, for example the volume
and traffic and its associated nuisance level.
Method of Payment bias - the payment medium used, e.g. taxes, cash payment, affects
the final bid. A study by Bateman et al. (1993) noted that changing from charitable
fund donation to income tax payment nearly doubled the willingness to pay value.
Respondents preferred the tax as it ensured everyone would pay (cited in Bateman,
CVM can be used to elicit both willingness to pay and willingness to accept values.
Major disparities between these two values have been revealed, although traditional
economic theory suggests that willingpess to accept and willingness to pay values
should be similar. Johanssen (1993) identifies divergences up to a factor of 10 (cited in
Bateman, 1994).

3. Stated Preference Techniques and Valuation

The term Stated Preference (SP) refers to a variety of individual techniques such as
trade-off analysis, transfer price analysis, and conjoint analysis. The techniques use the


results of direct interviews to model behaviour of individuals. ''Stated preference

techniques refer to a number of dzfferent approaches all of which use peoples
statements of how they would respond to dzfferent situations" (Pearmain et al., 1991).
The techniques use controlled experimental designs to construct a series of alternative
situations, individuals are then asked how they would respond to these situations if they
were faced in reality. In the context of the appraisal of road schemes SP techniques are
also used in the existing UK Trunk Road appraisal process to supply monetary values
of time to COBA.
However an extensive literature'review reveals very little research into the use of SP
techniques to place values on environmental impacts. In the main, such experiments
have focused on valuing recreational amenity, such as the study by Louviere et al.
(1987) and a study by Adamowicz et al. (1994) into the values of the attributes of a
recreational site. The literature review reveals only one study carried out into the use of
SP techniques for the valuation of the environmental impacts of road schemes. This
study was commissioned by the Department of Transport, UK, and carried out by JMP
(1996). The study of SP techniques was part of a larger remit to investigate a number of
methods for the monetary valuation of traffk nuisance. Within the context of the
research upon which this paper is based the SP technique refered to is 'Trade
Off 'analysis.

Advantages of SP over the Contingent Valuation Method (

The implicit nature of the valuation from SP experiments gives it a number of

advantages over CVM. The most important advantage is that strategic bias is unlikely
to occur. The complicated designs of SP experiments and the trade-offs offered make it
very difficult for the respondent to strategically influence the results of a SP
experiment. Magat et al. (1988) recognise this pattern in an evaluation of morbidity
risk. The study compares a SP approach with a CVM approach. They argue that the
contingent valuation approach may create incentives for respondents to state values
which are somewhat below their true reservation prices for the commodities being
valued, while the paired comparison (SP) approach eliminates these incentives to
understate preferences and thus it seems to provide more accurate measures of
willingness to pay (Magat et al., 1988). SP techniques should also be less open to
learning bias, for similar reasons, i.e. the complexity of the experimental format. If
learning should occur, it is likely that this would be a much slower process than with
the CVM method, as it is a more difficult system to master.

It is unlikely that starting point bias will occur. A SP experiment is designed with
various levels for each attribute (e.g. environmental factors and cost factors). The levels
are designed, piloted and tested to ensure that they are within boundary values, i.e.
respondents are facing a situation in which they will trade off one attribute against
Method of payment bias (i.e. if a tax or cash payment is proposed) and information bias
(i.e. the method of representation, description and amount of information) are equally
likely to apply to SP as to CVM. The impact of the type, representation, and amount of
information supplied to a respondent is vital to the validity of both SP and CVM


SP techniques, therefore, have all the advantages of the CVM when compared to the
indirect valuation techniques. However, they also have potentially significant
advantages over CVM, especially in terms of eliminating bias in response.

The Design of the Stated Preference Research

The road scheme chosen for the research was the A428 Bedford Western By-pass, the
scheme being particularly suitable because it impacted upon a large and relatively
dense population and also because some of the population experienced an
environmental loss whilst others experienced a gain. The following section details the
design of the Stated Preference research. The basic method employed was non-adaptive
'trade off In order to concentrate on the methodological aspects of the SP technique,
the experiments concentrated on a few environmental attributes only. This is in contrast
to the JMP study that investigated 12 different policy impacts. The smaller number of
attributes meant that the research could focus upon the SP methodology, analysis and
statistics in greater detail than is possible with a wider ranging survey. This simpler
format also allowed the SP experiments to be presented to the respondent as a pen and
paper exercise rather than using computer based methods.

The non-adaptive approach to the SP experiments requires that the SP designs have to
be tailored, so that respondents face scenarios showing the existing situation and
scenarios that they could reasonably expect in the future. This tailoring of the
experiment, used existing and perceived levels of attributes so that options could be
built around existing experience (Ortuzar and Willumsen, 1994). The tailored design
ensured a more realistic choice context, and with the aim of obtaining more reliable
results. In order to tailor the experiments in this way a range of pre-SP data collection
and analysis was required. These were included in the three main stages of the research

pre-SP data collection and analysis, and design of SP experiments;

SP Fieldwork
analysis of SP experiments and conclusions.

5.1 Pre-SP Data Collection and Analysis

Very little research has been conducted to identify people's salient attributes relating to
the impact of transport on the environment. These issues are also likely to differ
between geographical areas. A SP design required information that related directly to
respondents within a particular geographical area or socio-economic grouping, etc. This
ensured that the hypothetical SP choice scenarios were realistic and used existing or
perceived levels of attributes so that options were built around existing experience
(Ortuzar and Willumsen, 1994). An initial pre-SP data collection phase was necessary
to gather these base data. The initial data collection phase included discussion groups
and an attitudinal survey of residents of the study area.
5.1.1 Discussion Groups
Initially, a number of discussion groups were used to gain the required base
information for the SP scenarios. The discussion groups provided background
information, so that the hypothetical SP scenarios were based on realistic assumptions,


i.e. to ensure that the SP experiments did not contain elements that were not important
to the respondent. In order to construct a realistic set of SP choice situations it was
necessary to identify the environmental attributes that most affect or annoy people and
the levels of these impacts that affect households. It was also important to identify a
payment mechanism to which people could easily relate For this purpose, several
discussion groups were employed to discuss these issues, and to isolate the
environmental attributes for the SP survey. The discussion groups were designed to
discuss the following issues:
attitudes to the environmental consequences of transport:
which environmental impacts are most important?
how do people perceive differences in traffic levels and the impacts on the
attitudes to payment for environmental gain, or compensation for environmental
are people willing in theory to pay for a better environment?
do people connect payment (e.g. tax) with improvements in the environment?
what method of payment is most easily understood?
approximately how much are people willing to pay?
A secondary stage of pre-SP data collection was required to gather more detailed
information. The data from the discussion groups was used to design an attitudinal
survey to gain further information for the SP experiment.

5.1.2 Attitudinal Survey

The attitudinal survey used the results of the discussion groups to define a more
structured survey, in order to gain more detailed information from the local population.
For example, the environmental impacts defined in the discussion groups were placed
in a rating exercise so that the most important impacts could be identified.
The data gained from the discussion groups and attitudinal surveys were as follows:

definition of important attributes;

definition of attribute levels, particularly monetary levels;
information on how environmental impacts are perceived;
definition of method of payment.

5.2 The Stated Preference Survey

The survey was conducted at residences in the Kempston and Queens Park areas of
Bedford, with both areas experiencing reductions in traffic flow through the opening of
the by-pass. The attributes included in the SP experiment were:

Road Safety

Air Pollution



Road Tax
Journey Time to Central Bedford

The first two are environmental attributes carried forward from the discussioii groups
and attitudinal surveys. Journey Time was included as a control variable in order for
values of time obtained from the research to be compared with other studies for
validation purposes. Road Tax was included as the monetary attribute in the design.
The attribute levels are shown in Table One
Table One Attribute Levels

Air Pollution

Road Tax

Journey Time to Bedford

The experimental design together with the associated attribute levels is shown in Table
Table Two Experimental Design and Attribute Levels







5 mins
10 mins
10 mins
10 mins
5 mins
10 m i n s

Respondents were asked to respond to a discrete choice experiment between the

approximate present situation, the base card (shown on an orange card) and a potential
new situation (shown on a series of white cards). The new situations were always better
in terms of environmental attributes, but had an additional cost component.
Respondents were therefore, being asked to imply a willingness to pay for an


environmental improvement. In order to improve the reliability of the data respondents

were asked to consider the new situations as being due to a by-pass being built around
Bedford that would reduce the amount of traffic passing their house.

A functional measurement of rating responses was used to measure the utility of each
alternative. This required respondents to express the strength of their preferences on a
numerical scale. The points on this scale were defined by the following phrases:
Definitely prefer orange card
Probably prefer orange card
No preference both the same to me
Probably prefer white card
Definitely prefer white card
A majoi sment of the research was to investigate how responses to the SP might vary
with di- *mtforms of representation of the attributes. Four types of representation
were used:

A graphical method
Pictures (photographs)
Pictures (photographs and text)

A discussion of the results of this component of the research is outside the scope of this
paper and in the following section aggregate results relating to the total data set are

The Results of the Stated Preference Research

As previously stated the SP experiments were designed using the discrete choice
method. Respondents were asked to choose between two cards, representing a base
present situation and a hypothetical future situation. The responses were recorded using
a 5 point rating scale. This method allowed the use of multiple least squares regression
to calculate the utility model for each representation type. A logit model was used to
derive the parameters for the utility. model. The parameters, estimated by linear
regression, for the total dati set after removing non-trading and illogical respondents
are are shown in Table Three.
Table Three Utility Model Parameter Values -Full Data Set

R = 0.25

Respondents = 222

Significance F = 0.00

T Signif.


@ p4.05

Road Safety




Road Tax





21.12 / traffic

Journey Time




E5 1.97 I hour

Air Quality




El. 19 I % dec. in traffic





The journey time attribute was included in the model to validate the results obtained.
Therefore, the non-significance of this attribute had a significant effect on the
interpretation of the results, in the sense that it was not possible to gain a value of time
for comparison with other studies. Therefore, it is not possible to validate the
environmental utility model values in this way. However an indication of the validity of
the values can be gained by comparison with the CVM results from the attitudinal
survey and also with results from the JMP study. Table Four shows this comparison.

Table Four - Comparison with JMF' and Attitudinal Study

50% Decrease in Road Traffic

Value per Year

SP Results Total Data Set:

Air Quality


Road Safety




Attitudmal Study
Willingness to Pay


J M P Study


The results of the SP experiments compared well with the results from the attitudinal
survey willingness to pay questions. The total value for air quality and road safety was
similar to that of the CVM result. However, the CVM result may have been expected to
be higher due to the impact of strategic bias, people may have been expected to
overstate their willingness to pay. Also although air quality and road safety were the
most important factors in the attitudinal survey other important factors were also
recorded (these were not all included in the SP experiments, e.g. noise, vibration).
Therefore, it would be expected that the CVM willingness to pay value would be higher
than the SP valuations.
The J M P figure was almost double that gained in the SP experiments, two possible
reasons for this are:
the JMP study was an overall figure, therefore it 'may include other attributes
than air quality and road safety;
the difference in location, time of survey, road environment might have
on the results gained.




This paper has introduced SP techniques as a method for placing monetary values on
environmental impacts. It has shown that SP techniques are applicable to the area of
environmental valuation. Respondents were able to respond logically to a carefully
designed and controlled experiments and statistically significant monetary valuations
were gained for measurable levels of air quality and road safety. SP has therefore been
revealed as a potentially viable method for valuation of environmental impacts of road
transport. However, before the methodology can achieve the public and institutional
acceptability that it requires, for inclusion in an economic appraisal process, further
research is required to refine the methodology.
Problems and issues have arisen from the research upon which this paper is based and
these need to be resolved before SP methods can be used as a means of including
monetary valuations of environmental costs and benefits within an economic
framework of trunk road appraisal The main areas of concern that need to be addressed
in future research are:

The selection and particularly the measurement o f attributes and

attribute levels - one method may be to use attitudinal scales to relate
measurable changes in traffic levels or pollution levels to attitudes ( in the
Bias - further research is required into forms of bias as a whole, but particularly
in SP experiments the bias due to the impact of method of payment and further
investigation into the impact of information bias brought about by different
forms of representing attributes. Further research is required to identify the
pattern of this bias and why it occurs, as this has a significant impact on the
monetary valuations derived.

More case studies finally more case studies of monetary valuation of

impact via SP are required. Further studies, may resolve
some of
the issues raised in this research, but also perhaps more importantly
provide a
pool of monetary values for comparison and possible future use in road
investment appraisal.


Adamowicz, W., Louviere, J. and Williams, M. (1994). Combining Rescaled and
Stated Preference Methods for Valuing Environmental Amenities. Journal of
Environmental Economics and Management, 26, pp 271-292.
Alexandre, A. (1995). The Need to Reconcile Transport and the Environment. Global
Transport, Spring 1995, Chartered Institute of Transport.
Bannister, D. and Button, K. (1993). Transport the Environment undSustuinuble
Development. Spon: London.
Bateman, I. (1994). Contingent Valuation and Hedonic Pricing: Problems and
Possibilities. Landscape Research, 19(1), 1994.
Baughan, C. J. and Savill, T. A. (1994). Contingent Valuation Ouestions for Placing
Monetarv Values on Traffic Nuisance. Transport Research Laboratory Project Report
90 (unpublished).
Button, K. J. (1990). Environmental Externalities and Transport Policy. oxford
Review of Economic Policy, 6, pp 61-75.
Button, K. (1993). Transport the Environment undEconomic Policy, Elgar.

JMP Consultants Ltd. (1996). Monetarv Valuation of Traffic Nuisance. Department of

Transport, UK, (unpublished).
Johansson, P-0. (Spring 1990). Valuing Environmental Damage. Oxford Review of
Economic Policy, Vol. 6, No. 1.
Louviere, J. J., Schroeder, H., Louviere, C. and Woodworth, G. C. (1987). Do the
Parameters of choice Models depend on Differences in Stimulus presentation: Visual
versus Verbal Presentation. Advances in Consumer Research, Vol. 14.
Magat, W. A., Viscusi, W. K. And Huber, J. (1988). Paired Comparison and
Contingent Valuation Approaches to Morbidity Risk Valuation. Journal of
Environmental Economics and Management, 15, pp 395-411.
Ortuzar, J. de D. and Willumsen, L. G. (1994). Modelling Transport. 2*Edition,
Wiley, New York.

Pearmain, D., Swanson, J., Goes, E. and Bradley, M. (1991). Stated Preference
Techniaues: A Guide to Practice. 2dEdition, Steer Davies Gleave and Hague
Consulting Group.
Rendel Planning and Environmental Appraisal Group, University of East Anglia
(1992). Environmental Amraisal: A Review of Monetarv Evaluation and Other
Techniaues. Transport Research Laboratory Contractor Report 290.


SACTRA (March 1992). Assessing the Environmental Impact of Road Schemes.