IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
Using the choice experiment method for valuing
improvements in water quality: a simultaneous
application to four recreation sites of a river basin
Julie Poirier
CREST-LEI, GATE
Exiopol Summer School
July 12, 2010
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 1/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The Water Framework Directive (WFD)
European Directive 2000=60 of October 2000
Establish a Community framework for water protection
Main objective of the Directive
Build a water management policy at the European Union level
Achieve a `good ecological status' in all European waters by 2015
) Expects the adoption of programs of measures (=management
plans) for all river basins
Time extensions if costs > bene�ts
) Cost-Bene�ts Analysis to assess whether cost of measures
are disproportional
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 2/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
Economic approaches
Costs : computed by water experts
Benefits : methods for `valuing environmental goods and
services', based on the examination of individuals'
preferences
Main objectives of the study
Multi-attribute valuation of local residents' preferences for water
quality improvements at a speci�c river basin in France (which
consists of a system of four recreation sites)
Focus on the spatial dimension in the good that provides both
use and non-use values for the sites
) Choice experiment method with site-speci�c attributes
Estimate the non-market economic bene�ts of moving towards
good ecological status
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 3/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The choice experiment method (CE)Study area and data collectionSurvey design
Outline
1 Methods and Data
The choice experiment method (CE)
Study area and data collection
Survey design
2 Models of choice
The models
Choice experiments results
3 Discussion
Cost-Bene�t Analysis (CBA)
Further comparison : exclusion of protest bids
4 Concluding remarks
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 4/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The choice experiment method (CE)Study area and data collectionSurvey design
A few things you need to know
What is an environmental good ?
Example : air, water, biodiversity. . .
No monetary value but still a value which can take several forms
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 5/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The choice experiment method (CE)Study area and data collectionSurvey design
Literature
Study area = system of spatial units ) use of site-speci�c
attributes
,! An economic valuation of simultaneous management of
several recreational sites located in the same forest area
(Horne, Boxall, Adamowicz, Forest Ecology and Management,
2005)
Random parameters logit model to take into account
heterogeneity of preferences
,! An economic valuation of improvements to the ecology of
two rivers (River Wear and River Clyde) in the UK (Hanley,
Wright and Alvarez-Farizo, Journal of Environmental
Management, 2006)
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 6/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The choice experiment method (CE)Study area and data collectionSurvey design
General framework
Goal
Elicit information about environmental preferences from individuals
through the construction of an hypothetical (but realistic) market
involving an improvement (or a decline) in environmental quality.
Then, infer individual's WTP from those preferences.
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 7/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The choice experiment method (CE)Study area and data collectionSurvey design
Study area
A river basin, composed of a
coastal strip and its three wa-
tersheds, in Lower Normandy
(France)
Four outdoor recreation
sites, frequented primarily
for water-based
recreation
) Users and non-users
Heterogenous in their
ecological statusJ.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 8/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The choice experiment method (CE)Study area and data collectionSurvey design
Data
De�nition of a scope of in�uence for the study area : 15 kms
around each river and 80 kms from the coastline
Population to survey : a representative sample of the
households who live in this scope of in�uence
Quota-sampling approach (age and occupation of the head of
family, number of persons in the household, population of the
cities located in the de�nite scope of in�uence)
) 824 resident households surveyed using in-house interviews
in the summer of 2008
N.B. : Quite large sample compared with other studies.
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 9/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The choice experiment method (CE)Study area and data collectionSurvey design
Implementation of Choice Experiments
1 An environmental good : waters of the river basin
! A global system consisting of four distinct spatial units2 Four spatial, or site-speci�c, attributes
Two levels for each attribute : `status quo' level and `good'
level
3 A monetary attribute : an annual voluntary contribution
Four levels for the `price' attribute : 10e, 20e, 30e, 40e
4 Construction of 8 choice sets, each consisting of three
alternative water management regimes for the four sites : two
improvement alternatives and the status quo scenario
N.B. : Because of respondents' cognitive capacity, the 8 choice sets
were blocked in two versions of the CE, each containing 4 choice
sets.J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 10/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The choice experiment method (CE)Study area and data collectionSurvey design
Example of a choice set
Figure: A simpli�ed example of a choice set
Attribute Option A Option B Status quo optionCoastline Current bad status Good status Current bad status
River Touques Current moderate status Good status Current moderate status
River Dives Good status Current bad status Current bad status
River Vie Current poor status Good status Current poor status
Annual contribution 10e 30e 0e
Respondents are asked to choose their preferred alternative
from each of the choice sets.
Advantages of the CE approach :
Attributes can be quantitative or qualitative in nature
Provide independent values for each attribute
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 11/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The modelsChoice experiments results
Outline
1 Methods and Data
The choice experiment method (CE)
Study area and data collection
Survey design
2 Models of choice
The models
Choice experiments results
3 Discussion
Cost-Bene�t Analysis (CBA)
Further comparison : exclusion of protest bids
4 Concluding remarks
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 12/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The modelsChoice experiments results
General remarks
Based on the new approach to consumer theory (Lancaster,
1966)
Consistent with random utility theory (Marschak, 1960)
CE assumes that :
Individuals choose the alternative which maximizes their utility.
) Can apply probabilistic models to choices between the
di�erent alternatives available in each choice set.
A good is valued in terms of its attributes.
Always an attribute which takes the form of a proxy for price
) WTP estimates for changes in attribute levels can be
derived from marginal utility estimates.
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 13/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The modelsChoice experiments results
Multinomial models
Unordered multinomial outcomes that arise from individual choices
Assumptions
Individual i , alternative j
ARUM : Uij = Vj + "ij
Probability of `i ' choosing `j ' over alternative `k ' :
P(yi = j jC ) = P(Uij > Uik);8k 2 C
= P("ik � "ij < Vj � Vk); 8k 2 C
Multi-attributes choices ) P(yi = j jC ) expressed on terms of
logistic distribution
The "ij 's are IID Gumbel-distributed.
) P(yi = j jC ) =exp(Vj )P
h2C exp(Vh)
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 14/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The modelsChoice experiments results
Conditional logit (CL) model
All regressors are alternative speci�c.
Functional form of the utility function :
Uij = �j + X 0
j �x + P 0
j�p + "ij
Xj is a vector of site-speci�c attributes describing j
Pj is the monetary cost associated with alternative j
�j is the alternative speci�c constant (ASC), which is a
dummy variable that equals one when the status quo option
was chosen. It captures the e�ects of non-observable variables
that play a role in choice decisions.
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 15/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The modelsChoice experiments results
Random parameters (RP) logit model
Limitations of the CL model : assume homogeneous preferences
Functional form of the utility function :
Uij = �j + X 0
j �x + P 0
j�p + X 0
j �i + "ij
�i � N(0;��x )
) the combined error (X 0
j �i + "ij) are now correlated across
alternatives.
Assume that preferences relating to the four site-speci�c
attributes are heterogeneous while preferences towards the
price attribute are homogeneous.
The RP logit model :
allows for variation in preferences across individuals.
adjusts for error correlation across alternatives.J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 16/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The modelsChoice experiments results
Estimation results
Site-speci�c and price attributes have the expected signs.
The ASC has a positive sign ) `Cost to change'.
Parameters estimates increase in absolute value when we
estimate a RP logit model rather than a simple CL model.J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 17/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The modelsChoice experiments results
WTP estimates
Marginal WTP associated with an improvement in
the quality of any attribute `a' :
WTPa = ��a�p
All the implicit prices are positive.
`Naive' measure of welfare in the case of `good
ecological status' at all 4 sites :
52.1e per year and per household (with RP logit estimates)J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 18/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
The modelsChoice experiments results
Individual heterogeneity
We include the income of the respondent interacted with the price
attribute in the model in order to estimate implicit prices speci�c
to individuals'types.
WTP for each attribute vary with
income.
Eg. : Households whose income is
less that 750e are willing to pay
7.40e per year for good ecological
status at coastal waters.
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 19/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
Cost-Bene�t Analysis (CBA)Further comparison : exclusion of protest bids
Outline
1 Methods and Data
The choice experiment method (CE)
Study area and data collection
Survey design
2 Models of choice
The models
Choice experiments results
3 Discussion
Cost-Bene�t Analysis (CBA)
Further comparison : exclusion of protest bids
4 Concluding remarks
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 20/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
Cost-Bene�t Analysis (CBA)Further comparison : exclusion of protest bids
Tools and results
CBA : estimate the equivalent money value of the bene�ts and
costs to the achievement of a good ecological status in the
river basin
Net present value for the project : NPV =PT
t=0Bt�Ct
(1+r)t
NPV on the period of time 2010-2040Costs : 30-year period
Bene�ts : only entered in the accounts from 2015
NPV<0 ) Justi�cation for time delaysJ.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 21/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
Cost-Bene�t Analysis (CBA)Further comparison : exclusion of protest bids
Zero bids and status quo responses
33% of respondents always report the status quo option irrespective
of the choice set.
) Ancillary question in order to know the reason underlying this
choice behavior.
) Status quo responses can be categorized into two types :
True zero bids (=genuine zero bids) ) real zero value
False zero bids (=protest bids) ) true value for the good is
positive
,! Aversion to the principle of paying for environment
conservation or to the payment vehicle or to the change
) Restriction of the analysis to only individuals who stated real
bids (positive bids and true zero bids).
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 22/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
Cost-Bene�t Analysis (CBA)Further comparison : exclusion of protest bids
Estimates when excluding protest bids
All parameters have expected signs.All parameter estimates increase in absolute value, except ASC) This suggests that :
protest bids do a�ect the results.
respondents tried to signal their aversion to the change
through their protest bids.
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 23/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
Outline
1 Methods and Data
The choice experiment method (CE)
Study area and data collection
Survey design
2 Models of choice
The models
Choice experiments results
3 Discussion
Cost-Bene�t Analysis (CBA)
Further comparison : exclusion of protest bids
4 Concluding remarks
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 24/25
IntroductionMethods and DataModels of choice
DiscussionConcluding remarks
Extensions
CE enables to take into account both environmental and
socio-economic characteristics across sites.
Allow to compare costs and bene�ts, as imposed in the WFD.
CE studies are expensive and time-consuming
) Could be worth to set up a system of bene�ts transfer in all
Europe.
Future work : �nd a way to `treat' protest bids when running
the analysis.
J.Poirier(CREST) Using the choice experiment method for valuing improvements in water quality 25/25