DEFRA Bathing Waters Valuation of Benefits Report€¦ · 6/25/2002 · bathing water contact...

Valuation of Benefits to England and Wales of a Revised Bathing Water Quality Directive and Other Beach

Characteristics Using the Choice Experiment Methodology

Final report submitted to

Department for Environment, Food and Rural Affairs

by

Economics for the Environment Consultancy Ltd

25 June 2002

Valuation of Benefits to England and Wales of a Revised Bathing Water Quality Directive and Other Beach Characteristics Using the Choice Experiment Methodology

eftec i 25 June 2002

EXECUTIVE SUMMARY • The overall objective of this study was to find out people's willingness to pay (in

England and Wales) for changes in water quality and other beach attributes that are likely to arise from implementation of a revised EC Bathing Water Directive.

Study Methodology • The valuation technique used in the study was the choice experiment

methodology, which is one of a number of stated preference techniques used to find the economic value of environmental goods and services that are not reflected in actual markets. The choice experiment approach estimates individuals' preferences, in monetary terms for individual attributes and marginal changes in these by providing respondents with a number of choice sets and asking them to state their most preferred choice.

• The choice sets used in this study consisted of different beach scenarios described

by five attributes: water quality, advisory note system, litter/dog mess, safety and amenities and additional water charges per year, where litter/dog mess and safety and amenities are independent attributes, not related to the revision of the Directive.

• Water quality was described in two different ways, giving rise to two versions

of the questionnaire, A and B. In version A, water quality was described as the risk of suffering a stomach upset from bathing. Whereas, in version B, water quality was described as the number of days in the bathing season when it was considered unsafe to swim due to an elevated risk of suffering a stomach upset.

• Each of the choice sets consisted of three scenarios: the current situation on the

average beach in Britain, and two hypothetical scenarios, which represented improvements over the current situation.

• In each of the scenarios the attributes took on different levels. Water quality

took on three levels of either risk or number of unsafe days; advisory note system was either present or not; there was some or none litter/dog mess; and safety and amenities were either average or good. The different price levels for the extra water charges were £0, £3, £11, £15 and £25 per household per year.

• In line with standard practice, the choice experiment questionnaire designed for

this study consisted of the following four sections: (a) attitudes, opinions, knowledge and use; (b) valuation scenario; (c) debriefing and follow-up questions; and (d) socio-economic characteristics.

• The questionnaire was tested through the use of focus groups and a pilot survey

before the main survey was administered through face-to-face interviews to a representative sample of 809 people throughout England and Wales.


eftec ii 25 June 2002

• The summary of socio-economic characteristics of the respondents shows that the sample is representative of the population of England and Wales.

Results • 97% of the respondents had visited a beach in Britain sometime in their life.

Based on the number of visits in the last year, the average beach visitation rate is 5.69 visits per person per year.

• Walking, relaxing, sunbathing and picnicking were found to be the most

commonly practised beach activities, while 57% of respondents and their families never go swimming.

• Using both the responses to the questionnaire and epidemiological evidence, the

study estimated around 1.3 million cases of stomach upset every year in England and Wales could be associated with bathing in faecally contaminated bathing water. A second calculation based on respondents� personal assessment of cases of stomach upset revealed that this figure could be upwards of 2.2 million cases.

• 51% of the sample thought that the proposed advisory note system would be

very useful. 38% of the sample preferred to see a simple symbol indicating "do not bathe", while 51% preferred to see more detailed information.

• 78% of the sample indicated that they would still visit the beach if an advisory

note advised against going in the water because of poor water quality. This allays fears that the advisory note system would have a significant adverse effect on beach visitation.

• Respondents were willing to pay between £1.10 and £2.00 per household per

year for a 1% reduction in the risk of suffering stomach upset. For a 2.3% reduction in risk, which is the likely reduction due to revisions in the Bathing Water Directive, this equates to a minimum of £61 million per year for England and Wales.

• Respondents were willing to pay between about £5.6 and £13.7 per household

per year for the introduction of an advisory note system. This equates to a minimum of £134.4 million per year for England and Wales.

• Respondents were willing to pay between about £6 and £11 per household per

year for avoiding the presence of some litter / dog mess on the beach. This equates to a minimum of £144 million per year for England and Wales.

• Respondents were willing to pay between around £2.5 and £7.3 per household

per year for improving the standard of amenities and safety from average to good. This equates to a minimum of £60 million per year for England and Wales.

• A number of tests were undertaken to assess the validity of the results and

determine whether any potential biases were at play. These tests all show that the results are free from any bias, and are supported by several strong arguments for their validity.


eftec iii 25 June 2002

• The modelling results showed that all beach attribute coefficients are significant at the 1% level and have the expected signs across all versions and model specifications.

• Transferring the benefit estimates of this study to other countries and contexts

should be undertaken with caution.


eftec iv 25 June 2002

TECHNICAL SUMMARY In December 2000, the European Commission released a Communication (COM(2000)860) proposing a number of changes to the current Bathing Water Directive. The changes were a response to improved scientific knowledge, the need for more active quality management of bathing waters and a desire for improved public information. The revised Bathing Water Directive is expected to: • tighten microbiological standards; • harmonise beach management regimes across the EU; • require more public information on water quality; and • widen the scope of designated bathing waters to include waters used by non-

bathing water contact sports such as surfing, windsurfing and slalom kayaking. The relevant policy question in the context of this study is whether the potential benefits of the revision justify the costs arising from a revised Directive. Many of the benefits that the public currently enjoys at beaches, and may see improvements to through implementation of the revisions, are not captured in actual markets. Therefore, the policy question can only be answered by a study undertaken to elicit individuals' preferences for these improvements, which is the particular focus of the work presented here. However, most of the improvements and the impacts these will have on people's enjoyment of bathing waters and beaches are not known with certainty. What is known is the type of attributes of beaches that might be affected by the revision. Therefore, this study uses the so-called choice experiment methodology in order to estimate how much people value changes to those beach attributes. The choice experiment is part of the group of survey techniques, collectively referred to as stated preference techniques. By using this methodology it is possible to arrive at an implicit price and ranking of the various beach attributes. This document reports on the design, implementation and analysis stages of the study. This technical summary also follows the same order of stages, namely: 1. initial research: main objectives and literature review; 2. choice of survey method and valuation techniques; 3. choice of population and sample; 4. questionnaire design; 5. testing the questionnaire, which includes conducting focus groups, a pilot survey

and the main survey; 6. statistical and econometric analysis; 7. validity and reliability testing; and 8. aggregation. TS.1 Initial research As mentioned above, the overall objective of this study is to find out people's preferences for those beach attributes that are likely to be affected by the revisions to the Bathing Water Directive. Within this overall objective, the terms of reference of the study set out the specific tasks the study should address as:


eftec v 25 June 2002

• estimate individuals' willingness to pay for aspects of beach use in England and

Wales measured through a number of beach attributes; • discuss the scope for ensuring the validity and transferability of the estimated

values; and • produce an estimate of total benefits from the revision of the Bathing Water

Directive to England and Wales. As part of the initial research, a literature review was undertaken. The review shows that the economic literature on the recreational uses of beaches focuses mainly on water-based recreational activities such as angling, swimming and boating. Very few studies look at people's preferences for water quality improvements and none actually considers the changes that are likely to occur if a revised Bathing Water Directive is implemented. Section 2.2. presents the current evidence from the economic literature for background information. Some crosschecks between the existing literature and this study are presented in Section 5.2. TS.2 Choice of survey method and techniques Sections 1 and 2 of the main report put forward the arguments for the recommended use of a choice experiment methodology for this study. One such argument is that the choice experiment technique allows the relative value of a number of changes in beach attributes to be assessed, and that these values can be used to portray a number of different scenarios (in the absence of precise information about the expected outcome of the policy change). Adopting the choice experiment approach also means that the face-to-face interview method has to be used. Choice experiment questionnaires are too complex to be administered through alternative interview methods, namely, by telephone or mail. Therefore, the questionnaire designed for the study has been administered through one-to-one interviews that took place at respondents' homes. TS.3 Choice of population and sample The population of concern for this study is the population of England and Wales. This includes both those who make use of beaches and bathing waters as well as those who do not, but nonetheless may have preferences for quality improvements. To use the economic terminology, the population consists of both 'users' and 'non-users'. To ensure that the survey was representative, the sample was chosen randomly across England and Wales. The geographical selection of sampling points also ensured that both those who live in vicinity of beaches and those who do not were sampled. The sample size in any stated preference study depends on the complexity of the design as well as the different sub-groups within the population the study aims to survey. In this study, design complications meant that there had to be five versions of the questionnaire requiring at a minimum a sample of 800 respondents. In the event, the total number of completed surveys was 809.


eftec vi 25 June 2002

TS.4 Questionnaire design In line with standard practice, the choice experiment questionnaire designed for this study consists of the following four sections (see Section 3.1 of the main report for details). Section A (attitudes, opinions, knowledge and use): contains questions on the types and frequency of uses of British beaches, preferences for beach characteristics, knowledge of beach and water quality standards and management schemes and attitudes towards risks related to beach use. The main reasons for asking these questions were to: (i) help respondents explore their thoughts and attitudes; (ii) reveal the most important underlying factors driving respondents' preferences; (iii) provide tests of validity for the survey design and results; and (iv) gather some data for use in the aggregation of results such as total beach visits per person per year. Section B (valuation scenario): presents the respondents with the characteristics of a typical beach in Britain using five beach attributes described as they occur currently at the �average British beach� and under two hypothetical scenarios, which represent improvements over the current situation. These three choices are presented in 'choice cards' and each respondent was presented with the same eight choice cards. The attributes and the levels that they take, which were selected as a result of several rounds of tests of the questionnaire design (see below), are: • water quality (expressed as risk of suffering a stomach upset1 and days in the

bathing season that it is unsafe to swim due to poor water quality); • advisory note system2 (its presence or absence); • cleanliness of the beach (presence or absence of dog mess and litter); • quality of safety services and amenities (average or good); and • the cost of maintaining the beach scenario (payable through an increase in water

charges of £3, £11, £15 or £25 per household per year). The two different expressions of the water quality attribute led to two versions of the questionnaire: A and B, which are otherwise identical. In addition, different starting levels of the water quality attribute were presented in the current situation scenario to test whether a different 'starting point' had an effect on the respondents' choices. The results of this test are presented in Section 5.1. These starting points also represent the range of possible water quality levels across beaches in Britain (see Section 3.1.2 for more information on how these values were arrived at). These variations resulted in

1 The questionnaire defines 'stomach upset' as follows: �The possible symptoms of stomach upset include mild diarrhoea, indigestion with fever, nausea with fever, vomiting and diarrhoea. It is thought that the majority of cases of stomach upset that result from swimming in polluted seawater are at the milder end of the spectrum of symptoms. In fact, the chance of getting a stomach upset that would require a visit to the doctor, day(s) off work or admission to hospital would be no different if you did not bathe.� 2 COM(2000)860 on Developing a New Bathing Water Policy states that the [revised] Directive �should require the competent authorities to adopt new methods to actively inform the public about bathing water quality, including about all known factors that might influence that water quality. This information should be available at all times at the bathing zone.� Based on this description an advisory note system was defined in the questionnaire as a system of notices which would inform bathers of when it is considered safe or unsafe to bathe.


eftec vii 25 June 2002

five versions of the questionnaire that were tested in the main survey, as shown in Table TS.1. Table TS.1: Levels for the Attribute of Water Quality in Different Versions of the Questionnaire VERSION A VERSION B Chance of getting a stomach upset from

going in the water Number of days when it is unsafe to swim

(out of 150 days in the bathing season) Current

Situation Scenario

A Scenario

B Current

Situation Scenario

A Scenario

B A1 7/100 1/100 0.5/100 B1 16/150 5/150 1/150 A2 5/100 1/100 0.5/100 B2 10/150 5/150 1/150 A3 2/100 1/100 0.5/100 Section C (debriefing and follow-up questions): contains questions investigating the reasons why respondents were or were not willing to pay for improvements in each of the attributes and their general views on the scenarios presented. Section D (socio-economic characteristics): contains questions on income, education, household structure and so on. Information on these characteristics is used to draw a profile of the different types of respondents and to explain the individual variations in willingness to pay responses through econometric analysis. The full questionnaire can be found in Annex 1. TS.5 Testing the questionnaire The questionnaire was tested through focus groups, peer review, pilot survey and finally through the main survey (which is itself a large-scale test). Sufficient time was allowed within and between each test in order for lessons to be taken into account and revisions to be made. Prior to the focus groups, several discussions between the study team and the steering group resulted in the first draft of the questionnaire. Four focus groups, which are discussions conducted by a moderator, were held in London and Norwich in January 2002. As a result of these discussions, a number of key changes were made to the questionnaire. Peer review of the questionnaire and the pilot testing were implemented in tandem. The peer reviewer was Professor Nick Hanley from the University of Glasgow. His comments were insightful and useful in refining the final version of the questionnaire. Two 25-interview pilot surveys took place in February 2002. The pilot surveys helped to identify the questions that were found by respondents to be too long or complicated, and confirmed the need for a worked example in the valuation section as suggested by Professor Hanley. In general, the pilot study worked well with regression analysis producing coefficients that are significant at the 5 or 10% level.


eftec viii 25 June 2002

An overview of these tests is provided in Section 3.2 with details on focus groups and pilot surveys appearing in Annexes 2 and 3, respectively. TS.6 Statistical and econometric analysis Section 4 of the main report presents statistical summaries of the socio-economic characteristics of the sample and responses to the attitudes and beach use questions as well as econometric analysis of the choice questions. The latter results in benefit estimates for the marginal changes described for each attribute per household and per year. Sixty-two per cent of the respondents found the questionnaire interesting. The summary of socio-economic characteristics of the respondents showed that the sample is representative of the population of England and Wales. Ninety-seven per cent of the respondents had visited a beach in Britain sometime in their life. Based on the number of visits in the last year, the average beach visitation rate is estimated to be around 5.69 visits per person per year3, while the median rate is 3 visits per person per year. This compares with the average the 1998 UK Day Visits Survey (National Centre for Social Research, 1998) figure of 4 visits per year per person. Walking, relaxing, sunbathing and picnicking are the most commonly practised beach activities, while 57% of the sample said they never go swimming. Among the characteristics they were presented with, the respondents chose cleanliness of the beach, water that looks clear of foam and litter and clean changing facilities as the most important ones. A closer look at the data reveals that 51% of respondents considered that water quality was very important. Based on the responses to the questions about contact with water (number of times head is submerged under water), two rates of potential illness can be estimated. The first one (predicted rate of illness) used the number of swims data gathered from the survey in conjunction with the epidemiological data4 and estimated there could be 1.3 million cases of stomach upset per year in England and Wales associated with swimming alone, or 2.84 million for all water-based activities. The other (self-assessed rate of illness) was estimated using the responses to a question that asked the number of stomach upsets an individual suffered, which were believed to be caused by poor quality bathing water. This estimate is 2.2 million cases of stomach upset per year in England and Wales. The convergence of the two types of results shows that there was consistency between personal assessments of bathing water-related illness and predictions from the epidemiological literature. Responses to choice experiment questions (in the valuation scenario) were analysed using an indirect utility function. This function explains the satisfaction that an individual receives from the state of an average British beach by the attributes of alternative beaches, observable influences on the individual's choice and an error term that captures unobservable factors. Two types of models have been used in this analysis. First a simple conditional logit model was tested. However, the conditional logit model failed the test for compliance 3 This is the trimmed mean value (mean of the 10th to 90th percentile). 4 The average risk level used is 43 stomach upsets per 1000 exposures (4.3%), as estimated by the Environment Agency (2002) using the WHO approach (WHO, 2001) to disease burden estimation.


eftec ix 25 June 2002

with the Independence of Irrelevant Alternatives (IIA) property upon which the model critically relies. This property states that the relative probabilities of two scenarios being selected (from the choice set) are unaffected by the introduction or removal of other alternatives. Since this was not the case, other more advanced model specifications that do not rely on the IIA assumption were required. The model used for the second round of analysis was the nested logit model. The nested logit model does not rely on the IIA assumption. In applying the nested logit framework to the data, the choice problem is recast as a hierarchical nested structure. In this way, choices are analysed in two stages. The first stage of the model predicts whether or not respondents are willing to pay something towards improved beaches, and the second stage predicts which of the two beach improvement scenarios they prefer. Three variations on both models were tested (the results are shown in Table TS.2 below). The first is a basic model, which explains respondents� choices between beach scenarios solely as a function of beach attributes. The second model uses an alternative specific constant (ASC) for the third beach scenario in each choice set. This constant captures any variation in choices that is not explained by either the beach attributes or respondent-specific socio-economic variables. The third model includes socio-economic and attitude regressors specific to individual respondents. In total the following variables were used in the regression: ASC; chance of getting a stomach upset; days in the bathing season in which it is unsafe to swim; presence or absence of an advisory note system; presence or absence of litter and dog mess; the level of amenities and safety provisions; water rates; socio-economic class; age; gender; income; whether or not respondents engage in swimming or water sports often; and whether respondents thought that water quality was a very important factor in choosing a British beach. All models performed well. The results show that all beach attribute coefficients are significant at the 1% level and have the expected signs across all versions and model specifications. This implies that higher levels of water pollution are undesirable as is the presence of litter/dog mess; while increased amenity quality and the presence of an advisory note system have a positive impact on utility. And, of course, an increase in water charges had a negative impact on utility. Results show that richer respondents, those who are often in contact with water (via swimming or water sports) and those that consider water quality to be a very important determinant of their choice of a beach (in version A) are more likely to choose the most expensive beach improvement scenario (which also afforded the largest improvements). These results are consistent with expectations. Age and gender were not found to be significant determinants of choice. Using the coefficients estimated in the models, point estimates of willingness to pay for a change in one of the beach attributes can be calculated by estimating the marginal rates of substitution between the attribute of interest and the price attribute. This is the rate at which the respondent is willing to trade off money for improvements in beach attributes and is estimated by dividing the coefficient of the attribute by the coefficient of the price attribute. Table TS.2 shows the estimated range of mean WTP values for beach attributes for questionnaire versions A and B


eftec x 25 June 2002

and for both logit models. The ranges arise from the three different model variations tested within both conditional and nested logit models. The nested logit model with ASC35 and individual specific variables has the best fit, and the higher WTP results. The conditional logit model results are presented here for completeness; however, only the nested logit results should be employed. Table TS.2: Estimated Range of Mean Values for Beach Attributes Across Versions and Models (£ per household, per year)

Attribute

Definition

Conditional Logit Model

Range of Mean WTP

Nested Logit Model Range of

Mean WTP

Version A Version B Version A Version B Water A Value of decreasing

the chance of getting a stomach upset by 1 in 100 (over the range:

0.5/100-7/100)

0.7-1.2

_

1.1-2.0

_

Water B Value of reducing �unsafe to swim�

bathing day by 1 (over the range: 1- 16 days)

_

0.8-1.0

_

0.9-1.1

Advisory Value of having an advisory note system in place, advising on poor water quality

days

6.7-16.6

4.8-11.6

7.2-13.7

5.6-12.1

Litter Value of avoiding the presence of some litter

/ dog mess

6.1-11.8

5.9-9.8

6.8-11.1

6.0-10.0

Amenity Value of improving the standard of

amenities (toilets/showers) and

safety (life-guards and equipment) from average to good

3.0-8.7

3.4-6.1

2.5-7.3

3.4-6.2

These estimates indicate that for example, respondents were willing to pay between £1.10 and £2.00 per year for a 1 in 100 reduction in the chance of contracting stomach illness from British bathing water (over the range of risk reduction from 0.5 in 100 to 7 in 100). Interestingly, this value range is almost identical to the 90 pence to £1.10 that respondents are prepared to pay to avoid 1 day of poor water quality. Given that we do not know the health risk relationship between a 1 in 100 chance of suffering a stomach upset and an unsafe bathing day we cannot make any judgements about this

5 ASC3 refers to the constant associated with the third scenario in the choice sets. See Glossary for further details regarding Alternative Specific Constants.


eftec xi 25 June 2002

outcome. However, there is no reason why these two different measures water quality improvement should not provide the same measure of welfare. Table TS.2 also shows that the WTP to eliminate dog mess/litter and to have an advisory note system commanded the highest value ranging from £5.6 to £13.7. As these are household yearly values for all British beaches, even these larger figures seem to be well within the bounds of reality. A number of follow-up questions were asked after the valuation section. Those who chose the current situation in every choice set, about 35% of the sample, were asked their reasons for being unwilling to pay for improvements. About 63% of these were invalid or protest responses such as objections to paying higher water charges or belief that government should pay. These, which consist of 21% of the total sample, were excluded from the analysis. This exclusion does not lead to biased WTP estimates, as explained in Section 5. Those who chose Scenario A or B, i.e. some level of improvement from the current situation, on at least one of the choice sets were asked why they were willing to pay for improvements. The most popular reason was 'I would like to improve beach and bathing water quality for everyone to enjoy'. TS.7 Aggregation, Validity and Transferability Employing the most conservative (lower-bound) estimates from the nested logit model, we are able to calculate aggregate benefits of the range of beach improvements for England and Wales. These are presented in Table TS.3 per year as well as the present value of these benefits over a 25 year time horizon (both 3.5% and 6% discount rates are used). We can assume that the risk of stomach upset from bathing at beaches in the 2001 bathing season was 4.3% per swim (calculated using the WHO methodology). If the Commission suggested mandatory standard of 200 intestinal enterococci, at 95 percentile compliance, were to be adopted then we would see a reduction to 2.0% risk per swim, i.e. a 2.3% reduction in risk per swim at UK beaches. Therefore, using a 6% discount rate, we find the present value of this improvement to be about £776 million over 25 years. This compares to £1.7 billion for the introduction of an advisory note system, £1.8 billion for the avoidance of dog mess and litter and £767 million for the improvement of amenities from average to good.


eftec xii 25 June 2002

Table TS.3: Aggregate Benefits of Beach Improvements (per year and present value) Attribute £ /hh/yr £ million/yr* PV6

£, million (r=3.5%)

PV6 £, million (r=6%)

Change in risk of suffering a stomach upset (% reduction in risk) 1% 1.10 26.4 435 337 2% 2.20 52.8 870 675 2.3% 2.53 60.72 1,001 776 2.5% 2.75 66 1,088 844 3% 3.30 79.2 1,305 1,012 4% 4.40 105.6 1,740 1,350 Change in number of safe bathing days (increase in safe days) 1 0.90 21.6 356 276 5 4.50 108 1,780 1,381 9 8.10 194.4 3,204 2,485 11 9.90 237.6 3,916 3,037 15 13.5 324 5,340 4,142 Advisory note system 5.60 134.4 2,215 1,718 Avoiding the presence of some litter/dog mess

6.00 144 2,373 1,841

Improving the standard of amenities from average to good

2.50 60 989 767

* 24 million households in England and Wales. Validity of results Two kinds of validity test were applied to the results: content validity and construct validity, where validity refers to the degree to which a study succeeds in measuring the intended values by overcoming potential biases and the hypothetical nature of the valuation exercise. Further discussion is presented in Section 5, with statistical analyses reported in Annex 6. Content refers to whether the study asked the right questions in a clear, understandable, sensible and appropriate manner with which to obtain a valid estimate of willingness to pay. Many of the judgements about content validity are essentially subjective. However, the responses to a number of questions in the survey are useful for testing content validity (and indeed some serve this purpose only). Following are a number of positive indicators of content validity for this study: • only 9% of respondents considered the questionnaire to be unrealistic and only 6%

found it difficult to understand. Therefore the questionnaire created a believable hypothetical market.

• income non-response at 35% of the sample is a comparable rate to other studies. It

was also found that there are no significant differences in the non-response rate to the income question between sub-samples (it varies between 34% and 36%).

6 In calculating present values it is assumed that the benefits of improvements accrue from the end of the first year during which the improvements are introduced.


eftec xiii 25 June 2002

Thus, we can assume that the effect of income non-respondents on the WTP is uniform, and hence their inclusion (or exclusion) would not change the WTP estimates.

• 21% of the total sample gave protest (zero) responses. When these protest bids

were included in the econometric analysis, the models did not function well. A model was therefore tested to assess whether protest bids are random and to try to explain the variation in protest bids across the protest respondents. Based on this analysis, it can be concluded that the protest bids are a random occurrence and that we cannot say that any particular group was consistently providing protest answers. Thus, their exclusion from the WTP analysis is unlikely to lead to biased WTP estimates.

• question A-16, which elicited preferences for a variety of beach characteristics,

found that those attributes which scored highest (see Figure 4.3) were all included in the beach scenarios. This lends support for the design of the cards, in that all of the attributes chosen were also considered important by respondents.

• whether the levels of the price attribute were in fact representative of the correct

price bracket has implications for the validity of the WTP results. Overall, the results suggest that the chosen price vector was adequate as there does not seem to be a large concentration of bids at either end of the distribution, i.e. the majority of respondents choosing not to pay or accepting the highest price.

Construct validity is concerned with whether or not the results are consistent with expectations based on other studies, economic theory and intuition. The willingness to pay values from this study are lower than those found in previous studies. This can be explained by two factors: (i) this study measured marginal changes as opposed to the discrete changes measured in previous studies and (ii) this study sampled the national population rather than beach users, as in other studies. Beach users are expected to hold higher preferences for beach quality than the national population as a whole. The model results show that all the attributes have the expected signs (in that reductions in risk of illness increase welfare and increases in the costs of improvement decrease welfare). Other explanatory variables also have the expected effects on the model, e.g. those who thought water quality was very important were also willing to pay more. Finally, most analysis concentrates on the signs and statistical significance of the WTP estimates rather than answering the straightforward question of whether the magnitude of the WTP estimates seem plausible. The estimates here, which all fall in the £0.9-£13.7 per household per year bracket, are plausible figures given that WTP is for marginal changes. Therefore, it can be concluded that in addition to the more sophisticated statistical tests, the results also pass the �reality check�. Potential biases The types of potential biases tested and the results of these tests are reported below.


eftec xiv 25 June 2002

Concerns regarding scope, embedding and part-whole effects arise from "the frequent finding that WTP for a good is approximately the same for a more inclusive good" (Fisher, 1996; p19). These effects can only be avoided or at least reduced by good questionnaire design, since there is very little that econometric analysis can do other than detect their existence. The following discusses the three ways in which the above effects could have presented themselves in this study and how they were avoided by careful design of the questionnaire: • Geographically (related to scope): where a respondent values a good whose spatial

attributes are larger or smaller than the spatial attributes of the presented good. In order to avoid this effect, we designed the choice experiment for a typical beach in England and Wales, i.e. improvements to all beaches, rather than a single beach or beaches in a single region.

• Capture of benefits: where a respondent includes a broader or a narrower range of

benefits in valuing a good than intended. As Section 5.2.1 shows, the beach scenarios were defined in terms of all the attributes that the respondents have stated to be important in Section A of the questionnaire. Therefore, we can say that we have included all the attributes and hence the benefits of improvements to these attributes that could contribute to the WTP responses.

• Policy-package: where a respondent values a broader or a narrower policy

package than the one presented. All possibly significant changes to the selected beach attributes due to revisions to the Bathing Water Directive were included in the definitions of beach scenarios. Furthermore, the respondents were not told that the improvements they were presented were due to the revisions to the Bathing Water Directive. Therefore, no other part of the policy package, i.e. revisions to the Directive, could have entered the considerations of the respondents.

In order to test whether the starting points for the water quality attribute had an impact on the WTP estimates, each sub-sample was analysed separately. The results show that there is no strong and systematic evidence of starting point bias as none of the WTP estimates for sub-samples is statistically different at the 1% level. Transferability In terms of the transferability of the results, as in any benefit transfer, the benefit estimations found in this study should only be applied to similar: • environmental resources; • uses of environmental resources; • change in environmental resources; and • affected populations. This means that it would be acceptable to apply the results to other bathing waters aside from beaches, as these provide similar recreational experiences, unless there is a reason to suspect that inland bathing waters are significantly different than coastal bathing waters. Results are likely to be transferable to Scotland and Northern Ireland and possibly Ireland as the beach environment can be considered to be similar (although population characteristics may be significantly different, especially in


eftec xv 25 June 2002

Ireland). To a lesser extent the results are transferable to other countries with similar climate / beach use and population characteristics. However, the literature on transferability of values indicates that transferability between countries has always proved less reliable than transfers of values and benefit functions within countries. Clearly, using this reasoning they are not transferable to Mediterranean Member States. The results cannot be applied to water sports activities that are very different from bathing (as people who do water sports are likely to have different attitudes to risk) and they cannot reasonably be transferred to other health impacts of water quality other than stomach upset or to stomach upsets caused by other sources. TS.8 Conclusions A number of conclusions of the study that are of particular importance are described below: • Using both the responses to the questionnaire and epidemiological evidence, the

study estimated that there were around 1.3 million cases of stomach upset in England and Wales every year that could be associated with bathing in faecally contaminated bathing water. A second calculation based on respondents� personal assessment of cases of stomach upset revealed that this figure could be upwards of 2.2 million cases.

• Respondents revealed that the most popular activities undertaken on the beaches

in Britain are walking, relaxing, sun-bathing and picnicking, while 57% of respondents and their families never go swimming.

• The most important beach attributes were cleanliness of the beach, water that

looks clear from litter and foam, and amenities. 51% of the sample found water quality to be very important.

• The modelling results showed that all beach attribute coefficients were significant

at the 1% level and had the expected signs across all versions and model specifications.

• 51% of the sample thought that the proposed advisory note system would be very

useful. 38% of the sample preferred to see a simple symbol indicating "do not bathe", while 51% preferred to see more detailed information.



• Respondents were willing to pay between £1.10 and £2.00 per household per year

for a 1% reduction in the risk of suffering a stomach upset. Interestingly, this value range is almost identical to the 90 pence to £1.10 that respondents seemed to be prepared to pay to avoid 1 day of poor water quality.


eftec xvi 25 June 2002

• Respondents were willing to pay between about £5.6 and £13.7 per household per year for the introduction of an advisory note system. This equates to £134.4 million per year for England and Wales.


year for avoiding the presence of some litter / dog mess on the beach. This equates to £144 million per year for England and Wales.

• Respondents were willing to pay between around £2.5 and £7.3 per household per

year for improving the standard of amenities and safety from average to good. This equates to £60 million per year for England and Wales.


determine whether any potential biases were at play. These tests all showed that the results are free from any bias, and are supported by several strong arguments for different types of validity.

• The benefit estimates produced in this study have limited transferability, and

caution should be taken, especially in applying the results to other countries and contexts.


eftec xvii 25 June 2002

TABLE OF CONTENTS I. INTRODUCTION ..........................................................................................................................1

1.1 POLICY CONTEXT......................................................................................................................1 1.2 OBJECTIVES OF THE STUDY .......................................................................................................2 1.3 IMPLEMENTATION OVERVIEW AND REPORT STRUCTURE ..........................................................2

II. METHODOLOGY AND CONCEPTUAL OVERVIEW ...........................................................4

2.1 THE CHOICE EXPERIMENT TECHNIQUE .....................................................................................4 2.2 CHOICE OF POPULATION AND SAMPLE ......................................................................................5 2.3 LITERATURE REVIEW ................................................................................................................5

III. SURVEY DESIGN AND IMPLEMENTATION...................................................................12

3.1 SURVEY DESIGN......................................................................................................................12 3.2 TESTING THE QUESTIONNAIRE ................................................................................................23

IV. RESULTS..................................................................................................................................30

4.1 ATTITUDES TOWARDS THE QUESTIONNAIRE ...........................................................................30 4.2 ATTITUDES AND USES .............................................................................................................30 4.3 VALUATION RESULTS..............................................................................................................41 4.4 FOLLOW-UP QUESTIONS..........................................................................................................52

V. AGGREGATION, VALIDITY AND TRANSFERABILITY...................................................55

5.1 AGGREGATION ..........................................................................................................................55 5.2 VALIDITY TESTING.....................................................................................................................57

VI. CONCLUSIONS.......................................................................................................................66

REFERENCES .....................................................................................................................................68

GLOSSARY..........................................................................................................................................72

List of Tables Table 2.1: Overview of the results of previous evaluations of bathing water benefits (euro) ..................7

Table 2.2: Summary of results from georgiou et al (2000), (£ per year)................................................10

Table 3.1: Beach attributes and their levels: improvement scenarios.....................................................15

Table 3.2: Beach attributes and their levels: current situation................................................................15

Table 3.3: Levels for the attribute of water quality in different versions of the questionnaire...............17

Table 3.4: Sample breakdown by questionnaire version .......................................................................26

Table 3.5: List of sampling points for the main survey...........................................................................27

Table 3.6: Summary statistics of selected socio-economic variables .....................................................28

Table 4.1: Distribution and number of beach visits per year ......................................................................

Table 4.2: mean and median number of beach visits per person per year ..................................................

Table 4.3: Exposure to water from swimming and other recreational water sports (using 2001 as count year)........................................................................................................................................................34

Table 4.4: Number of times undertook activity by exposure category (using 2001 as count year)........35


eftec xviii 25 June 2002

Table 4.5a: Gross estimate for predicted illness (stomach upset)...........................................................36

Table 4.5b: Gross estimate of personal assessment of illnesses..............................................................37

Table 4.6: Behavioural responses to an advisory note system................................................................40

Table 4.7: Awareness of beach standards and schemes..........................................................................41

Table 4.8a: Conditional logit modelling results (version a: utility of reductions in the chance of contracting a stomach upset).............. .....................................................................................................47

Table 4.8b: Conditional logit modelling results (version b: utility of changes in the number of days it is safe to swim)............................................................................................................................................46

Table 4.9a: Nested logit modelling results (version a: utility of reductions in the chance of contracting a stomach upset) ........................................................................................................................................47

Table 4.9b: Nested logit modelling results (version b: utility of changes in the number of days it is safe to swim)....................................................................................................................................................48

Table 4.10: Household implicit mean willingness to pay for beach.......................................................49

Attributes: Values in £ per year (version a)............................................................................................49

Table 4.11: Household implicit mean willingness to pay for beach attributes: values in £ per year (version b)................................................................................................................................................50

Table 4.12: Estimated range of mean values for beach attributes across versions and models (£ per household, per year) ...............................................................................................................................51

Table 4.13: Reasons for not wanting to pay for beach improvements....................................................52

Table 4.14: Reasons for paying for beach improvements.......................................................................53

Table 4.15: Preferred scenarios for reducing risk of illness at beaches..................................................54

List of Figures Figure 1.1: The stages of analysis in a stated preference study ................................................................3

Figure 4.1: Attitudes towards the questionnaire .....................................................................................30

Figure 4.2: Beach uses............................................................................................................................32

Figure 4.3: Preferences for beach attributes ...........................................................................................33

Figure 4.4: Perceptions of beach recreation risks ...................................................................................37

Figure 4.5: Attitudes towards health and risk .........................................................................................38

Figure 4.6: Perceived usefulness of an advisory note system.................................................................40

Figure 4.7: Rating of beach standards and schemes ...............................................................................41

Figure 4.8: Attributes considered the most in the choice experiment.....................................................53


eftec xix 25 June 2002

ANNEXES ANNEX 1: FINAL QUESTIONNAIRE

ANNEX 2: FOCUS GROUPS

ANNEX 3: PILOT SURVEY

ANNEX 4: SENSITIVITY ANALYSIS FOR ANNUAL CASES OF STOMACH UPSET

ANNEX 5: WTP AND COST OF ILLNESS APPROACHES TO VALUING THE BENEFITS OF

ENVIRONMENTAL HEALTH RISK POLICIES

ANNEX 6: STATISTICAL ANALYSES

ANNEX 7: QUOTA SHEET


eftec 1 23 May 2002

I. INTRODUCTION 1.1 Policy Context The costs and benefits of improving bathing water quality have increasingly come under scrutiny in the UK and throughout Europe as the European Commission makes moves to revise the 25 year old Bathing Water Directive (76/160/EEC). The Directive lays down standards for the sanitary quality of waters used for bathing1 and is implemented in England and Wales by the bathing water regulations (SI 1991/1597). Currently, compliance with the Directive is assessed through a range of numerical quality standards which are specified at �I� (Imperative or Mandatory) values and �G� (Guideline) values for a range of physico-chemical, bacteriological and aesthetic criteria. EU Member States are legally obliged to set standards that bathing waters must meet, and these standards must be no less stringent than the �I� values. Whilst nineteen different parameters are listed, in practice only those for total and faecal coliforms are significant2. In December 2000, the European Commission released a Communication (COM(2000)860) on developing a new bathing water policy. The Communication was the first step in a long process of consultation to revise the Directive that is still ongoing. Reasoning behind proposed changes to the Directive includes improved scientific knowledge, the need for more active quality management of bathing waters and a desire for improved public information. The revised Bathing Water Quality Directive is expected to: • tighten microbiological standards; • harmonise beach management regimes across the EU; • require more public information on water quality; and • widen the scope of designated bathing waters to include waters used for water

contact sports such as surfing, windsurfing and slalom kayaking. Many of the benefits that the public currently enjoys at beaches, and may experience through implementation of a revised Directive, are non-market benefits. They are enjoyed for free, but are evident not least from the number of people that choose to visit the British coast each year. However, information on the magnitude of these benefits is still sparse. In fact, even reliable estimations of the number of visits to beaches are lacking. While a number of UK economic valuation studies have provided estimates of the welfare benefits of various aspects of beach recreation, only a handful have specifically sought to estimate the benefits of water quality improvements. And so far none has sought to

1 Bathing water is defined in the Directive as fresh or sea water in which bathing is explicitly authorised, or is not prohibited and is traditionally practised by a large number of bathers. However, for the purpose of this study, only sea water bathing waters were considered. 2 The mandatory microbiological quality requirements for faecal coliforms are that 95% of samples should not exceed the level of 2000 faecal coliforms per 100ml. The guideline requirement is much more strict suggesting that 80% of samples should not exceed 100 faecal coliforms per 100ml.


eftec 2 25 June 2002

identify the benefits of implementation of the proposed changes to the EC Directive to the population of England and Wales. This study was therefore commissioned by DEFRA to provide targeted information on the expected benefits of these proposals. 1.2 Objectives of the Study The aims of this study are to: • estimate individuals' willingness to pay for aspects of beach use in England and

Wales measured through a number of beach attributes; • discuss the scope for ensuring the validity and transferability of the estimated

values; and • produce an estimate of total benefits from the implementation of a revised

Bathing Water Quality Directive to England and Wales. The valuation technique recommended in the terms of reference to this study is part of a set of methods jointly referred to as 'choice modelling', which are stated preference techniques. Stated preference techniques (SP) use carefully structured questionnaires to elicit respondents� preferences for a given natural resource or environmental change, the value of which is not reflected in actual markets. For a more detailed discussion of the choice-modelling technique refer to Section 2.1. 1.3 Implementation Overview and Report Structure A number of research stages are involved in a stated preference study. These are specified in the best practice guidelines in eftec (2000) and illustrated in Figure 1.1 below. This study has adhered to these guidelines and the various stages are reflected in the report structure, as follows: • Section 2: Methodology and Conceptual Overview � This section covers stages

1-3 in the flowchart, which consist of the initial background research and literature review; identification of the questions being asked; the choice of survey method and valuation technique; and the choice of population and sample.

• Section 3: Survey Design and Testing � This section takes in stages 4 and 5 of the flow chart, covering the details of survey design, such as choice of valuation questions and format; and testing of the survey through focus groups, pilot studies and the consultation of experts, followed by the main survey.

• Section 4: Results � This section is the same as stage 6 in the figure, consisting of analysis of the survey results, including summary statistics of attitudinal responses, socio-economic characteristics and the econometric analysis of the valuation responses.

• Section 5: Aggregation of Results, Validity Testing and Conclusions � This section covers stages 7 and 8 of the flowchart and presents the aggregate benefits of the beach improvement across the population of England and Wales, as well as testing the validity and reliability of results.

• Section 6: Conclusions - Final conclusions drawn from the study are presented here.



Figure 1.1: The stages of analysis in a stated preference study The report also contains seven annexes containing the original questionnaire; focus group protocol, transcripts and report; pilot survey results; cases of illness sensitivity analysis; willingness to pay versus cost of illness calculations; statistical analyses; and the sampling quota sheet, respectively.

1. Initial research What question is being answered? What is the object being valued?

2. Choice of survey method and valuation technique

What is the survey method (e.g. face-to-face, mail, mix format)? Contingent valuation or choice modelling?

3. Choice of population and sample

What is the target population? What kind of sample should be selected?

4. Questionnaire design What form of question? What elicitation format? What payment vehicle (tax, price, donation etc)?

5. Testing the questionnaire Focus groups Redesign questionnaire Pilot survey(s) (pre-test) Redesign questionnaire Main survey

6. Analysis Responses entered into spreadsheet for statistical and econometric analysis

7. Validity and reliability testing

Do the results meet validity and reliability tests?

8. Aggregation Aggregating from the sample results to the target population



II. METHODOLOGY AND CONCEPTUAL OVERVIEW 2.1 The Choice Experiment Technique The choice experiment method is one of the choice modelling approaches that assume that any good can be described in terms of its attributes, or characteristics and the levels that these take (eftec, 2000). These methods seek to elicit respondents' preferences for different attributes varying at different levels, each with a price tag attached, from which willingness to pay (WTP) can be indirectly inferred (Hanley, Mourato and Wright, 2001). They are the preferred approach for valuing complex and multidimensional scenarios, especially to: • identify which attributes are significant determinants of the values people place

on non-market goods; • determine the implied ranking of these attributes amongst the relevant

population(s) (e.g. in the context of overall beach quality, how beach cleanliness is ranked in relation to the provision of amenities);

• estimate the value of changing more than one of the attributes at once (e.g. if a

management plan results in a reduction in water related illnesses but a deterioration in safety provisions); and

• as an extension of the above, estimate the total economic value of a resource or

set of quality improvements so long as this is covered within valuation scenarios. Choice modelling includes a number of methods based on the way the choice question is designed, namely: choice experiments, contingent ranking, contingent rating and paired comparisons. However, only two approaches (choice experiments and contingent ranking) are based on economic theory which allow the results to be interpreted as being equal to marginal (or total) values for use in policy development or appraisal. The choice experiment method estimates individuals' preferences, in monetary terms for individual attributes and marginal changes in them by providing respondents with a number of choice sets and asking them to state their most preferred choice. For the experiment to comply with the requirements of the economic theory, the respondents should be given the right to choose 'no change', i.e. one of the choice sets must be the current situation. The choice sets in this study consist of different beach scenarios described by five attributes: water quality, advisory note system, litter/dog mess, safety and amenities and additional water charges per year. Each respondent is given a number of these choice sets consisting of different levels of the attributes. At the end of the survey, the record of the most preferred choice sets enables the probability of an alternative scenario being chosen to be modelled in terms of the attributes used to describe the scenarios (Bennett and Blamey, 2001). The coefficients in this model can be used to estimate the rate at which respondents are willing to trade off one attribute with another. The rate at which respondents are



willing to trade off money for any given attribute is the ratio of the coefficients of that attribute and the cost attribute in the model. The trade-off estimated is also known as the implicit price or part worth of the attribute. The implicit price of an attribute is the marginal rates of substitution between the attribute of interest and the price attribute and equivalent to respondents� willingness to pay for that attribute. A choice experiment should be designed in such a way that model identification is assured and one can obtain the best (or �most efficient�) possible estimates of the parameters of the model of interest. This involves choosing the right number of choice sets in relation to the number of attributes and levels chosen. Based on the design catalogue (see US Department of Transportation, 1982, SPSS or other statistical packages), statistical experimental designs are used to select subsets of the total set of possible alternatives for use in the survey. Such fractional factorial designs are able to reduce the number of alternatives presented to respondents but there will be a concomitant loss in estimating power. The trade-off between these two factors needs to be considered during the design stage of the questionnaire. Section 3 details the design process adopted in this study. 2.2 Choice of Population and Sample The population of interest is the population of England and Wales. This includes both those affected adversely and beneficially and those who may hold use and / or non-use values. Since it is impossible for each member of the population to express their preferences within a single survey, a sample needs to be selected. It was therefore decided that the sample for this study should be randomly selected among the English and Welsh population, reflecting as best as possible the proportion of people in each gender, age group, and socio-economic group. The necessary sample size required for the choice experiment survey is dependent on the final number of attributes chosen, their levels and other design factors. Based on the final questionnaire design, which is discussed in the next section and which involves five versions of the questionnaire and five attributes at 2 to 4 levels each, a sample size of 800 respondents was considered to be the minimum allowable to support the design. In the event, the sample consisted of 809 respondents. 2.3 Literature Review A considerable body of applied literature has been developed in the past two decades on marine recreation for a variety of recreational sites and for different leisure activities, both in developed and developing countries. The focus of these research efforts has been on activities such as angling, swimming, boating and beach use. Table 2.1 reports a summary of studies looking at the value that beach users place on water quality changes most of which come from the US. The range of estimates is wide (Euro 0.74-227), which can be partly explained by differences in the valuation method used, in the pollution indicators adopted, in the percentage quality improvement being valued and in the units in which WTP is measured (per visit, per season, per year). When expressed as a percentage of income these values are typically found to be less than 1% of annual income.



However, even when the individual value attached to water quality improvement is found to be small, the aggregate value can be very significant, given the very large numbers of recreationists involved (Freeman, 1995).

Empirical evidence on the value of marine water quality improvements in Europe is even scarcer than for the US. The few existing estimates are extremely wide: for example, Sandstrom (1997) reports welfare changes of Euro 24-55 per trip for reductions in nutrient load in Swedish beaches while Georgiou et al, (1996) find an individual value of Euro 24-90 per year for water quality improvements in East Anglian beaches. By and large, the available estimates indicate that the aggregate value attached by recreationists to marine water quality improvements seem to be significant, even when individuals� WTP values are found to be small.

For activities like swimming, pollution may impose risks on human health. Therefore, if those risks are important and correctly perceived, then a significant share of the values that have been estimated in the literature is certainly related to recreationists� health concerns. Indeed, the available evidence suggests that �threats to health such as faecal coliform bacteria and PCB contamination are important attributes� (Freeman, 1995, p.400). Strikingly, very few economic studies have considered explicitly the health risks of bathing in polluted marine waters. Those few European studies that have examined these risks, and specifically in the context of the EC Directive on bathing waters are summarised in more detail below. These are: • �Valuing the benefits of coastal water quality improvements using contingent and

real behaviour� (Hanley, Bell and Alvarez-Farzio, 2001); • �Non-parametric and semi-parametric approaches to analysing payment ladder

contingent valuation data: bathing water quality improvements in Scotland� (Day et al, 2001);

• �Coastal bathing water health risks: developing a means of assessing the adequacy

of proposals to amend the 1976 EC Directive� (Georgiou et al, 2000); • �Improving the assessment of water related health impacts: evidence from coastal

waters in Portugal� (Machado and Mourato, 1999); and • �Determinants of individuals� willingness to pay for perceived reductions in

environmental health risks: a case study of bathing water quality� (Georgiou et al, 1998).



Table 2.1: Overview of the results of previous evaluations of bathing water benefits (Euro) Study Approach Location Stressor Quality change Value per visit Value/season/year

Barton (1998) Costa Rica Sewage, FC To swimable - 144-172.8

Brockstael et al (1987) RUM/TCM Boston, USA Oil, faecal coliforms, COD

10% improvement 0.74 14.88

Brocksteal et al (1987) RUM/TCM Boston, USA Turbidity, FC, COD 30% 1.956 47.04

Brocksteal et al (1987) CVM Chesapeake Bay - Unacceptable to acceptable - 190.8

Day et al (2001) CVM Ayr and Irvine, UK Sewage To EC mandatory - 10.25 � 16.82

Choe et al (1996) CVM Philippines Sewage To swimable - 14.4-28.8

Freeberg and Mills (1980) RUM/TCM Boston, USA Oil, total bacteria, colour 10% improvement n.a. 5.56

Georgiou et al (2000) CVM East Anglia, UK Sewage Meet revised BW standard 32-67

Georgiou et al (1996) CVM UK Sewage To EC standards - 24-90

Hanley, Bell and Alvarez (2001)

CVM Scotland Sewage Meet minimum EU standard 17-52

Le Goffe (1995) CVM France Eutrophication, sewage To swimable/shellfishable - 52.44

Machado and Mourato (1999) CVM/CR Portugal Sewage, FS Bad to average (EC mandatory) Average to good (EC guideline)

16.10 6.90

-

McConnel & Ducci (1989) CVM Barbados Sewage - - 16.8-227.2

Mantymaa (1999) CVM Finland Sewage, nutrients From swimable to drinkable - 48

McConnel & Ducci (1989) CVM Uruguay Sewage To swimable level - 22.2

Niklittschek and Leon (1996) CVM Chile Sewage To swimable - 127.2

Niklittschek and Leon (1996) TCM Chile Sewage To swimable - 168

Sandstrom (1997) RUM/TCM Sweden Nutrients 50% improvement 24-55.2 -

Zylicz et al (1995) CVM Poland Eutrophication To swimable - 21.6 to 85.2

Source: WRc (2001) Exchange rate 1$-=1.2 Euro. RUM: Random Utility Model; TCM: Travel Cost Method; CVM: Contingent Valuation Method; CR: Contingent Ranking Note: The shaded studies are summarised below.



�Valuing the benefits of coastal water quality improvements using contingent and real behaviour� (Hanley et al, 2001) This study uses a combined stated and revealed preference approach to value the improvements that result from compliance with the EU legislation on bathing water quality. It focuses on an area of Scotland which has consistently failed to meet standards under the current Bathing Water Directive. An on-site sampling based questionnaire asked beach visitors about their trips to the beach where they were sampled; the activities they generally participate in on the beach; their trips to other beaches in the area, and their perceptions of water quality at the beach where they were sampled, and at other beaches. The proposed improvement in water quality was described as an investment by West of Scotland Water which would guarantee that all bathing waters in the area would reach the standard. Respondents were asked whether this would cause them to visit the beach where they were being surveyed more frequently and how the activities they pursued on the beach might change. If visits were expected to increase people were asked by how many per year and whether this would be at the expense of trips to other beaches. A total of 414 responses were obtained. Over the whole sample an increase of 52 trips was predicted as a result of the improvement in water quality, from 3,954 trips per year without the improvement to 4,006 trips with the improvement. This amounts to 1.3 per cent of total predicted trips over the 249 useable responses in the survey. The implied increases in consumer surplus were £7.81 per person per year or £0.48 per visit. Combining these results with national data on average beach visit per person gave aggregate benefits of £1.25 million per annum. This is an underestimate of total benefits as it does not include the benefits of bathing water improvement to those who are currently non-visitors, but who would visit beaches if bathing water quality improved. �Non-parametric and semi-parametric approaches to analysing payment ladder contingent valuation data: bathing water quality improvements in Scotland� (Day et al, 2001) This study was funded by the ESRC under the Global Environmental Change programme. Survey work was undertaken in 1999, and data analysis in 2001. The main intention of the study was to estimate the benefits to local residents of improvements in coastal water quality, in terms of the EU Bathing Water Directive. Ayr and Irvine were chosen as the case study sites, on the basis of persistent failures of the local beaches to meet EU mandatory standards. A contingent valuation approach was used, utilising local council taxes as the payment vehicle. Residents were questioned in their homes by a market research firm, and sample sizes of 351 (Ayr) and 432 (Irvine) were obtained. A payment ladder method was used to collect WTP bids. A non-parametric approach was used to estimate WTP to improve local water quality such that no more failures of the EU mandatory standard occurred in the future, for both samples. Mean WTP was found to be significantly different between the two sites, being higher in Ayr at £10.43 per household per year than in Irvine (£6.36 per household per year). A semi-parametric



approach was used to estimate bid functions, which explained the variation in WTP. These showed for example that: • An extra £1000 of disposable income increases average WTP by £2.15 in the Ayr

sample, but by only 50p in the Irvine sample. • Age influences WTP positively at first then negatively, though it is not a

particularly significant regressor for the Irvine data. • Beach activity factors are not significant in determining WTP from the Ayr

sample. In Irvine two factors (swimming and beach recreation) are weakly significant and are positively signed. That is, the more a respondent participates in swimming or beach based recreational activities the more (on average) he or she would be willing to pay for improvements.

The data from this study are described as of high quality, as state-of-the-art econometric approaches and study designs were used, with a fair sized sample. However, it is not known how transferable the results are to other sites. �Coastal bathing water health risks: developing a means of assessing the adequacy of proposals to amend the 1976 EC Directive� (Georgiou et al, 2000) This study is the only identified study to deal specifically with a revision to the Bathing Water Directive that goes some way beyond the existing Directive. The revision examined in the study, however, is not the same as the currently proposed revision scenarios. Indeed precise information on the nature of the improvements was not given to the respondents because at that time it was not possible to interpret the standards in terms of a reduction in the risk of contracting gastro-enteritis. However the respondents were made aware of the fact that there would be some unspecified reduction in disease related to water quality. In the survey the respondents were in fact asked to say what they felt the new standard would deliver in terms of disease reduction. Twenty five per cent believed that the revision would eradicate the risk of disease, while the remaining 75% believed some level of disease would continue under the revision. In effect therefore the respondents were asked to value a reduction in disease incidence that they themselves considered to result from the improved quality. This was conditioned in relation to evidence from Fleisher et al (1998) study which stated that the risk of contracting gastro-enteritis was approximately 5% at beaches which met the mandatory standard. Analysis of the responses revealed that the overall expectation of the reduction in illness was approximately 65%, i.e. down from 5 in 100 to 2 in 100. Therefore the respondents were in fact valuing an improvement which would reduce the risk of illness by a perceived expectation of 3 cases in every 100. Three surveys were undertaken: while Norwich survey was a stratified sample of local residents, Lowestoft and Great Yarmouth surveys were of people in the vicinity of these beaches. Respondents were asked their WTP per year for all beaches in the Anglian region to comply with the new standards. The percentage willing to pay, the mean willingness to pay (of those who were willing to pay) and the mean willingness to pay for all persons are summarised in Table 2.2.



Table 2.2: Summary of Results from Georgiou et al (2000), (£ per year) Norwich Great Yarmouth Lowestoft

Full sample

Full sample

Local resident

Day tripper

Holiday maker

Full sample

Local resident

Day tripper

Holiday maker

% of sample willing to pay

83 64 49 63 69 71 58 79 74

WTP of those willing to pay

42.29 32.22 20.17 28.3 37.71 31.84 37.41 27.55 32.84

WTP of total sample

35.10 20.62 9.88 17.83 26.02 22.61 21.70 21.76 24.30

Personal values such as trustworthiness of government and importance of the directive to the respondent and household income had a positive effect in WTP. The nature of use made of the resource had a direct bearing on WTP, i.e. the valuation of a local resident was lower than that of a holidaymaker, while a day tripper�s valuation lay between that of resident and holiday maker. WTP rose in line with expectation of the reduction in risk of ill health. In the case where the expected reduction in perceived risk led to zero cases of illness, then WTP was equal to £40 across the three samples shown in Table 2.2. �Improving the assessment of water related health impacts: evidence from coastal waters in Portugal� (Machado and Mourato, 1999) This study used both contingent valuation and contingent ranking. In the contingent valuation survey 401 Lisbon residents using the beaches along the Estoril coast were randomly selected. In-person interviews were conducted on-site on 11 beaches with varying degree of water quality. Respondents were asked to value in terms of pain and inconvenience, three health episodes associated with bathing water pollution: gastro-enteritis, eye irritation and respiratory illness. Each symptom was described in terms of symptoms, restrictions on activities and duration. Respondents were asked to move up and down a payment card and specify the amount they were almost certain they would pay and not pay (specified certainty level). Mean WTP was calculated from these interval data. In the contingent ranking exercise 195 Lisbon residents using the beaches along the Estoril coast were interviewed in-person on-site on 11 beaches. Respondents were asked to rank (as sequential choices) four beaches varying in terms of water quality and cost of travelling there, and identical in all other aspects. Water quality was varied between good (blue flag), average and bad, while the cost varied between 1.50, 4.00 and 7.50 Euros. Ninety five per cent were able to choose their most preferred beach option, while 75% provided full ranks of all three sets of four beaches. Both methods elicit mainly use value, health and amenity related. The contingent valuation survey showed that among the three water-related symptoms (eye irritation, respiratory illness and gastro-enteritis), gastro-enteritis was considered to be the worst in terms of pain and inconvenience by 61% of respondents. Mean WTP per person to avoid an episode of gastro-enteritis was 38.80 Euros (standard error of 3.81 and median WTP of 12.50 Euros; reflecting a wide distribution). One third of the sample had zero WTP. Around half of these did not take the survey seriously or could not decide on a value, and were therefore excluded from the



analysis. The top ten WTP answers reporting unrealistically high WTP amounts (over a third of their monthly income) were also excluded. The Contingent Ranking model estimated the recreational/amenity use value per visit at 16.10 Euros for changing from a bad to a good quality beach, and 6.90 Euros for moving from average to good quality beaches. The mean expected WTP to avoid an episode of gastro-enteritis for the Contingent Ranking sample was 7.70 Euros. The study found that income had a positive effect on WTP, whilst age had a negative influence. Socio-economic variables such as sex, family size and education levels were not found to have a significant effect. Attitudinal factors had a significant effect. Those who were concerned about health implications generally had a higher than average WTP, while those that were not concerned had a lower than average WTP. �Determinants of individuals� willingness to pay for perceived reductions in environmental health risks: a case study of bathing water quality� (Georgiou et al, 1998) This study involved a combined economic and epidemiological investigation of coastal bathing water health risks in order to assess the adequacy of present standards of bathing water from both a physical/scientific and a social/public perception standpoint. A sample of 400 individuals were interviewed face to face at two beaches on the coast of East Anglia, namely Great Yarmouth Pier beach and Lowestoft beach. Great Yarmouth was chosen because it failed to meet the EC standard, Lowestoft because it passed. Therefore, two separate economic measures of welfare change were estimated: WTP for a gain in quality at Great Yarmouth, and WTP to avoid a loss at Lowestoft. However, as perceived/actual units of quality change were not constant between the two sites, there were no prior assumptions regarding the relative value of these measures. In the economic analysis, it was found that the mean WTP amounts were similar at the two sites (£18.19 at Great Yarmouth and £16.21 at Lowestoft). These measures reflect both theoretical issues concerning welfare gains and losses, and perceptual differences regarding the magnitude of quality provision changes across the sites as revealed in the pilot survey. There were notable differences in variables explaining WTP amounts both between the two sites, and across different user groups, namely holiday makers, day trippers and locals. Socio-economic and use variables were found to be significant in explaining stated WTP, as were additional factors, including the perception and acceptability of various risk issues. Attitudes to personal health, as ascertained by the Health Locus of Control questionnaire (Wallston and Wallston, 1978), were also found to be important. Due to the novel approach taken in the research, and relatively small sub-sample sizes, it was not possible to draw definitive conclusions about the relationships between stated WTP, attitudes to health and risk related factors.



III. SURVEY DESIGN AND IMPLEMENTATION Section 3.1 provides a brief overview of the survey questionnaire, the full text of which is presented in Annex 1. The design presented here and used in the main survey is the product of several rounds of testing ranging from expert review to focus groups, and pilot and main surveys. This process is summarised in Section 3.2 with detailed records presented in Annexes 2 and 3. 3.1 Survey Design In line with the standard practice, this choice experiment questionnaire consists of the following sections: • Section A: Attitudes, opinions, knowledge and uses: uses of British beaches,

preferences for beach characteristics, knowledge of beach and water quality standards and management schemes and attitudes towards risks related to beach use;

• Section B: Valuation scenario: a choice experiment to assess willingness to pay for four different beach attributes: water quality improvements, existence of an advisory note system, beach cleanliness (absence of litter and dog-mess), and provision of safety and amenities;

• Section C: Debriefing and follow-up questions: reasons why respondents were or were not willing to pay for improvements in each of the attributes and views of the scenarios presented; and

• Section D: Socio-economic characteristics: including income, education and household structure, and so on.

Section 3.1.2 shows that five versions of the questionnaire were administered to reflect the complexity of the questions this study is designed to answer. Variations in each version are confined to the valuation scenario (Section B). 3.1.1 Attitudes, opinions, knowledge and uses of British beaches Overall the aim of this section was four-fold: • to help respondents explore their personal thoughts and attitudes towards British

beaches, beach quality standards and the potential risks involved in water-based activities as a preparation for responding to the valuation section;

• to reveal the most important underlying factors driving respondents� attitudes for

supporting (or not) beach water quality and management improvements so as to aid in the interpretation of the valuation results;

• to provide tests of validity for the survey design and results; and • to provide some raw data for general use, and more specifically, for use in the

aggregation of the results, such as total beach visits per person per year and average exposure to risk of contracting a stomach upset.



A variety of question formats was used in this section: yes/no answers, Likert 5-point scales, multiple choice and open-ended questions. Information was elicited on the uses of British beaches in the last year visited. In this way it was possible to aggregate data on a yearly basis. This information included: • frequency and types of beach visits undertaken, i.e. as a local resident, day-tripper,

holiday-maker or other type of visitor; • participation in a number of different beach activities, from water-based activities

such as swimming to land-based activities such as walking and sun-bathing; and • exposure to water through water-based activities. Preferences for a number of beach attributes were sought with the use of a Likert scale. These included those attributes selected for the choice experiment as well as others, such as sandiness of beach, availability of water sports, and tidal conditions. The answers to this question were used as information in its own right and as a consistency check for WTP results. It is expected that the ranking of these attributes to be similar to the ranking of attributes according to WTP estimates. Experience of beach-related accidents and bathing water-related illness, as well as general perceptions of exposure to risk of these occurring was assessed through a series of questions. These asked: • information about previous cases of stomach upset as a result of poor water

quality at British beaches; • experience of near drowning or collision with jet-skiers; and • perceptions of the current level of risk of suffering any of the above. A stomach upset was described in the questionnaire with the use of a separate card (See Box 3.1), which was presented whenever stomach upset was mentioned. Box 3.1: Definition of a Stomach Upset Used in the Questionnaire The following describes the possible symptoms of a stomach upset that might result from swimming in polluted seawater.

mild diarrhoea

indigestion with fever

nausea with fever vomiting diarrhoea

It is thought that the majority of cases of stomach upset that result from swimming in polluted seawater are at the milder end of the spectrum illustrated above. In fact, the chance of getting a stomach upset that would require a visit to the doctor, day(s) off work or admission to hospital would be no different if you did not bathe.

Increasing severity of symptoms



This definition of stomach upset is supported by the findings of a publicly available report (WRc, 1994). The report concerned an epidemiological study of sea bathers in the UK. The report found that �At the post-exposure interviews at Southsea and Southend-on-Sea, subjects were asked if they had encountered health problems severe enough to cause them to visit their doctor, to lose days off work or normal activity or to seek hospital treatment... There were no significant associations between bathing and health either at each resort or as a whole." It further states, "These two findings are important, since they indicate that, neither from the results of the medical examinations nor from the rates of symptoms requiring medical attention or loss of normal activity, were any significant differences found between the bathing and non-bathing cohorts." Information was also sought on respondents� attitudes to a proposed advisory note system. These questions were about: • the perceived usefulness of the proposed system; • preferences for the type of information presented by the advisory notes; and • whether such a system would change the way respondents used beaches. Finally, awareness and perceived importance of EU and national standards and schemes for attainment and reward of good water quality and beach management practices were assessed through a single question at the end of this Section. 3.1.2 Valuation section The purpose of the valuation section was primarily to obtain respondents� willingness to pay for a number of beach improvements, including improvements in bathing water quality. The use of the choice-experiment method means that the results provide both individual valuations for each beach attribute as well as an implicit ranking of these. In the choice experiment, a series of choice cards were presented to each respondent. Each of these choice cards presented three possible scenarios for the average beach in Britain. The first scenario described the �current situation� at the average beach, which remained the same across all cards, and the other two scenarios (A and B) presented different levels of improvement over the current scenario. Respondents were asked to choose their preferred beach scenario out of these three in each card. Each of the improvement scenarios for the �average beach� was described by five different attributes, taking different levels, as described in Table 3.1 below. The experimental design used to create choice cards out of this group of attribute and levels is described below.



Table 3.1: Beach Attributes and their Levels: improvement scenarios Attribute No. of

levels Description of Levels in Improvement Scenarios

Water quality A (risk of suffering a stomach upset)

2 1/100, 0.5/100

Water quality B (unsafe bathing days)

2 Current number of unsafe days, 5/150, 1/150, where 150 is the number of days in a bathing season

Advisory note system 2 presence or absence

Cleanliness of the beach (litter and dog mess)

2 presence or absence

Quality of safety services and amenities

2 average or good

Cost of maintaining the beach scenario (payable through an increase in water rates)

4 £3, £11, £15 or £25 per household per year

In addition, for the description of the baseline current situation, the following levels were used (see Table 3.2). Table 3.2: Beach Attributes and their Levels: current situation Attribute No. of

levels Description of Levels in Current Situation

Water quality A (risk of suffering a stomach upset)

3 2/100, 5/100, 7/100

Water quality B (unsafe bathing days)

2 10/150, 16/150, where 150 is the number of days in a bathing season

Advisory note system 1 Absent

Cleanliness of the beach (litter and dog mess)

1 Present

Quality of safety services and amenities

1 Average

Cost of maintaining the beach scenario (payable through an increase in water rates)

1 £0 per household per year

Defining the attributes and levels In the case of water quality two different ways of describing this particular attribute were adopted, which gave rise to two different versions of the questionnaire. In version A the water quality attribute was described as the chance of getting a stomach upset from bathing. Stomach upsets are used as the main health impact of polluted water quality because this is the impact for which there is a reasonably sound and repeated epidemiological basis of evidence (notably Fleisher et al, 1993; Kay et al,



1994, see also Pruss, 1998 for a review). Whilst some epidemiological evidence exists for non-enteric illnesses (Fleisher, et al, 1996), the discussion in policy circles has mainly related to gastrointestinal symptoms (e.g. House of Lords, 1994)1. Furthermore the recent WHO (2001) report, Bathing Water Quality and Human Health, states that �The most frequent adverse health outcome associated with exposure to faecally contaminated recreation water is enteric illness, such as self-limiting gastro-enteritis�. This WHO report is expected to be the basis for new standards in the revised Bathing Water Directive. In version B water quality attribute was described as the number of days in the bathing season when it is unsafe to bathe. The first approach produces WTP for reducing the risk of suffering a stomach upset by 1 in 100, whilst the second produces WTP for an additional safe bathing day. Both are useful for the valuation of future beach management scenarios. While use of one or the other value will depend on the particular policy question being asked, Version A is used in this report when presenting aggregate benefits of revisions to the Directive to the population of England and Wales. Because at the time of designing the questionnaire the water quality of the current situation at the average beach in Britain was not known with certainty, a number of different baseline values were tested. This also allowed us to test whether or not a 'starting point bias'2 exists, whereby respondents value a change in risk or unsafe bathing days differently depending on how high the baseline risk is considered to be (See Section 5.2.3 for starting bias tests). Three starting points were tested for version A and two for version B, which resulted in five different versions of the questionnaire. The different levels of attributes for each of these versions are presented in Table 3.3.

1 Other illnesses from polluted bathing water include respiratory illnesses and ear and eye ailments. 2 While starting point bias usually refers to the price attribute, in this instance we apply it a change in risk.



Table 3.3: Levels for the Attribute of Water Quality in Different Versions of the Questionnaire

Version A Version B Chance of getting a stomach upset from

going in the water Number of days when it is unsafe to swim

(out of 150 days in the bathing season) Current

Situation Scenario

A Scenario

B Current

Situation Scenario

A Scenario

B A1 7/100 1/100 0.5/100 B1 16/150 5/150 1/150 A2 5/100 1/100 0.5/100 B2 10/150 5/150 1/150 A3 2/100 1/100 0.5/100 The three different risk levels for the starting points in version A of the questionnaire were agreed with the Steering Group, taking into account epidemiological work by Kay et al (1994), Fleisher (1998), and compliance rates of British beaches in 2001. The current mean excess risk of stomach upset from bathing in UK coastal waters was estimated by the Environment Agency (2002) to be 43 per 1000 exposures (4.3%). This figure was estimated using the WHO approach (WHO, 2001) to disease burden estimation3. The same methodology is also used to calculate the mean excess risk of stomach upset from bathing in UK coastal waters that comply with a proposed 95 percentile standard of 200 IE. This is estimated to fall to a figure of 20 per 1000 exposures (2.0%). Given this data, it was thought that a 7/100 risk is representative of poorer bathing water, 5/100 risk close to the average and 2/100 a good level of water quality. The values of 10 and 16 days per 150 days in the bathing season were chosen by DEFRA by examining bathing water results from across the country for the 2001 bathing season. The results were used to work out, using the number of high readings (those readings which exceeded a standard roughly equivalent to the Guideline standard under the existing EC Bathing Water Directive � similar to the Commission�s suggested new Mandatory standard for the revised Bathing Water Directive), how often an advisory note would have had to be posted over a sampling regime consisting of twenty samples. This result was scaled up to 150 bathing days (the bathing season). Sixteen advisory note days were thought to be representative of a poorer bathing water, whereas 10 advisory note days were thought to be more typical for the average beach. Whilst it would be useful to be able to find a relationship between the mean excess risk of stomach upset and the number of poor water quality days, this was not feasible within the resource constraints of the present study. COM(2000)860 on Developing a New Bathing Water Policy states that the revised Directive �should require the competent authorities to adopt new methods to actively

3 This takes the probability density function for IE exposure based on the average UK bathing water quality data for 2001, in combination with the relationship between water quality and excess probability of stomach upset from sea bathing found by Kay et al, (1994), to calculate the proportion of bathers likely to suffer stomach upset. The approach assumes that equal proportions of bathers use each bathing water. An additional factor was built into the basic WHO methodology in order to take into account the fact that less bathing takes place during periods of poor water quality (typically associated with poor weather conditions).



inform the public about bathing water quality, including about all known factors that might influence that water quality. This information should be available at all times at the bathing zone.� Based on this description an advisory note system was defined in the questionnaire as a system of notices that would inform bathers of when it is considered safe or unsafe to bathe. The payment vehicle chosen was water rates. This decision came out of focus group discussions covered in Section 3.2 below. Price levels were arrived at through a number of constraints. Zero is the starting point as there is no extra cost for maintaining the current scenario. The maximum price level was tested through focus groups and the pilot surveys. Price levels within the range had to meet the criteria of being non-linear, as a group. The total number of levels was determined by the constraints of the factorial experimental design of the choice cards. Whether these levels represent an adequate bracket is tested in Section 5.2.1. Designing the choice cards As discussed above, each beach scenario is made up of five attributes at different levels. The number of beach improvement scenarios that cover all possible combinations of attribute levels are determined by multiplying the number of levels that the attributes take in the hypothetical improvement scenarios (see Table 3.1). This implies a total of 64 beach improvement scenarios (2*2*2*2*4). Note that the current situation was not part of the experimental design. It was added to the pairs of improved beach scenarios to form triplets with a current situation and 2 possible improved scenarios4. Respondents are not able to cope with choosing simultaneously between such a large number of beach scenarios. Therefore a fractional factorial design, available in statistical software packages, was used to reduce the number of beach scenarios, while still maintaining the possibility of estimating �main effects� which are the effects of the attributes on the choices (which typically explain 80% of all the variation). The fractional factorial design produced 8 scenarios out of the original 64. This is still a large number of scenarios for respondents to be able to cope simultaneously. Next, choice cards were constructed to show to respondents using the 8 beach improvement scenarios. A number of considerations were necessary. • A choices from a set of 3 beach scenarios. Since the baseline has to be included

in the set, this implies grouping the 8 beach improvement scenarios identified above into pairs, to which the baseline is added.

• There are many possible ways of grouping the 8 beach improvement scenarios into pairs. Following a review of the literature, we adopted a random procedure to manually create 8 pairs of beach improvement scenarios from the 8 available scenarios, making sure there were no repeated pairs or pairs with dominated scenarios in them. Other procedures would have required larger samples.

• Adding the baseline to the 8 beach improvement scenario pairs created 8 triplets.

4 This is a limitation of the design, as the utility gained from purchasing beach improvements other than water quality were confounded with improvements to water quality, which always improved in scenarios A and B.



Each respondent was therefore asked to repeatedly choose their favourite beach scenario in each of 8 possible triplets. Effects of the various sub-samples The experimental procedure just described was repeated for each of the 5 sub-samples in our study described in Table 3.3 (A1, A2, A3 and B1, B2). Hence, we ended up with 5 sets of 8 cards each (mutually exclusive within versions A and B) which were then assigned to each of the sub-samples. The use of more than one design increased the efficiency of our estimates (Louviere et al, 2000). Administering the choice experiment The attributes and the scenarios were described to respondents by the interviewers and an example choice card was worked through with the respondents, in order to ensure they understood the task they were being asked to complete. This section of the questionnaire is presented in Box 3.2 below. 3.1.3 Debriefing and follow-up questions In this section respondents were asked a number of questions in order to learn more about the thought processes and motives for choosing the options they chose. These included two questions on why respondents were or were not willing to pay for improvements, where always choosing the current scenario is equivalent to not being willing to pay. These questions are important to identify invalid choices in the choice experiment, that is, choices that do not reflect the welfare change due to the choice selected. This includes those who choose not to pay because they object to the payment vehicle (e.g. water charges). Invalid answers are then removed from the data set before analysis is undertaken of willingness to pay. Detailed assessment of these responses is presented in Section 5.2.1. Another follow-up question asked whether respondents considered all attributes during the choice experiment or whether they focussed on one or two in particular, and if the latter, which ones. Also potentially influencing willingness to pay are respondents� attitudes to their health. A question here presented a series of statements about health and illness. Those respondents who have fatalistic views about health risks (e.g. nothing can be done to protect health), could be expected to be willing to pay less since they may not believe that improvements will reduce the risk to health. Finally, respondents were asked directly whether they would prefer to pay for water quality improvements or accept an advisory note system at no extra cost.



Box 3.2: Valuation Scenario � Description of Attributes and Worked Example I am about to show you a number of cards that describe three possibilities for the average state of beaches in Britain. In each card I am going to ask you to choose your preferred �average� beach. First of all, each �average� beach is described by a number of characteristics (SHOW CARD B-characteristic). (POINT TO EACH CHARACTERISTIC AND THEN READ CORRESPONDING DESCRIPTION IN TABLE BELOW) Table B-characteristic: CHARACTERISTIC DESCRIPTION

Average water quality

Average water quality is described as the risk of getting a stomach upset from going in the water.

Advisory notice system

This is the presence or absence of a system that advises against swimming on days when the water quality is worse than the average and hence the risk of getting a stomach upset is higher than the average.

Litter / dog mess

This is the presence or absence of litter and dog mess on the beach.

Safety & Amenities

This is the quality of the toilets and showering facilities, life-guards and life-saving equipment at the beach.

Additional water charges per year

£ This is the extra cost to all households in Britain for improving British beaches. Investments would have to be made to improve water quality and beach management, which would mean an increase in water charges that all households across the country would have to pay specifically to provide these services.

Let us look at an example of a choice card. (SHOW CARD A4-SIZED B-example). In each card the current situation on average in beaches across Britain is presented first. Then two other scenarios are shown, A and B, in which some improvements have been made, at some cost. In all cards the current scenario is the same. Let�s examine this first. (POINT TO CURRENT SITUATION ON CARD B-example) Some experts believe that at British beaches, on average, there is now a 7 in 100 chance of getting a stomach upset from going in the water (SHOW STOMACH UPSET CARD ALSO), so that 7 in every 100 bathers is expected to get a stomach upset. Currently there is no advisory notice system in operation on British beaches. There is some dog mess and litter and the



safety and amenities are of average quality. There is no extra cost to your household for maintaining this situation. (STILL SHOWING CARD B-example) In this example, in scenario A, average water quality is improved by reducing the risk of a stomach upset down to 1 in 100, so that only 1 in every 100 bathers on average get a stomach upset. In addition beach management measures are taken to clean up litter and dog mess. The other beach characteristics remain the same as the current situation. The extra cost of Scenario A is £11 a year. In Scenario B water quality and litter and dog mess are also improved to the same level as in Scenario A; and in addition an advisory notice system is in place and safety and amenities are improved to a good level. The cost of these improvements is £25 a year � or an extra £14 over Scenario A. The respondent in this worked example chose Scenario A because he was willing to pay an extra £11 a year for improved water quality and tidy beaches. However, he did not think that the additional improvements in Scenario B � the advisory notice system and the improvements in safety and amenities � were worth an extra £14 a year to his household. I would like you to consider a series of cards similar to this one, which show the current situation and two alternative scenarios. Please consider the 3 different beach scenarios carefully and take into account ALL of the characteristics � water quality, advisory notice system, litter/dog mess, amenities, life-guards and equipment, and cost (per year). The beaches are identical in all other aspects (e.g. access, sandiness, convenience, etc). Considering whether the cost of the additional measures is worth it to your household; please tell me, for each card, which scenario would you prefer to see in place in a typical British beach?



B EXAMPLE CARD Current Situation Scenario A Scenario B Average water quality (chance of getting a stomach upset)

7 in 100

1 in 100

1 in 100

Advisory notice system (advice against swimming on poor water quality days)

None

None

In place

Litter / dog mess Some

None

None

Safety & Amenities (toilets, showers, safety equipment and amenities)

Average

Average

Good

Additional water charges per year

£

£0

£11

£25

I would prefer: ! ✔ !



3.1.4 Socio-economic characteristics The purpose of this final section of the questionnaire was to collect information on socio-economic characteristics, which could be used to: • ascertain the representativeness of the survey sample relative to the population of

England and Wales as a whole; and • study how WTP for beach attributes varies according to respondents� socio-

economic characteristics. The survey collected data on: • sex; • age; • educational attainment; • socio-economic group; • presence of children in household; • employment status; • newspapers read; • household income; and • membership of any beach or environmental group. 3.2 Testing the Questionnaire The various designs of the questionnaire were tested through four distinct stages: • focus groups; • peer review; • pilot survey; and • main survey. The focus groups, pre-pilot and pilot stages of implementation were managed by eftec, and the main survey was conducted by Carrick James Market Research Ltd., a company with previous experience in conducting stated preference surveys. Prior to the first focus group, several versions of an initial questionnaire were drafted by eftec and discussed with the steering group. The questionnaire was drafted on the basis of prior experience in similar studies (see Section 2.3) and in line with best practice guidance in the field. Initial discussions took on board a number of considerations in the preparation of a draft questionnaire, such as the cognitive load that a respondent can be expected to bear when undertaking a choice experiment. Efforts were therefore made to reduce the initial list of proposed attributes, which came to nine. Findings of the literature were taken into account during these deliberations as were the considerations of efficient model estimation. Within and between each testing stage before the main survey, sufficient time was allowed to collate the lessons learnt and redesign the questionnaire accordingly. The remainder of this Section summarises the findings of the testing stages.



3.2.1 Focus Groups A focus group is a structured discussion group on a specific topic, facilitated by a moderator. Focus groups can be used to gain insights about issues of interest and particular methodological issues, including questionnaire design characteristics. More often than not, unexpected findings can be obtained from group discussions and interaction. Four focus group discussions were held in London and Norwich between 16 and 21 January 2002, conducted by members of the study team. Each group was made up of between six and 11 participants from a mix of socio-economic brackets, age groups and beach use experience. Summaries of the proceedings of each of the four focus groups are provided in Annex 2, along with a copy of the handouts provided and report of the results. The focus group discussions were structured to: • identify the most important beach characteristics; • identify the most appropriate payment vehicle; • assess which description of risk respondents understand better (i.e. days or health

risk); • obtain comments on design of the choice experiment (including the choice of

attributes and levels); and • test whether people were able to understand more difficult questions (such as

those relating to the advisory note system). Through the focus group discussions, we were able to ensure that the questionnaire was considered clear and credible. The results of the focus groups are summarised below. • Attitudinal and use questions: Overall many small changes were needed to the

questions in the attitudes and use section of the questionnaire to make these easier to both understand and to answer, and also to make the purpose of some questions clearer to avoid confusion. These ranged from adding a definition of beach visit in order to make this open to a wide interpretation and capture more users in the questionnaire by providing multiple choice rather than open-ended formats in some cases. Extra care was required in describing the advisory note system accurately and this wording was agreed at the focus groups.

• Attributes: One of the attributes that was in the initial list, namely �water sports

areas�, was found to be an unimportant factor in influencing people�s choice of beach. Therefore, it was eliminated from the list of attributes for the choice experiment. On the other hand the groups confirmed that water quality, advisory note system and dog mess and litter were important attributes.

• Levels: the selected range of levels for annual water charge increases was found

to be too wide, with £35 being considered too high. Other attribute levels posed no problem to the groups.



• Payment vehicle: the groups expressed distrust in the use of council tax as a payment vehicle and preferred water charges or a national tax.

• Valuation scenario: the use of scenarios depicting various alternatives for the

state of the �average beach� in Britain was considered acceptable and understandable by the groups.

3.2.2 Peer review The questionnaire was reviewed by Professor Nick Hanley from the University of Glasgow. His comments included a suggestion to add a worked example to the valuation scenario to aid people�s comprehension of the task at hand. He also suggested adding a travel cost component to the study and reducing further the number of attributes. These comments are further detailed in the pilot phase report in Annex 3. His comments along with the recommended changes emerging from the focus group report were discussed with the steering group and a number of changes were agreed in preparation for the main survey. 3.2.3 Pilot Study Once a fairly advanced draft of the questionnaire had been developed using input from the focus groups and detailed discussions with the steering group, this was subjected to a more formal and larger scale test by means of a pilot survey in February 2002. The pilot survey was managed by eftec, employing a number of interviewers, who were trained to administer the questionnaire. It was conducted on a sample of 50 people in various locations throughout London and Norwich. The pilot was divided into two phases to allow changes to be made half way through. The pilot served a number of objectives: • to identify any remaining problems in the wording of the questionnaire and the

formats used for answering each of the questions; • to test whether respondents could understand the choice experiment and whether

this was able to produce sensible results; and • to test how long it took to administer the questionnaire, and assess whether it was

too long. Interviewers were briefed after each phase of the pilot. Overall the survey went well. However, interviewers reported that respondents found the interview to be too long and sometimes lost interest as a result. Also some of the questions were difficult to understand. In order to address the first problem the length of the questionnaire was reduced with cuts made to questions that were not essential to explaining the results of the choice experiment and were not providing very useful additional information. These included those questions relating to the travel cost component of the study and a confusing question on the advisory note system, among others. To address the second issue a number of changes were aimed at improving clarity and credibility of the scenarios depicted in some questions. Following Professor Hanley�s



recommendation, a worked example scenario was added to make the choice experiment easier to administer and understand. The regression results revealed that the coefficients of the attributes were significant at the 5% or 10% level. Although there were some discrepancies in results, these were explained by the small sample size (25 per pilot phase). Selected results of the pilot study are discussed further in Annex 3. 3.2.4 Main Survey The sample numbers for each of the versions of the questionnaire are presented in Table 3.4 below. The table implies that a total of 488 people responded to Version A while 321 people responded to Version B. Table 3.4: Sample breakdown by questionnaire version

Version Sample size Percent A1 160 19.78 A2 164 20.27 A3 164 20.27 B1 160 19.78 B2 161 19.90

TOTAL 809 100 The main survey took place between 6 and 20 March 2002 at 49 different sites around England and Wales. A sample of 809 respondents was interviewed throughout these sites chosen among both coastal and inland locations. Depending on the question in the questionnaire, respondents were either representing themselves as individuals, or their family/household. Thus the sample is both representative of 809 individuals and 809 households. A list of the 49 sampling points is provided in Table 3.5. The refusal rate for interviews was 28 refusals for every 16 interviews. The survey was a random stratified survey designed to be representative of England and Wales on gender, age and socio-economic grade. The quota sheet used by interviewers to select a representative national sample is replicated in Annex 7. Table 3.6 presents a summary of selected socio-economic variables. As shown in the table, 55% of the sample are males and the average age is 41. Eleven per cent have not studied beyond primary school, 78% have completed secondary education or hold a professional degree (compared to 72.25% on average for England and Wales) and 7% hold a college or university degree. Forty-three percent of respondents are employed full-time, 13% work part-time, 8% are retired and 8% are self-employed. The Sun (17%), the Mirror (13%) and the Mail (13%) are the most popular newspapers; a distribution that compares very favourably to the average in England and Wales. Income is conceivably an important determinant of WTP for beach water quality and management. Average annual income is £21,234 per household, before taxes. This



compares favourably with the average annual income for England and Wales of £22,672. Table 3.5: List of sampling points for the main survey

District, County, Region 1 Newcastle-Upon-Tyne, Tyne & Wear, North East 2 South Tyneside, Tyne & Wear, North East 3 Langbaurgh-On-Tynes, Cleveland, North East 4 Stockton-On-Tees, Cleveland, North East 5 Wansbeck, Northumberland, North East 6 Doncaster, South Yorkshire, Yorkshire 7 Sheffield, South Yorkshire, Yorkshire 8 Kirklees, West Yorkshire, Yorkshire 9 Leeds, West Yorkshire, Yorkshire

10 Wakefield, West Yorkshire, Yorkshire 11 Great Grimsby, Humberside, Yorkshire 12 York, North Yorkshire, Yorkshire 13 Wigan, Greater Manchester, Lancashire 14 Manchester, Greater Manchester, Lancashire 15 Oldham, Greater Manchester, Lancashire 16 Rochdale, Greater Manchester, Lancashire 17 Stockport, Greater Manchester, Lancashire 18 Ellesmere Port & Neston, Cheshire, Lancashire 19 Leicester, Leicestershire, Midlands 20 Nottingham, Nottinghamshire, Midlands 21 Birmingham, West Midlands, Midlands 22 Coventry, West Midlands, Midlands 23 Hereford, Hereford & Worcester, Midlands 24 Shrewsbury & Atcham, Shropshire, Midlands 25 Stoke-On-Trent, Staffordshire, Midlands 26 Peterborough, Cambridgeshire, E. England 27 Norwich , Norfolk, Anglia 28 Ipswich, Suffolk, E. England 29 Hammersmith & Fulham, Greater London 30 Barnet, Greater London 31 Bexley, Greater London 32 Bromley , Greater London 33 Harlow, Essex, Greater London 34 East Hertfordshire, Hertfordshire, Greater London 35 North Hertfordshire, Hertfordshire, Greater London 36 Maidstone, Kent, Southern 37 Brighton, East Sussex, Southern 38 Fareham . Hampshire, Southern 39 Southampton, Hampshire, Southern 40 Ashford, Kent, Southern 41 Canterbury, Kent, Southern 42 Restormel, Cornwall & Isl. of Scilly, South West 43 Mid-Devon, Devon, South West 44 Plymouth, Devon, South West 45 North Dorset, Dorset, Southern 46 Sedgemoor, Somerset, Wales 47 Rhymney Valley, Mid Glamorgan, Wales 48 Taff-Ely , Mid Glamorgan, Wales 49 Neath, West Glamorgan, Wales



Table 3.6: Summary Statistics of Selected Socio-economic Variables

Survey sample

England & Wales

Total number of individuals/households 809 individuals/ households

52.9 million individuals (24 million households)

Demographic variables Males (%) 55 491 Average age 41 38 Age (using mid-point of intervals) 18-24 years old (%) 12 25-34 years old (%) 25 35-44 years old (%) 24

61

632

45-54 years old (%) 22 55-64 years old (%) 17

39

37

Education: Primary school (%) 11 18.53

O� levels (%) 54 20

A� levels or vocational training (%) 16 20

Professional Degree (%) 8 18

College or University (%) 7 18.5

Higher Degree (MSc, PhD etc.) (%) 2 4

Economic variables Household income (per year, using mid-points of intervals) 21,234 22,6724 Income non-response (%) 35 - Employment: Self-employed 8 85

Full-time employed (%) 43 48

Part-time employed (%) 13 16

Student (%) 4 1

Unemployed (%) 6 5.1

Housewife (%) 15 4

Retired (%) 8 2

Unable to work (%) 4 2

Socio-economic group: AB (%) 16 19.1

C1 (%) 30 28

C2 (%) 28.5

DE (%) 24 23.5

1 Population Trends - Spring 2002 2 These figures represent the proportion of the population in these age brackets where the total population of interest is between 16 and 64 (in order to compare with the survey age bracket). Original figures are: 16-44 is 41% of the total population and 45-64 is 24% of the total population (2000 figures). 3 Percentages are calculated for the population of England and Wales aged 16-44 and taken from Education and Training Statistics for the UK (Dept. for Education and Skills, 2001). 4 From the family expenditure survey: average for England and Wales. Adopted from the weekly income figures presented in Family Spending 2000-01, Office of National Statistics (2002). 5 All employment data for Dec 2001 - Feb 2002. From Labour Trends, Spring 2002. Percentages based on ages 16-64 for total population.



Table 3.6 continued�.

Other variables Newspapers read: The Sun (%) 17 206

The Mirror (%) 13 12 Daily Mail (%) 13 12 Local Paper (%) 14 - The Times (%) 4 3 The Guardian (%) 4 2 Don�t read newspapers (%) 21 47

Overall, the sample is representative of the population of England and Wales as the comparisons in Table 3.6 demonstrate. It is important to inspect whether there are differences in the observable characteristics of the various sub samples used in the study (A1, A2, A3, B1 and B2). This is because differences in the results across versions may be attributable to differences in the sample characteristics. Annex 6.1 contains the analysis of the socio-economic characteristics of each sub sample. The results show that there are no differences across the various sub-samples in any of the variables at the 5% level (although the result is only marginal in the case of age and gender between A1 and A2). The proportion of respondents who did not reveal their income was 35%. This result is comparable to percentages obtained in similar studies. Overall, 16% of all respondents belonged to socio-economic segment AB, 55% to segments C1 and C2, and 23% to group DE. See Section 5 for a test of whether this is likely to have an effect on WTP estimates.

6 From Social Trends, Spring 2002. Figures for 2000/1 for Great Britain.



IV. RESULTS This section presents an overview of the attitudes to the questionnaire (Section 4.1); attitudes and uses (Section 4.2); and valuation results (Section 4.3). 4.1 Attitudes Towards the Questionnaire As Figure 4.1 illustrates, the majority of respondents (62%) rated the questionnaire as interesting and only 9% considered the scenarios presented to be unrealistic. Moreover, only 6% of the sample found the questionnaire difficult to understand. These results indicate that the initial design and pre-test phases were successful in producing an effective survey instrument, which was credible and easy to understand. Figure 4.1: Attitudes Towards the Questionnaire

Note: percentages add up to more than 100% as respondents were allowed to give more than one answer.

4.2 Attitudes and Uses Section A of the questionnaire contained a number of questions relating to respondents� use of beaches in Britain, their preferences for different beach characteristics as well as their attitudes to beach related risk and health issues. 4.2.1 Beach uses and preferences Responses to questions A1-A3 reveal that the vast majority (97%) of respondents had visited a beach in Britain sometime in their life. A beach visit was defined loosely as spending any amount of time on or around a beach. Out of all respondents, 92% had visited a British beach in the last five year and 76% in 2001. This compares with the 1998 UK Day Visits Survey finding that 51% of the English population and 61% of the Welsh population visited the British coast in 1998 (National Centre for Social Research, 1999). Respondents were asked, in question A-4, to state the total number of days they spent visiting British beaches in the last year they visited. Table 4.1 summarises this information, showing the distribution of visits for each type of trip taken, i.e. as a local resident, day tripper, on a short break or on holiday in the year 2001. The

62

126

159

0

25

50

75

Interesting Too Long Difficult tounderstand

Educational Unrealistic/not credible



number of trips for each type of activity is comparable between both estimates. Local residents make up the majority of beach visits at 42% of the total. Table 4.1: Distribution and Number of Beach Visits per Year No of visits in 2001* % of total As a local resident 3,758 42% As a day tripper 2,122 24% On short breaks 743 8% On long breaks 1,109 12% Other 1,235 14% TOTAL

8,967 100%

* for the sample population of 809 Dividing the total number of yearly visits by the sample population produces an average number of beach visits of 11.08 per person in 2001 (see Table 4.2). This calculation reflects the fact that there are some �outlier� observations in the upper portion of the distribution, i.e. people who visited the beach every single day of the year. The 10% trimmed mean value (mean of the 10th to 90th percentiles) is equal to 5.69 visits. If we take the median number of visits then this is equal to 3 visits per person per year. These figures compare with the 1998 UK Day Visits Survey (National Centre for Social Research, 1998) which found an average of 4 visits per year per person in 1998. In addition the 1998 UK Day Visits Survey found that there were in excess of 240 million visits to the British Coast in 1998. If we multiply the trimmed mean in Table 4.2 with the population of England and Wales we arrive at an estimated total number of visits in 2001 of 301 million. Table 4.2: Mean and Median Number of Beach Visits per Person per Year

Beach visits per person in 2001

Mean

11.08

Trimmed mean 5.69

Median 3

Figure 4.2 shows that the sample stated that walking, relaxing, sun-bathing and picnicking are the most commonly practised activities on the beach (question A-5). Activities involving contact with seawater, such as swimming and water sports, on the other hand are much less frequently practised: 57% of respondents and their families never go swimming, while 78% never participate in water sports. These results are supported by focus group discussions in which many people admitted to finding the British beaches too cold for swimming.



Figure 4.2: Beach Uses

0

10

20

30

40

50

60

70

80

90

%

OftenSometimesRarelyNever



Figure 4.3: Preferences for Beach Attributes

1 2 3 4 5

convenience of location, access and parking

water quality

tidal conditions

water which looks clear and free from litter and foam

sandiness of beach

water sports options

cleaniness of beach no litter/ dog mess)

life guards and life-saving equipment

clean toilets, showers and changing facilities

few people, not congested

restaurants, kiosks, pubs, shops and amusements

information warning you when it is not safe to swim due to tidal currents

having a quality award

Not Important Very Important



Figure 4.3 illustrates the results of a Likert rating of various beach characteristics (question A-16). The chart shows average ratings chosen across the sample and demonstrates that the most important characteristics for respondents were the cleanliness of the beach, followed by water that looks clear of foam and litter, then changing facilities. A closer look at the data reveals that 51% of respondents responded that water quality was very important, i.e. they selected 5 on the Likert scale. Weather was a significant �other� beach characteristic identified by respondents as being important. 4.2.2 Exposure to risk of illness (stomach upset) In order to assess the exposure of the population to the risk of suffering a stomach upset from potentially poor water quality, respondents were asked how many times they had participated in a number of different water-based activities and how often they tended to put their head under water in the process (question A-8). Tables 4.3 and 4.4 present results for each recreational water sport activity for the year 2001. Table 4.3 presents the percentage of the sample that undertook each category of water sport in the year 2001 as well as the breakdown of this total percentage according to the four different exposure categories respondents identified. Table 4.3 also reveals that only about 20% of the population went swimming in 2001. Table 4.3: Exposure to Water from Swimming and Other Recreational Water Sports (using 2001 as count year)

TOTAL of respondents

who did activity in

2001

Often

Some-times

Rarely Never

How often do you tend to put head under water when swimming in Britain?

20.4% 6.43% 4.94% 2.10% 6.92%

How often do you tend to put your head under water when surfing/windsurfing/ water-skiing in Britain?

2.35% 1.36% 0.87% 0% 0.12%

How often do you tend to put your head under water when diving/ snorkelling in Britain?

2.1% 1.85% 0.25% 0% 0%

How often do you tend to put your head under water when jet-skiing/canoeing in Britain?

1.48% 0.62% 0.49% 0% 0.37%

How often do you tend to put your head under water when boating/sailing in Britain?

4.33% 0.49% 0.74% 0.62% 2.47%



Table 4.4 presents a calculation of the total number of times the sample undertook each category of water sport for the year 2001, and then divides each total according to the type of exposure level that is expected for each water sport category. From Table 4.3, we see that about 6.43% often put their head under water when swimming, accounting for a total of 460 swims/dips where the head was submerged (in Table 4.4). Table 4.4: Number of Times Undertook Activity by Exposure Category (using 2001 as count year)

TOTAL number of times

undertook activity in 2001

Often puts head under water

Some-times

puts head under water

Rarely puts head

under water

Never puts head

under water

swim/dip 1,291

460 448 37 346

surfing/windsurfing/ water skiing

311

232 71 0 8

diving/snorkelling 127 121 6 0 0

jet-skiing/canoeing 150 124 20 0 6

boating/sailing 373 74 34 13 252

All recreational water activities

2,252 1,011 579 50 612

Dividing the number of swims/dips by the total sample produces a mean number of swims/dips of 1.6 per person per year (found by dividing the total number of swims, 1,291, by the sample size, 809). When applied to the population of England and Wales (52.9 million) this results in a total number of swims/dips in 2001 of 84.64 million. However, when the head is not submerged in the water during a dip/swim, a bather is not considered to have been exposed to enteric pathogens (Kay, 1994). Therefore in order to calculate the number of exposures, only those swims/dips for which the head was submerged should be counted. The most conservative calculation would be to assume that only those who stated that they often put their head under water should be counted in the analysis. This results in the calculations presented in Table 4.5a below.



Table 4.5a: Gross Estimate for Predicted Illness (stomach upset) (1)

No. of swims in 2001

(2) Exposure Rate (=(1)/809)

(3) Risk of suffering stomach upset

Number of cases of stomach upset per year (=(2)x (3)x 52.9 million*)

For swim/dip 460 0.57 0.043 1.30 million

(95% confidence interval)

(0.331 � 0.798)

(0.75 � 1.81 million)

For all water sports activities

1011 1.25 0.043** 2.84 million


(0.635 � 1.864)

(1.44 � 4.24 million)

* Population of England and Wales. ** This figure, which represents the average water quality on British beaches, relates to risks related to swimming/dips only, and may or may not be correct for all water sports. A sensitivity analysis of this result is provided in Annex 4, making different assumptions about the number of exposures and the average risk level. Some public health professionals have reason to believe that frequent swimmers will develop immunity to gastro-enteritis. If this were indeed the case, then the total number of illnesses would be smaller, as repeat swims above a certain number would no longer increase the risk of becoming ill. However, as there is no epidemiological evidence yet to this effect, this caveat has not been incorporated into the analysis here. The above result can be compared with responses to questions in the survey about personal experience of bathing water-related illness. These questions found that 8.16% of respondents (66 persons) believed they had suffered a stomach upset as a result of poor quality bathing water in Britain at some time in the past. While 2.59% of respondents (21 persons) believed they had suffered a stomach upset in the past year (since 2001) as a result of poor water quality in British bathing water. The total number of cases of stomach upset believed by respondents to have resulted from poor water quality in the last full year (2001) was 34. As well as being asked the total number of cases of stomach upset to have resulted from poor water quality in the year 2001, respondents were also asked the total number of days that they suffered these symptoms. The average duration of illness amongst those reporting stomach upsets that they believed was due to poor bathing water quality was 4.1 days. This figure is identical to that reported by Fleisher et al (1998), and hence is considered to add further credibility to the personal assessments and reporting of bathing water illnesses found in the present study. Using the total number of cases of stomach upset data we can calculate the number of stomach upsets per year in England and Wales, based on personal assessment. This calculation is shown in Table 4.5b below.



Table 4.5b: Gross estimate of personal assessment of illnesses (1)

Number of cases of stomach upset from poor water quality in 2001

(2) Number of cases per person per year (=(1)/809)

Number of cases of stomach upsets per year in England and Wales (=(2)x 52.9 million*)

Gross estimate of personal assessment of illnesses

34

0.042

2.2 million


(22.34 - 45.65) (1.461 million � 2.99 million)

* Population of England and Wales. The overlap in the confidence intervals between the two results in Tables 4.5a and 4.5b shows that there is a degree of consistency between personal assessments of bathing water related illness and predictions from the epidemiological literature, as well as an internal consistency in the way that people answered these two sets of questions. These consistencies lend weight to the results and the methodology used. In addition, if, instead of using the more conservative result of the �often submerges head�, we use the �sometimes submerges head� result (see Annex 4) the convergence of the two results is even greater. As there is no sound scientific evidence that the risk levels derived for swims/dips can be applied to all water sports, comparison of this result with the personal assessment of risk is discouraged. Also, a certain amount of error should be expected in the self-reporting of stomach upsets, which is due to a number of well-recognised factors. For example, respondents may not fully remember events that happened in the past. This inability to remember may lead to errors of omission (inability to recall an event that has occurred), telescoping (compressing time so that an event is remembered as occurring more recently than it actually did) and creation (recalling an event that did not actually occur). A respondent may have recalled the stomach upset correctly, but may not know if it was due to a hamburger eaten on the beach that day or the poor bathing water quality. Question A-12 aimed to discover how respondents rated the risk of suffering a stomach upset and how this compared with other beach-related risks. Figure 4.4 shows that while the majority of respondents did not think there was a high risk of getting a stomach upset from swimming in British bathing waters, they considered this risk to be higher than near-drowning or collision with jet-skiers. Figure 4.4: Perceptions of beach recreation risks

2.65

2.22

1.77

1 1.5 2 2.5 3 3.5 4 4.5 5

Stomach upset

Near drowning

Collision with jet-ski

Not at risk at all High risk



4.2.3 Attitudes to health risk In question C-5 respondents were presented with a series of statements relating to general health and health-risk attitudes and were asked to what extent they agreed or disagreed with each statement. Figure 4.5 presents these results. Figure 4.5: Attitudes towards health and risk a. �I am very careful with my health�

b. �No matter what I do, if I�m going to get sick I will�

c. �I know there are other things I can do to improve my health but I just can�t be

bothered.�

2

6

24

34

35

0 5 10 15 20 25 30 35 40

Disagree strongly

Agree strongly

%

18

20

23

20

19

0 5 10 15 20 25 30 35 40

Disagree strongly

Agree strongly

%

30

23

23

16

8

0 5 10 15 20 25 30 35 40

Disagree strongly

Agree strongly

%



d. �I am only really concerned about serious illnesses, I can cope with the rest�

e. �If I take care of myself I can prevent illness from happening.�

These results show that the majority of respondents feel that they have control over their health and can prevent illness from happening. This is contrary to feeling that outside forces decide whether or not you will get ill. It is expected that those who feel they have less control over their health, i.e. those who are fatalistic, may be less willing to pay for reductions in health risks from bathing waters. Alternatively if one considers the locus of that internal control over health extends to control over paying for cleanup, then this may translate to being more willing to pay for health risk reductions. The degree of fatalism expressed in these answers is used in validity testing (see Section 5). 4.2.4 Preferences for an advisory note system The results of several questions aimed at assessing the public�s desire for and potential reactions to an advisory note system are summarised here. When presented with a description of the proposed advisory note system (question A-13), the majority (51%) of respondents thought that it would be very useful, while only 4% thought it would be not useful at all (see Figure 4.6).

14

17

26

23

20

0 5 10 15 20 25 30 35 40

Disagree strongly

Agree strongly

%

9

10

29

30

22

0 5 10 15 20 25 30 35 40

Disagree strongly

Agree strongly

%



Figure 4.6: Perceived Usefulness of an Advisory Note System

When asked �Which sort of information would you prefer to see on an advisory notice board?� (question A-14) 38% of the sample responded that they would prefer to see a simple symbol indicating �do not bathe�, whilst 51% responded that they would prefer to see more detailed information. Presented with a scenario where an advisory note advises against swimming on the day that the respondent visits a particular beach (question A-15), about 78% of respondents said they would still visit the beach if an advisory note advised against going in the water because of poor water quality (see Table 4.6). This allays fears that the advisory note system would have a significant adverse effect on beach visitation, although arguably the some 19% of people who decide not to stay at the beach as a result of the sign may still represent a notable economic loss. The most common response was that they would still visit the beach, but avoid contact with the sea (about 46% of respondents). Table 4.6: Behavioural Responses to an Advisory Note System Statement Percentage of respondents

agreeing with the statement Would still visit the beach as typically I do not go in the sea

27.94%

Would still visit the beach but avoid contact with the sea

45.49%

Would still visit the beach and use the sea 4.33% Would turn around and not visit that particular beach 18.54% Don�t know 0.12% 4.2.5 Attitudes towards beach standards and schemes The importance that respondents attach to various standards and schemes promoting good beach and water quality may have an influence on their willingness to pay for improvements in these. For example, those who believe these to be important, may be expected to be more likely to have a positive WTP for improvements. Figure 4.7 below provides an overview of rating responses to this question.

12

4

3

11

18

52

0 10 20 30 40 50 60

Don't know

Not useful at all

2

3

4

Very useful

%



Figure 4.7: Rating of Beach Standards and Schemes

The results show that most respondents thought that all of these schemes were generally quite important. Not much differentiation is evident between the schemes. This is likely because for the most part many respondents were unaware of the scheme or standard (see Table 4.7). An exception is the Blue Flag scheme, of which 81% were aware. Table 4.7: Awareness of Beach Standards and Schemes

Scheme/ Standard % of sample aware of standard or scheme

Blue Flag 81 %

Seaside Award 59%

Good Beach Guide 55% European Union bathing water standards 49%

Local authority poster scheme 49%

Green Coast 34% 4.3 Valuation Results 4.3.1 The model The standard economic model that underlies the choice experiment method used in this study is based on the following indirect utility function Uij that represents the satisfaction that individual i receives from the state of an average British beach j: Uij = Vij+ εij = b1 WATERij + b2 ADVISORYij+ b3 LITTERij+ b4 AMENITYij+ b5

PRICEij+εij (1) where Vij is a deterministic element which is a linear index of the attributes of the j different alternative beaches; and εij is a stochastic element which represents unobservable influences on individual choice. In this model, the beach attributes are

1 1.5 2 2.5 3 3.5 4 4.5 5

EU bathing water standards

Blue Flag

Green Coast

Seaside Award

Good Beach Guide

Local authority poster scheme

Not at all important Very important



water quality (WATER), presence of an advisory note system (ADVISORY), presence of litter and dog mess (LITTER), quality of amenities and safety provisions (AMENITY) and the cost to the household of achieving a certain level of beach quality (PRICE). The coefficients b1, b2, b3 and b4 describe the influence of the various attributes on Uij. Formally, the probability that a particular beach scenario is chosen by a respondent i can be expressed as a conditional logit (CL) model:

∑=≠∀>

jij

igihig

VVghUUP

)exp()exp(),( (2)

This model can be estimated using maximum likelihood procedures and the log likelihood function which is shown below; where yij is an indicator variable which takes a value of one if respondent i chose beach scenario j and zero otherwise:

∑ ∑∑

== =

=

N

i

J

j J

jij

ijij

V

VyL1 1

1

])exp(

)exp(log[log (3)

Three different conditional logit models were estimated for each version of the questionnaire (Bennett and Blamey, 2001; Louviere et al, 2000). The first model is a basic choice model which explains respondents� choices between beach scenarios solely as a function of their attributes (as in equation 1 above). The second model includes an alternative specific constant (ASC) for the third scenario in each choice set. ASCs capture any variation in choices that is not explained by either the choice attributes or respondent-specific socio-economic variables. Unlike other ASCs that were tried and subsequently dropped from our estimation, the constant for the third beach scenario in each choice set remained highly significant across all models. What distinguishes the third scenario in each choice set is that it corresponded always to the most expensive improvement option. Finally, the last CL model also includes socio-economic and attitude regressors, specific to individual respondents. This specification allows us to introduce respondent heterogeneity, i.e. differences between individuals, into the model. Individual-specific variables however cannot be introduced into the model on their own as they would drop from the estimation since they do not vary across choices. Hence, these variables have to be interacted either with the beach attributes or with the ASCs. In this report we present the results of interactions with the ASC. However, the conditional logit model failed the test for compliance with the Independence of Irrelevant Alternatives (IIA) property upon which the model critically relies. This property states that the relative probabilities of two beach scenarios being selected (from the choice set) are unaffected by the introduction or removal of other alternatives. Since this was not the case, another more advanced



model specification that does not rely on the IIA assumption was required. The model used for the second round of analysis was the nested logit (NL) model. The Nested Logit (NL) model avoids the need to rely on IIA by modelling choices in a hierarchical nested structure. This device allows error terms across choices within each �nest� to be correlated with one another, although choices across �nests� are still assumed to be uncorrelated. In order to apply the NL framework to the present data, the choice problem needs to be recast as a hierarchical nested structure. In fact, it seems perfectly possible that the respondents could be organising the choice problem in two stages, first whether or not they are willing to pay something towards improved beaches, and second which of the two beach improvement scenarios they prefer. This would imply a nesting structure along the following lines.

CHOICE

In a two tier choice structure, the probability of choosing a particular alternative k out of the n second stage scenarios, conditional on having selected a particular alternative j out of the m first stage scenarios, can be expressed as indicated in equation (4). The logarithm of the denominator of this expression is known as the inclusive value (I), because it summarises the information about the alternatives included in this lower nest. Inserting this inclusive value as an explanatory variable in the first stage of the decision tree yields the expression for the unconditional probability of choosing scenario j out of the m first stage scenarios, given in equation (5). Inclusive value, I, is reported in Annex 6.

jkI

Xikjkb

nikXnbikXjkb

jkP|exp

)|exp(

)exp(

)|exp()|( ==

∑ (4)

P jc X I

c X Ij ij k j

m ij k jm

( )exp( )

exp( )/

/

=+

+∑ρ

ρ (5)

No payment Current situation

(baseline alternative)

Some payment Scenarios A or B

Improvement Scenario A

Improvement Scenario B

STAGE 1:

STAGE 2:



Three variations on the nested logit model were tested, as with the conditional logit model. The first explains respondents� choices solely as a function of beach attributes; the second model includes an alternative specific constant (ASC) for the third scenario in each choice set; and the third model also includes socio-economic and attitude regressors, specific to individual respondents. In terms of regressors, three different types of variables were included: i) beach attributes; ii) socio-economic characteristics of respondents; and iii) variables related to uses of and attitudes towards British beaches. Care must be taken in including this latter category as these variables tend to be correlated with socio-economic factors creating a problem of endogeneity. The variables are defined as follows: C3 Alternative specific constant for scenario B (most expensive

improvement, and third option) Water A Chance of getting a stomach upset (e.g. 1 in 100) Water B Days in the bathing season in which it is unsafe to swim (e.g. 1 in 150) Advisory Presence of an advisory note system advising against swimming on poor

water quality days (dummy variable) Litter Presence of litter and dog mess (dummy variable) Amenity Dummy variable indicating whether the level of amenities and safety

provisions present is good or average Price Water rates Sex Gender dummy (1 if male) Age Age Inc Household Income (Yearly) Socio Dummy variable indicating whether respondents were in socio-

economic segments AB/C1 or C2/DE Often Dummy variable showing if respondents engage often in swimming or

water sports at British beaches Wqimp Dummy variable showing if respondents thought that water quality was

a very important factor in choosing to visit a British beach Tables 4.8a and 4.9b present the estimation results of both conditional and nested logit models for both versions of the questionnaire. All nested logit models performed well. The results show that all beach attribute coefficients are significant at the 1% level and have the expected signs across all versions and model specifications. This implies that higher levels of water pollution are undesirable as is the presence of litter/dog mess; while increased amenity quality and the presence of an advisory note system have a positive impact on utility. And, of course, the higher the water rates, the lower the welfare. The alternative specific constant is negative and also highly significant. As noted above, the socio-economic and attitude variables used in the model interacted with the ASC for the third beach scenario in the choice sets. The interactions therefore show the effect of these variables on the probability of choosing the most expensive beach improvement scenario. Results show that richer respondents, those who are often in contact with water (via swimming or water sports) and those that consider water quality to be a very important determinant of their



choice of a beach (only in the model for version A) are more likely to choose the most expensive beach improvement scenario (which also afforded the largest improvements). These results are consistent with expectations. Age and gender were not found to be significant determinants of choice. Protest bids are responses to the valuation question which do not give the respondent�s genuine WTP because their reasons for not bidding (or bidding at an excessively high level) do not reflect the welfare change from the good being considered. For example, when the reason for not bidding is because the respondent considers that the government should pay, this does not reflect a zero value, but a protest against the payment vehicle. Protest bids are generally considered invalid and hence have been excluded from the analysis (See Section 5 for further analysis).. Table 4.8a: Conditional Logit Modelling Results (Version A: Utility of Reductions in the Chance of Contracting a Stomach Upset)

CL Model + ASC + Interactions

Coeff. (s.e.)

P-value

Coeff. (s.e.)

P-value Coeff. (s.e.)

P-value

C3 _

_ -0.072 (0.071)

0.000 -1.579 ( 0.246)

0.000

Water A -0.094 (0.013)

0.000 -0.049 (0.014)

0.001 -0.028 (0.017)

0.097

Advisory 0.506 (0.057)

0.000 0.621 (0.059)

0.000 0.687 (0.072)

0.000

Litter -0.465 (0.056)

0.000 -0.499 (0.057)

0.000 -0.488 (0.070)

0.000

Amenity 0.230 (0.054)

0.000 0.338 (0.057)

0.000 0.361 (0.069)

0.000

Price -0.076 (0.004)

0.000 -0.046 (0.005)

0.000 -0.042 (0.006)

0.000

Sex*C3 _

_ _

_ -0.152 (0.112)

0.176

Age*C3 _

_ _

_ 0.002 (0.005)

0.700

Inc*C3 _

_ _

_ 0.0000328 (3.89E-6)

0.000

Often*C3 _

_ _

_ 0.596 (0.163)

0.000

Wqimp*C3 _

_ _

_ 0.281 (0.112)

0.012

Log-Lik -3122.01 -3069.79 -2074.53 Pseudo R2 0.07 0.09 0.10 N respondents

409 409 260



Table 4.8b: Conditional Logit Modelling Results (Version B: Utility of Changes in the Number of Days it is Safe to Swim)


Coeff. (s.e.)



P-value

C3 _

_ -0.735 (0.091)

0.000 -0.825 (0.268)

0.002

Water B -0.061 (0.007)

0.000 -0.047 (0.007)

0.000 -0.049 (0.009)

0.000

Advisory 0.390 (0.075)

0.000 0.561 (0.079)

0.000 0.563 (0.095)

0.000

Litter -0.474 (0.069)

0.000 -0.472 (0.070)

0.000 -0.537 (0.085)

0.000

Amenity 0.275 (0.068)

0.000 0.295 (0.070)

0.000 0.269 (0.084)

0.001

Price -0.081 (0.005)

0.000 -0.048 (0.006)

0.000 -0.056 (0.008)

0.000

Sex*C3 _

_ _

_ 0.024 (0.138)

0.862

Age*C3 _

_ _

_ -0.007 (0.006)

0.228

Inc*C3 _

_ _

_ 0.0000103 (5.03E-6)

0.040

Often*C3 _

_ _

_ 0.783 (0.156)

0.000

Wqimp*C3 _

_ _

_ 0.181 (0.137)

0.187


249 249 170



Table 4.9a: Nested Logit Modelling Results (Version A: Utility of Reductions in the Chance of Contracting a Stomach Upset)

NL Attributes only NL attributes + ASC3

NL attributes + ASC3 + individual-specific

variables

Coeff. (s.e.)



P-value

C3 _

_ -0.715 (0.074)

0.000 -1.031 (0.099)

0.000

Water -0.1977 (0.034)

0.000 -0.066 (0.024)

0.007 -0.052 (0.022)

0.017

Advisory 0.695 (0.074)

0.000 0.672 (0.080)

0.000 0.630 (0.082)

0.000

Litter -0.654 (0.074)

0.000 -0.544 (0.075)

0.000 -0.510 (0.076)

0.000

Amenity 0.240 (0.062)

0.000 0.354 (0.061)

0.000 0.341 (0.060)

0.000

Price -0.097 (0.006)

0.000 -0.049 (0.006)

0.000 -0.047 (0.006)

0.000

Socio*C3 _

_ _

_ 0.276 (0.093)

0.003

Often*C3 _

_ _

_ 0.727 (0.134)

0.000

Wqimp*C3 _

_ _

_ 0.195 (0.092)

0.035


383 383 383



Table 4.9b: Nested Logit Modelling Results (Version B: Utility of Changes in the Number of Days it is Safe to Swim)

NL Attributes only NL attributes + ASC3

NL attributes + ASC3 + individual-specific

variables

Coeff. (s.e.)



P-value

C3 _

_ -0.750 (0.096)

0.000 -0.929 (0.126)

0.000

Water A -0.079 (0.013)

0.000 -0.060 (0.014)

0.000 -0.051 (0.014)

0.000

Advisory 0.514 (0.105)

0.000 0.643 (0.108)

0.000 0.591 (0.113)

0.000

Litter -0.548 (0.083)

0.000 -0.528 (0.087)

0.000 -0.494 (0.091)

0.000

Amenity 0.314 (0.076)

0.000 0.328 (0.079)

0.000 0.308 (0.078)

0.000

Price -0.092 (0.008)

0.000 -0.053 (0.008)

0.000 -0.050 (0.080)

0.000

Socio*C3 _

_ _

_ 0.184 (0.115)

0.109

Often*C3 _

_ _

_ 0.596 (0.137)

0.000

Wqimp*C3 _

_ _

_ -0.363 (0.116)

0.755


249 249 249

4.3.2 Willingness to Pay Results Using the coefficients estimated in the models above, point estimates of the willingness to pay for a change in one of the beach attributes can be calculated by estimating �implicit prices�. These are marginal rates of substitution between the attribute of interest and the price attribute. It is straightforward to show that, for the linear utility index described in equation 1, these implicit prices are given by:

ice

Attribute

bb

WTPPr

−= (6)

The coefficient bPrice gives the marginal utility of income and is the coefficient of the cost attribute. Tables 4.10 and 4.11 contain estimates of household mean willingness to pay for each attribute level (per year) and respective standard errors for versions A and B of the questionnaire. Table 4.12 summarises the findings. Findings for both the conditional logit and the nested logit model are presented, although only the nested logit results are recommended for use.



Table 4.10: Household Implicit Mean Willingness to Pay for Beach Attributes: Values in £ per Year (Version A)


Basic model + ASC + Interactions

WTP (s.e.)

P-value WTP (s.e.)

P-value WTP (s.e.)

P-value

Water A 1.242 (0.162)

0.000 1.068 (0.272)

0.000 0.678 (0.373)

0.070

Advisory 6.670 (0.694)

0.000 13.621 (1.746)

0.000 16.559 (2.612)

0.000

Litter 6.130 (0.717)

0.000 10.951 (1.550)

0.000 11.771 (2.134)

0.000

Amenity 3.033 (0.708)

0.000 7.423 (1.455)

0.000 8.700 (2.038)

0.000

Nested Logit Model


WTP (s.e.)

P-value WTP (s.e.)

P-value WTP (s.e.)

P-value

Water A 2.040 (0.277)

0.000

1.337 (0.397)

0.000

1.113 (0.376)

0.003

Advisory 7.166 (0.625)

0.000 13.655 (1.715)

0.000 13.535 (1.729)

0.000

Litter 6.756 (0.640)

0.000 11.066 (1.513)

0.000 10.966 (1.545)

0.000

Amenity 2.478 (0.650)

0.000 7.204 (1.447)

0.000 7.334 (1.449)

0.000

Note: The NL model with the ASC3 and individual specific variables (2nd model) has the best fit (and higher WTP results). The WTP for water quality is very stable across all NL model specifications. The mean WTP values for �Advisory�, �Litter� and �Amenity� are all higher in the �+ASC� and �+Interaction� models than in the basic �CL model�.



Table 4.11: Household Implicit Mean Willingness to Pay for Beach Attributes: Values in £ per Year (Version B)



WTP (s.e.)

P-value WTP (s.e.)

P-value WTP (s.e.)

P-value

Water B 0.758 (0.077)

0.000 0.973 (0.139)

0.000 0.870 (0.140)

0.070

Advisory 4.819 (0.862)

0.000 11.630 (2.090)

0.000 10.090 (2.057)

0.000

Litter 5.856 (0.851)

0.000 9.784 (1.786)

0.000 9.622 (1.865)

0.000

Amenity 3.389 (0.839)

0.000 6.113 (1.582)

0.000 4.821 (1.593)

0.003

Nested Logit Model


WTP (s.e.)

P-value WTP (s.e.)

P-value WTP (s.e.)

P-value

Water A 0.862 (0.096)

0.000 1.129 (0.203)

0.000 1.019 (0.204)

0.000

Advisory 5.593 (0.885)

0.000 12.109 (2.064)

0.000 11.790 (2.115)

0.000

Litter 5.971 (0.795)

0.000 9.947 (1.747)

0.000 9.847 (1.776)

0.000

Amenity 3.423 (0.785)

0.000 6.183 (1.534)

0.000 6.136 (1.563)

0.000

Note: The NL model with the ASC3 and individual specific variables (2nd model) has the best fit (and higher WTP results). The WTP for water quality is very stable across all NL model specifications. Version B produces more stable estimates than version A. In Version B, only the WTPadvisory is different between the basic attributes only model and the other two. The mean WTP values for �Advisory�, �Litter� and �Amenity� are all higher in the �+ASC� and �+Interaction� models than in the basic �CL model�. These estimates indicate that for example, respondents are willing to pay between £1.10 and £2.00 per year for a 1 in 100 reduction in the chance of contracting stomach upset from British bathing water (nested logit results in Table 4.10). Interestingly, this value range is similar to the 90 pence to £1.10 that respondents are also seem to be prepared to pay to avoid 1 day of poor water quality (nested logit results in Table 4.11). Given that we do not know the health risk relationship between a 1 in 100 chance of suffering a stomach upset and an unsafe bathing day we cannot make any judgements about this outcome. There are no prior expectations as to what values these two measures of water quality improvement should take and the two attributes



are to be interpreted and applied differently in policy analysis. In brief, there is no reason why these two different goods should not provide the same measure of utility. Table 4.12 summarises the WTP estimates for both versions and models of the analysis. It shows that the WTP to eliminate dog mess/litter and WTP to have an advisory note system commanded the highest values ranging from £5 to £17. As these are household yearly values for all British beaches, even these larger figures seem to be well within the bounds of reality. Table 4.12: Estimated Range of Mean Values for Beach Attributes Across Versions and Models (£ per Household, per Year)

Attribute

Definition


Range of Mean WTP (£)

Nested Logit Model Range of

Mean WTP (£)

Version A Version B Version A Version B Water A Value of decreasing

the chance of getting a stomach upset by 1 in 100 (over the range:

0.5/100-7/100

0.7-1.2

_

1.1-2.0

_

Water B Value of reducing �unsafe to swim� days

by 1 day (over the range: 1- 16 days)

_

0.8-1.0

_

0.9-1.1

Advisory Value of having an

advisory note system in place, advising on poor water quality

days

6.7-16.6

4.8-11.6

7.2-13.7

5.6-12.1

Litter Value of avoiding the presence of some litter

/ dog mess

6.1-11.8

5.9-9.8

6.8-11.1

6.0-10.0

Amenity Value of improving the standard of

amenities (toilets/showers) and

safety (life-guards and equipment) from average to good

3.0-8.7

3.4-6.1

2.5-7.3

3.4-6.2



4.4 Follow-up Questions This section summarises those questions which followed the valuation section of the questionnaire and sought to understand the thought processes of respondents whilst answering the choice cards. These are questions C-1 to C-3. It appears from the survey results that 35% of respondents (285 people) voted for the current situation in each choice card. By asking these respondents why they were unwilling to pay for improvements it was possible to assess that of these 63% were protest bids (those shaded in Table 4.13) � or 21% of the total sample. Protest bids are invalid because they do not reflect people�s welfare change from the good being considered. These bids have therefore been removed from the final analysis reported in Section 4.2. Further analysis shows that this removal does not cause a bias in the WTP estimates (see Section 5.2 for further discussion). Table 4.13: Reasons for not wanting to pay for beach improvements

Statement % of sample

I can�t afford to pay 14.07

The improvements are not important to me 9.63

I seldom use the beach 11.11

I never use the beach 2.59

I object to paying higher water rates 34.07

The government or council should pay for this 20.74

I don�t believe the improvements would actually happen 7.78 Note: the percentages are calculated from the subset of 285 respondents who stated zero WTP.

Question C-3 asked those who chose Scenario A or B (beach improvement scenarios) in at least one of the choice sets why they were willing to pay for improvements. The total number of respondents who were willing to pay (at least once) for beach quality improvements was 66.6%. About 34% were acting in part upon altruistic motives and some 18% were concerned about the welfare of future generations, whilst about 33% were only considering their own and their family�s use of beaches (see Table 4.14). Over 4% of the respondents who were willing to pay for improvements, said they did so because 'it is a good cause'. Although this could be classified as 'warm glow' and hence an invalid reason, the wording used here is not sufficient for this conclusion to be drawn. It could be that those who chose this scenario thought paying for these improvements was a particularly good cause rather than just any other good cause. Therefore, and also because it is such a small proportion of respondents, these have been included as valid answers in the analysis in Section 4.2.



Table 4.14: Reasons for paying for beach improvements

Total economic value

Statement % of sample

use I would like better beaches and facilities for myself and my family to enjoy

19.08

use I would like to protect my own and my family�s health 14.31

use and altruistic

I would like to improve beach and bathing water quality for everyone to enjoy

33.90

bequest I would like to improve beach and bathing water quality for future generations to enjoy

17.72

existence I would like to improve beach and bathing water quality to protect marine wildlife

6.98

�warm glow� It is a good cause 4.77

None of the above 3.24

Figure 4.8 reports the results of question C-2 that asked respondents which attributes they have considered most when making their choices in the choice sets. Of the respondents willing to pay for improvements about a third considered all the attributes when making their choices. For those who focussed on one or more attributes, dog-mess and litter was the most important (almost 48% of these respondents), whilst the advisory note was considered the least (about 16%). The answer to this question does not determine the importance of the attributes mentioned. In fact, a respondent could consider an attribute that they find the least important in order to be able to reject the scenarios that contain more of that attribute. Therefore, the results reported in Figure 4.8 should not be compared to the beach characteristics of importance or indeed the rankings of the WTP estimates. Figure 4.8: Attributes considered the most in the choice experiment

Note: the percentages do not add to 100% because respondents were allowed to select more than one answer

When asked whether or not they would like to see improvements to bathing water quality at extra cost to them, 51% of respondents replied that they would prefer this

32.63%

18.92%

13.54%

47.68%

16.14%

15.20%

all attributes

water quality

advisory note

dog-mess & litter

safety & amenties

cost



over a system of advisory notices and no quality improvements at no extra cost (see Table 4.15). It is important not to compare this result with the percentage of respondents willing to pay for beach improvements (66.6%), as water quality was only one of the improvements being offered in the choice cards. It should also not be compared to WTP estimate for advisory note system as presented in the choice sets. In addition, there is no clear policy message from this result, which essentially sees the sample undecided between the two choices. Therefore, the WTP results offer a better indicator for policy design and appraisal. Table 4.15: Preferred Scenarios for Reducing Risk of Illness at Beaches In order to reduce the risk of illness from bathing at British beaches which of these options would you prefer? Improvements to seawater quality (through better wastewater treatment and reduced pollution from agricultural sources). At extra cost to your household and all households in Britain through increased water rates.

51%

Keep seawater quality at the current level, but be advised by an advisory notice posted on the beach when it is unsafe to swim. At no extra cost to you.

48%

Don�t know

1%



V. AGGREGATION, VALIDITY AND TRANSFERABILITY Section 5.1 below presents the range of WTP results aggregated across the population of England and Wales, and the present value of these benefits assuming two different discount rates, calculated over a 25 year period. Section 5.2 presents the outcome of a number of tests on the validity of the study results, including content and construct validity, as well proof of the robustness of the results in relation to certain well recognised biases. Finally Section 5.3 discusses the transferability of the results to different contexts and countries. 5.1 Aggregation Section 4 presented the results of the survey in terms of WTP per household per year for marginal changes in each of the attribute. What are possibly of more use for policy purposes are the aggregate benefits for the population of England and Wales. There are two steps to the aggregation process. The first is aggregation across the population. Given that the sample is representative of the population of England and Wales (see Table 3.6), this aggregation is straightforward, namely, per household results multiplied with the number of households in England and Wales. In addition, the marginal WTP results, i.e., 1% reduction in risk of stomach upset and 1 day reduction in number of days unsafe to swim, can be aggregated across different potential reduction levels. Given that a linear model is used in explaining utility or benefit received from beach improvements, a linear aggregation process is used. Thus, for example, the marginal WTP result of £1.10 per household per year per 1% reduction in risk of stomach upset becomes £2.20 per household per year per 2% reduction in risk. The second column of Table 5.1 presents £ WTP per household per year for five levels of risk reduction, five levels of reduction in unsafe bathing days, and the three remaining attributes. The third column shows the aggregation across England and Wales using the total number of households. The last two columns present this result in terms of present value (PV) over 25 years using discount rates of 3.5% and 6%, respectively. WTP for the first expression of water quality (version A) measures the benefit of reducing the risk of suffering a stomach upset due to contact with polluted bathing water at British beaches. It does not include any other potential health impacts of such contact. WTP for a reduction in the number of days on which bathing waters are unsafe to swim is also linked to the risk of a suffering stomach upset and not to any other potential health impacts since only stomach upsets were discussed in the survey. In both cases, the most conservative estimates from the preferred statistical model were used for aggregation, as is the usual practice in stated preference studies. This conservative approach is adopted in order to deflect criticism regarding possible respondent overbidding due to the hypothetical nature of the survey.



We can assume that the risk of stomach upset from bathing at beaches in the 2001 bathing season was 4.3% per swim (calculated using the WHO methodology). If the European Commission�s suggested mandatory standard of 200 intestinal enterococci at 95 percentile compliance were to be adopted then we would see a reduction to 2.0% risk per swim, i.e. a 2.3% reduction in risk per swim. Therefore, using a 6% discount rate, we find the present value of this improvement to be about £776 million over 25 years. This compares to £1.7 billion for the introduction of an advisory note system, £1.8 billion for the avoidance of dog mess and litter and £767 million for the improvement of amenities from average to good. Table 5.1: Aggregated WTP results (£) Attribute £ /hh/yr £ million/yr* PV**

£, million (r=3.5%)

PV** £, million (r=6%)

Change in risk of suffering a stomach upset (% reduction in risk) 1% 1.10 26.4 435 337 2% 2.20 52.8 870 675 2.3% 2.53 60.72 1,001 776 2.5% 2.75 66 1,088 844 3% 3.30 79.2 1,305 1,012 4% 4.40 105.6 1,740 1,350 Change in number of safe bathing days (increase in safe days) 1 0.90 21.6 356 276 5 4.50 108 1,780 1,381 9 8.10 194.4 3,204 2,485 11 9.90 237.6 3,916 3,037 15 13.5 324 5,340 4,142 Advisory note system 5.60 134.4 2,215 1,718 Avoiding the presence of some litter/dog mess

6.00 144 2,373 1,841

Improving the standard of amenities from average to good

2.50 60 989 767

* 24 million households in England and Wales. ** Over a 25 year period. WTP for an advisory note system measures the benefit of having this system in place. The benefit could stem from an increase in publicly available information as well as a reduction in exposure to risk. WTP for the litter / dog mess attribute shows the benefit of reducing the presence of litter / dog mess on British beaches from 'some' to 'none'. WTP for the amenities attribute shows the benefit of improving the quality of toilets, showers, safety equipment and amenities from 'average' to 'good'. Other possible benefits If the aggregated WTP results are to be used in a cost-benefit analysis of the revisions to the Bathing Water Directive, two other benefits may potentially be considered. One of these is actual expenditure of beach visitors (e.g. food, accommodation, shopping and so on). Such expenditures made currently are likely to be transfer payments, i.e.



the activities would have taken place elsewhere in England and Wales. Thus, there would be no net increase in spending across the country. Although they can legitimately be added to an economic impact analysis, they should not be included in a cost-benefit analysis. However, if we think that improvements to bathing water quality could attract new visitors to the affected areas (foreign tourist or residents choosing to stay in England and Wales rather than go abroad), these expenditures can be included in cost-benefit analysis. The second benefit type could be the avoided cost of illness including medical expenses and unpaid sick leave. In fact, WTP to avoid the risk of suffering a stomach upset and the cost of illness of stomach upset episodes are two different ways of measuring the same benefit. WTP is an ex-ante measure based on individual's own preferences, while cost of illness is an ex-post measure using actual expenditures. When evaluating health hazards that strike with some degree of randomness, so that no-one could predict exactly who will actually suffer from the hazard (or benefit from its prevention), as in the case of stomach upsets in this study, then WTP is the most appropriate approach. For further detail on this issue, see Annex 5. This survey only mentioned stomach upset as the health impact of polluted bathing water. It is possible that respondents were already informed of other health impacts and considered these as well as stomach upset when making their choices. We believe this to be unlikely, and hence if the economic cost of other health impacts are known, these could be added to the results here. Finally, potential ecological improvements due to improved bathing water quality are not captured (at least explicitly) by any of the WTP estimates since these have not been included as a separate attribute in the choice experiment. Table 4.14 shows that just under 7% of the respondents willing to pay for bathing water quality improvements have done so 'to protect marine wildlife'. It is not possible to measure the quantitative impact of this motivation on the resulting WTP estimates. 5.2 Validity Testing This section assesses whether or not the questionnaire has provided an unbiased and transparent vehicle to give respondents the best chance to elaborate about their preferences and to get as close as possible to the values they would affirm in the light of experience. Validity refers to the degree to which a study succeeds in measuring the intended values by overcoming potential biases and the hypothetical nature of the valuation exercise. There are two kinds of validity test: content validity and construct validity, discussed in Sections 5.2.1 and 5.2.2. Potential biases are discussed in Section 5.2.3. 5.2.1 Content Validity Content validity refers to whether the study asked the right questions in a clear, understandable, sensible and appropriate manner with which to obtain a valid estimate of willingness to pay. Many of the judgements about content validity are essentially subjective. However, the responses to a number of questions in the survey are useful



for testing content validity (and indeed some serve this purpose only). Following are a number of positive indicators of content validity for this study. Attitudes to the survey As reported at the beginning of Section 4, only 9% of respondents considered the questionnaire to be unrealistic and only 6% found it difficult to understand. This is an indicator that, for the most part, the survey design was successful in creating a credible hypothetical market. Non-response A sign of low content validity is a high percentage of respondents refusing to answer one or a group of questions. This is known as item non-response in the literature. In this study the only high item non-response has been for the income question, which 35% of respondents refused to answer. Although high, this percentage is typical of other surveys the study team has undertaken in the UK. Comparison of this percentage between this and other surveys is made difficult by the fact that majority of survey reports or articles do not report this figure. The simplest way to deal with those who refused to answer the income question is to exclude them from the WTP analysis. This approach was adopted in estimating the conditional logit model (see Section 4). The problem with this approach is that if income non-respondents form a particular group, excluding their responses from the analysis would result in biased WTP estimates. Statistical analysis in Annex 6 (Section A6.3) shows that there are no significant differences in the non-response rate to the income question between sub-samples (it varies between 34% and 36%). Thus, we can assume that the effect of income non-respondents on the WTP is uniform, and hence their inclusion (or exclusion) would not change the WTP estimates. In other words, even when these income non-respondents are included in the analysis, resulting marginal WTP estimates will not be significantly different than when they are excluded. Nevertheless, the nested logit model (the version with ASC and interactions) included these respondents by using the information about the socio-economic group they fall into instead of the missing income information. The result that both conditional logit (which excludes income non-respondents) and nested logit models (which includes these respondents) produce similar WTP estimates is another evidence that income non-response was a random occurrence. Protest bids As reported in Section 4, 35% of the sample were not willing to pay for improvements to beach attributes. Of these, 21% (of the total sample) were protest bids meaning that these respondents� refusal to pay was not an indication of a true zero WTP. This percentage is in line with what is expected from surveys of this type. Protest bids are particularly difficult to deal with in econometric analysis. The usual approach is to exclude them from the analysis of WTP responses. The danger with this approach is that if protests bids are not random, i.e. if a particular group of



respondents have consistently given protest responses, their exclusion from the analysis would result in biased WTP estimates. In this study, when protest bids were included in the analysis, the models did not function well. This indicates that protest bids in this survey were a random occurrence. A further test of whether protest bids are random would be to try to explain the variation in protest bids across the protesting respondents. If a separate model can explain this variation, then protest bids cannot be said to be random. Annex 6 (Section A6.4) shows the results of a logit model that tests for this. The model attempts to explain the probability of a respondent stating a protest bid by gender; socio-economic group of the respondents; respondent finding the valuation scenario not credible; respondents' frequency to swim or do water sports; and a scale of agreement to the (fatalistic) health risk statement �no matter what I do, if I am going to get sick I will�. The results of the model show that there was a tendency for protesters to be women, wealthier (only significant at 10% level), and fatalists in relation to health risks. They also tended to think the valuation scenario was not credible and to swim or do water sports often. However, although this model overall is significant, only 7.5% of the variance can be explained (Pseudo R2=0.075). In addition, not many variables were found to be significant. Based on this analysis, it can be concluded that the model was not able to explain the variance in the protest bids and that these are therefore a random occurrence, i.e. we cannot say that any particular group was consistently providing protest answers. Thus, their exclusion from the WTP analysis is unlikely to lead to biased WTP estimates. The choice of attributes Question A-16, which elicited preferences for a variety of beach characteristics, found that those attributes which scored highest: water quality, cleanliness of beach, amenities, and information telling people that it is not safe to swim (see Figure 4.3), were all included in the beach scenarios. This lends support to the design of the beach scenarios, in that all of the attributes chosen were also considered important by respondents. The only attribute which was found important by respondents (again according to the responses to question A-16) but did not figure in the beach scenario was "water that is clear from litter and foam", but this can be assumed to be included within both water quality and beach cleanliness attributes. The choice of attribute levels The selection of the levels for the two expressions of the water quality attributes (risk and days) is discussed in Section 3. The other three attributes, advisory note system, litter / dog mess and amenities can only be expressed as two levels (absence / presence; some / none and average / good, respectively). Section 3 also shows how the price levels were selected. Whether the levels of the price attribute are in fact representative of the correct price bracket has implications for the validity of the WTP results. If the levels were too high, we would have seen the majority of respondents choosing the scenarios with the lowest price level. If the levels were too low, we would have seen the opposite trend,



i.e. the majority of respondents would have chosen the scenarios with the highest price level. The way to test whether the respondents behaved in either one of the two ways described above would be, in each sub-sample, to examine the proportion of respondents who chose the beach scenarios associated with the highest price. The highest price was £25 and it appeared in four out of each of the eight choice sets that each respondent faced. Overall the highest price appeared in 20 different beach improvement scenarios across the five versions of the questionnaire. If the proportion of respondents choosing those scenarios is high it suggests that the bid range used might not have been large enough to capture an important part of respondents� valuations (in which case we would have what is known as a �fat-tail� problem). The results reported in Annex 6 (Section A6.6) show that in general the proportion of respondents choosing the highest priced beach scenario is low. In addition, • on average, 12% of the respondents accepted the highest bid; • but the proportion of respondents accepting to pay the highest price for beach

improvements is distributed unevenly across the various beach scenarios, depending on the precise attributes of the proposed improvement; and

• the proportion of respondents paying £25 for low and moderate scenarios of improvement is as low as 4% while as much as 21% of respondents accept to pay £25 for a large overall improvement across all attributes.

This suggests that for some 21% of the sample we may not have captured the total value. In the case of the best scenario of improvement using a higher price level could have captured more of their valuation. However, had we used a higher price than £25 we would most likely have lost information on respondents� preferences for more moderate scenarios of improvement.

Overall, the results suggest that the chosen price vector was adequate as there does not seem to be a large concentration of bids at either end of the distribution, i.e. respondents choosing not to pay or accepting the highest price. An extension of this research could examine the impact of using a larger number of price levels together with a larger price range with more price levels1. This would require larger samples. Supporting results A number of cross-checks are possible within the survey to see whether respondents� responses support each other, which is a reflection of the credibility of the results. For example, when presented with a description of the proposed advisory note system (question A-13), the majority (51%) of respondents thought that it would be very useful, while only 4% thought it would be not useful at all. This supports the valuation results, which show that implicit WTP for the advisory note is high compared to other attributes, possibly in order to improve the level of publicly available information.

1 Note that at focus group discussions, £35 was found to be too high by the participants.



Good practice A final test of content validity is that the questionnaire followed best practice guidelines of stated preference studies (eftec, 2000). 5.2.2 Construct Validity Construct validity is concerned with whether or not the results are consistent with expectations. There are two types of construct validity: • convergent validity, which asks whether the estimates are consistent with other

studies or other external evidence from the marketplace; and • expectations based validity which asks whether estimates are consistent with

economic theory, intuition and prior expectation. Convergent Validity In convergent validity testing, no measure can automatically claim to be superior in terms of being a naturally closer approximation of the value held by respondents. Likewise, the result that two approaches deliver similar or logically related measures does not mean that those measures are valid; instead they may be equally invalid. Nevertheless, it is clear that a large and unexpected difference between estimates would show that at least one measure is invalid or two different questions are being addressed. The estimates of this study contrast to some extent with the much larger estimates from some previous studies in the area of water quality improvements (Mourato and Machado, 1998; and Georgiou et al, 2000). The closest estimation of the water quality attribute comes from Hanley et al (2001), who estimate the WTP to be £7.81 per person per year for water quality improvements using the travel cost method. Given that this study measures WTP for only a 1% reduction in risk, compared to a step change in risk in the Hanley et al study, these estimates do not seem to be significantly different. However, two types of studies (beach user - on site and general population) are not directly comparable. On-site studies usually measure a discrete change in quality (e.g. compliance with standard) rather than the marginal changes investigated in this study. In addition, on-site studies usually are about a particular beach or all beaches in a particular area, while the estimates here are for a typical British beach encompassing all beaches in England and Wales. Of note is the fact that water quality, the one factor that attracts the highest political attention and where most of the previous socio-economic research work has tended to concentrate, is only one amongst the many factors that contribute to the well-being gained from visiting British beaches. This finding is in line with similar results (using a different stated preference approach) for marine waters in Portugal (Machado and Mourato, 2002, in press).



Expectations Based Validity Tables 4.8a,b and 4.9a,b show that all the attributes have the expected signs (in that reductions in risk of illness increase welfare and increases in the costs of improvement decrease welfare). Other explanatory variables also have the expected effects on the model, e.g. those who thought water quality was very important were also willing to pay more. The fact that versions A and B produce statistically similar estimates for all common variables is another affirmation of convergent validity, as is the fact that the model was able to find significant estimations for all of the attributes. Finally, most analysis concentrates on the signs and statistical significance of the WTP estimates rather than answering the straightforward question of whether the magnitude of the WTP estimates seems plausible. The estimates here, which all fall in the £0.9-6 per household per year bracket, are plausible figures given that WTP is for marginal changes. Therefore, it can be concluded that in addition to the more sophisticated statistical tests, the results also pass the �reality check�. 5.2.3 Potential biases This section lists the potential biases that were tested for during the analysis along with the results of these tests. Scope, embedding and part-whole The terms, scope, embedding (or nesting) and part-whole are often used interchangeably however there are important distinctions to be drawn between them (Carson and Mitchell, 1995): • Scope concerns a change in just one argument of a multivariate utility function.

For example, a change in the scope of a good occurs when the quantity of that good increases or decreases, say, when we move from considering improvements on one beach to improvements on all beaches in a region.

• Embedding concerns changes to two or more arguments within a multivariate utility function. For example, when we move from considering improvements to bathing water quality alone to improvements to bathing water quality and the provision of amenities on beaches and indeed other attributes.

• Part-whole effects: here when a set of goods (the 'parts') are valued individually the sum may exceed that for the same set of goods valued together (the 'whole'). If the part goods have differing attributes then this can be thought of as an embedding phenomenon. However, if all parts were the same good (possibly of differing sizes) then this could be classed as a scope effect.

Concerns regarding scope, embedding and part-whole effects arise from "the frequent finding that WTP for a good is approximately the same for a more inclusive good" (Fisher, 1996; p19). These effects can only be avoided or at least reduced by good questionnaire design, since there is very little that econometric analysis can do other than detect their



existence. The following discusses the three ways in which the above effects could have presented themselves in this study and how they were avoided by careful design of the questionnaire: • Geographically (related to scope): where a respondent values a good whose

spatial attributes are larger or smaller than the spatial attributes of the presented good. In order to avoid this effect, we designed the choice experiment for a typical beach in England and Wales, i.e. improvements to all beaches, rather than a single beach or beaches in a single region.

• Capture of benefits: where a respondent includes a broader or a narrower range

of benefits in valuing a good than intended. As Section 5.2.1 shows, the beach scenarios were defined in terms of all the attributes that the respondents have stated to be important in Section A of the questionnaire. Therefore, we can say that we have included all the attributes of importance and hence the benefits of improvements to these attributes that could contribute to the WTP responses.

• Policy-package: where a respondent values a broader or a narrower policy

package than the one presented. All possibly significant changes to the selected beach attributes due to revisions to the Bathing Water Directive were included in the definitions of beach scenarios. Furthermore, the respondents were not told that the improvements they were presented were due to the revisions to the Bathing Water Directive. Therefore, no other part of the policy package, i.e. revisions to the Directive, could have entered the considerations of the respondents.

Starting point bias The questionnaire was designed in five versions, specifically to test whether there would be a starting point bias2. In other words, in version A, we tested whether the value of a reduction of a 1 in 100 chance of suffering a stomach upset differs significantly depending upon the risk level in the current situation scenario, i.e., 7%, 5% and 2%. In version B, we tested whether the value of a reduction of one day in the number of days on which it is unsafe to swim differs significantly depending on the number of 'unsafe to swim' days in the current situation scenario, i.e. 16 and 10 days. The results presented in Section 4 are for the entire sample. In order to test whether the starting points (and hence the reduction ranges) for the water quality attribute had an impact on the WTP estimates, each sub-sample (receiving each of the five versions of the questionnaire) is analysed separately. If WTP estimates for sub-samples are significantly different from each other, then starting point bias can be said to exist. The full detailed results of the analysis are reported in Annex 6 (Section A6.5). In summary, there was no strong and systematic evidence of starting point bias as none of the WTP estimates for sub-samples is statistically different at the 1% level although there are three differences at the 5% level: • None of the (WTP for) attributes is statistically different between A1 and A2 at

the 5% level;

2 While starting point bias usually refers to the price attribute, in this instance we apply it a change in risk



• None of the attributes is statistically different between A1 and A3 at the 5% level, except WTP for risk reduction which is not different only at the 1% level; and

• None of the values is statistically different between A2 and A3 at the 5% level, except WTP for risk reduction which is not different only at the 1% level.

• None of the values is statistically different between B1 and B2 at the 5% level, except WTP for advisory note system which is not different only at the 1% level;

5.3 Transferability The transferability of the results of this study to another policy context should follow the standard criteria for benefits transfer. In other words, the benefit estimates of the study should only be applied to similar: • environmental resources; • uses of environmental resources; • change in environmental resources; and • affected populations. Following these criteria, four types of benefits transfer are possible. Transfer across environmental resources The WTP estimates found in this study are for coastal bathing waters alone in England and Wales. It can be assumed that inland bathing waters provide sufficiently similar benefits to make transfer of these estimates a viable option. Although landscape and other amenities inland are bound to be different than those along the coast, these are not affected by the revisions and hence are not relevant to the analysis. Transfer across the uses of bathing waters The WTP estimates found in this study are for bathing. Since the sample was representative of the general population of England and Wales, it also captured some who engage in water sports and hence reflected their preferences within the average WTP estimate. However, this estimate cannot be used as a proxy for the WTP of the sub-group of those who engage in water sports (if this was of policy interest). There are three reasons for this. First, the risk brackets used in this survey did not include the risk of stomach upset due to water sports. Therefore, if it is believed that risk of suffering a stomach upset associated with water sports is higher than that for general bathers, the WTP estimates here cannot be extrapolated beyond the risk bracket of the study. Second, the demographics of those who participate in water sports may be quite different from the demographics of ordinary bathers in the population (as well as their respective WTP). Third, the attitudes to risk could be different between those who engage in water sports and general bathers. For example, water sport enthusiasts may be risk lovers and hence may have lower WTP for improvements to bathing water quality.



Transfer across the impacts of bathing water quality Poor bathing water quality is associated with a number of impacts on human health including stomach upsets, ear infections and so on. The WTP estimates found in this study are explicitly linked to a marginal reduction in the risk of suffering stomach illness. The results are not conducive to transfer to other health impacts of bathing water quality or from stomach upsets caused by bathing water quality to stomach upsets caused by other sources. Transfer across countries The WTP estimates reported here are for the population of England and Wales. Considering the Europe-wide relevance of the revisions of the Bathing Water Directive, whether these estimates can be transferred to other EU countries is an important question. The estimates are likely to be transferable to Scotland and Northern Ireland and possibly Ireland as the beach environment can be considered to be similar (although population characteristics may be significantly different). To a lesser extent the results are transferable to other countries with similar climate and/or beach use and population characteristics. Clearly, using this reasoning they are not transferable to Mediterranean Member States. Even for countries similar to England and Wales it is worth adding the warning that the literature on transferability of values indicates that transferability between countries has always proved less reliable than transfers of values and benefit functions within countries.



VI. CONCLUSIONS A number of conclusions of the study that are of particular importance are described below: • Using both the responses to the questionnaire and epidemiological evidence, the

study estimated that there were around 1.3 million cases of stomach upset in England and Wales every year that could be associated with bathing in faecally contaminated bathing water. A second calculation based on respondents� personal assessment of cases of stomach upset revealed that this figure could be upwards of 2.2 million cases.

• Respondents revealed that the most popular activities undertaken on the beaches

in Britain are walking, relaxing, sun-bathing and picnicking, while 57% of respondents and their families never go swimming.

• The most important beach attributes were cleanliness of the beach, water that

looks clear from litter and foam, and amenities. 51% of the sample found water quality to be very important.

• The modelling results showed that all beach attribute coefficients were significant

at the 1% level and had the expected signs across all versions and model specifications.

• 51% of the sample thought that the proposed advisory note system would be very

useful. 38% of the sample preferred to see a simple symbol indicating "do not bathe", while 51% preferred to see more detailed information.



• Respondents were willing to pay between £1.10 and £2.00 per household per year

for a 1% reduction in the risk of suffering a stomach upset. Interestingly, this value range is almost identical to the 90 pence to £1.10 that respondents seemed to be prepared to pay to avoid 1 day of poor water quality.

• Respondents were willing to pay between about £5.6 and £13.7 per household per

year for the introduction of an advisory note system. This equates to £134.4 million per year for England and Wales.


year for avoiding the presence of some litter / dog mess on the beach. This equates to £144 million per year for England and Wales.

• Respondents were willing to pay between around £2.5 and £7.3 per household per

year for improving the standard of amenities and safety from average to good. This equates to £60 million per year for England and Wales.




determine whether any potential biases were at play. These tests all showed that the results are free from any bias, and are supported by several strong arguments for different types of validity.

• The benefit estimates produced in this study have limited transferability, and

caution should be taken, especially in applying the results to other countries and contexts.



REFERENCES Barton, D. (1998), Applying NOAA Panel Recommendations to Contingent Valuation Studies in Developing Countries � A Case Study of Coastal Water Quality in Costa Rica, Discussion Paper #24/98, Department of Economics and Social Sciences, Agricultural University of Norway. Bennett J. and R. Blamey (2001) The Choice Modelling Approach to Environmental Valuation, Edward Elgar Publishing, Northampton, UK. Brocksteal, N.E., Hanemann, W.M. and Kling, C.L. (1987) �Estimating the value of water quality improvements in a recreational demand framework�, Water Resources Research, Vol 23 No 5, 951-960. Brockstael, N.E., McConnell, K.E. and Strand, I.E. (1989) �Measuring the benefits of improvements in water quality: The Chesapeake Bay�, Marine Resource Economics, Vol 6, 1-18. Carson, R.T. and Mitchell, R.T., (1995) Sequencing and nesting in contingent valuation surveys, Journal of Environmental Economics and Management, (28), 155-173. Choe, K. Whittington, D. and Lauria, D. (1994) �The Economic Benefits of Surface Water Quality Improvements in Developing Countries: A case Study of Davao, Philippines�, Land Economics, 72(4), 519-537. Countryside Recreation Network (1996), UK Day Visits Survey 1994, CRN Publications. Day B., Hanley N. and Bergland O. (2001) Non-parametric and semi-parametric approaches to analysing payment ladder contingent valuation data: bathing water quality improvements in Scotland, Working paper, Economics Department, University of Glasgow. eftec (2000), Guidance on Using Stated Preference Techniques for the Economic Valuation of Non-Market Effects, for DTLR, published by Edward Elgar (forthcoming 2002). Environment Agency (2002), �WHO Standards for Bathing Waters�, Internal Mimeo, DEFRA, UK. Feenberg, D. and Mills, E.S. (1980) Measuring the Benefits of Water Pollution Abatement, Academic Press. Fisher, A.C. (1996) �The conceptual underpinnings of the contingent valuation method�, in Bjornstad, D.J. and Kahn, J.R. (eds) The Contingent Valuation of Environmental Resources: Methodological Issues and Research Needs, Edward Elgar, Cheltenham, 19-37. Fleisher, J.M., Jones, F., Kay, D., Stanwell-Smith, R., Wyer, M., and Morano, R. (1993), �Water and Non-Water Related Risk Factors for Gastro-enteritis among



Bathers Exposed to Sewage Contaminated Marine Waters�, International Journal of Epidemiology, 22, 698-708. Fleisher, J.M., Kay, D., Salmon, R.L., Jones, F., Wyer, M.D, and Godfree, A.F. (1996), Marine Waters Contaminated with Domestic Sewage: Nonenteric Illnesses Associated with Bather Exposure in the United Kingdom", American Journal of Public Health, 86, 1228-1234. Fleisher, J.M., Kay, D., Wyer, M. and Godfree, A.F. (1998), "Estimates of the Severity of Illnesses Associated with Bathing in Marine Waters Contaminated with Domestic Sewage", International Journal of Epidemiology, 27, 722-726. Fisher, A.C. (1996) The conceptual underpinnings of the contingent valuation method, in Bjornstad, D.J. and Kahn, J.R. (eds) The Contingent Valuation of Environmental Resources: Methodological Issues and Research Needs, Edward Elgar, Cheltenham, 19-37. Freeman, A.M. (1995) The Measurement of Environmental and Resource Values: Theory and Methods, Resources for the Future, Washington D.C. Georgiou, S., Bateman, I., Langford, I. and Day, R. (2000) �Coastal bathing water health risks: assessing the adequacy of proposals to amend the 1976 EC Directive�, Risk, Decision and Policy, 5, 49-68. Georgiou, S., and Langford, H.I., Bateman, I.J., and Turner, R.K, (1998) �Determinants of individual's willingness to pay for perceived reductions in environmental health risks: a case study of bathing water quality�, Environment and Planning A, 30, 511-594. Hanley N., Bell, D. and Alvarez-Farizo, B. (2001) �Valuing the benefits of coastal water quality improvements using contingent and real behaviour�, Environmental and Resource Economics. Hanley, N., Mourato, S. and Wright, R. (2001) �Choice Modelling Approaches: A Superior Alternative for Environmental Valuation?� Journal of Economic Surveys 15, 433-460. House of Lords (1994-5), Select Committee on the European Communities, 1st Report, Bathing Water, HL Paper 6-I; and 7th Report, Bathing Water Revisited, HL Paper 41, HMSO, London. Kay, D., Fleisher, J.M., Salmon, R.L., Jones, F., Wyer, M.D., Godfree, A.F., Zelenauch-Jacquotte, Z. and Shore, R. (1994), "Predicting Likelihood of Gastro-enteritis from Sea Bathing: Results from Randomised Exposure", The Lancet, 344, 905-909. Le Goffe, P. (1995) �The benefits of improvements in coastal water quality: a contingent approach�. Journal of Environmental Management, 45, 305-317.



Louviere J.J., D.A. Hensher and J.D. Swait (2000) Stated Choice Methods: Analysis and Application, Cambridge University Press, Cambridge. Machado, F. and Mourato, S. (2002) Evaluating the Multiple Benefits of Marine Water Quality Improvements: How Important Are Health Risk Reductions? Journal of Environmental Management. Machado, F. and Mourato, S. (1999) �Improving the assessment of water related health impacts: evidence from coastal waters in Portugal�, CSERGE Working Paper GEC 99-09, University of East Anglia, Norwich, UK. Mantymaa, E. (1999) �The free selection of reference points and the willingness-to-pay�, In M. Boman, R. Brannlund and B. Kristrom (eds), Topics in Environmental Economics, Kluwer Academic Press. McConnell, K. and Ducci, J. (1989) �Valuing Environmental Quality in Developing Countries: Two Case Studies�, Paper Presented to Applied Social Science Association, Atlanta, Georgia. National Centre for Social Research (1999) Leisure Day Visits: Summary of the 1998 UK Day Visits Survey, London. Niklitschek, M. and Leon, J. (1996) �Combining Intended Demand and Yes/No Responses in the Estimation of Contingent Valuation Models�, Journal of Environmental Economics and Management; 31(3), 387-402. Pruss, A. (1998), "A Review of Epidemiological Studies from Exposure to Recreational Water", International Journal of Epidemiology, 27,1-9. Sandstrom, M. (1997), Recreational benefits from Improved Water quality: A Random Utility Model of Swedish Seaside Recreation, Working Paper 121, Stockholm School of Economics. US Department of Transportation (1982) Guide to Forecasting Travel Demand With Direct Utility Assessment, Washington D.C. Wallston, K.A., Wallston, B.S. and DeVellis, R. (1978), �Development of the Multidimensional Health Locus of Control (MHLC) Scales�, Health Education Monographs, 6, 160-170. WHO (2001), Bathing Water Quality and Human Health, WHO, mimeo - limited distribution. WRc (1994), Health Effects of Sea Bathing (WMI 9021) � Phase 3, Final report to the Department of Environment, Report no: DOE3412/2. WRc (2001), Economic Evaluation of the Bathing wAter Quality Directive 76/160/EEC and of its Revision (contract no. ENV.B.2/ETU/200/0022): Benefits Report (12333/R3), European Commission, Directorate General Environment.



Zylicz, T., Bateman, I.J., Georgiou, S., Markowska, A. Dziegielewska, D. Turner, R.K., Graham, A. and Langford, I.H. (1995) Contingent Valuation of Eutrophication Damage in the Baltic Sea, CSERGE Working Paper GEC 95-03, University of East Anglia, Norwich, UK.



GLOSSARY Alternative Specific Constants (ASCs): These are modelling constants that are created to represent alternative choices in the choice experiment, so that if three choices are presented in a choice card two of the alternatives must be associated with an ASC. ASCs represent new �attributes� in the model and have the purpose of taking up any variation in choices that cannot be explained by either the attributes or the socio-economic variables. Altruistic value: Altruism is the desire to secure an enhancement of the wellbeing of others. Altruistic economic value is the willingness to pay on the part of individual A to ensure that individual B secures some gain in wellbeing. Benefits transfer: An approach which makes use of previous valuations of similar goods at a study site and, with any necessary adjustments, applies them to produce estimates for the same or similar good in a different context, known as the policy site. What is transferred may be a mean WTP, with or without some adjustment for changed conditions (e.g. different income levels), or a benefit function. Bequest value: Bequest values measure people's willingness to pay to ensure their heirs and future generations will be able to use the resource in the future. Bequest values are an example of non-use values. Choice experiment: A form of choice modelling in which respondents are presented with a series of alternatives and asked to choose their most preferred. Choice modelling: encompasses a range of stated preference techniques, including choice experiments, contingent ranking, contingent rating, and paired comparisons. CM approaches describe an asset in terms of its attributes, or characteristics, and the levels that these take, and may be used to determine which attributes are significant determinants of value; their implied ranking; the value of changing them; and the total economic value of a resource or good. Choice set: A set of alternatives presented to respondents, usually in a choice experiment context, where they are asked to choose their most preferred. Coefficient of the attribute: The coefficients in the logit models represent the relative effect of, say, an attribute on the probability of the respondent choosing a particular scenario. For example, a positive sign means that respondents are more likely to choose a scenario that contains more of that attribute. Coercive payment vehicles: These are payment mechanisms involving a degree of compulsion, for example taxes, rates, fees, charges or prices. May be contrasted with non-coercive instruments such as donations. Compensating variation: The compensating variation of a price fall (rise) is the sum of money that, when taken away from (given to) the consumer, leaves him/her just as well of with the price change as if it had not occurred. Thus, initial utility is held constant.



Construct validity: This examines whether the relationships between measures produced by a CV study and other measures are in accordance with expectations. Examples include predictors from economic theory, and empirical regularities in the form of associations with other variables which seem intuitively correct and which hold across a large number of studies Consumer surplus: is the difference (or the net gain) between the price actually paid when purchasing a good or service and the price the consumer would have been willing to pay for the same good or service. Content validity: This assesses whether the CV study asked the right questions in a clear, understandable, sensible and appropriate manner with which to obtain a valid estimate of the construct (say maximum WTP for a specific good) under investigation. Contingent ranking: A form of choice modelling in which respondents are required to rank a set of alternative options. Each alternative is characterised by a number of attributes, which are offered at different levels across options. Respondents are then asked to rank the options according to their preferences. Contingent rating: A form of choice modelling exercise in which respondents are presented with a number of scenarios and are asked to rate them individually on a semantic or numeric scale. Convergent validity: A process of content validity in which the measures obtained from a given CV study are compared with other results to see if they produce similar answers, or answers that vary in a predicted manner. Dichotomous choice: An elicitation format in which respondents are faced with only two response alternatives, such as yes/no, agree/disagree, or vote for/vote against. Sometimes, a �don�t know� option is also included to avoid forcing respondents into artificially choosing one of the answers. Direct use value: Where individuals make actual use of a resource for either commercial purposes or recreation. Discounting: is the process of expressing future values in present value terms which allows for the comparison of cost and benefit flows regardless of when they occur. The present value of a future flow of benefit or cost will be lower than the future value because of discounting. There is no a priori correct way to discount future gains and losses, although exponential discounting is most widely used. Stated preference techniques may be used to derive discount rates. Dummy variable: is a variable that takes the value of 1 when the subject of the variable is present and value of 0 when it is absent. It is used for discrete variables, which are not measured in continuous units but take a value of either/or. For example, gender is entered into the analysis as a dummy variable, e.g. male - 1, female - 0. Another example is from the beach attributes is that of advisory note system, in scenario where the system is present, the dummy variable takes the value of 1 and vice versa.



Economic value: The monetary measure of the wellbeing associated with the change in the provision of some good. It is not to be confused with monetary value unless the latter is explicitly designed to measure the change in wellbeing, nor with financial value which may reflect market value or an accounting convention. As Freeman (1993), notes the terms 'economic value' and 'welfare change' can be used interchangeably. Equivalent variation: The equivalent variation of a price fall (rise) is the sum of money that, when given to (taken from) the consumer leaves him/her just as well off without the price change as if it had occurred. Thus, it preserves the post-change utility level. Existence value: The value that people put on the existence of a resource, even when they have no intention of ever using the resource. Existence values are part of non-use values. Expectation-based validity: A type of content validity in which CV measures are related to other constructs in a manner which is consistent with prior expectations (theoretical and intuition and empirically driven expectations). Focus groups: A focus group is an interview conducted by a moderator among a small group of respondents in an unstructured manner. Factional factorial design: a design, formulated by a programme, available in statistical software packages, which can reduce the number of all possible combinations of attributes and levels to a manageable number for the purpose of building scenarios, while still maintaining the possibility of estimating �main effects� which are the effects of the attributes on the choices (which typically explain 80% of all the variation). Focus group: is a structured discussion group on a specific topic, facilitated by a moderator. Health locus of control: is a development from the health and psychology literature (Wallston, Wallston and DeVellis [1978]). Here people�s perceptions of control over their own health status, irrespective of the truth of those perceptions, are generally believed to be major determinants of their health related behaviour, as well as whether they become ill or not. Hypothetical bias: The possibility that stated preference estimates may be biased upwards due to the hypothetical nature of the payment commitment. Incentive compatibility: The questionnaire design objective of constructing a valuation mechanism in which truth telling and utility maximisation coincide. Indirect utility function: A function that describes household utility (or wellbeing) usually in terms of how much utility it can derive from income, given the prices of goods and, say, the level of provision of a non-market good.



Likert scale: A type of itemised rating scale where respondents are asked to indicate a degree of agreement or disagreement with statements about the object. Logit specification: The logit models used for choice modelling explain the probability of respondents choosing a particular scenario as a function of the attributes of the scenario, respondent-specific characteristics and ASC (see above). See above for the meaning of the coefficients of attributes in logit models. Likelihood calculations: The estimated probability of each respondent making a particular scenario choice (or stating a particular WTP response) is known as their likelihood contribution and is denoted li. Multiplying the likelihood contributions would give us the total likelihood of observing these WTP values given our selected values for the parameters. We can write: ∏=

iilL

where L is known as the likelihood function and measures the total probability (predicted by the model at the given parameters) that respondents will have provided the WTP responses recorded in the CV survey. It frequently proves more convenient to work with the log of the likelihood function, a function that, not all together surprisingly, is termed the log likelihood function:

∑=i

ilL lnln

Of course, if we select different values for the parameters we might find that our model predicts a higher total likelihood (and hence log likelihood). We would assume that these values for the parameters are closer to the true values that we are trying to estimate. Indeed, using a technique known as maximum likelihood estimation, it is possible to establish (with the help of computer software) the set of parameters that best fit the data; that is, to calculate the set of parameters that maximise the total likelihood of observing the reported WTP values. Marginal rates of substitution: is the rate at which a respondent is willing to trade off one attribute with another. WTP of an attribute estimated through choice experiments is in fact the marginal rate of substitution between that attribute and money (the price or cost attribute). This is why, WTP is estimated by dividing the coefficient of the attribute with the coefficient of the price or cost attribute. Non-use value: The value placed on a resource by people who are not current users of that resource and who do not intend to use the resource themselves. See altruistic, bequest and existence values. Open-ended format: A straightforward elicitation format which asks respondents to state their maximum willingness to pay (or minimum willingness to accept). Option value: The value that people place on having the option to use a resource in the future even if they are not current users.



P-value: shows the probability that a coefficient would assume a value greater than or equal to the observed value (reported coefficient) strictly by chance. Passive use values: Where society benefits from using a resource indirectly � e.g. the watershed and carbon sequestration functions of a forest. Protest bid: A response to a valuation question which does not give the respondent�s genuine WTP (or WTA), but either a zero value or an unrealistically high (or low) value. Pseudo R2: This is a measure of the explanatory power of the entire model and calculated as:

0

max2

lnln

1L

LR −=

Where ln Lmax is the value of the log likelihood function from the estimated model with covariates and ln L0 is the value of the log of the likelihood function from the unparameterised model. The pseudo R2 statistic takes values between 0 and 1 where a value of zero suggests that the included covariates do nothing to explain the distribution of WTP in the sample. In general, the larger the value of the pseudo R2 statistic the greater the explanatory power of the model. Reference-dependent utility: The idea is that an individual�s utility is defined in relation to that individual�s reference point, normally the status quo. Sample frame population: A list of the target population from which the sample will ultimately be drawn, e.g. all dwelling units in a city, all visitors to a site, all households with a telephone. Starting point bias: When the final valuation estimate shows dependence on the starting point used. Theoretical validity: The consistency between WTP estimates derived from stated preference studies and underlying theoretical expectations (e.g. fewer people will agree to higher price bids). Total economic value: Total economic value of an environmental resource is made up of, i) use values and ii) non-use values. Use values are composed of a) direct use value, b) indirect use values and c) option values, whilst non-use values are made up of a) altruistic, b) existence values and c) bequest values. Use value: The value placed on a resource by users of that resource. See direct use value, indirect use value and option values. Utility: originally thought of as a number measure of a person�s happiness, utility is used in economics as a way of describing consumer preferences (through utility functions), where a more preferred choice set is said to provide a higher utility. Utility difference model: An indirect utility model constructed to estimate the difference in utility that results from provision of the non-market good. This method



relies on specification of the underlying utility functions and is then used to derive specific forms of bid functions. Validity: refers to the degree to which a study measured the intended quantity. Welfare: (or social welfare) is the sum of individual utilities. The maximisation of social welfare is the goal of welfare economics, a branch of economics concerned less with how the economy works, but more with how well it works. Willingness to accept compensation: The amount of the monetary compensation people are willing to accept to tolerate a loss or forgo a gain. Willingness to pay: The amount of the money people are willing to pay to avoid a loss or for a gain.

DEFRA Bathing Waters Valuation of Benefits Report€¦ · 6/25/2002 · bathing water contact...

Documents

Transcript of DEFRA Bathing Waters Valuation of Benefits Report€¦ · 6/25/2002 · bathing water contact...