Analysis of the economic impacts of water
eutrophication: Lake Macha case
Dissertation
Ondřej Vojáček
Supervisor: Doc. Ing. Petr Šauer CSc.
Prague, March 2010
Declaration
I hereby declare that I have made my dissertation titled “Analysis of the economic impacts of
water eutrophication: Lake Macha case” on my own and that I have listed all the sources and
reference literature used.
Prague, on
______________________________
signature
This dissertation could not have been made without having spent several years in the inspiring
setting of the Department of Environmental Economics at the University of Economics
Prague. I would therefore like to express my thanks to all the Department staff. My thanks go
specifically to Doc. Ing. Petr Šauer CSc. for his exceptionally helpful approach throughout
my presence at the Department and involvement in projects that have given me the
opportunity to grow professionally, and Ing. Jan Melichar PhD. for his know-how and time,
which he was always generously willing to give me and without which this paper could not
have been made. In addition, I would like to thank my whole family and my girlfriend for all
that they have done for me in my life.
Table of Contents
Introduction .............................................................................................................................. 6
1. Water eutrophication – a topic for environmental economics research and policy....... 9
1.1. Eutrophication as an environmental problem.................................................................. 9
1.2. Water eutrophication world-wide.................................................................................. 11
1.3. Economic consequences of water eutrophication ......................................................... 12
1.4. Water eutrophication in the Czech Republic................................................................. 15
1.5. Sources of phosphorus in the Czech Republic .......................................................... 17
1.5. Lake Macha – case study .............................................................................................. 20
1.5.1. Sources of phosphorus in Lake Macha .................................................................. 21
1.5.2. Overview of water quality improvement measures................................................ 22
1.5.3. Lake Macha – Water condition .............................................................................. 22
1.6. The research questions .................................................................................................. 22
1.7. The research methodology ............................................................................................ 23
1.8. The choice experiment .................................................................................................. 24
1.8.1. History of choice experiment ................................................................................. 24
1.8.2. The economic rationale behind choice experiment................................................ 25
1.8.3. Pros and cons of choice experiment ....................................................................... 26
1.9. The economic grounding of welfare measurement ....................................................... 27
1.9.1. The impacts of price changes on welfare ............................................................... 28
1.10. Further analysis ........................................................................................................... 31
1.11. Summary ..................................................................................................................... 32
2. The choice experiment design and sample characteristics ............................................. 34
2.1.Qualitative preliminary survey at Lake Macha .............................................................. 34
2.1.1. International literature review ................................................................................ 36
2.2. Quantitative preliminary survey.................................................................................... 38
2.2.1. Assigning attribute levels ....................................................................................... 39
2.2.1.1. Beach overcrowding........................................................................................ 40
2.2.1.2. Water quality ................................................................................................... 40
2.2.1.3. Beach facilities ................................................................................................ 42
2.2.1.4. Entrance fee..................................................................................................... 42
2.2.2 The status quo choice; scenario setting (design) ..................................................... 42
2.3. The choice experiment design....................................................................................... 43
2.3.1. Making the experimental design ............................................................................ 44
2.3.2. Making the choice sets ........................................................................................... 47
2.3.3. The final questionnaire structure............................................................................ 48
2.3.4. The study population and the study area ................................................................ 49
2.3.5. Sampling Strategy .................................................................................................. 50
2.8. The data description ...................................................................................................... 52
2.8.1. Sample characteristics ............................................................................................ 57
2.8.2. Sample characteristics: attitudes and knowledgeability......................................... 63
2.9. Summary ....................................................................................................................... 69
3. Data analysis: theoretical discussion ................................................................................ 71
3.1. Economic grounding for the discrete choice models .................................................... 71
3.2. Theoretical discussion of discrete choice models ......................................................... 72
3.2.1. Multinomial logit model......................................................................................... 73
3.2.2. Nested logit model ................................................................................................. 75
3.2.3. Probit model ........................................................................................................... 78
3.2.4. Random parameters logit model ............................................................................ 79
3.3. Discrete choice models: Empirical application ............................................................. 81
3.3.1. Description of variables.......................................................................................... 84
3.3.2. Multinomial logit model......................................................................................... 84
3.3.3. Nested logit model ................................................................................................. 86
3.3.4. Probit model ........................................................................................................... 88
3.3.5. Random parameters logit model ............................................................................ 89
3.3.6. Model comparison.................................................................................................. 91
3.3.7. Willingness to pay analysis .................................................................................... 92
3.4. Summary ....................................................................................................................... 94
4. Cost-benefit Analysis of the Measures for Water Quality Improvement at Lake
Macha ...................................................................................................................................... 95
4.1. CBA History.................................................................................................................. 96
4.2. Description of Phosphorus Balance in the Lake ........................................................... 97
4.2.1. Supply along with phosphorus-containing sediments ............................................ 97
4.2.2. Point sources of pollution....................................................................................... 98
4.2.3. Lake bottom sediments........................................................................................... 98
4.3. Corrective Measures for Water Quality Improvement - Estimation of Costs.............. 99
4.3.1. Cofferdam............................................................................................................... 99
4.3.2. Sewerage .............................................................................................................. 100
4.3.3. Lake bottom.......................................................................................................... 101
4.3.3.1. Chemical Precipitation by Means of Aluminous Salts (PAX)...................... 101
4.3.3.2. Lake Fisheries ............................................................................................... 103
4.3.3.3. Sediment Extraction ...................................................................................... 104
4.4. Results of the Measures Taken.................................................................................... 105
4.5. Valuation of Benefits .................................................................................................. 106
4.6. Cost-benefit analysis ................................................................................................... 107
4.6.1. Sensitivity analysis ............................................................................................... 110
4.6.2. Monte Carlo simulation........................................................................................ 112
4.7. Summary ..................................................................................................................... 118
Conclusions ........................................................................................................................... 120
References ............................................................................................................................. 125
Annexes ................................................................................................................................. 133
Introduction
Water eutrophication is a term describing excessive saturation of water with phosphorus and
nitrogen. Water mostly becomes eutrophicated in standing water bodies (reservoirs). The
saturation of water with these chemicals entails many negative phenomena (chiefly excessive
presence of cyanobacteria in the water), which significantly limit the possibilities for
exploiting the water, be it as drinking water, industrial water, or for recreation. These
restrictions imply large costs to the society, whether direct, such as costs of drinking water
treatment, or indirect, such as reduced human utility due to decreased exploitability of the
water for bathing and other water sports.
Since the proportion of eutrophicated water bodies has been growing globally, the issue is
becoming a topic of global importance for ecological sciences and epidemiology as well as
for environmental economics and economic policy. According to some sources, up to 80% of
the Czech Republic’s water reservoirs are eutrophicated. Most of the reservoirs are quite
unique cases, which is why the economic impacts of water eutrophication on summer
recreational utility are analysed on a case study, allowing the researcher to work in a required
depth. Lake Macha was chosen as the research site, as it is an area greatly exploited for
summer recreation as well as seriously affected with long-term excessive eutrophication. At
the same time, the site is not too distant from Prague, meaning that the costs of the research
conducted on the site could be proportionate.
The author of the present thesis believes that water eutrophication does not receive sufficient
attention of social science in the Czech Republic. The broader objective of the thesis is
therefore to contribute to resolving the water eutrophication issues, chiefly to discussing its
economic aspects. Specifically, it deals with an economic view of the recreational utilisation
of the water – the effect of the water eutrophication on the (economic) preferences of holiday
makers who may be affected by the water quality during their holidays.
Several narrower goals were defined based on the above broader objective. They are as
follows:
(a) Does the rate of water eutrophication affect people’s recreational utility?
(b) If so, to what extent is people’s recreational utility affected by water eutrophication?
(c) To what extent does water eutrophication affect people’s (recreational) utility compared to
other characteristics of the recreation site (such as beach overcrowding, entrance fees, etc.)?
(d) Do people perceive water quality chiefly in terms of its (un)wholesomeness, or do they
discern even such parameters as water limpidity knowing that the water is wholesome?
Based on the above research questions, three hypotheses for this thesis were defined:
1. Water quality is the factor most significantly affecting people’s recreational utility from
the chosen site. The first hypothesis thus focuses on the core of the problem of the
water eutrophication – recreational utility relationship.
2. The confirmation or rejection of the above hypothesis is independent of the choice of the
discrete choice econometric model. The second hypothesis concentrates on the applicability
of different model types for the data analysis, thus attempting to contribute to the current
global debate concerning these methods and models.
3. Measures taken at Lake Macha in order to improve the water quality are economically
efficient. This efficiency is the focus of the third hypothesis.
The answering of the defined research questions and the verification of the first formulated
hypothesis are best allowed by the so-called conjoint analysis techniques, specifically the socalled choice experiment method. It is currently one of the most advanced methods in
environmental economics. The thesis applies the method in modelling demand for summer
recreation at Lake Macha. In order to apply it, data collection from the studied target
population – here, visitors to the studied recreational site – was necessary. That was why data
collection was conducted directly on Lake Macha beaches in the summer of 2007, following
several stages of preliminary research.
Analysis of collected choice experiment data employs so-called discrete choice models; they
were employed in solving our research problem as well. The application of these models
poses a question of robustness of estimates in the models, which consequently affect the
obtained estimates of change in recreational utility. The second hypothesis focuses on the
robustness of the results of different types of models.
Having determined the impact of impaired water quality on people’s recreational utility (i.e.,
answered the first and second hypotheses), and knowing the measures to reduce the
eutrophication taken in the study area and the costs thereof, it was possible to proceed to
verifying the hypothesis concerning the efficiency of water quality improvement measures.
The cost-benefit analysis method was employed here.
The structure of this thesis is as follows:
The first chapter introduces the reader to water eutrophication as a crucial topic in
environmental economics with growing economic consequences to human society worldwide.
The chapter identifies the chief reasons for water eutrophication with a special focus on the
Czech Republic. In addition, the chapter brings a description and justification of the research
questions that the thesis tries to answer afterwards. The chosen methodology for handling the
research questions is then identified and discussed, and brought into context with welfare
economics theory. The final part of the chapter characterises the Lake Macha site, including
the scope of the environmental problem, its causes and main steps taken to resolve it.
The second chapter deals with the qualitative research that preceded the preparation of the
choice experiment as such; moreover, it pays attention to the preparation of the choice
experiment, the preliminary survey at Lake Macha, the data collection, the data collection
strategy, and the basic characterisation of the choice sample.
The third chapter brings the choice experiment data analysis. It specifies the economic
grounding of the choice experiment, and the particular discrete choice models are discussed
and applied. The welfare measures are also calculated.
The fourth chapter deals with the cost-benefit analysis of the provisions for improving Lake
Macha waters and undertakes some further steps in the cost benefit analysis such as the Monte
Carlo simulation and optimization model.
1. Water eutrophication – a topic for environmental economics
research and policy
This chapter introduces the reader to water eutrophication as a crucial topic in environmental
economics with growing economic consequences to human society worldwide. The chapter
identifies the chief reasons for water eutrophication with a special focus on the Czech
Republic. In addition, the chapter brings a description and justification of the research
questions that the thesis tries to answer afterwards. The chosen methodology for handling the
research questions is then identified and discussed, and brought into context with welfare
economics theory. The paper deals with the Lake Macha site as an example case of analysing
the water eutrophication problem; it is described here, including the scope of the
environmental problem, its causes and main steps taken to resolve it, which are dealt with in
the subsequent chapters (above all, Chapter 4).
1.1. Eutrophication as an environmental problem
Water eutrophication is a phenomenon which causes problems in water bodies, especially in
standing waters. In its original sense, water eutrophication means saturation of water with
nutrients – particularly organic compounds – which are released to the water from different
sources, e.g., through natural release of nitrogen and phosphorus and other minerals from the
soil, as well as decomposition of dead organisms in water. These nutrients (particularly
phosphorus and nitrogen) are subsequently used by green aquatic organisms for their growth.
To this degree, eutrophication of rivers, lakes and other water bodies is a natural phenomenon
which exists practically in all waters (so-called natural eutrophication).
Nowadays, the term water eutrophication is more or less used in the strict sense of the word,
that is, the excessive growth of green aquatic organisms (cyanobacteria and algae), which is
caused by the surplus of anthropogenic organic compounds in the waters. In this sense, water
eutrophication is an environmental problem caused by humans (hereinafter, anthropogenic
eutrophication), which has been accelerating in the last few decades due to massive amounts
of nutrients (phosphorus and nitrogen) from different human activities, mainly from
agriculture and municipal sewage discharges. For this reason, it is more intensively discussed
and dealt with in environmental policy.
The main effect of anthropogenic eutrophication is excessive growth of the phytoplankton in
the water (green algae and cyanobacteria). The presence of large quantities of these organisms
in water has a negative impact on water quality both during their lifetime – through their
living processes – and after death, when they decompose (decay).
Live phytoplankton cause primarily organoleptic (smell) defects to drinking water, and their
mechanical properties decrease the treatability of drinking water (clogging filters). Some
phytoplankton species (cyanobacteria) produce toxic substances with proven significant
negative impacts on human health. Cyanobacteria have similar negative impacts on bathing
water too. Moreover, they reduce the aesthetic aspect of such water, which is especially poor
after algal and cyanobacterial bloom occurs.
After death, phytoplankton sink to the bottom, where they decompose. This process consumes
oxygen dissolved in the water. That often leads to a total depletion of oxygen in the aquatic
environment and a change from aerobic processes (i.e., processes in the presence of oxygen)
to anaerobic ones. This development threatens the vital functions of higher life forms both
with an absence of oxygen and because an anaerobic environment produces toxic gases such
as ammonia, methane, hydrogen sulphide, and nitrites.
These phenomena may result in the death of fish and other organisms living in the water.
Under a lasting oxygen deficiency, the dead animals may decompose anaerobically. In
exceptional cases, massively eutrophicated watercourses or water bodies turn into putrid
gutters.
In addition, if cyanobacteria decompose, they release the toxins bound inside their cells. The
quality of the aquatic environment is thus further worsened from the public health
perspective. In less severely eutrophicated waters, it is “only” the sensory properties of the
water that are impaired, and sometimes toxic substances form with an adverse impact on the
aquatic life forms as well as human beings (conjunctivitis, skin eruptions).
1.2. Water eutrophication world-wide
Water eutrophicated has become a serious problem principally since the latter half of the 20th
century, when intensive and blanket fertilisation of farmland began, and when the human
population began growing exponentially (Kočí et al., 2000). The accumulation of inorganic
nutrients in waters has grown to such dimensions in certain countries that hydrological
research there focuses primarily on these issues. Eutrophication has particularly negative
impacts on drinking water reservoirs, where phosphorus accumulates in sediments (mud) at
the bottom.
A more detailed review of the extent of eutrophication world-wide can be found in Kočí et al.
(2000), for instance. Among other things, he says that “the eutrophication issue is by no
means local and concerns not only Central Europe… today, it is literally a world-wide
problem… Almost all major European rivers – the Seine, the Danube, the Elbe – are
eutrophicated…” (Kočí et al., 2000, p. 7).
In England, 84% of the Sites of Special Scientific Interest show signs of eutrophication.
Frequent summer anoxic crises have led to the disappearance of almost all benthic microfauna
in Venice (Tagliapietra, 1998). Lake Balaton (Biro, 1997) as well as the Arendsee in Germany
(Findlay et al., 1998) are eutrophicated. In Sweden, about 14,000 (out of the 90,000) lakes are
eutrophic (Bernes, 2000). A similar situation occurs practically throughout Europe.
Non-European examples include Lake Apopka in Florida, which was clean and had a lush
macrophytic vegetation until the early 1950s (Bachmann et al., 1999). The US lakes of
Mendota in Wisconsin and Travis in Texas suffer from a similar situation; the Great Lakes
have also been experiencing serious problems (Nicholls, 1998). Other examples have been
documented in Chile, Brazil and Africa (both in lakes and man-made reservoirs). Numerous
cases are found in China, Korea and Australia.
Unfortunately, eutrophication does not only concern freshwater; it affects saltwater too. Some
seriously eutrophicated areas are found in the Baltic Sea (e.g., the Gulf of Finland)
(Soederqvist, 1998) and in the North and Mediterranean Seas (along the shores of France)
(Menesguen, 1999).
At the European and Czech levels, the high importance of clean water to society and economy
is also reflected in the current Czech and European legislation, which imposes a statutory
obligation of greater control of phosphorus levels in freshwater ecosystems. For the sake of
completeness, the four principal EU-level legislative regulations (EU Directives), dealing
with the general control of phosphorus and nitrogen, are listed below:
a) the Water Framework Directive (2000/60/CE);
b) the Urban Waste Water Treatment Directive (1991/271/EEC);
c) the New Bathing Water Directive (2006/7/EC); and
d) the Nitrate Directive (1991/676/EEC).
Recently, new European legislation on bathing water has been adopted1. The “New Bathing
Water Directive” (2006/7/EC) updates the measures of the previous Directive 1976/160/EEC
and the Member States have to comply with its stricter and more ambitious requirements.
These directives are not analysed in any more depth or commented in the following; only
some statistics which are reported in compliance with them are quoted.
1.3. Economic consequences of water eutrophication
From the above facts concerning cyanobacteria, their life cycle and the world-wide extent of
the problem, it is straightforward that water eutrophication affects the environment and human
society in a number of ways. It has a number of negative consequences to the human society.
They can be listed as follows:
•
It influences the quality of water supplies and results in increased cleanup costs (e.g.,
clogging of water supply filters);
•
It impacts on the human health, which also has a number of economic consequences,
e.g.:
o rising medical and care-giving costs;
o work loss such as lost personal income and lost productivity;
•
It has negative effects on recreational activities, e.g.:
o decreases aesthetic quality;
1
Directive 2006/7/EC, concerning the management of bathing water quality and repealing Directive
1976/160/EEC
o influences commercial and sports fishing;
o increases the defensive and averting expenditures of individuals.
A body of epidemiological literature deals with the impacts of polluted water on human health
(e.g., WHO, 2001; Ackman, 1997; Figueras 2000; Bahlaoui et al., 1997).
A massive amount of economic studies also exists that strive to figure out welfare changes
resulting from dirty water. Recently, great interest in marine protected areas has been
observed. While biologists stress both stock enhancement effects and various non-use values
such as biodiversity (Novaczek, 1995), economists have so far mainly focused on various use
values2, such as stock enhancement and increased harvests (Sanchirico and Wilen, 2000),
willingness to pay for drinking water improvement solutions (Ahmad et al., 2005),
maintaining a certain level of flow in rivers (Garrod and Wills, 1998), controlling water
pollution (Feenberg and Mills, 1980; Choe et al., 1996; Abou-Ali and Carlsson, 2004;
Whittington et al., 1988, 1990, and 2002), assessing the environmental values of water supply
options (Blamey, et al., 1999), etc. The non-use values were estimated, e.g., for wetland
conservation (Birol et al., 2005; Milon and Scrogin, 2005; Brouwer and Bateman, 2005) and
for estimating the economic value of improvements in river ecology (Hanley et al., 2003).
Many other field were studies have been conducted, focusing not only on acquiring values as
such, but also on method refinement, especially in the recent years.
However, these studies are not the subject of interest in this thesis. The focus here is on a
review of studies dealing with the economic values of water quality (especially water
eutrophication) in relation to the benefit changes from recreation. In the following chapter this
analysis is even further narrowed to choice experiment studies focusing on the topic.
The variety of the studies is large and also the values that recreationists place on water quality
in particular studies vary a lot. The range of estimated WTP is wide. Most of the studies come
from the US. The values start at EUR 0.74 and go up to EUR 227 for saline water. This wide
range of values can be partly explained by differences in the valuation method used, the
pollution indicators adopted, percentage quality improvement being valued, and especially the
units in which WTP is measured (per visit, per season, per year). However, when expressing
2
Garrod and Willis (1999) is the exception in this regard as the study estimates both use and non-use values.
these values as a percentage of income, they are typically less than 1% of the annual income,
which makes the values quite plausible. As Freeman (2003) highlights, in spite of low
individual WTP values attached to water quality improvement, the aggregate value can be
very significant regarding the large numbers of recreationists involved.
Several studies estimate recreational benefits from improvements in marine water quality,
applying the contingent valuation method, random utility models, or the travel cost model
(Freeman, 2003). Other studies, for example Binkley et al., (1978), determine water-based
recreation as an environmental asset with two possible scenarios: (i) WTP to avoid having the
water quality at their favourite site decline to the lowest level possible; and (ii) WTP to have
the water quality at their favourite site improved to the highest level possible. Responses to
three questions were obtained in the survey conducted as part of the present study: (1) How
much would the admission fee to your favourite site need to increase for you to switch some
trips to your second favourite site?; (2) What is your willingness to pay to avoid extreme
pollution at your favourite site?; and (3) What is your willingness to pay for improvement of
the water quality to the highest level at your favourite site? The median value was USD 1.24
per trip for Question 3. Green and Tunstall (1991) valued river quality improvements (using
CVM) in 12 recreational river corridors throughout the UK. They used a three-level scale of
water quality: good enough for water birds, good enough to support many fish and vegetation,
and good enough to be safe for children to paddle or swim in. They found the mean individual
willingness to pay for water quality improvement in one river (values calculated only for
individuals willing to pay a positive amount) to be 1,203 pence per year (53%); 135-166
pence per month (56%-59%) (values are reported in 1987 UK pounds). Parsons et al., (2003)
measured the economic benefits of water quality improvements to recreational uses in six
north-eastern states of the USA (applying the random utility maximization model) using
secondary data. The purpose of the study was to estimate the economic benefit associated
with recreational water quality improvement in 6 states in the USA3. All lakes, rivers, and
coasts in the above states were included in the analysis. They estimated separate random
utility maximization models for fishing, boating, swimming, and viewing using data from the
1994 National Survey of Recreation and the Environment4. The authors valued the impact of
changes in water quality on recreational water use. They assigned an index of high, medium,
and low water quality in terms of biological oxygen demand, dissolved oxygen, total
3
4
New Hampshire, Maine, Vermont, Massachusetts, Rhode Island, and Connecticut
The data set includes 20,925 river sites, 2,975 lake sites, and 1,231 coastal sites.
suspended solids, and faecal coli forms5. The authors estimated that improving all sites to the
"medium" water quality index increases welfare by USD 0.04 for boating; USD 3.14 for
fishing; and USD 5.44 for swimming. They estimated that improving all sites to the "high"
water quality index would increase welfare by USD 31.45 for viewing; USD 8.25 for boating;
USD 8.26 for fishing; and USD 70.47 for swimming per person and trip. Similarly, Carson
and Mitchell (1993) used the contingent valuation method to study the value of clean water at
a national level and categorizing the water quality as usable for boating, fishing, and
swimming. Georgiou et al., (2000) valued bathing water improvements from a health risk
perspective.
Hanson and Hatch (1995) is a revealed preference study using the hedonic property price
model and single site travel cost method (TCM).Both models valued the effects of permanent
changes in water level on lakefront property and recreational use values of Lake Martin
reservoir in Alabama. The study was of the “resource extraction” type, focusing on permanent
changes to reservoir water management policies with respect to length of time at summer fullpool water level, and winter drawdown. Regarding the lake property value change expected,
the change in total aggregate use expenditures (1995 USD million) for Lake Martin for
permanent changes to summer full-pool water levels was expected to reduce lake property
values by 9.8%, i.e., by USD 13,000/foot decrease. A 5-foot drop in full-pool level would
result in a 47% loss in LPV or by USD 88,000. (Hanson and Hatch, 1995)
1.4. Water eutrophication in the Czech Republic
The numbers of excessively eutrophicated water reservoirs in the Czech Republic are not
reported specifically; neither is the welfare loss from connected with the worsened
recreational possibilities. The website příroda.cz says, “The numbers of water reservoirs with
massive occurrence of cyanobacteria and algal bloom are ever growing. It was 66% of the
water reservoirs in the present-day Czech Republic in the 1960s; it was nearly 80% in 1999.”
(Vencl, 2005)
5
High water quality means biological oxygen demand is less than 1.5 mg/l; total suspended solids is less than 10
mg/l; dissolved oxygen is greater than 0.83%; and fecal coliforms are less than 200 MPN/100ml. Medium water
quality allows for biological oxygen demand less than 4 mg/l; total suspended solids is less than 100 mg/l;
dissolved oxygen is greater than 0.45%; and fecal coliforms are less than 2000 MPN/100ml. All water outside
the medium range is considered "low quality".
Based on Directive 2006/7/EC (The new Bathing Water Directive, 2006) the Czech Republic
has the obligation to inform the public about the bathing water quality. This quality is by far
not only related to excessive eutrophication; nevertheless, it still provides some information
about eutrophication as such. Member States present the European Commission with a report
on “Bathing water quality results” in May every year, which is a component of the report on
the water quality in the given year (EC, 20076, pp. 2-3).
The section of the report related to the Czech Republic (last available version for 2007) states,
among other things, that “72.9% of the freshwater bathing waters in the Czech Republic met
the mandatory values in 2007. This is a decrease compared to the previous year. The rate of
compliance with the guide values, however, increased slightly to 54.3%. The number of noncompliant bathing waters (8.5%) was the same as in 2006. The number of prohibited bathing
waters increased to 25 (13.3%).“ Furthermore, the report says, “Water quality problems were
most frequently related to mass proliferation of cyanobacteria. The WHO recommendation
was adopted for the limit value of the “cyanobacteria” indicator, i.e. a three-level water
quality assessment with the ban imposed if a visual inspection reveals the presence of water
bloom. In the 2008 bathing season were 15 bathing bans (mostly due to cyanobacteria)…
Unsatisfactory water quality was identified on certain bathing sites and during certain part of
the bathing season. This was mostly due to cyanobacteria. (EC, 20077, pp. 2-3)
The “Report on the state of water management in the Czech Republic in 2008” (MZe, 2008)
provides a similar picture of water eutrophication in the Czech Republic. The report states, “A
number of water reservoirs had eutrophicated waters in 2008… Major problems with water
quality occurred throughout the year in the following drinking water reservoirs and those used
partly for drinking water: Hamry, Křižanovice, Vrchlice, Seč, Lučina, Vír, Fryšták, Hubenov,
Mostiště, Znojmo, Boskovice, Bojkovice, Ludkovice, Opatovice, Nová Říše, and Koryčany;
and those not for drinking water: Les Království, Rozkoš, Mšeno, Pařížov, Fojtka, Skalka,
České údolí, Hracholusky, Brněnská přehrada, Horní Bečva, Bystřička, Novomlýnské nádrže,
Luhačovice, Křetínka, Moravská Třebová, Jevišovice, Oleksovice, Plumlov, Těrlicko, and
Olešná… In summer, water was evaluated as less suitable or unsuitable for recreation in some
non-drinking water reservoirs (such as Orlík, Seč, Rozkoš, Skalka, Hracholusky, Luhačovice,
Brněnská přehrada, Baška, Těrlicko, and Olešná).”
6
7
Last publicly available version
Last publicly available version
Concerning bathing water quality in 2008, the report states: “The most common water quality
problems are related to the massive occurrence of cyanobacteria, in some places leading to
annual bathing bans. In the 2008 bathing season, public health protection authorities
monitored 259 sites used for bathing, out of which 129 were swimming sites operated in the
open, and 130 were bathing areas8. Based on laboratory analyses, public health protection
authorities issued a ban on 19 sites in the Czech Republic in the 2008 bathing season9 (within
that, 7 bathing sites in the open and 12 bathing areas). Water quality termed unsuitable for
bathing occurred on 24 of the sites, and impaired water quality, unsuitable for sensitive
individuals, who are recommended to take a shower in drinking water after bathing, was
identified on 72 sites in the Czech Republic.” (MZe, 2008, p. 18) A simple calculation shows
that 44% of the monitored sites used for bathing suffered from impaired water quality in
2008. The report identifies eutrophication as the main cause of the poor water quality. As for
the total bathing bans, the report identifies excessive cyanobacterium occurrence as the only
reason for its declaration. The WHO recommendation was adopted in the Czech Republic as
the threshold for the cyanobacterium indicator10.
1.5. Sources of phosphorus in the Czech Republic
The main reason for the anthropogenic eutrophication in the Czech Republic is the excessive
import of phosphorus into the waters despite the fact that phosphorus is not the only chemical
element necessary for the development of so-called primary producers (i.e., organisms
capable of producing organic substances from inorganic ones). The reason is that there is
usually an abundance of other elements (so-called nutrients, i.e., carbon, oxygen, hydrogen,
nitrogen, potassium, calcium, magnesium, and sulphur) in the aquatic environment. Only
phosphorus tends to be insufficient, which is why its concentration in the ecosystem tends to
8
Surface waters used for human bathing (so-called bathing areas) are defined in Act no. 254/2001 Coll. and
their list and definition are subject to Regulation no. 159/2003 Coll. These sites are not facilities, have no
operators, but due to the fitness of their water quality, they are used for bathing by many people. The
responsibility for inspecting water quality in these bathing areas is borne by public health protection authorities;
the scope and frequency of inspection is subject to Regulation no. 135/2004 Coll. There are 130 such sites in the
Czech Republic.
9
In the 2007 bathing season there were 25 bathing bans (mostly due to cyanobacteria) (Bathing water quality
annual report, 2007, p. 3).
10
The WHO recommendation was adopted for the limit value of the “cyanobacteria” indicator, i.e. a three-level
water quality assessment with the ban imposed if a visual inspection reveals the presence of water bloom.
(Bathing water quality annual report, 2007, p. 3)
be the limiting factor for the development of entire plant populations. If it becomes
overabundant for various reasons, it usually leads to an overproduction of phytoplankton.
The Integrated Pollution Register, for instance, says, “The stechiometric nitrogen to
phosphorus (N:P) ratio of 16:1 is optimal for biomass production. In most of the Czech
Republic’s reservoirs, the N:P ratio is high above 16, which is why phosphorus is the limiting
factor in eutrophication. To eutrophicate, water also requires warm weather and enough
sunshine.” (Integrovaný registr znečištění, 2010)
Phosphorus concentrations in the order of hundredths of mg/l are significant to algal bloom.
The threshold phosphorus concentrations at which the prolific growth of green organisms
starts cannot be determined generally, as it depends on the conditions in the reservoir or lake,
and on the overall water composition. A rough limit for phosphorus concentration in water
may be defined at 0.01 mg/l.
Phosphorus is thus a critical nutrient for the origination of anthropogenic eutrophication (rise
of the cyanobacterium and algae). For this reason, the main effort regarding eutrophication
treatment is concentrated on reducing the phosphorus discharges coming from human
activities.
There are very few data available on the phosphorus balance in the Czech Republic. The most
reliable and commonly recognized are probably the following data shown in Table 1-1. The
table summarizes the main anthropogenic phosphorus sources in the Czech Republic. These
sources are as follows (their volumes will be discussed further):
•
The use of industrial fertilizers in crop production followed by phosphorus and
nitrogen fertilizer runoff to water bodies;
•
Insufficient disposal of organic fertilizers in livestock production;
•
Insufficient treatment of sewerage waters;
•
Polyphosphate use in synthetic detergents (e.g., liquid dishwashing detergents).
Table 1-1: Phosphorus balance (produced and discharged) in 2004 (in tonnes/year)
Produced
Discharged
Point sources total
6,815
1,839
Within that, municipal sources
6,298
1,617
industrial sources
517
222
Treatment efficiency, point sources
73
municipal sources
74.3
industrial sources
57.1
2,377
Drain in border points
Difference: Drain in border points – Point sources total = Minimum contribution of areal
538
77.4
“Share” of point sources in drain in border points [%]
municipal sources in drain in border points [%]
68
industrial sources in drain in border points [%]
9.3
Phosphorus content in laundry detergents sold (2003)
1,678
430.8
Phosphorus content in washing-up detergents (estimated)
1
0.3
Phosphorus content in laundry and washing-up detergents
1,679
431.1
Share of phosphorus from laundry and washing-up detergents in phosphorus
26.7
26.7
in municipal wastewater [%]
“Share” of phosphorus from laundry and washing-up detergents in phosphorus drain in
18.1
Source: Nesměrák (2006)
It is clear from the table that over 60% of the water contamination (under 7 thousand tonnes
of P) comes from point sources of pollution. Within that, over 53% comes from municipal
sources of pollution11. Only approx. 7% is from industrial sources12 (Nesměrák, 2006).
Based on the difference between the amounts of phosphorus in boundary points and the
pollution discharged from the point sources into the collectors, the minimum contribution of
areal and diffused sources of pollution to the total wastewater contamination with phosphorus
can be estimated13. However, it has to be kept in mind that this contribution to the
contamination is calculated as if rivers had no self-purifying capacity. This means that the
calculated table value is the lower limit of the pollution contribution; the real value will likely
be higher: at least twice according to VUV T.G. (Nesměrák, 2006)14. It follows from the
aforesaid that areal and diffused sources contribute about one half less to collector
11
Depending on the diet, a human body produces about 1.08 to 1.65 g of phosphorus a day. The amount of
phosphorus from detergents roughly equals the amount generated by humans in some countries, i.e., approx. 1.5
g a day. Over 3 g of phosphorus per capita enter the sanitary sewers every day.
12
The phosphorus elimination efficiency parameters for WWTP shown in the table are calculated as the
difference between the discharged and generated pollution. On average, this value is under 77%. (Nesměrák,
2006)
13
The contribution of areal and diffused sources of pollution also significantly depends on the annual hydrous
capacity (effect of runoff and leaks).
14
The table value was multiplied by two based on the information on its undervaluation. The figure thus
obtained (1,164=582x2) was subsequently the basis for all other calculations.
contamination than point sources. Farming, i.e., field fertilisation and subsequent fertiliser
runoff into collectors, is the major polluter in the areal and diffused sources category. The
elution of phosphorus from soil contributes only a small portion to the total phosphorus
content in the Czech Republic’s waters.
Municipal sources of pollution (municipal wastewater treatment plants, dispersed human
settlements) therefore contribute approximately 60% to the contamination of Czech
watercourses with phosphorus. This phosphorus arrives in the wastewater chiefly from human
bodies and liquid detergents used in households.
A part of the phosphorus is retained from sanitary sewerage in biological wastewater
treatment plants by incorporating it in the biomass of the freely generated sludge, so that not
all the phosphorus and nitrogen compounds enter the collectors. The rest of the phosphorus
compounds remaining in the water after biological treatment can also be removed: by
chemical coagulation in so-called tertiary wastewater treatment (Hally, 2002). In order to
remove phosphorus content from wastewater thoroughly, a wastewater treatment plant
therefore has to be equipped with a so-called third stage of treatment15.
1.5. Lake Macha – case study
As already mentioned, water eutrophication particularly severely affects standing waters,
including drinking water reservoirs, where phosphorus accumulates in the sediments (mud) at
the bottom. Lake Macha is one of the surface water bathing sites in the Czech Republic which
suffers from low water quality in recent years. It is situated in the Liberec Region. It lies
almost 100 km north of Prague. Its size is approximately 305 hectares and the average longterm flow rate is 0.563m3/s. It lies in a tourist district and is crucial for the tourism in the
district; at the same time, it is part of the Českolipsko SPA and Jestřebsko-Dokesko SAC
within the NATURA 2000 system of EC protected sites. Extremely valuable wetlands are
located in its immediate vicinity, including the Novozámecký rybník National Nature Reserve
and the Swamp National Nature Monument, which are crucially affected by the Lake Macha
waters.
15
The third stage of treatment is currently being added to municipal WWTP in the Czech Republic in connection
to Council Directive 91/271/EEC on Urban Wastewater Treatment and the approached chosen by the Czech
Republic in respect to implementing the Directive.
The poor water quality of Lake Macha is caused by high phosphorus content in the water
(water eutrophication), which has caused significant cyanobacterium occurrence. As was
already discussed, high cyanobacterium occurrence in water has a great negative influence on
the water quality and also induces significant economic costs.
The problem of eutrophication of the Lake water escalated in 2004, when the swimming was
banned and the beaches closed already in June. Because of the bathing ban, the revenues from
tourism decreased and caused economic problems to many businesses in the tourist-oriented
region (Doksy, 2007).
In reaction to the problems caused by the cyanobacteria, the town of Doksy (which lies by the
Lake and benefits the most from the tourism) started to perform some activities to prevent the
2004 situation from repeating. A complex analysis of the phosphorus balance (especially its
sources) and water quality provisions made at Lake Macha is made in Chapter 4 (CB
Analysis); they are only briefly summarized on the following pages to provide a rough picture
of the problem.
1.5.1. Sources of phosphorus in Lake Macha
Several reasons for the high phosphorus content in the water column can be identified in the
study site. In particular, the Lake bottom sediments contain a great store of nutrients, most of
which originated in the latter half of the 20th century, when Lake Macha was utilised for
intensive fishery and fertilised with superphosphates (fishermen threw approx. 3,800 kg of P
in the Lake in 1950-1957)16. The use of phosphorus-containing laundry detergents and the
subsequent inadequate wastewater treatment definitely also played a role in the nutrient
saturation with phosphorus (the use of phosphate laundry detergents was finally banned in the
Czech Republic only in 2006). Moreover, the Lake is relatively shallow (average depth of
about 2 metres), which significantly contributes to water stirring along with the numerous
holiday makers and carp family of fish digging up the Lake bottom. The Lake being so
shallow also causes its rapid warming through, which again creates good conditions for
phytoplankton (algal and cyanobacterial) growth. Moreover, the Lake continually receives
water with sediment seriously enriched with phosphorus.
16
Given the mesotrophic water quality of Lake Macha already back then, it is probable that most of the
phosphorus was locked in the sediment.
1.5.2. Overview of water quality improvement measures
Some of the activities which the town of Doksy started to carry out were focused on
preventing future phosphorus discharges into the Lake (proper treatment of wastewater
flowing into the Lake). Then the Doksy inlet (which contains a considerable part of the
phosphorus-contaminated sediment) was separated from the rest of the Lake; also, PAX 18
was applied to the water. PAX 18 – a method of phosphorus precipitation by means of
aluminous salts – has been applied several times since 2005 to ensure at least certain water
quality during the recreational season.
1.5.3. Lake Macha – Water condition
Table 1-2 below summarizes the situation in the bathing seasons of particular years at
particular beaches along Lake Macha. As is evident from the table, the water quality at Lake
Macha has improved significantly since 2005 after the interventions by the town of Doksy.
Table 1-2: Water quality at Lake Macha beaches in particular years
(according to the Directive 1976/160/EC and the New Bathing Water Directive 2006/7/EC)
Staré Splavy
Doksy
Klůček
Borný
2004
Banned or closed
throughout the season
Banned or closed
throughout the season
Banned or closed
throughout the season
Banned or closed
throughout the season
2005
2006
2007
2008
Good
Good
Excellent
Excellent
Excellent
Excellent
Good
Good
Excellent
Excellent
Excellent
Good
Poor
Excellent
Excellent
Excellent
Notes:
Poor = Not compliant with mandatory values
Good = Compliant to mandatory values
Excellent = Compliant to guide values
Source: Bathing water quality annual report, Bathing water results 2007, online:
http://www.eea.europa.eu/themes/water/status-and-monitoring/state-of-bathing-water1/bathing-water-data-viewer
1.6. The research questions
In light of the above mentioned situation at Lake Macha, the main motivation of the case
study was to find out what influence the water quality changes have on the recreational utility
of Lake Macha users. Secondly, it was to establish the other determinants influencing the
vacationists when deciding what resort to visit (in this project, recreation stands for waterside
recreation in the Czech Republic in the summer season – swimming, windsurfing, and beach
sports). And finally, we intended to determine the importance of the water quality in
comparison with other recreational aspects of the beaches.
Another motivation for the research was to evaluate the economic effectiveness of the
measures taken at Lake Macha in order to reduce the water eutrophication.
1.7. The research methodology
Cost-benefit analysis – a method commonly used for this type of analysis (Boardman et al.,
2006) – was employed to evaluate the economic effectiveness of the measures taken at Lake
Macha in order to reduce the water eutrophication. The analysis is made in Chapter 4, where
the methodology is also described.
Different research methods were considered for answering the first set of the research
questions. Previous research done at Lake Macha was done by Ščasný et al., (2006). In this
study, the contingent valuation method was used. The targeted population in the case study
included residents living in four surrounding villages (39% of respondents) and the
households living in the two towns on the shores of Lake Macha. Water quality was classified
on a five-level scale of total phosphorus content. Hypothetical water quality improvement
from the current quality level by 1 class or 2 classes, respectively, was considered as the
contingent product. Welfare change due to provision of the contingent product was measured
by compensation surplus derived from the residents’ willingness to pay. The payment vehicle
for WTP used in the scenario was an increase in the sewage fee paid by the respondent’s
household or, for households not connected to the public sewerage system or water supply, an
increased price of emptying septic tanks and treating the wastewater. The mean WTP
estimated ranged from CZK 461.3 to CZK 782 (in real 2005 CZK), depending on the model
used.
A different research methodology had to be used for answering the research questions in our
case study, in respect of both the target group and the method used, as well as the method of
estimating changes in demand. The objective was to focus on the target group who use the
Lake for recreation and determine the chief drivers behind the demand. Finally, the choice
experiment method was chosen as it fits the best regarding the research questions.
1.8. The choice experiment
Choice experiment is commonly included among so-called stated preference methods, which
are usually applied in order to establish the value of environmental goods. A body of literature
exists on the methodological problems of the stated preferences methods (e.g., Slavík, 2007;
Cordato, 2004; Šímová, 2007). The choice experiment eliminates some of these drawbacks
(see below); at the same time, it is the most developed of the stated preference methods at
present. Choice experiment belongs among choice modelling techniques, also known in
literature as conjoint analysis (CA). Its applicability is also much broader than that of other
stated preference methods. Choice experiment as such originates from Lancaster’s consumer
theory (Lancaster, 1966) and the random utility model, RUM (McFadden, 1974; Phaneuf,
2005).
Basically, we can identify several reasons (Bateman et al., 2002) why the choice experiment
method best fits answering the research questions:
•
choice experiment may provide information about which attributes significantly affect
(constitute) the value associated by people with non-market goods;
•
it allows the inference of the order of importance of these attributes within a relevant
(study) population;
•
it allows the economic valuation of a change in the level of an attribute or multiple
attributes simultaneously (which can be further used to analyse the effectiveness of
expenditures on changing the attribute level).
1.8.1. History of choice experiment
Choice experiment (CE) was first developed decades ago, and its origins are more closely
related to psychology, where psychologists attempted at creating such discrete choice models
that would allow them to predict decisions of individuals in certain situations (Thurstone,
1927; Luce, 1959). As Håkan and Olsson say, “Later these ideas were refined by economists
and linked to the characteristics theory of value (Lancaster, 1966 in Håkan and Olsson, 2004)
and the random utility theory (Manski, 1977 in Håkan and Olsson, 2004).” The foundations of
choice modelling thus date from the 1960s and are ascribed to Lancaster (1966). These
techniques are applicable in empiric analysis. Their application is most frequent in marketing
studies (SAS Institute Inc., 1997; Warren et al., 1994) and in transportation literature
(Henscher, 1994). They have recently been utilised in other areas too, such as economic
valuation of environmental goods (Louviere, 2001). A review of some of the applications of
choice experiment can be found in Hanley et al., (2001). A review of applications in
modelling demand for water-based recreation using the choice experiment method is
contained in Chapter 2.
1.8.2. The economic rationale behind choice experiment
The principal idea behind choice experiment is that individuals choose from various product
baskets the one that brings them the most utility. The product is described with several
attributes, which show differing levels. The price is always one of the attributes. A forest, for
example, may be characterised with species diversity, age structure, and recreational
possibilities. A river may be described with chemical water quality, biological quality, and
appearance. A fishpond may be described using attributes such as surface area, depth, water
quality, biodiversity, bottom type, etc. In this sense, bus transport is characterised with costs,
time, and comfort. Changing the levels of the attributes of a given good results in the
provision of a different product. This is based on the “new consumption theory” characterized
by Lancaster (1966).
Choice experiment aims at valuating the change in the attributes of a given good. Individuals
are assumed to know their preferences and to make choices maximizing their utility, while
these preferences are not fully known to the researcher. Individuals are also assumed to
consider the full set of offered alternatives in the choice situation17. Based on the random
utility framework and welfare economics, it is then possible to calculate welfare estimates for
various changes in the levels of the different attributes. As already mentioned, these
characteristics of CE are mostly used in marketing studies (SAS Institute Inc., 1997; Warren
et al., 1994) and in traffic behaviour studies (Henscher, 1994).
Among choice modelling techniques, only choice experiment (CA) and contingent ranking
are CA techniques that are in line with economic theory, permitting the application of the
results of these methods in cost-benefit analysis (CBA).
17
This has recently been questioned by Campbell (2007).
For choice experiment application, four reasons can be identified that allow estimating the
changes in utility closely linked to the economic grounding of the welfare measurement (see
below):
•
CE allows the respondent to make a trade-off between changing attribute levels and
costs related to such changes;
•
the respondent may choose the status-quo alternative, i.e., zero change in
environmental quality, thus no additional costs;
•
applicability of econometric methods in line with the rational choice theory;
•
the compensating and equivalent surplus related to a change in environmental quality
can be estimated.
1.8.3. Pros and cons of choice experiment
Compared to the contingent valuation (CVM) method, for instance, the advantage of choice
experiment is that the multi-dimensional aspects of a proposed change can be reflected.
Another significant advantage is that the willingness to pay is estimated indirectly in a CM
study. The respondents do not express their willingness to pay directly, but rather score
points, compare or choose from alternatives. This minimises certain distortions arising from
applying the CVM (where respondents may choose either a strategic answer or one that they
think is socially desirable) (Bateman et al., 2002).
Choice experiment offers each respondent a set of alternative products, from which they
choose the most preferred alternative. The decision-making is similar to the real market,
where the consumer chooses from two or more products with similar attributes, differing only
in their levels (examples may include two identical mobile telephones differing only in their
battery life and price). Like on the real market, the respondent in a choice experiment has the
option to choose product A, or product B, or neither.
In addition, CE allows better measurement of the threshold value of attribute changes in
environmental goods. Compared to the travel cost method, the advantage of choice
experiment is that it can also study the levels of attributes which do not exist in reality, so
reflecting even the non-use values of environmental goods.
Moreover, a significant advantage of so-called unlabeled CE18 over labeled CE19 is their more
probable adherence to the independent and identical distribution condition20 (see Chapter 3).
The reason is that if an alternative is assigned a name (in labeled CE), the respondents’ ideas
may be influenced regardless of the actual experiment. For example, if a transport alternative
is labeled as “Aeroplane” instead of “Alternative 1”, the label influences the respondents’
ideas. They may associate attributes such as comfort, short travel time, etc., with the term
aeroplane subconsciously. However, that is in contradiction to the independent and identical
distribution condition (see below under IID).
The cognitive requirements on the respondents’ decision-making are the principal
disadvantage of choice modelling. The respondents repeatedly make decisions among
different alternatives, which may be limiting in respect of absorbing more information. The
complexity of choices may be a problem for the respondents. If the design contains too many
attributes as well as relations among the attributes, it may result in respondents losing interest,
or trying to making their decisions simpler, which may distort the final results (for a brief
discussion of the number of decision-making operations, see Chapter 2). As with a CVM
study, value estimates may depend on the quality of the entire study.
The disadvantages of CE include the fact that estimation of the total value of an
environmental good (not the case in the present analysis) has to take into account that the total
value is a sum of the values of all the attributes of the good. However, an important attribute
may be missing in the design. This is highlighted in some transportation studies, for example,
where the resulting value corresponding to the overall change is lower than the sum of the
attributes that were estimated using CE (Steer Davies Gleave, 2000).
1.9. The economic grounding of welfare measurement
Expressing environmental values belongs to the consumer theory. The consumer demand
economic theory is described briefly in this chapter (Freeman, 2003; Kolstad, 2000). The
reason for classification under the consumer theory is the fact that environmental problems
require a compromise (trade-off) between using resources for ordinary goods or for
18
The CE case at Lake Macha
“Experiments using generally applicable names for alternatives are called unlabeled experiments. If a concrete
name (e.g., car) is used, the experiment is referred to as a labeled experiment.” (Hensher et al., 2005, p.112)
20
i.e. random variables are distributed identically and independently
19
environmental protection (Kolstad, 2000). In the consumer’s perspective, it is this
compromise that the demand curve shows, i.e., how much the consumer is willing to give up
on (or pay) for a certain level of an environmental good. It is therefore consumers’
willingness to pay for the protection of a certain environmental good that is studied. Nonmarket valuation methods are employed to identify the values; the choice experiment is
grounded among them as a stated reference technique.
It is important to note that an individual demand curve (generated based on maximising an
individual’s utility) is the basic unit of this analysis. The demand function expresses the
relation between the amount and the price of the demanded good at a consumer’s given
income. The demand curve expresses consumer preferences, which are the basic element of
an economy (Vojáček, 2007). The demand curve is a useful tool for analysing the private
goods market and important in analysing preferences for environmental goods. The nonexistence of a market is a difficult point in constructing a demand curve for an environmental
good, where the quantity of the environmental good consumed at various prices cannot be
observed.
Another important concept in analysing consumer behaviour is the consumer utility function
u(z, q), where z and q are two goods. This utility function yields the amount of utility that the
consumer receives for various combinations of the goods z and q.
1.9.1. The impacts of price changes on welfare
How do we evaluate the impact of changes in the price of an environmental good on the
consumer’s welfare? We follow Freeman (2003), Kolstad (2000), Markandya et al. (2002),
and Ward and Beal (2000). Let us assume that the price of the good z changes from pz0 to pz1.
If the prices pz0 and pz1 correspond to consumptions z0 and z1, respectively, then the impact of
the price change on the consumer’s welfare corresponds to the change in the consumer
surplus that is related to the change from z0 to z1.
This indicator of the welfare change entails the problem that the income effect is connected to
the substitution effect. A more favourable valuation of the price change effect can be found
below the compensating demand curve. The price change shifts the consumer from one utility
level to another utility level. Since utility is constant for the compensating demand function, it
is a question which compensating demand function (utility level) to use. One of the
compensating demand functions is related to the initial price; the other one, to the final price.
Both alternatives can be used, but the result will differ moderately. The amount of money that
we give the consumer in compensation for the price increase, i.e., return the consumer to the
original utility level, is the compensating variation (CV). The amount of money the consumer
is willing to pay in order to avoid the price increase is the equivalent variation (EV). CV and
EV are positive for a price increase, and negative for a price decrease. Both the indicators can
also be expressed using the expenditure function, which indicates what income level the
consumer demands to achieve a given utility level. Either may be measured in a choice
experiment, depending on the specific design of the study. In the Lake Macha case (see
below), EV is measured in the event of any of the attributes changing for the worse (compared
to the initial level), and CV in changes for the better.
These concepts can be depicted graphically too. Chart 1-1 shows Marshall’s demand curve for
the good z, EF. Reducing the price from pz0 to pz1 causes the consumed quantity to rise from
z0 to z1. The income remains constant. The consumer enjoys higher utility after this price
change. The chart also shows two compensating demand curves: one crosses point E, the
other one crosses F. These curves correspond to the original utility level and the utility level
after the price change. The area AEFB below Marshall’s demand curve reflects the consumer
surplus related to the price change. The area AEGB represents the compensating variation
corresponding to the price change. The area ADFB represents the equivalent variation of the
price change. Chart 1-1 also shows that the consumer surplus is higher than the compensating
variation for a price decrease (assuming this is a normal good). Nevertheless, it is lower than
the equivalent variation. The situation is opposite for a price increase (EV ≤ CS ≤ CV). CS, CV
and EV are roughly equal for minor price changes.
Chart 1-1: Compensating variation, equivalent variation,
and consumer surplus of a price change
pz
pz0 A
E
D
F
pz1
B
G
z0
z1
z
Source: Kolstad (2000)
Compensating and equivalent variations can also be used to estimate the welfare change when
the quality of an environmental good changes. Let us follow Freeman (2003), Kolstad (2000),
and Markandya et al. (2002). Let us assume that the consumer consumes an initial level q0.
They dispose of an income y and are on a utility level U0. Let us change the quantity of the
environmental good to q1. The consumer shifts to the utility level U1.
What is the value of the change in the environmental good? The value of this change can be
estimated from the amount of the money that has the same impact on the consumer’s utility as
the environmental change. The amount of money that keeps the consumer at the original
utility level at which they were prior to the change in the environmental good is called the
compensating surplus (CS). The amount of money that shifts the consumer to a new utility
level which they would reach in the event of a change in the environmental good is called the
equivalent surplus (ES).
Let us assume a change from q0 to q1 that makes the consumer’s situation worse. Let us
compensate the consumer for this change: let us give them an amount of money that will shift
them back to the original utility level. In this case, it is a compensating surplus. The
equivalent surplus is the amount of the consumer’s benefit that has to be taken away to shift
them to a new utility level. The consumer would be at that utility level if the environmental
change occurred.
CS and ES can also be depicted graphically. Chart 1-2 shows the amount of the goods z and q
that can be consumed. z is an ordinary good, and q is an environmental good. If the price of z
is pz and the income is y, then the budgetary restraint is y = pz z.
Chart 1-2: Compensation and equivalent surplus connected to a change in an
environmental good
C
z
A
ES
pz
y = pz z
CS
pz
D
B
q0
q1
q
Source: Kolstad (2000)
As shown in Figure XX, the budgetary restraint is a curve parallel to the axis q. The reason is
that the given good q is unpaid for. The consumer is at the utility level U0 at q0. Increasing q
to q1 increases the consumer’s utility to U1. The compensating surplus is the monetary value z
that returns the consumer to U0 at a new level of the environmental good q1. The equivalent
surplus is the monetary value z that shifts the consumer to U1 instead of a change in q.
1.10. Further analysis
The following chapters concentrate only on Lake Macha case study, except for the
dissertation conclusions. The second chapter deals with the qualitative research that preceded
the preparation of the choice experiment as such; moreover, it pays attention to the
preparation of the choice experiment, the preliminary survey at Lake Macha, the data
collection, the data collection strategy, and the basic characterisation of the choice sample.
The third chapter deals with the choice experiment data analysis. It specifies the economic
grounding of the choice experiment, and the particular discrete choice models are discussed
and applied. The welfare measures are also calculated. The fourth chapter deals with the costbenefit analysis of the provisions for improving Lake Macha waters.
1.11. Summary
Water eutrophication is a phenomenon which causes problems in water bodies, especially in
standing waters. This chapter highlights the fact that eutrophicated reservoirs are not sporadic
cases, but that the problem is much more extensive (approx. 80% of the water reservoirs in
the Czech Republic are excessively eutrophic). It is also shown that not only are the Czech
Republic’s waters eutrophicated: it is a global problem.
Phosphorus is identified in this chapter as the chief environmental stressor. The phosphorus
content in the water is thus a limiting factor for the development of negative processes
associated with water eutrophication, chiefly massive growth of cyanobacteria, which are
toxic and prevent virtually any other uses of the water. Point sources of pollution were
identified to be the primary sources of phosphorus in the Czech Republic (60%, of which over
53% are municipal sources). Approximately 30% is a contribution of areal and diffused
sources (particularly field fertilisation and the subsequent fertiliser runoff into collectors).
The consequences of water eutrophication and the quantity of affected sites indicate the
significant economic dimension of the problem. It is clear that clean water is becoming an
increasingly scarce good for both recreational and other purposes.
Many epidemiological studies world-wide deal with the consequences of water eutrophication
to human health. Since the issue has an economic dimension, significant attention is paid
world-wide to studying the impacts of water quality and water management on welfare
changes. Some of the study areas are indicated here; special attention is paid to studies
dealing with impacts on recreational utility.
Lake Macha is identified as one of the water reservoirs that have suffered from excessive
eutrophication in the long run. It is important both to the regional economy (due to its high
visitation rates in the bathing season) and because it is part of an ecologically valuable
ecosystem of the Českolipsko Special Protection Area and the Jestřebsko-Dokesko Special
Area of Conservation within the NATURA 2000 system of EC protected areas. In addition,
extremely valuable wetlands, crucially affected by the Lake Macha waters, are situated in its
immediate vicinity.
Given the long-term tradition of summer recreation by Lake Macha, the enormous visitation
rate of the water body, and the serious ecological problems that the area suffers from, the area
was chosen for the case study of modelling demand for summer recreation.
The choice experiment method was chosen as a suitable method for answering the defined
research questions, and set in the context of the macroeconomic consumer theory and welfare
measurement theory. The method is also discussed in respect of its advantages and
disadvantages, chiefly in relation to the formulated research questions.
In order to answer an additional research question concerning the evaluation of the
effectiveness of measures taken at Lake Macha, cost-benefit analysis was selected as a
suitable method; Chapter 4 deals with this analysis.
2. The choice experiment design and sample characteristics
This chapter deals, firstly, with the crucial stages in preparing the choice experiment study
and, secondly, data description, basic sample characteristics and respondents attitudes are
described and analysed here.
The discussion of the choice experiment design starts with the qualitative research, which
preceded the preparation of the choice experiment as such, focusing on the selection of
attributes, their levels and their description for respondents. This stage of the preliminary
survey aimed at identifying the demand parameters (attributes) and selection of their levels. A
review of studies using choice experiment methodology in waterside recreation is also part of
this subchapter.
Moreover, the chapter pays attention to the quantitative preliminary survey directly on Lake
Macha beaches and the conclusions derived from it. A discussion of the quantitative
preliminary survey is followed by a description of the making of the experimental design and
the choice sets. The chapter proceeds with a description of the final questionnaire used in the
main data collection, a specification of the study population, and a discussion of the sampling
strategy.
The last part of the chapter is devoted to the basic sample characteristics, data description and
an analysis of respondents’ attitudes.
2.1.Qualitative preliminary survey at Lake Macha
As discussed in the previous chapter, various types of methods and techniques currently used
in environmental economics for economic valuation of environmental goods were considered
for use in estimating the recreational demand at Lake Macha. Choice experiment was chosen;
its theoretical basics and economic justification were discussed in the previous chapter.
Prior to the quantification and econometric estimate of the summer season recreational
demand at Lake Macha, a survey on (or detection of) the recreational demand determinants
had to be carried out first. The qualitative research methodology (Hendl, 2005) was employed
in this stage of the research. This stage included the identification of relevant attributes of the
non-market good that is the subject of study. Attributes potentially taken into consideration
are such that are believed to aptly describe the good surveyed or expected to significantly
affect people’s preferences towards the good surveyed (for a forest, for instance, these may be
tree density, tree species composition, tree health, overall forest biodiversity, etc.);
furthermore, attributes that are relevant in respect of the policy assessed (for water, this may
be improved water transparency resulting from implementation of certain measures) as well
as attributes that have proven important in previous similarly oriented studies, or attributes
that have arisen from empirical preliminary surveys in the form of in-depth interviews or
focus groups. As this is a key part of CE preparation, this stage of the survey requires special
attention. The resulting values acquired using the CE method are sensitive to the overall CE
design used. The selection of attributes, their levels, the way the questionnaire is presented to
the respondent (e.g., using photographs) etc., may have a crucial influence on the result
(Bateman et al., 2002, pp. 273-274). That is why the following subchapter is dedicated to a
detailed presentation and discussion of the preliminary survey stage, during which relevant
demand determinants for Lake Macha were identified.
The stage comprised several steps:
•
an international review of studies dealing with similar issues (see below);
•
group and one-to-one interviews;
•
consideration of other attributes depending on the nature of the research problem and
environmental measures taken the effectiveness of which was going to be assessed in
the study;
•
since the costs of using the good had to be one of the attributes, the possible ways of
specifying the costs of using the good were examined (e.g., travel costs related to
travel to Lake Macha, entrance fee to the beaches, etc.).
The attributes for the Lake Macha choice experiment were first suggested by the researchers
intuitively. Subsequently, they were supplemented based on the survey of relevant literature
(see below). The first version of the questionnaire was compiled in this way, excluding the
choice experiment but including questions concerning the ranking of possible waterside
recreational demand determinants by their importance to the respondent (for the two
questions, see Annex 1, Questions 16 and 17) with an opportunity to add attributes that the
respondent missed in the offer yet considered to be important. A group interview and eight
one-to-one interviews were conducted subsequently. The interviews were recorded, and the
key segments concerning the attributes (chiefly the water attribute) were transcribed. These
interviews gave rise to findings that were used in drafting the first version of the questionnaire
to include the choice experiment.
2.1.1. International literature review
As was discussed in the first chapter, a considerable body of research exists in the field of
water economics. These studies focus, e.g., on evaluation of different environmental
management options, stock enhancement, harvest increase, evaluation of drinking water
policies, water pollution control, etc. In Chapter 1, attention was paid mainly to studies
dealing with the economic values of water quality in relation to changing benefit from
recreation. In this subchapter, the literature review is even narrower, focusing only on studies
using the CE methodology.
Unfortunately, there are not many choice experiment studies devoted entirely to the
recreational value of water quality in the world. Most of the studies which somehow deal with
the issue are connected with angling activities and not with bathing/swimming activities. The
reasons for that may be various, but it is not the intention of this chapter to discuss them. The
only CE studies that concentrate on recreational benefits connected with water quality are
reviewed in the following paragraphs. I suggest labelling the studies where water quality
substantially influences the population’s recreational value
studies”. For an overview of these studies, see Table 2-1.
as “water amenity valuation
Authors
Table 2-1: Water amenity valuation studies
Water
Study name
Attributes
attribute
Terrain, Fish Size, Fish
Catch rate, Water quality,
Facilities, Swimming,
Beach, Distance to site,
Combining Revealed Standing water feature,
Adamowicz, W., and Stated Preference Standing water fish species,
J. Louviere, and Methods for Valuing Standing water boating,
M. Williams,
Environmental
Running water feature,
1994
Amenities
Running water fish species
Heterogeneous
Bathing water quality,
Preferences for
Biodiversity levels,
Marine Amenities: A Cod catch per trawling hour,
Cost (Swedish kronor) for
Choice Experiment
an individual for each
Eggert, H. and B. Applied to Water
alternative
Olsson, 2004
Quality
Valuation of Benefits
to England and Wales
Department for
of a Revise Bathing
Environment,
Water Quality
Average water quality,
Directive and Other
Advisory notice system,
Food and Rural
Brach Characteristics Litter / dog mess,
Affairs by
Using the Choice
Economics for
Safety & Amenities ,
the Environment Experiment
Additional water charges per
Consultancy Ltd Methodology
year
Water attribute
description
water quality
swimming
beach
Good water quality
Bad water quality
Percentage of sites
violating the quality
standard:
- 12 % (baseline),
bathing water - 10 %,
quality
- 5 %;
Two different
expressions of the
attribute:
- risk of suffering a
stomach upset
- days in the bathing
season that it is unsafe
to swim due to poor
Average
water quality water quality
Source: Own analysis
The improvements in the water quality are mostly specified as:
•
some “degree of water quality improvements” (e.g., bad water quality, good water
quality);
•
changes in wildlife habitat;
•
changes in some of the ecosystem attributes (flora and fauna – fish, plants and
invertebrates, local livelihood, ecological function, rare and endangered species
occurrence);
•
aesthetics (represented by the amount of litter in the river);
•
quality of bank sides (in terms of vegetation and level of erosion);
•
risk of suffering a stomach upset;
•
days in the bathing season that it is unsafe to swim due to poor water quality.
There have so far been five well-known choice experiment water amenity valuation studies,
namely:
a) the freshwater recreation study previously mentioned (Adamowicz et al., 1994);and
b) the study of heterogeneous preferences for marine amenities (Eggert and Olsson,
2004);
c) the study of the marine water improvement amenities in United Kingdom done by
Economics for the Environment Consultancy in 2007.
As obvious from above, these findings should be considered in the phases of searching for the
demand determinants. Moreover, it follows from the text that some of the attributes were used
in modelling demand while others were not. Some of the attributes proved to be unimportant
during the subsequent stages of the qualitative preliminary survey; others, during the
quantitative preliminary survey; some could not be included due to the growing cognitive
requirements on the respondents were they included as well as the need for restricting the
number of choice sets to one (given the limited funding possibilities, limiting the maximum
number of questionnaires).
2.2. Quantitative preliminary survey
Important attributes were identified and their levels set tentatively based on the qualitative
preliminary survey. The first version of the questionnaire to include the CE was then tested on
Lake Macha beaches in June 2007. The description of the attributes and their levels used in
the qualitative preliminary survey constitutes Annex 2; following is a plain list: the attributes
were beach overcrowding, water quality, water facilities, the presence of a medical attendant
on the beach, and the entrance fee. An example site choice card constitutes Annex 3.
Respondents on the beaches did a choice experiment, supplemented with questions whether
any of the attribute failed to play a role to them when deciding on the site to choose, and a
question whether they had missed any attribute when imagining the sites they had been
choosing from (for both the questions, see Annex 4).
This part of the preliminary survey gave rise to several useful findings, based on which one
attribute – the presence of a medical attendant on the beach – was left out of the choice
experiment. It also became evident that people would like to be informed whether any
sporting equipment could be hired on the site, whether there were any pastimes for children
on the beaches, and what the Lake surroundings were like, whether they were any bad
weather trip options, etc. Unfortunately, these attributes could not be included in the choice
experiment due to the choice sample size. The reason is that the minimum required sample
size grows exponentially with the number of attributes. At the same time, the failure to
include these attributes would not pose any threat to the applicability of the choice
experiment. The reason is that if these attributes are not included on any of the cards, they will
not affect the people’s decisions in any decision-making situation. In econometric terms, the
utility from excluded attributes becomes part of the unobservable (stochastic) component in
the random utility theory.
2.2.1. Assigning attribute levels
Once attributes are selected, they need assigning levels that they may attain in the choice
experiment. The basic requisite on the attribute levels is that they be realistic and cover the
entire expected range of the respondents’ preferences (Hensher et al., 2005). At the same
time, the attribute levels should be realistic and credible. Failing that, scenarios may be
rejected by the respondents.
One way to set the attribute levels is to take the existing current attribute level as the default
level, and supplement it with a lower (worse, or less preferred) and a higher (better, or more
preferred) alternative. As a result, this allows estimation of both the increment and decrement
in utility when the attribute level changes. Another way is to set the maximum and minimum
levels of the attribute and the required or otherwise expected or important level of the that
attribute (such as one that is expected as a result of implementation of a certain environmental
policy).
To estimate economic variables (welfare change, willingness to pay, compensating demand,
etc.) in a choice experiment, special attention has to be paid to the levels of the price attribute,
however it is defined. It has to be kept in mind that if the values are set too low, respondents
will always accept them. Too high values, on the contrary, will be rejected. This may cause
the price coefficients to be too low or zero, which will be reflected accordingly in the derived
economic indicators.
These approaches were combined in the Lake Macha choice experiment. The preliminary
survey on the Lake Macha beaches had shown that respondents had protested against neither
too low nor too high levels of attributes. The levels were set as follows.
2.2.1.1. Beach overcrowding
The three initial levels, labelled “high, medium, low” were reduced to two levels. The
attribute was renamed to “beach overcrowding” and could attain the values “YES / NO”. The
levels were reduced for two reasons:
•
the number of attributes and their levels had to be reduced to allow only 1 choice set
(see below for detailed discussion); and
•
overcrowding defined as “YES” or “NO” seemed easier for the respondents to grasp in
the preliminary survey(see below for detailed discussion).
2.2.1.2. Water quality
It was a topic of many discussions how to describe the water quality attribute to the
respondents, so that they valued what the researchers were interested in. Initially the attribute
had five levels. The five-point scale, used by the T.G.M. Research Institute in water quality
assessment, proved to be too complex. The respondents were unable to differentiate between
the slight nuances in water quality. Since the research focuses on assessing quality of water
affected by cyanobacteria (water eutrophication) and measures against cyanobacteria, the
water quality assessment was covered by three levels of the attribute:
Table 2-2: Water quality attribute
Water quality
No algae
No cyanobacteria
LEVEL 1
Swimming convenient
Clean water
Visible algae
No cyanobacteria
LEVEL 2
Swimming convenient
Slightly polluted water
Strong algal occurrence
Cyanobacterial occurrence
LEVEL 3
Swimming inconvenient
Polluted water
Source: Own analysis
As can be seen in Table 2-2, the water quality attribute was characterized in terms of three
categories, namely:
•
algal occurrence,
•
cyanobacterial occurrence, and
•
the possibility of swimming in the water21.
The reasoning behind this attribute scale also arose from the discussions with water
environmentalists, from which it follows that a big difference exists between algal occurrence
and cyanobacterial occurrence in the water. With a certain level of simplification, it can be
summarized that whilst algae in the water cause mostly only lower in-depth water visibility,
the cyanobacteria are toxic and dangerous to human health with their metabolic processes22.
Level three thus corresponds to the water quality in 2004, or the potential water quality in
case no measures to reduce the Lake water eutrophication were taken. Level two then
corresponds to the water quality in Lake Macha since 2005, i.e., water convenient for bathing
(without, or with acceptable cyanobacterial occurrence, and low transparency).
Due to the good graphic presentation of the attribute, it was found out in the pilot surveys that
respondents did not have any cognitive problems understanding the attribute.
This attribute design is furthermore confirmed by experience of Ščasný et al., (2006) who
classified water quality in Lake Macha with a five-level scale of total phosphorus content. The
majority of the respondents stated the same WTP for a one-class and two-class water quality
improvement, i.e., the study found out that 39% of the respondents did not differentiate
between the scope of the improvement. (Ščasný et al., 2006, p. 12).
What we also achieved by this attribute design was that respondents revealed their preferences
of water in-depth visibility by favouring Level 1 to other levels of the attribute while
respondents revealed their preferences for cyanobacteria by favouring Level 1 or Level 2 to
Level 3.
21
The opportunity to swim was included in the attribute description to be sure that even respondents not familiar
with the cyanobacteria issue can differentiate between “dirty water” (i.e., water with serious cyanobacterial
occurrence) and “slightly polluted water” (i.e., water containing cyanobacteria unobservable by the naked eye).
22
Preliminary survey findings also supported this scale setting. For example, asked what they understood as
water quality, two respondents replied, “I don’t make a difference between water quality and appearance:
appearance derives from quality and vice versa, appearance always testifies to water quality.” Another
respondent replied that they did not “see a difference between water quality and appearance”.
In order to assess the effectiveness of measures taken to reduce eutrophication in Lake Macha,
the preference of Level 1 over Levels 2 or 3 is particularly important (estimates of these
values are the default value for estimating benefits in Chapter 4: CB analysis).
2.2.1.3. Beach facilities
The three initial levels of the attribute “Presence of sanitary facilities and refreshments” were
reduced to two, namely “YES” or “NO”. The reasons were the need to make only one choice
set and the requirement on the maximum possible sample size (given the limited data
collection capacity) and the maximum number of choices presented to each respondent (see
below for discussion, i.e., factorial design and choice sets).
2.2.1.4. Entrance fee
As already mentioned, special attention needs to be paid to the price attribute levels in order
to estimate the economic variables in applying the CE. Here, the price was defined as the
beach entrance fee per person per day. The initial entrance fee rates were tentatively set at
CZK 10, 50, 100, and 200 in the preliminary survey. The preliminary survey showed that
CZK 10 was too low a value (respondents did not respond to the fee level, or excluded the fee
from their decision grounds). In contrast, the CZK 200 fee proved to be too much – it resulted
in the card being rejected by the respondents. In order to cover the entire range of expected
preferences while reducing the number of attribute levels to three, we tested fee levels of CZK
40, 80, and 150 in the next round. These settings of the fee levels proved to be adequate. The
entrance fee was neither considered unimportant nor excessively protested against by most of
the respondents. The description of the final version of the attributes and their levels used in
the final survey constitutes Annex 5.
2.2.2 The status quo choice; scenario setting (design)
One of the most important moments in the choice experiment study design is the description
of the choice scenario, that is, a situation which is challenging the respondent when deciding
among offered choice alternatives. In the quantitative preliminary survey, the respondents
were offered the following scenario:
“Please imagine a situation where you are shown an information leaflet in which these 5
characteristics are used to describe different water bodies for recreation and you may choose
among them. Which of these 2 sites would you choose for your trip today? Would you choose
Site 1 or Site 2 or would you stay at home?”
This scenario proved to be inappropriate. The reason was that the status quo alternative was
presented as “staying at home”. People protested against such a scenario, because in some
situations they were forced to choose between a site that they felt strongly opposed to for
some reason (e.g., the water was polluted in both alternatives) and staying at home (while in
fact they meant to go on holiday). Some respondents therefore replied, “What would I be
doing at home? I’m leaving at any cost, albeit to a site less preferable than where I am now
(Lake Macha).”
Given these findings, the scenario was reformulated as follows:
“Please imagine a situation where you are shown an information leaflet in which these 4
characteristics are used to describe different water bodies and reservoirs in the Czech
Republic and you may choose among them. Which of these 2 alternatives would you choose
for your beach stay today? Would you choose Alternative 1 or Alternative 2 or would you
choose neither of them? Remember that you can only choose from these alternatives.”
This scenario proved to be acceptable and was therefore used in the main data collection.
People were not forced to choose any of the recreational alternatives and were also left room
to choose neither and yet not stay at home. This situation is much better at capturing a real life
situation, where one is not confined to choosing from two recreational sites, but when neither
suits one, one simply searches on until one finds a suitable site.
2.3. The choice experiment design
Having identified the recreational demand attributes and set and tested their levels, the next
stage in preparing a choice experiment study is to compile the attributes into products (cards)
comprising these attributes, and subsequently, to compile these products in so-called choice
sets. Afterwards, the designer has to integrate the choice experiment needs in the final
questionnaire structure, define the study area and the target population, and the appropriate
sampling strategy has to be selected and tested. All these stages are discussed in this
subchapter.
2.3.1. Making the experimental design
Having identified the recreational demand attributes and set and tested their levels, the
attributes have to be compiled into “products” (e.g., Site 1 is characterised with clean water,
low overcrowding, good facilities, etc.). Statistical design theory has to be employed here to
combine the attribute levels in several alternative scenarios, or several alternative products.
The result is the generation of certain profiles (products), which are then shown to the
respondents to choose from. If all possible products are generated by combining the attributes,
the result is a so-called complete factorial design. Its advantage is that it allows a complete
estimate of the impact of changes in the attribute levels on the study subject’s decisions and
utility.
However, the practical implementation of this design would require the respondents to
compare several dozen alternatives. In the case of the final version of the questionnaire used
at Lake Macha, this would entail asking respondents to compare 2 x 3 x 2 x 3 = 36 products.
In practice, it is therefore desirable to reduce the number of alternatives (products) somehow.
This has led to the development of statistical techniques allowing a certain reduction in the
number of alternatives while retaining certain required properties of the design.
These reduced designs are typically referred to as partial factorial designs (Louviere et al.,
2000), or experimental designs. The number of products (profiles, alternatives) is reduced
using these techniques. Experimental design was originally developed in experimental and
agricultural research, and constitutes an effective way of selecting a certain subset of
combinations within a complete factorial design for subsequent use in an experiment. The
orthogonal main-effects plan was used in the choice experiment at Lake Macha. The main
effects are defined as responses generated when moving from one level of a given attribute to
the next, whilst holding the level of all other attributes constant. When the orthogonal maineffects plan is used, the utility function is estimated based on the assumption that any higher
order or interaction terms are insignificant and that the consumer preferences depend upon
individual attribute levels alone and are not influenced by the combination of levels offered
across different attributes.
The advantage of the orthogonal main-effects plan is that each variable has zero correlation to
any other variable. This property permits identification of the impact of change in any of the
attributes on the respondent’s choice in the subsequent data analysis. The disadvantage of the
design is that it does not allow measurement of so-called interactions. It is based on the
assumption that the effect of each attribute is independent of the value attained by any other
attribute. Interaction effects can thus not be accounted for (revealed) using this design.
In the choice experiment conducted at Lake Macha, a possible interaction effect may be the
fact that valuation of water cleanness may depend on beach facilities, for example: people
may expected that they will spend more time on an equipped beach, meaning spending more
time in the water as well, thus caring more for the water quality. That is why the value of
water quality improvement with a simultaneous improvement in beach facilities may be
greater than the mere sum of the two said improvements as such.
Another example of a possible interaction effect is that people assess water quality, among
other things, by how many people are bathing in it. If beaches are “overcrowded”, people take
this overcrowding as an indicator of water quality, which is why they value overcrowding and
water quality in a similar fashion – their impacts on respondents’ utility from recreation
correlate.
There are also partial factorial designs that allow measurement of interaction effects too, but
no such design was applied at Lake Macha. Their application is more suitable where one
intends to confirm or reject a hypothesis on a certain interaction. Literature says that in a welldone choice experiment, over 80% of respondents’ behaviour can be explained by main
effects (i.e., without interactions), that is, even using an orthogonal factorial design, for
instance (Bateman et al., 2002).
The majority of contemporary statistical packets allows generation of partial factorial designs.
SPSS software was used to generate the design for the Lake Macha preference survey. The
experimental design specifies attribute levels as figures. These have to be assigned specific
values (meanings, interpretations) subsequently. The experimental design generated by SPSS
for the Lake Macha case is shown in Annex 6.
The partial factorial design is generated four times in Annex 6. The reason is the subsequent
generation of the choice sets – see the next subchapter – and the problem of so-called
dominant alternatives (see below too), which was manifested in the preliminary survey stage.
As mentioned above when discussing the Beach Facilities attribute, the numbers of attributes
and their levels had to be reduced in order to be able to make only one combination of choice
sets, which could be shown to each respondent. If that had failed, the requirements on the
choice sample would increase substantially. The actual project funds and time allowance
sufficed for some 300 interviews (eventually, 333 questionnaires were collected and 331 were
valid). If it had become necessary to develop more than one combination of choice sets due to
the numbers of attributes and their levels, each of those combinations would have had to be
covered by a required number of interviews (i.e., if K is the number of choice set
combinations and 300 is the number of interviews required to cover one choice set
combination, then K x 300 is the number of interviews required to cover all the choice set
combinations with a required number of interviews). The partial factorial design which had to
be made in the quantitative preliminary survey with the choice cards and levels shown in
Annex 2 consists of two combinations of eight choice sets each. Such a design would double
the requirements for the number of interviews.
The combination of 9 choice sets (with the option to choose Sites 1 or 2 or neither) made for
the final data collection was tested at Lake Macha again in early July 2007. Each respondent
was shown 9 choice sets (or, made nine subsequent choices between products). Choosing in
nine choice sets posed no problem to the respondents since the cards were clearly organised
and the decision-making problem was not too complex (four attributes, out of which two had
two levels, and two had three levels). As Bateman et al. (2002, p. 265) say, “The fewer the
number of attributes and levels, the more the number of choice tasks that can be allotted to
each person. In general, respondents should not be asked to undertake tasks that are too
difficult or complex because they may not perform them reliably and/or may resort to
shortcuts or haphazard answers”. Various analyses concerning the number of choice sets per
respondent can be found in various studies. For instance, Carlsson and Martinsson (2007, p.1)
state, “Our results indicate that neither the number of choice sets nor the design of the first
choice set has a significant impact on estimated marginal willingness to pay….” On the other
hand, Smith and Desvouges (1987) found that “ranking sets of between 4 to 6 elements yield
the most consistent answers, with more than 8 becoming too complex for most respondents to
handle”. (Smith and Desvouges, 1987 in Bateman et al., 2002, p. 265) The pilot surveys at
Lake Macha showed that respondents were able to cope with nine choice triplets each.
In some cases, a mechanical application of the experimental design may generate profiles
(products) that are not very credible. As a result, respondents may not take the interviews very
seriously. An example for the Lake Macha experiment might be a well-equipped, not
overcrowded beach, clean water and a low entrance fee. The alternatives offered to
respondents have to be credible (viable) and realistic. Respecting that, choice sets need to be
made based on generated factorial designs.
2.3.2. Making the choice sets
Having generated a partial factorial design, one must consider whether the entire set of
alternatives can be shown to respondents to decide. One does come across studies where
respondents were presented with up to 26 pair comparisons, but the figure does not exceed 10
pair comparisons in most papers.
When respondents are presented with a higher number of subsequent decisions, the question
arises whether they can grasp the decision-making problem properly and whether their
responses will be consistent. In part, that depends on the nature of the decision-making
problem (whether it concerns public transport or the extent of wildlife preservation in Alaska,
for example – with the latter problem unlikely to be very familiar or even of much interest to
them). A hypothesis may even be formulated that the more attributes and attribute levels a
decision-making problem has, the fewer decisions each respondent should be asked to make.
In the Lake Macha survey, the numbers of attribute levels were reduced until a single
combination of choice sets with an acceptable number of decisions could be made.
Respondents chose between sites nine times in a sequence. Had but one level been added to
any of the attributes, it would result in 16 choice sets (instead of the eventual 9), which would
have been too much to present the entire choice set to a respondent. That was the case of the
preliminary survey, when the entire set of alternatives corresponding to the factorial design
was split into two blocks (choice set combinations) with eight decisions each, and each
respondent was then presented either one of the two blocks. As mentioned, this solution, if it
had been applied in the actual data collection, would have had impacted on the total choice
sample size.
Moreover, a mechanical application of the experimental design may result in a so-called
dominant alternative problem, where one of the alternatives the respondent is to choose
between is better than the other one in all the attributes. In such a case, the respondent in fact
does not face a choice among Sites 1 and 2 and the status quo, but only between the dominant
alternative and the status quo. This problem has to be resolved using one of the following
methods: assign a different meaning to the numeric values shown in the partial factorial
design; change the attribute levels as such; or change the pairing of the factorial designs in the
pair choices for respondents. The last resolution method was used in making the choice cards
for Lake Macha. The choice sets used in the final survey are shown in Annex 7.
2.3.3. The final questionnaire structure
After the qualitative research, literature review and several in-depth interviews made in
Prague focusing on waterside recreation, a pilot version of the questionnaire was prepared.
The pre-test of the first version of the questionnaire was carried out on the Lake Macha
beaches in June 2007. Based on the findings (discussed above), the questionnaire was revised
and the final version was prepared. Then, two pilot surveys (about 10 respondents per pilot)
were carried out in late June in order to improve and finalize the questionnaire and to test the
sampling strategy in the terrain (see below). The final survey was carried out in July and
August. The questionnaire was designed and pre-tested for ease of answering. The
questionnaire was proposed to allow interviews to be completed in 15 minutes in order to
avoid respondent fatigue.
The final version of the questionnaire had the following structure. It was divided into six
thematic sections. The first section was devoted to the recognition of the repetitions in lake
visits and the frequency of the individual’s visits in the 2006 and 2007seasons. The second
part of the first section focused on the current trip (for trips longer than one day). The third
part of the first section dealt with the current trip (for one-day trips). The second section of the
questionnaire focused on the description of the current trip. Respondents were asked how
many people were on the trip together, whether they swam during their beach stay, etc. The
third part of the questionnaire focused on the choice experiment, which was discussed above.
The fourth section consisted of attitude questions regarding water quality, the information that
the respondents had about cyanobacteria in Lake Macha, about the reasons for their
occurrence, and further questions regarding water quality improvement sources and some
other attitude questions (see questionnaire in Annex 8). In the fifth section, socio-economic
information about respondent was inquired, e.g., education, employment status, municipality
and region of residence, the number of persons in the respondent’s household (and children
within those), the respondent’s year of birth and net monthly income. The last section
contained debriefing questions for the respondent and also for the interviewer, e.g., how
seriously had the respondent taken the interview, whether they had been influenced by
anyone, where and in what weather the interview had taken place, etc.
2.3.4. The study population and the study area
In order to apply a choice experiment, it is necessary to obtain relevant information from the
study population. In the Lake Macha case, we were interested in the preferences of the visitors
to Lake Macha, and for this reason the study population was the population of the visitors to
the paid beaches, which amounts approximately to 5/6 of all visitors to Lake Macha;
foreigners and children were excluded from the target population.
The rest of the visitors approach Lake Macha from many other diffused points surrounding it.
They possibly have different preferences (they may be, e.g., citizens of Doksy), but
considering the maximum possible sample size (around 300 questionnaires), we would have
approximately 50 additional questionnaires in the sample with perceptible time and monetary
costs. For these reasons, we finally decided to focus on the population of visitors of the four
paid beaches visited by most of the visitors to Lake Macha. The collection took place directly
on Lake Macha beaches, meaning it was an on-site collection. Localization of these four paid
beaches is shown in Chart 2-1 below.
Chart 2-1: Lake Macha study area
Notes:
1- Doksy Main Beach
2 - Staré Splavy Aquapark
3 - Borný Camp Site
4 – Klůček Beach
Source: Mapy.cz, online: http://www.mapy.cz/#mm=ZP@x=133599616@y=137709312@z=13
2.3.5. Sampling Strategy
During the first monitoring of the recreation site, Lake Macha was charted and the sampling
strategy was considered. Data were collected in a so-called selective survey. This is a method
of data collection in which data are collected in a certain standardised way from a specified
target group (here, visitors to the paid beaches by Lake Macha). In this case, the method of
standardisation was the above discussed questionnaire, made for the purposes of this survey.
Among other things, a selective survey is characterised by the data being collected from a
specified number of individuals who are part of a larger (surveyed) known target population
(here, the population of visitors to Lake Macha).
An important aspect of conducting a selective survey is the selection of units subsequently
included in the statistical survey. This aspect is referred to as the selection plan (Hendl, 2006).
The best selection plan for a statistical survey is the so-called random selection, characterised
by each unit of the population having a known likelihood of being included in the selection.
Random selection is expedient chiefly because it minimises the likelihood of a systemic error.
In this case, however, random selection could not be made as the numbers of visitors differ
among the beaches on which the collection took place. Therefore, the probability of each
visitor to a certain beach of being included in the selection was unknown.
Since a simple random selection could not be made on the site, the strategy chosen was the
Stratified Random Selection method. This selection method may be applied where the
population comprises heterogeneous subpopulations. In such a case, the population can be
divided into these subpopulations and then conduct a simple random selection for each group.
The results for all the groups then make up the selection. Following this data collection logic,
the population of visitors to Lake Macha could be divided into subpopulations of visitors to
the individual paid beaches. These beaches are characterised by differing visit rates (visitor
numbers per day) and cannot be expected to be visited by people with identical preferences
(the beaches differ in their nature). Therefore, 4 different subpopulations can be assumed to
exist by Lake Macha. Inside each of these populations, a method for making a simple random
selection can be identified. The researcher thus faces the question of how to ensure the
precondition of each visitor to a certain beach being equally likely to be included in the
selection (i.e., be interviewed).
In order to ensure this precondition, each beach was divided into several sectors (of roughly
equal sizes). Interviewers were supposed to choose the sector for making interviews
randomly: by tossing a die (the number on the die was the number of the sector on a map in
which they were to choose a respondent). They were supposed to choose a specific respondent
in their sector by counting every fifth or tenth (whether every fifth or tenth was to be specified
on the morning of the interview day depending on the presence of visitors) from the farthest
corner of the sector (i.e., the place they would walk to when entering the sector).
The Main Beach had 12 sectors. A road divides the paid beach area into two parts: the sandy
beach towards the lake, and the area of kiosks, restaurants, sporting grounds, changing
cubicles, etc., across the road from the sandy beach. Each of these areas was divided into six
sectors. The interviewers were given maps of their beaches with sectors marked and
numbered, dice to toss, and interviewing instructions (see Annex 7).
This data collection method was tested on several questionnaires during the preliminary
survey. It was found to be satisfactory. However, the weather was bad (rainy, cold) on the
first collection day, and there were few people on the beaches. There was hardly anyone on
the smallest beach, the Klůček. Given the expanse of the beaches, the interviewers would
spend a lot of time walking from one end to the other. Therefore, we decided to abandon this
method of guaranteeing randomness and exploit the back-up field sampling strategy: so-called
random walk.
This respondent selection method consisted in the interviewers walking up and down the
beach and intercepted every fifth person they counted with the question whether they were
willing to complete a questionnaire with them. If approved, an interview was made; if not, a
refusal was noted on the refusal card. This collection method proved to be satisfactory and
less time consuming. During the second data collection, we followed the same strategy to
guarantee the randomness of respondent choice.
2.8. The data description
As mentioned, the survey on the Lake Macha beaches was carried out from July to August
2007. Respondents were intercepted on each of the four beaches randomly and interviewed by
trained interviewers face to face. The interviewing started at 10 am every day. The survey
resulted in a total of 333 completed questionnaires. Two of them were excluded from the
analysis because the interviewed respondents were employees at the beaches and not
vacationists. Their responses were outstandingly different from the others. In this sense, they
were evaluated as extreme values23. The interviewed individuals had immediate experience
with the water quality and other qualities of the beaches.
23
Statistical tests for treating outlying values exist. Despite the fact that statistical literature mentions that their
usefulness is disputable, three of them were used in the data analysis so that the “(statistically) outlying values”
can be excluded from the sample. These included the rule saying that we can exclude the value being outside the
interval: arithmetic mean +/- 3 x standard deviation (s), both statistics counted without the suspicious value
(Hendl, 2006). Then also the “interquartil range” rule (i.e., the median 50% values in the set) – if the vale is
more than 3/2Q away from the lower or upper quartil, it can be regarded as outlying. And finally also the so
The data collection was done in two waves:
•
Saturday 21 July – Sunday 22 July; and
•
Tuesday 14 August – Wednesday 15 August.
The absolute and relative numbers of questionnaires per collection is shown in Table 2-3
below.
Table 2-3: Absolute and relative numbers of questionnaires per collection
Collection
Absolute number
questionnaires
of
Collection 1
156
47%
Collection 2
177
53%
Total
333
100%
Percentage
Source: Own analysis
A total of 47% of the questionnaires was collected in the first wave; 53% in the second wave.
The first collection was made by seven interviewers; the other one, with the help of six
interviewers. Part of the interviewers were students participating in the project; part were paid
temporaries. All the interviewers had been trained and underwent a demonstration interview.
The data were retyped into the data matrix. Then they were checked for errors using the
descriptive statistics. These checked data are used for the following sample characteristics.
The number of interviews during the days is shown in the following chart. As can be seen,
most of the interviews were done between 11 am 4 pm.
called “median absolute deviation (MAD)” was used, that is, if a certain value’s distance from the median is
more than 5xMAD, then it is an “outlying observation”.
Chart 2-2: Numbers of interviews during the day
Absolute Frequency
Numbers of interviews during the day
80
70
60
50
40
30
20
10
0
P
10
11
12
13
14
15
16
17
18
19
Hour
Source: Own analysis
The structure of the interviews by the particular beaches is as follows: 46.2% interviews were
done on the main beach called Doksy, which is the most visited beach by the Lake; 23.4% on
the Staré Splavy beach and aqua park, which is also very busy; approximately the same
number of interviews were done in Borný campsite and beach (21 %). Only 9.3% interviews
were done at Klůček, which is the smallest and also the least visited beach by the Lake.
The structure of the interviews by the particular beaches is also shown in Chart 2-3 below.
Chart 2-3: Sample structure by the beach
Absolute Frequency
Name of the beach
180
160
140
120
100
80
60
40
20
0
Doksy Main
Beach
Klůček Beach
Staré Splavy
Aquapark
Source: Own analysis
Borný Camp Site
The cloud cover and weather during interviews are shown in Chart 2-4 below. Most of the
interviews were made in dry weather– 306 interviews (93.3 %). Only 22 interviews (6.7 %)
were made in rainy weather. The structure of the interviews by the weather conditions is
shown in Chart 2-5 below.
Chart 2-4: Weather during interviews
Weather during interviews
No precipitation
Drizzle
Shower
Rain
Fog
Storm
Hail
Source: Own analysis
The following chart shows the interview structure by cloud cover. Most of the interviews
were made in somewhat cloudy weather (37.5%). The numbers of interviews made in cloudy,
overcast and cloudless weather were approximately identical: about 20% for each type of
cloud cover.
Chart 2-5: Cloud cover during interviews
Cloud cover during interviews
23%
19%
Cloudless
Somewhat cloudy
Cloudy
Overcast
20%
38%
Source: Own analysis
There were a mere 60 refusals to undergo the interview; within that, 20 respondents explained
their refusal with not having the time; 20 said they did not feel like it; 11 were asleep; and 7
stated no reason. The random selection also produced 26 children and 65 foreigners, who
were not the target population, so no interviews were made with them. Six respondents said
they had already completed the questionnaire with a different interviewer. Sixty-six per cent
of non-responses were women, while only 44% were men. The number of non-responses thus
amounted to 18%.
Chart 2-6: Numbers of refusals
Numbers of refusals
No reason
No time
Not feel like it
Asleep
Nudist
Child
Foreigner
Other reasons
Source: Own analysis
2.8.1. Sample characteristics
Unfortunately, to the author’s knowledge, no sociologic study exists that would provide
information on visitors to Lake Macha beaches and hence it is not possible to make any
comparison of the sample at Lake Macha in this case study with the target population.
The demand analysis and the CB analysis are this based on the assumption that the stratified
random collection used as the sampling strategy indeed gave the researchers a random sample
of respondents, and for this reason, we assume that the sample is also representative. Also, the
low number of non-responses (only 18%) enables us to make this assumption. For these
reasons, the following sample characteristics can be considered to be the general approximate
characteristics of the target population – i.e., summer visitors to the four paid beaches by Lake
Macha, Czech residents.
The age structure of the sample is shown in Chart 2-7 below.
Chart 2-7: The age structure of the sample
Respondents’ age
0-20
20-30
30-40
40-50
50-60
60-70
70-80
Source: Own analysis
The average age (37 years) is nearly the same as the median age (35 years). The average age
of the men in the sample is 36 years, while the average age of the women is 38 years. The
youngest respondent was 16 years old (only one from the sample under 18 years of age); the
oldest respondent was 74 years old. The other sample statistics are shown in Table 2-4.
Table 2-4: Respondents’ age – basic momentous characteristics
Age
Number of people in Numer of children in
the household
the households
Average
37.09
2.79
0.34
Median
35.00
3.00
0.00
Modus
Standartd
deviation
Minimum
Maximum
23.00
13.58
4.00
5.56
0.00
5.30
16.00
74.00
1.00
8.00
0.00
4.00
Source: Own analysis
70% of the sample was not at Lake Macha for the first time, and nearly the same percentage
of the visitors – namely, 71%, were visiting Lake Macha on a multiple-day trip.
Chart 2-8: Repetitiveness of visits to Lake Macha
Repetitiveness of the visit
29%
Not first time
First time
71%
Source: Own analysis
Chart 2-9: Type of current trip
Current trip
29%
One-day trip
Multiple-day trip
71%
Source: Own analysis
The relative distribution of the sample by the length of the trip (one-day vs. multiple-day
trips) by the respondents’ place of residence is shown in Charts 2-10 and 2-11 below. It
follows from the charts that more visitors on one-day trips came to Lake Macha from nearby
areas, while people on multiple-day trips came from more dispersed places across the country.
The highest percentage came to Lake Macha from Central Bohemia, Liberec Region, Prague,
and Ústí nad Labem Region. The other regions were represented only a little or not at all.
Chart 2-10: One-day trips by place of residence
One-day trips by place of residence
Prague
Central Bohemian Region
South Bohemian Region
Plzeň Region
Ústí nad Labem Region
Liberec Region
Hradec Králové Region
Vysočina Region
South Moravian Region
Source: Own analysis
The structure of the places of residence for multiple-day trips is different and more diverse.
The highest percentage came to Lake Macha from Prague, Ústí nad Labem, Central
Bohemian, and Liberec Regions. The other regions were also represented in the sample,
among them mostly Hradec Králové Region. The rest of the regions was also represented with
approximately the same percentage.
Chart 2-11: Multiple-day trips by place of residence
Multiple-day trips by place of residence
Prague
Central Bohemian Region
South Bohemian Region
Plzeň Region
Karlovy Vary Region
Ústí nad Labem Region
Liberec Region
Hradec Králové Region
Pardubice Region
Vysočina Region
South Moravian Region
Olomouc Region
Zlín Region
Moravian-Silesian Region
Source: Own analysis
The education structure is shown in Chart 2-12 below. Most of the respondents had a
secondary school education, while almost ¼ were skilled workers. 1/5 were university
educated people. Less than 1/10 of the respondents stated elementary education.
Chart 2-12: Education structure
Highest attained education
Not stated
Elementary education
Skilled workers
secondary school
education (with leaving
examination)/advanced
vocational training
University education
Source: Own analysis
The economic status of the sample is as follows: Most of the sample (i.e., less than 54% of the
respondents) were full-time employees; 12% were self-employed. Students and pensioners
alike made up 12% each. The other respondent groups are represented by under 4%.
Chart 2-13: Economic status of the sample
Economic Status
Full-time empoyer
Part-time empoyer
Self-employed, entrepreneur
Unemployed
Student, Apprentice
Pensioner
In household
Maternity leave
Working student / Working
pensioner
Source: Own analysis
The income distribution in the sample population is as follows (Chart 2-14):
Chart 2-14: Respondents’ net monthly income
Respondents’ net income
70
60
40
30
20
10
0
0
-5
5.
50 .50
0
1
-7
7.
00 .00
0
1
-8
8.
.5
50
0
1
-1 0
10
0
.5
01 .50
-1 0
13
3
.0
01 .00
-1 0
15
5.
.5
50
01
0
–
18
18
.0
01 .00
-2 0
24
4
.0
01 .00
-3 0
5.
35
00
W
.
0
0
ith
01
ou
a
to
ví
ce
w
n
in
co
W
m
ill
e
no
tr
ep
ly
Absolute frequency
50
Source: Own analysis
The average net income of an individual in the sample was CZK 14,176 (2007 CZK). The
median income was CZK 14,250. The highest proportion of the sample were respondents with
an income of CZK 18 to 24 thousand. The second largest income group was that with CZK
10,501 – 13,000. The lowest income category was stated by almost 7% of the respondents; on
the other hand, the highest income category was stated by almost 2% of the respondents. The
category “without own income” was stated by almost 5% of the individuals.
The results are summarized in Table 2-5 below:
Table 2-5: Respondents’ net monthly income
Average Median
Modus
Standartd
deviation
Net Income
14,176
14,250
21,000
Source: Own analysis
7,848
2.8.2. Sample characteristics: attitudes and knowledgeability
What is important for subsequent characteristics concerning water quality is whether the
respondents are in contact with the water during their beach stay. The following table shows
that 87% of the respondents swam in Lake Macha during their stay, while only 13% of the
interviewees did not (see Chart 2-15).
Chart 2-15: Swimming in Lake Macha
Swimming in Lake Macha
Do not swim
Swim
Source: Own analysis
People who did not swim largely comprised adults who were only supervising bathing
children and people spending time by the refreshment kiosks, reading, and looking around.
Respondents were asked to rate bathing water quality at the beach where they were sampled
on a 5-point scale ranging from very good to very poor (1 to 5 on the scale). Chart 2-16 shows
that most of the visitors considered the water in Lake Macha to be slightly polluted (54.3%),
while only 20% considered the water to be clean, and only 15% thought the water was
polluted. 1.2% thought the water was very clean, and only 2.4% considered the Lake water to
be seriously polluted. The analysis also revealed that there were not any significant
differences between average men’s and women’s ratings of the water quality.
Chart 2-16: Rating of the water quality in Lake Macha
Rating of the water quality by respondents
200
180
160
140
120
100
80
60
40
20
0
I can´t judge
Very clear
water
Clear water
Slightly
Polluted water
Seriously
polluted water
polluted water
Source: Own analysis
In Chart 2-17 below, the responses for very clean and clean water were grouped in the clean
water category; the same was done for polluted and seriously polluted water. The figure
shows prevailing water quality rating by respondents as slightly polluted, which to a high
degree corresponded to the real water quality.
On the other hand, there is an imperfect match between the perceived water quality and the
biological monitoring results (see Table 1-2 in Chapter 1). There may be several reasons for
this: people may judge water quality using a wide variety of cues, whilst biological
monitoring results are based on a very limited range of criteria; also, the water quality
changes during the season, so it is dynamic, while the water quality reports (listed in Table 12 in Chapter 1) are made for a certain day (they are static).
The people’s inability to differentiate water quality on a five-point scale corresponds, to some
extent, with the findings of Ščasný et al. (2006, p. 12; see above), who classified water quality
in Lake Macha using a five-level scale and found out that 39% of the respondents did not
differentiate between the scope of the improvements. That is to say, people are mostly able to
differentiate between clean, slightly polluted and dirty water but they are not able to
differentiate between very clean and clean water on the one hand, and polluted and seriously
polluted water on the other hand. Despite the fact that the respondents in Ščasný et al. (2006,
p. 12) expressed their water quality ratings as, e.g., seriously polluted water, the WTP for
improvements from seriously polluted water was the same as from the polluted water level.
Chart 2-17: Rating of the water quality in the Lake
Rating of the water quality in Lake Macha
60
50
Percentage
40
30
20
10
0
Cannot judge
Clean water
Slightly polluted
water
Polluted water
Source: Own analysis
Despite the serious problems with the cyanobacterial occurrence in Lake Macha, there were
still 25% of people who had not heard of the cyanobacteria in the Lake, as shown in Chart 218:
Chart 2-18: Percentage of the sample who had heard of the cyanobacteria in
Lake Macha
Have you heard of the cyanobacteria in Lake
Macha?
No
Yes
Source: Own analysis
Respondents were also asked about who they thought should most contribute to the funding
for improving the Lake water quality. The attitudes varied considerably; their distribution is
shown in Chart 2-19 below.
Chart 2-19: Financing of water quality improvements
Funding improvements in Lake Macha water quality
120.00
Absolute frequency
100.00
80.00
60.00
40.00
20.00
0.00
Holiday
makers
Inhabitants of Nearby farming
nearby villages and industrial
operations
Czech
Republic’s
population
Beach and
campsite
operators
Others
(specify)
Source: Own analysis
As follows from the chart, some 20% of the respondents mentioned nearby farming and
industrial operations; about 17% mentioned the entire Czech population; another 17% thought
the beach and campsite operators should contribute the most. The other options (i.e., holiday
makers and residents of the villages surrounding Lake Macha) made up less than 9%. It was
quite interesting that almost 30% of the respondents gave answers different from those offered
in the questionnaire. The prevailing answers among those 30% were the following: „Doksy
municipality“, „municipalities“, „the town “, „the state“, or a combination of those.
Respondents were also asked during the interview to express their agreement with the
following statements read out by the interviewer:
•
The beaches are too overcrowded;
•
The toilets and showers on the beaches are in perfect shape;
•
before I visit Lake Macha beaches, I try to find out about the water quality in the Lake.
The answers are shown in Chart 2-20 below.
Chart 2-20: Agreement with statements
Absolute frequency
Statement agreement
180
160
140
120
100
80
60
40
20
0
Cannot judge
Certainly agree
Rather agree
Rather disagree
Certainly disagree
Beaches are too overcrowded
Toilets and showers on the beaches are in perfect condition
Before I visit Lake Macha beaches, I find information about the water quality
there
Source: Own analysis
It arose from the interviews that:
•
most of the people disagreed with the statement that the beaches were overcrowded
(this corresponded with the low WTP for overcrowded beaches);
•
people were almost equally split between those who were satisfied and those who
were not satisfied with the quality of the toilets and showers on the beaches;
•
most of the people did not try to find out information about the water quality in Lake
Macha before they went to visit it.
These findings are even more apparent if the categories “certainly agree” and “rather agree”
and “rather disagree” and “certainly disagree” are joined with categories “agree” and
“disagree”. The following chart shows the distribution of agreement with the statements by
the newly established categories.
Chart 2-21: Agreement with statements
Statement agreement
250
200
150
100
50
0
Cannot judge
Agree
Disagree
Beaches are too overcrowded
Toilets and showers on the beaches are in perfect condition
Before I visit Lake Macha beaches, I find information about the water quality
there
Source: Own analysis
2.9. Summary
When applying the choice experiment method, the correct choice of the demand determinants
(attributes in choice experiment terms) and their levels is of special importance for a valid
design of the choice experiment and its subsequent application. For this reason, the choice
experiment design procedure was in the centre of attention in this chapter.
First, the qualitative preliminary survey was discussed; the relevant literature review focusing
on attributes used, one group interview, and several one-to-one interviews were conducted
based on it. The results of this preliminary survey stage gave rise to attributes relevant to
summer waterside recreation demand modelling, and then a first version of the choice
experiment study and of the questionnaire were compiled.
These were subsequently tested on-site during the quantitative preliminary survey. Based on
the results of this preliminary survey stage, some of the attributes were left out, and the levels
of others were adjusted. The chapter proceeded with the design of choice experiment cards
and choice sets. Some of the problems related to their making, chiefly with respect to the
cognitive requirements on respondent decisions during data collection, were discussed as
well. A brief description of the final questionnaire version was also made in the chapter. The
chapter proceeded with the specification of the study population and discussion of the
sampling strategy. As a sampling strategy for the final survey, the random walk was applied
as a method of ensuring randomness of collection of the statistical units.
The second part of the chapter was devoted to the sample description. The survey resulted in a
total of 333 completed questionnaires; two of them were set aside from the sample because
they did not satisfy the requirements for the final population in question. Selected sample
characteristics were then described and some of them were discussed in more detail.
Unfortunately, to the author’s knowledge, no sociologic study on Lake Macha visitors exists,
and hence it was not possible to make any comparison of the sample at Lake Macha in this
case study with the target population. The sample characteristics did not show any suspicious
results. For example, people on one-day trips came to the lake from destinations nearer by
than those on multiple-day trips; most of the respondents were in the age group of 20 to 40
with the average age being 37 years; almost ¾ of the visitors were not at the lake for the first
time; most of the people had secondary school education, while only 20% were university
educated; most of the people were full-time employees, while the second most numerous
group of the people were students; the respondents’ average income was CZK 14 thousand;
87% of the respondents swam in Lake Macha during their visits. A rather pleasant finding is
that more than 50% of the respondents rated the current water quality as “Slightly polluted
water”, which corresponded to the reality (according to WHO 2001 ratings). Quite
surprisingly, most of the people disagreed with the statement that they found out information
about water quality in Lake Macha before they went to visit it. Most of the people also
disagreed with the statement that the beaches were overcrowded, which corresponded with the
low WTP for overcrowded beaches (see Chapter 3).
3. Data analysis: theoretical discussion
The main objective of this chapter is the core of the economic (and welfare measure) analysis
– the choice experiment data analysis. First, the economic background of the discrete choice
models is explained. Then, various types of discrete choice models are compared and
discussed with respect to their ability to model discrete choice data and with respect to the
welfare measure values they provide the analyst with; their limitations and conveniences are
mentioned. Furthermore, the discussed models are deployed on the Lake Macha choice
experiment data. In this subchapter, the models are discussed with respect to their fitness,
estimated values, practical results, their limitations, conveniences and advantages in terms of
information they provide the analyst with. Welfare measures based on the model estimates are
also presented in this chapter.
3.1. Economic grounding for the discrete choice models
The choice experiment method is one of the non-market methods for the economic valuation
of natural resources. Discrete choice models are used for modelling of the choice experiment
data. The research in this area in economics began in the 1970s. Since then, both the
multinomial logit and probit models have been widely used in transportation, economics,
marketing and many other areas to study both revealed and stated preference data. Recently,
the research in this field has paid special attention to the error term of the models in an effort
to solve some of the problems of the discrete choice models and to make them more flexible.
Discrete choice modelling in economic theory complies with Lancaster’s new approach to the
individual utility maximization problem in consumer theory (Lancaster, 1966) and with the
random utility theory (McFadden, 1974). According to Lancaster’s approach to consumer
theory, consumers derive their utility not from the product as such, but from the
characteristics/attributes by which the product can be described.
The random utility theory (Manski, 1977; Phaneuf, 2005) then postulates that utility is a latent
construct that exists in the consumer’s mind and cannot be observed directly. It further
assumes that this latent utility can be partitioned into two components: a systematic or
representative utility (V) and a random, unexplainable component (ε). This random
component arises both because of the randomness in the individuals’ preferences and because
the attributes do not cover all of the individuals’ preferences. If we consider random sampling
of the respondents, then ε can be interpreted as a random term. Because of this random
component, the problem is inherently stochastic and an individual’s preferences cannot be
understood perfectly. It naturally leads to formulation of expressions for probability of choice.
Based on repeated observations of choices, one can examine how the levels of various
attributes affect the probability of choice. Furthermore, the random utility theory assumes a
utility maximization principle, i.e., if an individual chooses one alternative over another, then
the utility from the chosen alternative is greater than that from the unselected alternative.
3.2. Theoretical discussion of discrete choice models
The obvious objective in discrete choice modelling is to analyze the individual’s choice in
relation to the characteristics (attributes) of the product (e.g., choice of a transportation mode in
relation to its price, quality, comfort, etc.).
A decision-maker chooses among a set of J options. A dependent variable Y, a discrete
variable with a countable number of J values, represents the outcome of the decision. The
goal of the analysis is to understand what variables influence this choice, and to what extent.
The utility of the alternative for the decision-maker i can be expressed as a linear combination
of
the
observed
(non-random)
factors
[X i1 , X i 2 ,..., X iH ] = x′i
with
parameters
β′ = [β 0 , β1 ,..., β H ] , and the unobserved, random factors (εij), j = 1, 2, …, J. These factors
together represent the utility
U ij = Vij + ε ij , j = 1, 2, …, J.
(1)
If the decision-maker chooses the alternative which brings the greatest utility to them, then the
probability of the choice of the alternative j over j’,
π ij = P(Vij + ε ij > Vij′ + ε ij′ ) = P(ε ij′ − ε ij < Vij − Vij′ ) ,
is the cumulative distribution function of a random variable ε ij′ − ε ij = ε ijj∗ ′ . Different discrete
choice models are obtained from different assumptions about this probability distribution.
3.2.1. Multinomial logit model
The most widely used discrete choice model, a multinomial logit model (MNL), is derived
under the assumption that each ε ij , j = 1, 2, …, J, in (1) has a so-called Gumbel (or type I
extreme value) distribution with the cumulative distribution function
F (ε ij ) = exp[ − exp(ε ij )]
and with the variance of π2/6. If these random variables are distributed identically and
independently (IID) and follow the Gumbel (type I extreme value) distribution, then their
difference follows the logistic distribution (McFadden, 1974; Agresti, 2002)
F (ε ijj∗ ′ ) = 1 + exp(−ε ijj∗ ′ ) 
with a zero mean and with the variance of π2/3. As can be proven, the probability of choice of
the alternative j by the individual i is then
π ij =
exp(Vij )
∑ exp(Vij )
j
=
exp( x′ij β)
,
∑ exp( x′ij β )
(2)
j
where xij denotes the values of the H explanatory variables for subject i and response choice j.
Since the choice probability depends only on the difference in utility, not on the level of
utility, the values of any utility can be normalized to zero; e.g., for the first alternative
x′i1β = 0 , this alternative (2) can be expressed as
π i1 =
1
, and then
∑ exp( x′ij β)
j
x′ij β = ln
π ij
, j = 2, …, J.
π i1
(3)
There is a so-called logit on the right-hand side of (3).
The problem with the MNL model arises from the IID assumption. The odds of choosing an
alternative j over an alternative j′ do not depend on the other alternatives in the choice set or
on their values of the explanatory variables:
π ij exp( x′ij β )
=
= exp[(x ij − x ij ′ )′β] .
π ij ′ exp( x′ij ′β )
This direct consequence of the IID assumption – independence of irrelevant alternatives (IIA)
– is expressed as a proportionate shift: an increase in the probability of one alternative
reduces the probabilities for all the other alternatives by the same percentage. If the IIA
property holds, it is possible, for example, to reduce a number of choice alternatives without
influencing the relations among the remaining ones. It is unrealistic in some applications. The
key IID assumption is that the errors are independent of each other. However, unobserved
factors related to different alternatives might be similar and hence the random component
might be dependent. Then the assumption of independence can be inappropriate.
The IIA condition is usually tested with the Hausman-McFadden test (McFadden et al., 1976;
Hensher et al., 2005). Nowadays, the hypothesis of this test is commonly specified as
constraints on the parameters of the more general model. For the calculation of the test
statistic, each alternative is separately excluded from the model, and the parameters for
restricted and unrestricted models are estimated as well as their variance-covariance matrices.
The test criterion is chi-square distributed with the degrees of freedom given by the number of
estimated parameters.
Although the IID/IIA conditions may be worrying, any unrealistic assumption about the error
term is likely to be of small consequence if the amount of information in the unobserved
component is minimal. The richness of information in Vij captured in attributes depends in
particular on the proper implementation of the design and pre-test stages of the choice
experiment.
The maximum-likelihood method is applied to the model parameter estimates. To estimate the
parameters of the model, a sample of n decision-makers is obtained and their choices are
surveyed. To measure how well the model fits the data, the goodness-of-fit statistics on the
basis of the log-likelihood function are usually used (e.g., Agresti, 2002). There exist many
such statistics (Pecáková, 2007); the one most used in literature on discrete choice modelling
is McFadden’s statistic:
DMF =
ln L0 − ln LE
,
ln L0
where L0 is the likelihood of the intercept-only model and LE is the likelihood of the estimated
model. The interpretation of this statistic is not the same as that of the R-squared statistic in
the linear regression and usually its values are low; fortunately, an unambiguous relationship
between them exists that provides better interpretation (Domencich and McFadden, 1975),
where pseudo R-squared values between the range of 0.3 and 0.4 can be translated as an Rsquare of between 0.6 and 0.8 for the equivalent linear model.
For the model comparison, the log-likelihood ratio statistic, so-called deviance, is normally used
(Hensher et al., 2005; Agresti, 2002). It is the statistic for testing the null hypothesis that the
restricted model (R) holds against the alternative that the more general, unrestricted model (U)
holds:
D = −2(ln LR − ln LU ) .
It has an approximately chi-square distribution with degrees of freedom equal to the difference in
the number of parameters between both the compared models.
Wald tests are used most commonly for hypotheses about the significance of the single
parameters; however, sometimes likelihood ratio tests are recommended instead (Hosmer and
Lemeshow, 2000). The Wald test is known to have low power and it can be biased where
there are insufficient data.
3.2.2. Nested logit model
If the IIA does not hold, it is necessary to consider a choice model that is less restrictive.
Recently, much research effort has been concentrated on relaxing the strong IID and IIA
assumptions associated with error terms. The generalized extreme value (GEV) model allows
correlation in unobserved factors over alternatives; the unobserved portions of utility ( ε ij ) for
all alternatives jointly have a generalized extreme value distribution. The nested logit model
(NL model) is the most widely used member of the GEV family of models.
The choice alternatives are structured into several (K) groups (so-called nests) B1, B2, …, BK.
IIA holds within each nest, but it does not hold for alternatives across nests. The vector of
unobserved utility ε′i = [ε i1 , ε i 2 ,..., ε iJ ] has a generalized extreme value distribution with the
cumulative distribution function
λk
 K 
 
F (ε i ) = exp  −∑  ∑ exp(−ε ij / λk )   .
 k =1  j∈Bk
 

The parameter λk is a measure of the degree of independence in unobserved utility among the
alternatives in nest k; full independence among all the alternatives in all nests (λk = 1) reduces
the nested logit model to a multinomial logit model.
The probability of choice for the alternative j ∈ Bk is now


exp(Vij / λk )  ∑ exp(Vic / λk ) 
 c∈Bk

π ij =
λk
K 

∑  ∑ exp(Vic / λk ) 
k =1  c∈Bk

λk −1
.
(4)
The unobserved factors have the same correlation for alternatives within a nest and no
correlation for alternatives in different nests. As can be substantiated from (4), IIA holds
within each of the K nests of alternatives but not across nests.
If the observed utility (Vij) is decomposed into an invariable part for all alternatives within a
nest (Wk) and a part that varies across alternatives within a nest (Yij),
U ij = Wik + Yij + ε ij , j ∈ Bk , k = 1, 2, …, K,
then the probability of choice (4) can be written as a product of two standard logit
probabilities: the probability resulting from the choice among nests – the upper model (it
depends on both the mentioned parts of utility) and the conditional probability resulting from
the choice among the alternatives within the nest – the lower model (it depends on Yij only).
Thus,
π ij =
exp(Wik + λk I ik )
⋅
exp(Yij / λk )
∑ exp(Wic + λc I ic ) c∈∑B exp(Yic / λc )
K
c =1
.
k
The quantity Ik (so-called inclusive value – IV – or inclusive utility of nest Bk) that enters as an
explanatory variable into the upper model,
I k = ln ∑ exp(Yic / λc ) ,
c∈Bk
brings in the information from the lower model: it is the log of the denominator of the lower
model. The term λkIik expresses the utility expected from the choice among the alternatives in
nest Bk. Its parameter λk can be used to test whether the correlation structure of the nested
model differs from the multinomial logit model.
The principle of the test is that if the IV parameter does not differ statistically from 1, then a
nested logit collapses into a multinomial logit model. The significant values of the test
statistic (with an approximately standard normal distribution) justify nested structures
(Louviere et al., 2000). A number of researchers have independently shown that the IV
parameter for the lower model is the ratio of the scale parameter of the upper model to the
scale parameter of the lower model; in the NL model the scale parameter is introduced in the
variance of the unobserved effects for each alternative (the variance is an inverse function to
the scale). This is real progress; nevertheless, the NL model cannot be identified without
imposing an additional restriction. One possibility is that the researcher constrains the IV
parameter to be the same for all (or some) nests, indicating that the correlation is the same in
each of these nests (Train, 2003). That is also the approach applied in the NL model
estimation in this paper.
The NL model enables us to model choices in a hierarchical structure. These are sometimes
interpreted as a sequential decision-making process, that is, that the respondents decide first
on the nest and then on the particular alternative within the nest. However, this decision-
making process is not necessary for the nested logit model application. In other words, the
“NL tree structures are determined so as to accommodate differences in variance or scale that
exist in the unobserved components of the utility expressions (i.e., on econometric and not
behavioural grounds)” (Hensher et al., 2005). All the parameters of a nested model can be
estimated by standard maximum likelihood techniques again.
3.2.3. Probit model
The probit model provides an alternative way to fix the problem of the limitations of the
multinomial logit model, especially regarding the IID and IIA properties (Hausman, 1978). As
Train suggests, “the (multinomial) logit model is limited in three important ways. It cannot
represent random taste variation. It exhibits restrictive substitution patterns due to the IIA
property. And it cannot be used with panel data when unobserved factors are correlated over
time for each decision-maker… probit models deal with all three.” (Train, 2003)
The basic assumption of the probit model is that the unobserved utility components are joint
normally distributed with the density
f (ε i ) =
1
(2π )
J /2
Ω
1/ 2
exp [ −0,5ε′i Ωε i ] ,
with a mean vector of zero means and a known covariance matrix Ω. The choice probability
of the alternative j can be then expressed as
π ij = F (ε i ) = ∫ f (ε i )dε i .
εi
With a full covariance matrix, various patterns of correlation and heteroskedasticity can be
accommodated as needed, so that the IID and IIA are relaxed. However, the probabilities of
choice can be expressed only in the form of integrals and they must be evaluated numerically
through simulation. Also, the model interpretation is not as straightforward and intuitive as
the logit models. The linear combination of observed factors – the representative utility – in
this model is a probit, i.e., a percentile of the normal distribution.
The assumption of the normally distributed error term ε ij is appropriate in most cases;
however, situations may and do occur where it is not the case, since the normal distribution
has a zero density on both sides. These are mostly cases where it is inappropriate that the
estimated coefficients should have a positive value, e.g., price coefficients where “the model
necessarily implies that some people have a positive price coefficient” (Train, 2003). As Train
suggests. “using the lognormal distribution would be more appropriate, however yet cannot be
accommodated within probit.”
3.2.4. Random parameters logit model
Any random utility model can be approximated by a mixed logit model (random parameters
logit model, or RPL model). It is not restricted to normal distributions like the probit;
nevertheless, it is more flexible in the treatment of the variances and correlations of the
random component. The RPL model used for discrete choice data analysis overcomes the two
major limitations of the MNL model, i.e., the IIA property and the limited ability of the
previous models to explicitly account for heterogeneity in data (Train, 2003).
To be able to take into account correlations among the error components of different choice
alternatives, the model introduces into the utility function an additional stochastic element that
may be heteroskedastic and correlated across alternatives (Train, 2003).
The utility of the decision-maker i from the alternative j is specified in the mixed logit model
as
U ij = x′ij βi + ε ij i = 1, 2, …, n, j = 1, 2, …, J;
where xij are observed variables that relate to the alternative j and the decision-maker i, βi is
a vector of coefficients of the observed variables for the decision-maker i representing
individuals’ tastes; εij is a random term with an IID extreme value distribution.
In contrast to a standard logit model, all the coefficients (of the variables xij ) or some
coefficients (of some variables z ij ) vary across decision-makers in the population with a
density f(β
β ). They are considered to be random and can be decomposed into their means α
and deviations µi. Then
U ij = x′ij α + z′ij µ i + ε ij i = 1, 2, …, n, j = 1, 2, …, J.
The unobserved portion of utility with the error component z′ij µ i can be correlated among
alternatives and/or heteroskedastic for each individual (in the case of a zero error component,
we obtain the standard logit model).
The mixed logit choice probabilities are conceived as a mixture of the logit function evaluated
for different values of parameters β with f(β
β) as the density of the mixed distribution. The
density f(β
β ) is specified as continued and in particular normal, lognormal, uniform, triangular
or any other distributions are used. The applicable distribution is given by expectations about
decision-makers’ behaviour in the particular application.
Then, the mixed logit choice probabilities can be expressed as integrals of standard logit
probabilities over a density of parameters evaluated for different values of β by the density
f(β),
π ij =
exp ( x′ij β )
∫ ∑ exp ( x′ β ) f (β) dβ .
ij
j
To specify the distribution of the coefficients, an estimate of its parameters is necessary. For
that reason, two sets of parameters are used in the mixed logit model: parameters which enter
the logit formula, and parameters which describe the density. The first type of parameters has
an interpretable meaning as representing the tastes of individual decision-makers; the second
parameters describe their distribution across decision-makers.
Problems with model estimation via convergence problems may and do occur in the RPL
model. The estimation problems differ depending on the distribution of the variables used. In
the case of a normal distribution being imposed on the parameters, respondents with a reverse
sign compared to the anticipated sign occur in the results. The triangular distribution, which is
restricted on both sides (compared to the normal distribution), can then be an alternative for
the analyst since it is a proxy for the normal distribution. The triangular, normal and uniform
distributions can be constrained as well, and thus, unacceptable signs on the random
parameters can be avoided (Hensher and Greene, 2003). The uniform distribution can be
applied alternatively when dummy variables are used in the RPL model. The log-normal
distribution can be a suitable alternative as well since it does not induce the problem with the
unexpected signs; however, it is often problematic for estimation. As Train and Sonnier
(2003) state, “one reason is that the parameters of log-normal distribution are hard to estimate
with classical procedures, since the log-likelihood surface is highly non-quadratic.”
The mixed logit is well suited to simulation methods for estimation. The greatest value of this
model occurs in using each parameter with other linked parameter estimates. The mean
parameter estimate for a variable, an associated heterogeneity in its parameter and the
standard deviation of the parameter estimate represent the utility of this variable associated
with a specific alternative and individual.
3.3. Discrete choice models: Empirical application
The main objective of this subchapter is to deploy the models discussed in the previous
subchapter on a practical example of summer waterside recreation – the preference analysis of
the summer holiday-makers on the Lake Macha beaches in the summer of 2007. The
following pages are devoted to the empirical application, comparison and discussion of the
appropriateness of the discrete choice models discussed in the previous section. The models
are deployed in the same order as they were discussed. The models are discussed with respect
to their fitness, estimated values, practical results, and advantages in terms of information
they provide the analyst with. Welfare measures based on the model estimates are also
presented here.
As was discussed in detail, the discrete choice data for the case study were gathered using the
choice experiment method, which is one of the choice modelling approaches that are
consistent with economic theory (Bateman et al., 2002). The application of the choice
experiment method is linked to a sociologic survey. In the survey, several different products
(alternatives) described in terms of their attributes are offered to the respondents. One of the
attributes is always the price of the product or a similar measurement of its value (e.g., travel
distance to a certain place, a tax increase); the “opt-out/status quo” alternative is also offered,
meaning no change at no costs/preserving. The products (alternatives) are described by the
same attributes, but the levels of the attributes vary (e.g., water quality can be the attribute of
a product called recreation; the levels of the attribute are good or poor water quality). The
consumers then make a “trade-off” between the changes in the attribute levels and the prices
of the alternatives.
The crucial part of the choice experiment application is the appropriate choice of the
attributes. The attributes should characterize the product of interest as well as possible, so that
as much as possible of the individual’s preferences is covered by the attributes; in other
words, so that the random component of the utility
ε ij
in the RUM is minimized. The
attributes and their levels used in the final sampling are plotted in Table 3-1 For a detailed
discussion of the choice of the attributes and the choice experiment design, see Chapter 2.
Table 3-1.: Choice experiment attributes and their levels
Attribute
Attribute Levels
Yes
No
Beach Overcrowded
Clean water
Slightly polluted water
Polluted water
Water Quality
Yes
No
Beach Facilities
CZK 40 (EUR 1.6)
CZK 80 (EUR 3.2)
CZK 150 (EUR 6)
Entrance Fee
Source: Own analysis
Each respondent was asked to choose one out of the three possible alternatives; that is, among
Site 1, Site 2 and the opt-out option.
The structure of the choices in the particular choice sets is shown in Chart 3-1 below.
Chart 3-1: Structure of the choices in the particular choice sets
Structure of the choices in the particular choice sets
350
Absolute frequency
300
250
200
150
100
50
0
Q1
Q2
Q3
Q4
Q5
Q6
Q7
Q8
Q9
Number of the choice
Site 1
Site 2
Opt out
Source: Own analysis
Most of the “opt-out” choices were made in questions 7, 9 and 4 (in this order). Reasons for
selecting the “opt-out” options in choice set 7 are shown in the following chart.
Chart 3-2.: Reasons for the “opt-out” option in question 7
Reasons for the „opt-out“ option in the choice set 7
Absolute frequency
200
180
160
140
120
100
80
60
40
20
0
Polluted
water
Few people
Generally
Too many
people
No
equipment
Site 1
Source: Own analysis
Site 2
Too
expensive
The most frequently mentioned reasons for the “status–quo” option were “dirty water”, “too
expensive entrance fee” and “too many people”. Several respondents also protested against
lack of facilities on Site 1 beaches.
3.3.1. Description of variables
The explanatory variables used in the modelling exercise were as follows; see Table 3-2.
Table 3-2: MNL model; Explanatory variables
Variable
Crowd
Clear
Dirty
Noteq
Fee
Opt_out
Description of variable
Overcrowded beach
Water pollution level – clear
Water pollution level – dirty
No facilities at the beach
Entrance fee
The dummy for the opt-out alternative
Variable values
1=yes; 0=no
1=yes; 0=no
1=yes; 0=no
1=yes; 0=no
40; 80; 150
0 = no; 1 = opt-out
Source: Own analysis
In the discrete choice models, it is not the absolute magnitude of the utility that is measured, but
rather the changes in the utility in comparison with the basic scenario (basic levels of attributes).
Thus, for example, the estimated parameter for Clean water (“Clean”) can be interpreted as an
indication of the relative change in the people’s preferences in comparison with the attribute
level “Slightly polluted water” (the basic attribute level). This coefficient can also be expressed
as a relative change in the probability of selecting a certain site when water quality changes from
slightly polluted to clean water and all other attribute levels remain unchanged. The formulae for
converting the estimated parameter to the mentioned percentage change differ depending on the
model used and are neither discussed nor applied in this chapter.
3.3.2. Multinomial logit model
First, the basic multinomial logit model was estimated. The following table shows the parameter
estimates, analyzed by the Nlogit4 software.
Table 3-3: MNL model; Parameter estimates
Standard
Variable
Coefficient Error
b/St.Er. P[|Z|>z]
-0.419***
0.680
-6.159 0.000
CROWD
0.079
8.639 0.000
0.686***
CLEAR
-2.438***
0.095
-25.703 0.000
DIRTY
-1
.
991***
0
.
090
-22.115 0.000
NOTEQ
0.001
-14.335 0.000
-0.011***
FEE
0.122
-22.463 0.000
OPT_OUT -2.750***
Note: ***, **, * = Significance at 1%, 5%, 10% levels.
Source: Own analysis
The MNL model performed well. All the parameters of the model were significant at the 1%
level and had the expected sign; that is, the people showed a positive preference towards less
crowded beaches, they preferred clean water to dirty water, and they perceived dirty water as a
very negative factor for the utility they get from their recreation. They also had negative
preferences towards beaches which are not equipped, and preferred lower entrance fees. The Log
likelihood function was –2305.512. The model was statistically significant.
However, the multinomial logit model failed to pass the Hausman and McFadden test of the IIA,
which is crucial for its validity. The Hausman and McFadden test is specified as follows:
q = [bu − br ]_ [Vr − Vu ]−1 [bu − br ], where
bu is a vector of parameter estimates for the unrestricted model,
br is a vector of parameter estimates for the restricted model,
Vr is the variance–covariance matrix for the restricted model, and
Vu is the variance–covariance matrix for the unrestricted model.
As the p-value in Table 3-4. for the Hausman and McFadden test indicates the hypothesis that
the data comply with the IIA assumption (at an alpha equal to 5%) was rejected with a chisquare value of 832.96 and 5 degrees of freedom. As Louviere et al. (2000) state, “when
violations (of the IIA assumption) occur, the cross-substitution effects observed between pairs
of alternatives are no longer equal given the presence or absence of other alternatives within
the complete list of available alternatives within the model”.
Table 3-4.: The results of the Hausman and McFadden test for IIA of the MNL
Hausman test for IIA
ChiSqrd [5 d.f.] = 832.96
Pr(C>c)
=
0.00
Note: Excluded choice is “OPT_OUT”
Source: Own analysis
As has been discussed in the theoretical part of this chapter, when the MNL model does not
comply with the IIA assumption, the nested logit (NL) model provides the analyst with an
alternative as it represents a partial relaxation of the IID and IIA assumptions in the MNL
model.
3.3.3. Nested logit model
In order to apply the NL model framework to the Lake Macha data, the choice problem needs
to be reformulated as a hierarchical nested structure. In the Lake Macha data, this seems to be
possible, in the way that the respondents could be organizing the choice problem in two
stages: first, whether they are willing to visit any of the sites offered by the interviewer; and
then, which of the two sites to visit. This leads to the following nested tree structure (Chart 33):
Chart 3-3: The tree structure for the nested logit model
Respondent’s choice
I want to visit neither of the
sites
I want to visit a site
CHOICE: Site 1 or 2
NO_CHOICE alternative
Site 1
Source: Own analysis
Site 2
The estimated nested logit model, according to the depicted hierarchical tree structure, is
shown in Table 3-5 below.
Table 3-5: Nested logit model estimates
Variable
CROWD
CLEAR
DIRTY
NOTEQ
FEE
OPT_OUT
Standard
Error
0.680
0.079
0.095
0.090
0.001
0.122
Coefficient
-0.419***
0.686***
-2.438***
-1.991***
-0.011***
-2.750***
b/St.Er.
-6.159
8.639
-25.703
-22.115
-14.335
-22.463
P[|Z|>z]
0.000
0.000
0.000
0.000
0.000
0.000
IV parameters
tau(b|l,r)
sigma(l|r) phi(r)
0.895***
0.656
13.645
0.000
CHOICE
0.656
13.645
0.000
NO_CHOICE 0.895***
Note: ***, **, * = Significance at 1%, 5%, 10% levels
Source: Own analysis
This model again performed well. All the parameters of the model were significant at the 0.1%
level and they had the same signs and sizes as in the MNL. In specifying the NL model, the
correlation between the “choice” and “opt-out” variables needs to be restricted, so that the
correlation problem between the nest and the explanatory variable is avoided.
The log-likelihood of the model was –2304.386. Comparing our models with the log-likelihood
ratio test (the test criterion value 2.252 and critical chi value of 5.99)24, we cannot reject the null
hypothesis that the two models are identical. As the log-likelihood ratio test indicates (see Table
3-6), the NL model does not significantly improve the model estimation and does not contribute
to the explanation of the data variability.
Table 3-6: Log-likelihood ratio test
DF
ˇ
-2 x Log-likelihood Chi Critical
Model
Log-likelihood
DF difference
difference
Value
Conditional logit model
-2305.512
5
-2.252
5.99
Nested logit model
-2304.386
7
2
Source: Own analysis
24
−2 x (log-likelihood of the restricted model − log-likelihood of the unrestricted model) ~ χ2
This is also confirmed by a test of the IV parameters of the NL model. This test is used to
determine whether the nests help to explain people’s choices. The analysis of the IV parameters
runs in two steps (Hensher et al., 2005). The first is to find out whether the parameter is not equal
to zero (i.e., dividing the estimated IV parameter by its standard error and comparing the
resulting value to the critical value of the normal distribution), which is evidently not our case
(see in Table 5 the p-value for the CHOICE and NO_CHOICE attributes). The second step is to
determine whether the variable is statistically different from one, and thus, whether the nests help
explain people’s choices. If this is not the case, the NL model collapses into a single branch,
which is equivalent to a MNL model; in other words, the nested logit model is reduced to the
multinomial logit model (see the previous chapter). This occurs in the case of the Lake Macha
data. The value of the Wald test is -1.601 and thus we cannot reject the null hypothesis that the
IV parameter is statistically different from 1 (at the 95% confidence level of the normal
distribution; compare the test statistic to the critical value of ±1.96). Both the log-likelihood ratio
test and the test of the IV parameters indicate that the NL model is no better than the MNL
model in explaining individuals’ choices.
3.3.4. Probit model
Another possibility which provides the analyst with more flexibility compared to the MNL
model is the probit model, where it is assumed that the random component is normally
distributed. This model does not rely on the IIA assumption as the MNL model does. It is not
used in discrete choice data analysis as frequently as the NL or MNL models. The reason for this
may be that the model estimation is time-consuming and problems with the log likelihood
maximization occur.
The estimated probit model for the Lake Macha data is shown in Table 3-7 below.
Table 3-7: Probit model estimates
Variable
CROWD
CLEAR
DIRTY
NOTEQ
FEE
OPT_OUT
Coefficient
Standard Error b/St.Er.
P[|Z|>z]
-0.419***
0.0525944
-6.357
0.000
0.686***
0.0592831
8.921
0.000
-2.438***
0.0858123
-20.152
0.000
-1.991***
0.0790409
-17.659
0.000
-0.011***
0.0005922
-14.449
0.000
-1.979***
0.0989481
-20.007
0.000
Note: ***, **, * = Significance at 1%, 5%, 10% levels
Source: Own analysis
The log likelihood of the model is –2302.676. The parameters estimated using the probit model
do not differ much from those in the MNL model; see the model comparison at the end of the
chapter (Table 7). Also, the log-likelihood ratio test with a test value of 5.672 and a critical value
of 5.99 indicates that the probit model does not significantly improve the model fit and thus is no
better in explaining data variability than the MNL or NL models.
3.3.5. Random parameters logit model
Recent research in discrete choice data modelling has paid special attention to the random
parameters logit (RPL) models (mixed logit models) (Train, 2003). As discussed in more detail
in the previous chapter, these models relax the IID assumption in terms of the covariances;
however, “all are of open-form solution and as such require complex analytical calculations to
identify changes in the choice probabilities through varying levels of attributes and sociodemographic characteristics” (Louviere et al., 2000; Train, 2003). Compared to the models
deployed up to now, the application of the RPL model brings the analyst extra benefits in data
analysis. The RPL model enables us to determine whether heterogeneity in the data exists and
also to identify possible sources of this heterogeneity. In this regard, it provides the analyst with
similar possibilities as the latent class models.
The best random parameter logit model specification out of the many possible specifications for
the Lake Macha data is shown in Table 3-8 below.
Table 3-8: Random parameters logit model estimates
Variable
CROWD
CLEAR
DIRTY
NOTEQ
FEE
Coefficient
Standard Error
b/St.Er.
P[|Z|>z]
-0.603***
0.090
0.000
-6.713
0.828***
0.096
0.000
8.652
-2.649***
0.131
0.000
-20.204
-2.163***
0.112
0.000
-19.232
-0.013***
0.001
0.000
-12.474
Non-random parameters in utility functions
0.147
-20.552
0.000
OPT_OUT -3.02757
Derived standard deviations of parameter distribution
0.263
0.004
0.997
NsCROWD 0.001
0.795***
0.206
3.862
0.000
NsDIRTY
0.221
0.035
0.972
NsCLEAN 0.008
0.158
2.281
0.225
NsEQUIP 0.360**
0.003***
0.001
3.394
0.007
NsFEE
Note: ***, **, * = Significance at 1%, 5%, 10% levels
Source: Own analysis
The model is significant at the 1% level. All the parameters of the model are significant at the
1% level and have the expected signs. The estimated parameters slightly differ from the MNL
model estimates. In comparison with the MNL model, people had more negative preferences
towards crowded beaches, unequipped beaches and polluted water. They also had more positive
preferences towards clean water. The log likelihood function is –2290.49. The log likelihood test
with a test value of 30.044 and a critical value of 11.07 indicates that the model is statistically
better than the MNL model (and also the NL and probit models).
In the output of the RPL model shown in Table 3-8, all variables except the opt-out variable are
specified as random, drawn from a normal distribution. The interpretation of the random
parameters is much the same as in the MNL model; however, the mean of the random parameter
is the average of the parameters drawn over the R replications from the chosen distribution (the
normal distribution in our case). In comparison with the MNL model, there are five additional
variables in the output window. These are derived standard deviations of parameter distribution
calculated over each of the R draws and as such relate to the extent of the dispersion around the
mean of the parameter.
Insignificant parameter estimates for the clean water and crowdedness of the beaches indicate
that the dispersion around the mean is statistically equal to zero. That suggests that all
information about the people’s preference towards these variables is captured in the estimated
mean. However, this is not the case for dirty water, entrance fees, and beach facilities, where the
estimated standard deviations of the parameters are statistically significant. This suggests that
there exists heterogeneity in the parameter estimates across the sample population (around the
mean parameter estimate). It can be interpreted that different individuals have different
preferences that differ from the mean estimate for the sample population.
In the data analysis, several characteristics of the sample population were analyzed as possible
sources of the heterogeneity in the preferences. These were the respondents’ income, their
repeated visits to the Lake, a dummy variable indicating whether the people swim in the water
during their stays on the beach, the respondents’ sex, and the beach where the people met the
interviewer.
The data suggest that the source of the heterogeneity in the case of the dirty water variable may
be partly explained by the respondents’ income (people with a higher income perceive dirty
water more negatively) and also by the different preferences of visitors to different beaches
(visitors to the main beach in Doksy were less sensitive to water quality). These results indicate a
statistically significant interaction of the Dirty x Income variables and the Dirty x Doksy
variables, both at the 5% level. The source of the heterogeneity for the facilities variable was the
respondents’ sex (men had more negative preferences towards unequipped beaches) and the
respondents’ income (people with a higher income were less sensitive to unequipped beaches);
these findings indicate statistically significant interactions of the Facilities x Sex variables and
the Facilities x Income variables, both at the 5% level. Finally, the sources of the heterogeneity
for the fee variable were the repeated visits to Lake Macha (significant at the 1% level), that is,
people who visit the Lake repeatedly are more sensitive to the beach entrance fee levels;
swimming in the Lake (significant at the 5% level), that is, people who swim in the Lake during
their visits are more sensitive to the entrance fee levels; and the income (significant at the 1%
level), i.e., people with a higher income are more sensitive to the entrance fee levels (for model
estimates see Annex 9).
3.3.6. Model comparison
The model comparison is shown in Table 3-9 below. This parameter estimate overview
confirms the previous finding that the MNL, NL and Probit models yield principally the same
values of the estimated parameters despite the fact that they differ in terms of their
assumptions (see the previous chapter). All the models exhibit a decent fit as they achieve
approximately an R-squared of 0.3, which represents an equivalent of 0.6 – 0.8 for the linear
regression model. The parameters estimated with the RPL model do not differ in terms of
their signs, but they differ slightly in terms of their sizes. The difference is the greatest for the
dirty water and unequipped beach variables (see Figures 2-6); however, one has to keep in
mind that both these variables were estimated as random and their standard deviations are
significant at the 1% level (5% for the facilities variable); thus, the values shown in Table 7
present only the mean values of the parameter estimates.
Table 3-9: Model comparison (t-statistic in brackets)
OPT_OUT
MNL
-0.419
(-6.159)
0.686
-8.639
-2.438
(-25.703)
-1.991
(-22.115)
-0.011
(-14.335)
-2.75
(-22.463)
NL
-0.419
(-6.159)
0.686
-8.639
-2.438
(-25.703)
-1.991
(-22.115)
-0.011
(-14.335)
-2.75
(-22.463)
PM
-0.419
(-6.357)
0.686
-8.921
-2.438
(-20.152)
-1.991
(-17.659)
-0.011
(-14.449)
-1.979
(-20.007)
RPL
-0.603
(-6.713)
0.82837
-8.652
-2.64859
(-20.204)
-2.16329
(-19.232)
-0.0132
(-12.474)
-3.02757
(-20.552)
Log likelihood
-2305.512
-2304.386
-2302.676
-2290.49
Adjusted R2
0.287
0.316
0.300
0.304
CROWD
CLEAR
DIRTY
NOTEQ
FEE
Observations
2997
2997
2997
2997
Note: ***, **, * = Significance at 1%, 5%, 10% levels
Source: Own analysis
3.3.7. Willingness to pay analysis
The derived estimates of the parameter values for the particular models can be used to derive
welfare changes (here, a measure of willingness to pay, or WTP) caused by the changes in the
levels of particular attributes. This chapter compares the WTP results derived from the
estimated models. The estimation of the welfare changes caused by changes in the attribute
levels can by calculated as follows:
WTP∆X = β / γ , where
β
is the variable coefficient, and γ is the marginal utility of income (fee variable
coefficient) (Hanemann, 1984). Thus, the point estimates of WTP for a change can be derived
by calculating the marginal rates of substitution between the change in a given attribute and
the price attribute, that is, by dividing the coefficient of the attribute by the coefficient of the
entrance fee attribute. This is the rate at which the respondent is willing to trade off money for
improvements to the beach (facilities, crowdedness, water quality) attribute. The following
figure presents a comparison of the marginal WTP for the attribute level changes.
Chart 3-4: Marginal willingness to pay for a change in the attribute level;
Model comparison (Euro PPP 2007)
Willingness to pay - model comparison
6
PM
4
MNL
2
EURO (PPP 2007)
0
-2
-4
MNL
PM
NL
CROWD
RPL
NL
RPL
CLEAR
DIRTY
NOTEQ
-6
RPL
-8
-10
MNL
RPL
NL
MNL
PM
-12
NL
PM
-14
Variables
MNL
NL
PM
RPL
Note:
MNL = Multinomial logit model;
NL = Nested logit model;
PM = Probit model;
RPL = Random parameters logit model
Source: Own analysis
As apparent from Chart 3-4, there are only small differences in the willingness-to-pay estimates
caused by differences in the estimated parameters for the particular models.
3.4. Summary
The main objective of this chapter is the core of the economic (and welfare measures) analysis
– the choice experiment data analysis.
First, the economic background of the discrete choice models is explained. Then, the various
types of discrete choice models are compared and discussed with respect to their ability to
model discrete choice data and with respect to the welfare measure values they provide the
analyst with; their limitations and conveniences are mentioned. The focus was on the
multinomial logit model, the nested logit model, the probit model and the state of art – the
random parameters logit model. Special attention was paid to the crucial assumption for the
multinomial logit model validity, i.e., independence of irrelevant alternatives (IIA) and the
implications for choice experiment data analysis that follow for the analyst from violating the
assumption.
The chapter then proceeded to the deployment of the discussed models on the Lake Macha
choice experiment data. The multinomial, nested and probit models brought about exactly the
same results regarding the estimated model coefficients, the log-likelihood value and also
WTP estimates. The random parameters logit model yielded slightly different results and
information for the analyst. It was also used for the heterogeneity identification in the data set
and its sources.
4. Cost-benefit Analysis of the Measures for Water Quality
Improvement at Lake Macha
This chapter considers the economic costs and benefits associated with eutrophication
prevention measures undertaken at the Lake Macha site. These measures were undertaken to
improve water quality to the required standard for bathing waters. The economic benefits are
compared to the costs in order to indicate whether the eutrophication reduction measures
undertaken up to date could be justified by economic efficiency measures. Specifically, the
chapter undertakes a cost-benefit analysis (further CBA; CB analysis) of eutrophication
reduction measures for Lake Macha (Boardman et al., 2006). The net social benefit of the
project is determined relatively to the status quo. The project is appreciated as effective if the
net social benefits (NSB) are positive. The net social benefit of the project (NSB) is counted
as follows:
Net social benefits (NSB) = social benefits (SB) – social costs (SC)
The following CB analysis is done “in medias res”, that is, during the lifetime of the project.
The rationale behind CB analysis is an attempt to answer the question whether any of the
expenditures that have been made at the Lake represent effective and efficient use of
resources. The cost-benefit analysis (CBA), which balances the social costs and benefits, is an
important economic tool in the assessment of alternative options in decision-making
processes. Decision-makers should balance the opportunity costs of resources used for
environmental protection with the social welfare benefits assigned to better environmental
quality. As Georgiou and Bateman state, “Regulators and governments have to balance the
public desire for better environmental quality with the opportunity costs of any actions.”
(Georgiou and Bateman, 2005, p. 431). The benefits calculated using the choice experiment
(especially improvements), can be related to costs in a standard cost-benefit analysis
framework to provide policy guidance for decision-makers.
CBA of any pollution control and abatement policy is of special importance for public
programs which involve high expenditures with great environmental impacts. Expressing all
costs and benefits in monetary terms, CBA enables to quantitatively rank alternative options
of an intended project or alternative policies. The aggregate value of a project is measured by
its net social benefit (NSB), which equals the social benefits minus the social costs. We
should adopt the project if its NSB is positive (Boardman, 2006).
4.1. CBA History
Cost-benefit analysis is around 70 years old if we date its first practical application to water
resource developments in the USA in the 1930s. CBA was applied initially to efficient
allocation of water resources, dealing with the issue of water quantity. In recent years, CBA
has been applied to water quality improvements and concentrated on wider goals of water
ecosystem services – recreation, fisheries, biodiversity and general amenities (see e.g. EisenHecht et al., 2002; Dubgaard et al., 2005; Georgiou and Bateman, 2005; Saz-Salazar et al.,
2009).
As already mentioned, the freshwater bathing area - Lake Macha - belongs to the bathing sites
in the Czech Republic which suffer from strong water eutrophication with the corresponding
consequences – cyanobacteria in the water. A bathing ban has been issued at the Lake several
times because the Lake suffers from low water quality. The low water quality is caused by
high phosphorus content which enables significant cyanobacterium occurrence. This has led
to a strong decrease in revenues from tourism and caused economic problems to many
businesses in the tourism-oriented region (Doksy, 2007).
Unfortunately, no study providing a comprehensive analysis of the condition of the aquatic
ecosystems in the Robečský brook basin, which are of a strong relevance to Lake Macha
water quality, is available at the time of writing this dissertation. Measures taken at Lake
Macha aimed at reducing eutrophication to a tolerable level should not be regarded isolatedly.
The trophic burden of the entire area under management of the ANCLP, known as the
Kokořínsko PLA, should be assessed as a whole. Measures taken at Lake Macha until 2009
were based on component studies, developed based on component commissions of various
stakeholders (Doksy municipality, ANCLP, Lake Macha charity). According to available
information, no comprehensive study assessing the trophic load of the entire area impacting
on Lake Macha water quality is being developed at present. If such a study were available, it
would make it clearer whether the measures taken so far have been the best solutions to the
problem in place.
Because of the unavailability of such a study, the following sub-chapter briefly describes and
summarizes available data about:
•
the implemented anti-eutrophication measures and their costs; and
•
the results the above brought and their benefits in monetary terms.
This sub-chapter strives to give a firm foundation for the subsequent cost-benefit analysis of
these measures.
4.2. Description of Phosphorus Balance in the Lake
Theoretically, the chief water quality indicator is its trophic potential, given by the nutrient
content. However, in Lake Macha (and other stagnant water bodies in the Czech Republic) the
excessive additional nutrients make the amount of bio-available phosphorus in the water the
decisive factor for water quality. Its reduction necessarily results in reduced growth of algae
and cyanobacteria contained in the water column during peak growing seasons25. That is why
all the measures taken at Lake Macha so far have focused on reducing the content of
phosphorus freely available to aquatic organisms from the water column. The phosphorus
sources in the Lake can essentially be divided into three groups:
1) supply along with phosphorus-containing sediments (transported to the Lake from its
tributaries);
2) point sources of pollution (transported to the Lake from its tributaries); and
3) phosphorus release from Lake bottom sediments.
4.2.1. Supply along with phosphorus-containing sediments
Based on approximate calculations of the quantity of sediment transported to the Lake
annually and the theoretical rate of phosphorus content in the sediment, the estimate is 0.06
mg P/l of water. Given the above quoted atrophy threshold (approx. 0.01 mg P/l and less), the
amount of phosphorus transported to the hydrographical network along with the sediment is
well above the threshold. This approximate estimate shows that the erosion processes in the
25
Achieving this condition would naturally result in improved quality of water leaving the Lake, which would
pose significant benefits to areas downstream of Lake Macha.
catchment area are a very significant source of nutrients contributing to cyanobacterium and
algal growth in Lake Macha (ČVUT, 2009).
4.2.2. Point sources of pollution
The chief identified problems in the area of point sources of pollution were in the town of
Doksy proper. The town lies south-east of Česká Lípa, and has a permanent population of
5,000 and 318 recreational buildings. The town comprises urban residential development,
family houses, and public amenities. Moreover, recreational facilities, hotels, hostels, and
chalets are situated along Lake Macha. Doksy is crossed by the Robečský brook, and contains
the Čepelský pond, covering some 4 ha, as well as fishery store-ponds.
The town of Doksy has a sewerage network which is not entirely charted and in a
dissatisfactory condition. There are two types of sewerage in the town of Doksy: a storm
water sewer (managed by Doksy municipality) and a sanitary sewer (managed by
Severočeské vodovody a kanalizace, a.s., Teplice). Storm water is discharged into the Doksy
stream which flows into Lake Macha. The wastewater is led away to the sewage treatment
plant in Staré Splavy and subsequently discharged into the Doksy stream below Lake Macha.
The problem in the town of Doksy is that some of the houses are connected with their sanitary
sewers to the storm water sewer and vice versa. These houses have had to be reconnected, so
that the sewage water from them stops flowing into the Doksy stream and the Doksy inlet.
The beginning of this process of reconnecting started in early 2005, but has not finished yet.
The reasons are a lack of money and information about the existing problems with the
sewerage system. Nevertheless, the problem has been resolved at least partly.
4.2.3. Lake bottom sediments
The problem of highly nutritive sediment in Lake Macha was conditioned historically, when
the Lake was intensively fertilised by fishers in 1950-1957. An estimated 50 tonnes of
superphosphate (7.5% P) were thrown into the Lake in that period. This represents some 3.8
tonnes of P.26 This phosphorus supply has undoubtedly enriched Lake Macha sediments
significantly.
26
Compare figures provided by the Czech Association of Laundry Detergent Manufacturers: little under 1,700 t
of phosphorus were sold in the Czech Republic in 2003 in laundry detergents (excluding household detergents).
Given the high nutrient content in the Lake sediments, the fish population is another
significant factor. In particular, the carp family of fish (such as the Carp Bream) and
especially the Carp dig up to 30 cm deep in the sediments while searching for food. Their
digging of the bottom contributes substantially to phosphorus release from the sediment,
which encourages algal bloom occurrence.
4.3. Corrective Measures for Water Quality Improvement Estimation of Costs
As mentioned above, all the activities following the bathing ban issued in 2004 focused on
searching for the sources of the problem; after the identification of phosphorus as being the
main stressor, all the subsequent activities have concentrated on lowering the phosphorus
content in the water column. Structured according to the phosphorus sources, these measures
are as follows:
4.3.1. Cofferdam
As mentioned above, an important supply of phosphorus comes into the Lake with the
deposits from erosion processes. Based on this finding, the Doksy inlet has been separated so
that it can work as a protective advance check for highly eutrophicated sediments, which
would otherwise enter the Lake directly. If they entered the Lake, they would have a high
potential to negative affect the eutrophication processes. In this respect, the advance check,
formed by building the cofferdam, is seen as very effective (Doksy, 2007). The cofferdam
cost CZK 1.3 million to build, and was funded by the ANCLP. The ANCLP also built a
biological filter at the inflow point to the Doksy inlet so that the Doksy inlet can also serve as
a bio-filter thanks to the specific vegetation planted.
The cofferdam in the Doksy inlet now separates the Robečský brook inlet area from the rest
of Lake Macha. The cofferdam works with a good efficiency, catching the sediment brought
in by the Robečský brook from farmland and upstream ponds or Doksy municipal sewers.
This is evident from the fact that the sediment thickness in the inlet has grown over time. That
is why the sediment needs to be extracted regularly, at approx. 5-year intervals. The CBA
accounts for this alternative in future. Based on the costs of extracting 1 m3 of sediment
between CZK 170 and 350 (Public Administration official server, 2010) and given the
unavailability of more exact data for the site in question, the analysis works with the average
value of CZK 260 per m3 of sludge. The dimensions of the Doksy inlet are about 150 by 200
metres, i.e., roughly 30,000 m2. Given the average sediment thickness of 65 cm, the amount
of sludge to extract would be 19,500 m3. The costs are thus CZK 5.07 million every 5 years.
4.3.2. Sewerage
As mentioned above, the problems of the Doksy sewerage system started to be handled in
2005. Doksy started monitoring activities due to the potential contamination from sanitary
sewerage. These activities have detected that the situation in the urbanised zone is unclear. All
the sewerage in Doksy and the campsites around Lake Macha was built in the 1960s; its
condition is dissatisfactory. The survey of the Robečský brook in the urbanised zone of Doksy
has shown that it receives contaminated water even without rain. The routing of some of the
storm sewers is not known. Some of the storm sewers can be assumed to discharge into the
sanitary sewerage, or at least connect to it. The need to cleanse the sewers, examine them with
cameras, and make detailed passports for both the sanitary and storm sewerage is thus
evident. In addition, the identification of the piping condition should be followed by
renovation of any non-compliant sections. According to Liberec Regional Authority data
(Liberec Regional Authority, 2009) over CZK 24 million should be invested in renovation of
Doksy sewerage, including CZK 8.4 million in renovation of earthenware sewers between
2004 and 2006. The following project was a renovation of the PVC and PE sewers between
2007 and 2010, using total investment funds of CZK 15.9 million. The time distribution of the
investment costs in CZK millions is shown in Table 4-1 below.
Table 4-1: Time distribution of investment costs in CZK millions
(in real values)
Doksy Municipality
New investment
Sewerage system renovation
Other renovations
Total costs
0.00
24.35
0.00
2003
0.00
0.00
0.00
2004
0.00
2.81
0.00
2005
0.00
2.81
0.00
2006
0.00
2.81
0.00
2007
0.00
3.98
0.00
2008
0.00
3.98
0.00
2009
0.00
3.98
0.00
2010
After 2011
0.00
0.00
3.98
0.00
0.00
0.00
Source: Own based on Liberec Regional Authority data (online: http://zivotni-prostredi.kraj-lbc.cz/page1923)
Besides the said investment in Doksy, investment is planned in Staré Splavy and other
settlements near Lake Macha (Obora, Okna and Tachov). In Staré Splavy, the renovation and
extension of the wastewater treatment plant, to cost CZK 35 million, and a renovation of the
sewerage, to cost CZK 9.3 million, are scheduled for 2011-2013. Minor investment of about
CZK 20 million in the sewerage is planned in the other settlements. These investment
measures are not considered in the CBA since they do not affect the Lake water quality
directly. Nevertheless, they do affect the quality of water and ecosystems in the aquatic
system downstream of the Lake.
4.3.3. Lake bottom
As mentioned above, the problem of highly eutrophic Lake water has to be resolved primarily
in respect of further nutrient supply to the Lake. On the other hand, the high phosphorus
content in the Lake water is partly conditioned historically by the high phosphorus content in
its sediments. For a comprehensive solution, therefore, the phosphorus has to be prevented
from escaping from the bottom sediments, and its content in the water itself (chiefly in the
summer season) has to be reduced. One of the ways to achieve that is to employ chemical
precipitation by means of aluminous salts.
4.3.3.1. Chemical Precipitation by Means of Aluminous Salts (PAX)
Aluminium sulphate, aluminium chloride, and polyaluminium chloride (PAX, PAC) are
eligible for practical application of chemical phosphorus precipitation by means of aluminous
salts. Given the normally sufficient inhibition capacity of water in eutrophic water bodies,
aluminium sulphate is the most economic option. Where the bottom sediments need to be
covered over, implying that higher doses of the preparation have to applied, polyaluminium
chloride application is more suitable as it causes less pH reduction than aluminium sulphate.
Both the above coagulating agents are non-toxic chemicals used in massive amounts in
drinking water treatment. In lake care, their purifying effect is surpassed by their phosphorus
coagulation effect. Having been added to water, the aluminium hydrolyses while forming
aluminium hydroxide flakes as well as insoluble complexes with phosphorus. These
compounds are no longer a nutrient source. When the aluminium hydroxide flakes sediment,
they additionally remove dispersed impurities, algae and cyanobacteria from the water
column. The water thus becomes much cleaner (Klouček et al., 2005). As stated by Cooke and
Welch: “… the content of phosphorus released in an internal cycle from the sediment was
reduced by 80% on average after the chemical application” (Cooke and Welch, 1999, p. 4).
State of the art in phosphorus precipitation by means of aluminous salts denotes that it is an
economically effective method with positive short-term and long-term effects. The method
was developed in Sweden in the 1960s and 70s. However, as Carlsson and Sten-Ake quote, “it
was used incorrectly with low dosage or in lakes mainly dominated by external loads.
Because of this the method wasn’t successful” (Carlsson, 2005, p.1). Further applications
were refined and it has been applied, e.g., in the USA since the late 1970s as a restoration
method in several lakes with good results. In its manual on restoration, the US Environmental
Protection Agency (EPA) judged the method of precipitation as „being cost effective with
both long and short-term positive effects at a low risk. This makes it the most effective
restoration method for eutrophic lakes” (EPA, 1990 in Carlsson, 2005, p.1).
The method was later used with good results for example in Sweden, e.g., in Lake Turingen,
Lejondal and Bagarsjön in Nacka; in Finland in lake Kirkkojärvi; in Poland – in Lakes Dlugie
and Gleboczek; in the USA in Lake Morey and Green Lake; in Germany in Barleber See, and
others.
The precondition for a good and long-term effectiveness of this method is that all external
phosphorus sources have to be reduced. As a result of reduced phosphorus content in the
water column, the phytoplankton production decreases and water limpidity improves.
Reduced phytoplankton production leads to reduction in the organic matter bottom sediment.
That in turn reduces the oxygen consumption in summer and winter, resulting in an overall
better oxygen balance of the water body. Experience shows that it may take up to several
years for the oxygen content in water, particularly at the bottom, to stabilise at an adequate
level throughout the summer season. Initially, at least the anoxic periods and anoxic areas are
reduced. The underwater vegetation thrives due to improved water purity and limpidity, and
macrophytes proliferate. The water body has substantially improved living conditions with
increasing oxygen content: the lake restores an ecological equilibrium. (Klouček et al., 2005)
As a result of improved water purity, reduced growth of toxic cyanobacteria, and settlement
of higher fishes, the recreational value for both swimming and fishing increases significantly.
A lake in equilibrium also has a great importance as a landscape element (Carlsson, 2005).
As for the effective period, a number of case studies are available showing high rates and
long-term nature of the effects of the method, typically for 8-15 years. The effective period is
longer in deep and stratified lakes than in shallow reservoirs. However, the crucial importance
of external supply of phosphorus for the longevity of the effect has to be stressed.
Application of the Method at Lake Macha
Since the situation at the Lake required a response and given the high effectiveness of the
method discussed, it was decided in 2004 to apply PAX 18 (polyaluminium chloride) in the
Lake in 2005. It was then applied again in 2006, 2007 and 2008. PAX 18 application costs
approximately CZK 3.5 million each year. There are several reasons for such a frequency of
repetition. Firstly, the phosphorus supply into the lake had not stopped at the time of
application, which is the main condition for the fruitfulness of the PAX application. Secondly,
a very low concentration of the active substance was applied to the water. According to the
international experience, the concentration of the active substances should range between 5
and 25 mg Al/l. At the Lake, the minimum concentration (5mg Al/l) was applied, which
amounts to a total of 250 t of PAX-18 (this dose was determined based on laboratory testing
with respect to the availability of funds for Doksy municipality and other funders) (Kemwater
ProChemie, 2010). Since the effective substance is at such a low concentration, a single
application does not cover the sediments with a layer of treated sediment, which would
continue to prevent phosphorus from escaping from the deeper layers: this was one of the
reasons for repeated application.
PAX was not re-applied in 2009 (funding reasons, comprehensive site study in progress). In
spite of that, the water quality remained at a level that allowed recreation throughout the
summer season. The CBA therefore assumes that the specified low doses of PAX will be
applied every second year.
4.3.3.2. Lake Fisheries
The fisheries in the Lake are currently balanced so that the bottom is not dug up by the fish
and the sediment is not resuspended. That was not always the case. The estimated cost of
changing the fish population in 2007-2009 is CZK 1.5 million, evenly distributed across the
period.
The anaerobic black sediment is thus broadly enclosed under a layer of relatively clean sand.
The fishery in the pond is currently managed extensively, i.e., without any supply of
unwholesome substances (fertilisation, fodder additions) by the fishery manager (Rybářství
Doksy, 2010).
This is attested by the very low total fish biomass caught during the fishing-out on 3
December 2008 (only 43 kg of fish per hectare caught); the herbivore-predator ratio was 3:1.
This condition is assessed as extraordinarily well-balanced with a minimum impact on
increasing the trophic load of the Lake waters. For the sake of clarity, it must be said that the
fish biomass was under 2.5 g per 1 m3 of water in 2008; the impact of this type of fish
population on sediment formation can therefore also be regarded as entirely negligible and not
exactly quantifiable (ČVUT, 2009).
In terms of equilibrium, the fishing-out of Lake Macha extracted approx. 75 kg of phosphorus
by extracting the fish biomass and fish escaping (15,000 kg of fish x approx. 5 g P/kg =
75,000 g P). For the sake of comparison, the fish biomass in carp production ponds with semiintensive management prior to fishing-out is typically 50 – 150 g per 1 m3, sometimes even
greater (ČVUT, 2009).
4.3.3.3. Sediment Extraction
Originally, it was intended to remove the sludge from the lake bottom. This measure was
considered necessary to resolve the problem of Lake Macha eutrophication. Its costs were
originally estimated to be between CZK 150 and 300 million, depending on the method used
(Doksy, 2007). However, nowadays this seems to be quite counterproductive, especially
thanks to the high phosphorus inflow from the streams as discussed above, and also because
of the good efficiency of the PAX application and fish habitat change, which have largely
resolved the problem of phosphorus release from the sediment to the water column. Under
these conditions, the sludge removal appears unnecessary. Its result could be even negative,
especially in light of the findings of Vodní díla TBD, a.s., whose survey of the Lake bottom
and sediments (amounts, distribution and quality) showed that the bottom under the sediment
has a very broken relief. That means, among other things, that efforts to extract the sediment
completely would create numerous places that cannot be drained.
4.4. Results of the Measures Taken
All the above steps taken to reduce the trophic level of Lake Macha water have led to the
desired stabilisation of the situation, but not its definitive and absolutely satisfactory
resolution. The following results have been observed since the PAX 18 application in 2005:
•
water limpidity has increased from 0.3 m to 0.7 m;
•
the total cyanobacterium concentration has decreased from 90,000 to 30,000 cellules
per ml;
•
the total chlorophyll a decreased by over 50% in 10 days after PAX application; and
•
good water quality was maintained throughout the summer season with no restrictions
to recreation27. (Kemwater ProChemie, 2010)
Accompanying measures – i.e., repairs of at least the worst sections of Doksy sewerage and
changes in the fish population – were essential for maintaining the results on at least a limited
scale and preventing the problem from escalating again as much as it did in 2004. Concerning
the effect on the Lake ecosystem, not even the slightest longer-term inhibiting impact on nontarget organisms in Lake Macha has been observed after four years of PAX-18 application;
quite to the contrary, numerous plant species not seen prior to the application have thriven.
Water analyses done in the Lake recently have shown that the water quality is “good” for
recreational purposes, the bottom is stabilised, and the water column is saturated with oxygen
in its entire depth down to the (ČVUT, 2009)28. Based on this analysis, an approximate
aggregation of the cost side of the CBA can be performed. Table 4-2 and Chart 4-1 below
summarise the costs of the above measures; the costs are in real values.
27
PAX application has had no negative impacts even on other ecologically valuable areas (Swamp,
Novozámecký rybník), which have been continuously monitored and their values are within the normal range:
no negative impact of the treatment has been detected.
(Agency for Nature Conservation and Landscape Protection of the Czech Republic official website, cit:
12.5.2007, online: http://www.ochranaprirody.cz/?cmd=page&type=107&lang=cs&query=M%E1chovo+jezero)
28
In the opposite case, anaerobic processes might occur in the layer of water above the bottom, causing release
of phosphorus usable for aquatic organisms into the water column, which would negatively affect water quality
chiefly in the recreation season.
Table 4-2: Costs of the water quality measures at Lake Macha
(in CZK million real values)
CZK million
Doksy sewerage system renovation
Predatory fish habitat change
Bio-filter construction and sludge removal
PAX 18 application
Total
2004
2.8
0
0
0
2.8
2005
2.8
0
1.3
3.5
7.6
2006
2.8
0
0
3
5.8
2007 2008
4.0
4.0
0.5
0.5
0
0
3
3
7.5
7.5
2009 2010
4.0
4.0
0.5
0
0 5.07
3.5
0
4.5 12.6
2011
0.0
0
0
0
0.0
2012
0.0
0
0
3.5
3.5
2013
0.0
0
0
0
0.0
2014
0.0
0
0
3.5
3.5
2015
0.0
0
5.07
0
5.1
2016 2017
0.0
0.0
0
0
0
0
3.5
0
3.5
0.0
2018 2019
0.0 0.0
0
0
0
0
3.5
0
3.5 0.0
2020
0.0
0
5.07
3.5
8.6
Source: Own analysis
Chart 4-1: Progress of costs of water quality measures up to 2020
(in CZK million real values)
Costs of the measures
14
12
Mill CZK
10
8
6
4
2
0
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
Years
PAX 18 application
Predatory fish habitat change
Bio-filter construction and sludge removal
Doksy sewerage system renovation
Source: Own analysis
4.5. Valuation of Benefits
In an attempt to value the welfare changes coming from the described measures taken at the
Lake, all the social costs and benefits should be taken into account. The analysis of the benefit
side in this thesis can be considered to be somewhat restricted; however, the analysis itself
(see below) shows that this can be quite easily defended (see below).
The benefit analysis side can be viewed as incomplete in several respects. Firstly, it is the lack
of data regarding indirect use values such as camping, walking, cycling, changes in aesthetic
impacts. These values can be considered to be only partly included in the values coming from
the choice experiment survey. It is assumed for this reason that the real benefits brought about
by the anti-eutrophication measures are rather underestimated in the benefit analysis.
Secondly, there are also other impacts of the measures, such as recovery of the ecosystems
surrounding the Lake that are not accounted for in the analysis at all. Thirdly, there can also
be some negative welfare effects which are not accounted for (e.g., more crowded sites
negatively perceived by the residents). However, the research done by Ščasný et al., (2006)
and Škopková (2007) among residents of the town of Doksy showed a positive willingness to
pay for the water improvements.
I argue in the following CB analysis that even with these underestimated benefits, the net
present value of the undertaken anti-eutrophication measures is highly positive and for this
reason it is not necessary to proceed to any other more complicated and somehow
controversial steps such as using benefit transfer for estimating non-use values (potential use,
non-use, bequest values) coming from the anti-eutrophication measures.
As discussed above, all the activities at Lake Macha have aimed at keeping the water quality
at a satisfactory level during the summer season (that is, concentration of the cyanobacterium
lower than 100,000 cellules per millilitre) without negatively influencing the environment.
The results achieved by the measures correspond to the water quality improvement from
“dirty” on the choice experiment cards to “slightly polluted water” (no cyanobacteria29). This
corresponds to the average welfare increase for the visitors of 200 to 221 CZK/day/person
according to the model used.
4.6. Cost-benefit analysis
In the CBA the lifetime of the treatments is assumed to be until 2020, during which period all
relevant impacts on both the cost and the benefit side will be counted. The reason for this is
that cost-side measures have shrunken to PAX-18 application every second year and
removing sediments from the bottom above the protective cofferdam; the uncertainty
concerning the ecological status of the entire aquatic ecosystem grows with the time horizon
moving away; the visitation rates are also questionable due to the leisure facilities being
developed in the relatively nearby Most District (substitute), possible climate change, etc.
29
In fact there were cyanobacterium in the water all the time however they were not observable by the average
visitor and they did not represented any health danger. For this reason the given approximation as presenting
“slightly polluted water” as water without cyanobacterium is acceptable simplification for a medium informed
people. This corresponds to WHO bathing water classification (World Health Organisation (2001).
In benefits, the inflation rate and wage increase since 2005 were taken into account. All the
benefits were discounted. To discount the benefits and costs, a social discount rate of 5.5%
was used, which is recommended by the European Commission for the Cohesion Countries
(EC, 2008). The numbers of the adults visiting the Lake between 2005 and 2009 come from
the visit rate statistics ASTER (2009). The numbers of visitors between 2009 and 2020 are
counted as the average visitation in previous years (130 thousand adults per year and season).
The inflation rate is taken from the Czech Statistical Office data and Czech National Bank
projections (CZSO, 2010; CNB, 2010). The same data sources apply to the wage increase.
The inflation rate and wage increase since 2012 are assumed to be at 2.5%. The benefits were
expressed in 2007 CZK real values and the costs were taken as the real values; the values of
the costs and benefits are thus comparable throughout the time horizon.
Table 4-3 and Charts 4-2 and 4-3 below summarize the CBA results.
Table 4-3: Benefits and Net social benefits (in CZK million real values)
2005
Number of visitors (BAU Scenario)
Benefits per person and day (wage increase included)
Benefits per person and day (wage increase and inflation rate
included)
Average wage incease
Inflation rate
Benefits per year (CZK million in real terms)
Net social value (CZK million not discounted)
Net social value (CZK million discounted)
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2019
2020
181,429
121,069
99,261
123,708
130,000
130,000
130,000
130,000
130,000
130,000
130,000
130,000
130,000
130,000
130,000
186.7
192.3
200.0
203.8
205.1
207.4
214.3
219.6
225.1
230.7
236.5
242.4
248.5
254.7
261.1
267.6
183.2
3.0
1.9
21
187.6
4.0
2.5
34
200.0
4.3
2.8
24
216.6
1.9
6.3
22
207.2
0.7
1.0
26
212.4
1.1
2.4
28
219.0
3.3
2.2
28
225.1
2.5
2.5
29
230.7
2.5
2.5
30
236.5
2.5
2.5
31
242.4
2.5
2.5
32
248.5
2.5
2.5
32
254.7
2.5
2.5
33
261.1
2.5
2.5
34
267.6
2.5
2.5
35
274.3
2.5
2.5
36
13.6
13.6
28.2
26.8
16.7
15.0
14.0
11.9
21.1
17.1
15.1
11.5
28.5
20.6
25.8
17.7
30.0
19.5
27.2
16.8
26.4
15.5
28.8
16.0
Source: Own analysis
Chart 4-2: Flow of costs and benefits (in CZK million real values)
Flow of costs and benefits
35.00
30.00
25.00
20.00
CZK Million
2018
115,612
15.00
10.00
5.00
0.00
-5.00 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
-10.00
-15.00
Costs
Benefits
Source: Own analysis
33.1
17.4
30.4
15.2
34.8
16.4
27.1
12.1
Chart 4-3: Net social benefit (in CZK million real values)
Net social benefit
30.0
CZK million
25.0
20.0
15.0
10.0
5.0
0.0
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
Net social benefit
Source: Own analysis
As is obvious from the above table and charts, the discounted benefits largely outweigh the
costs of the water quality improvement measures. The discounted flow of net benefits during
the lifetime of the project is displayed in Chart 4-3. For the defined assumptions, the net
social benefits amount to CZK 263 million in the given time horizon (until 2020).
4.6.1. Sensitivity analysis
The sensitivity scenario analysis was then performed in order to identify the critical variables
for the value of NSB. The table below presents scenarios for each critical variable in the
extreme range of values.
Table 4-4: Net social benefits in 2007 million and scenario modeling
Scenario
Visitation
WTP
Costs
Disc. Rate
NSB
Pessimistic
39,000
40
86.5
0.0825
-41.4
Base
130,000
200
75.9
0.055
263.3
Optimistic
221,000
360
65.3
0.0275
969.1
Notes:
Visitation - number of visitors per year,
WTP – visitors’ willingness to pay,
Costs – aggregated costs of the anti-phosphorus measures in the time horizon,
Disc. Rate - the discount rate,
NSB - net social benefits
Source: Own analysis
Pessimistic scenario. It is represented by a combination of “high” costs and “low”
benefits.
Most probable or base scenario. This represents a combination of the benefits and
costs with the highest probability of being observed.
Optimistic scenario. Contrary to the pessimistic scenario, it is a combination of “low”
costs and “high” benefits.
As already mentioned, the anti-phosphorus treatments are highly beneficial and even more so
when considering the optimistic scenario; it yields net social benefits amounting to CZK 969
million. However, in the pessimistic scenario, the treatments are not socially profitable. The
net loss amounts to CZK 41 million. However, when interpreting these figures, attention
should be paid to the particular values in the scenarios. In my opinion, both scenarios
(pessimistic and optimistic) are not very plausible regarding the WTP values and in the
pessimistic scenario, also the visitation rate is extremely low.
To see which critical variables strongly influence the results of the CBA, a sensitivity
(elasticity) analysis for all critical variables was performed. The ranges (minimum and
maximum values) are presented for those variables in Table 4-5. The same values as in
scenario analysis were used.
Table 4-5: The range values of critical variables
Scenario
pi,min
pi,ref
pi,max
Visitation
39,000
130,000
221,000
WTP
40
200
360
Costs
65.3
75.9
86.5
Disc. Rate
0.0275
0.055
0.0825
Notes:
pi,min - the minimum value of a critical parameter,
pi,ref - the central value of a critical parameter,
pi,max - the maximum value of a critical parameter
Visitation - number of visitors per year,
WTP – visitors’ willingness to pay,
Costs – aggregated costs of the anti-phosphorus measures in the time horizon,
Disc. Rate - the discount rate
Source: Own analysis
Sensitivity analysis determines the net social benefits for the suggested project by varying the
individual parameters. This is done for each variable separately, with all other variables
remaining constant. The results are displayed in the sensitivity analysis diagram (Chart 4-4).
Chart 4-4: Sensitivity analysis diagram
Sensitivity analysis: Spider diagram
NSB/NSBref
2.00
1.50
1.00
0.50
0.00
-1
-0.8
-0.6
Visitation
-0.4
WTP
-0.2
Costs
0
0.2
Dis. Rate
0.4
0.6
0.8
1
pi/pi,ref
Source: Own analysis
Critical factor analysis
As is evident from the sensitivity analysis results (Chart 4-4), the NSB are sensitive to
variation in:
•
the value of visitors’ willingness to pay for water quality improvements; and
•
the numbers of visitors in particular years.
In other words, the sensitivity analysis shows us that the NSB are sensitive to the variation in
benefits. Lowering or increasing the WTP value by 10% would lead to a change in the NSB
of about 12 %. Lowering or increasing the visitation rate by 10% would lead to a change in
the NSB of approximately 8%. The other two analyzed factors, i.e., costs and discount rate,
have a negligible influence on the net social benefit.
4.6.2. Monte Carlo simulation
It is quite intuitive that particular values of both the identified factors with a great influence
on the NPV (i.e., WTP and visitation rate) in particular years are not fixed. For this reason, we
can consider them to be stochastic and not deterministic variables somehow distributed (with
a given distribution). Assuming this, we can then simulate the value of the dependent variable
(here, NPV) on randomly generated values of the stochastic variables (here, WTP and
visitation rate). This procedure can be done e.g. using Monte Carlo simulation (Rabl and
Zwaan, 2009). “Monte Carlo analysis is a risk modelling technique that uses statistical
sampling and probability distributions to simulate the effects of uncertain variable on model
outcomes” (Boardman et al., 2006).
However, to run the simulation, the type and parameters of the distributions have to be
estimated. This can be done for the WTP because as the best fit of the mixed logit model (see
previous chapter) showed, both the “fee” and “dirty water” parameters are random with
normal distribution30. These parameters are needed for calculating the WTP.
For the variable “number of visitors”, the type and parameters of the distribution cannot be
estimated due to the very short visitation rate data series. For this reason, the following Monte
Carlo simulation was done only for the WTP variation.
The simulation was done using the publicly available MS Excel Visual Basic supplement
made at The University of Texas in Austin (Jensen, 2004). The simulations were done by
iterated calculation with 1,000 iteration with random input generation based on the parameters
derived from the mixed logit model (see Table 4-6).
Table 4-6: Random parameter distribution assumptions for WTP calculation
Dirty
Fee
Mean
Standard deviation
-2.6486
-0.0132
0.7949
0.0034
Source: Own analysis
In the simulation, one thousand iterations were used to simulate the WTP measures. The
values (using the Limdep software) are plotted in the histogram below (Chart 4-5). The data
show that the WTP is more or less log normally distributed, which is in accordance with the
theory.
30
As was discussed in the case of the “dirty water”, “entrance fees”, and “beach equipment” the estimated standard
deviations of the parameters were statistically significant. The standard deviations relate to the extent of the
dispersion around the mean of the parameter. This suggests that there exists heterogeneity in the parameter estimates
across the sample population (around the mean parameter estimate). It can be interpreted in a way that different
individuals have different preferences that differ from the mean estimate for the sample population.
F req u en cy
Chart 4-5: The willingness to pay simulation
(in CZK million real values)
4. 291
106. 113
207. 936
309. 758
411. 581
513. 403
615. 225
717. 048
WTP
Source: Own analysis
Based on these simulated WTP values, the model for the NPV computation was created and
the net present value was simulated. The results of this simulation are shown in the histogram
below (Chart 4-6) .
F req u en cy
Chart 4-6: The net present value simulation
(in CZK million real values)
- 47. 671
111. 198
270. 067
428. 936
587. 805
NETPRESE
Source: Own analysis
746. 674
905. 543
1064. 411
As visible from the histogram, the vast majority of the simulated net present values is
positive; the probability of the NPV to be negative is very low. The distribuion as such
provides us with the indication of the variability and robustness of the NPV estimated which
is quite satisfactory regarding ploted results.
Based on these interesting findings, the lowest average WTP for the respondents was modeled
so that the NPV remains positive (the lowest possible WTP was modelled in a way to retain
the NPV positive). Holding other assumptions constant, the model showed the lowest possible
average WTP at an amount of CZK 34 in 2007. All the results of the model under the
assumption of this level of WTP are depicted in the table and charts below.
Table 4-7: Results for the zero Net social benefits
2005
115,612
Number of visitors (BAU Scenario)
Benefits per person and day (wage increase
included)
Benefits per person and day (wage increase and
inflation rate included)
Average wage incease
Inflation rate
Benefits per year (CZK million in real terms)
Net social value (CZK million not discounted)
Net social value (CZK million discounted)
2006
181,429
2007
121,069
2008
99,261
2009
123,708
2010
130,000
2011
130,000
2012
130,000
2013
130,000
2014
130,000
2015
130,000
2016
130,000
2017
130,000
2020
130,000
33.0
34.3
34.9
35.2
35.6
36.7
37.7
38.6
39.6
40.6
41.6
42.6
43.7
44.8
45.9
31.4
3.0
1.9
3.6
32.2
4.0
2.5
5.8
34.3
4.3
2.8
4.2
37.2
1.9
6.3
3.7
35.5
0.7
1.0
4.4
36.4
1.1
2.4
4.7
37.5
3.3
2.2
4.9
38.6
2.5
2.5
5.0
39.6
2.5
2.5
5.1
40.6
2.5
2.5
5.3
41.6
2.5
2.5
5.4
42.6
2.5
2.5
5.5
43.7
2.5
2.5
5.7
44.8
2.5
2.5
5.8
45.9
2.5
2.5
6.0
47.0
2.5
2.5
6.1
-3.97
-3.97
0.03
0.03
-3.33
-2.99
-3.80
-3.23
-0.09
-0.07
-7.82
-5.98
4.88
3.54
1.52
1.04
5.14
3.35
1.77
1.09
0.33
0.20
2.04
1.13
5.68
2.99
2.32
1.16
5.97
2.82
-2.46
-1.10
Chart 4-7: Flow of costs and benefits in particular years (in CZK million real values)
Flow of costs and benefits
8.00
6.00
4.00
2.00
CZK Million
2019
130,000
32.0
Source: Own analysis
0.00
-2.00
2018
130,000
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
-4.00
-6.00
-8.00
-10.00
-12.00
Costs
Benefits
Source: Own analysis
Chart 4-8: Flow net social benefit (summed up = 0)
(in CZK million real values)
Net social benefit
6.00
2.00
20
08
20
09
20
10
20
11
20
12
20
13
20
14
20
15
20
16
20
17
20
18
20
19
20
20
20
07
-2.00
20
06
0.00
20
05
CZK million
4.00
-4.00
-6.00
-8.00
Net social benefit
Source: Own analysis
4.7. Summary
In this chapter, cost-benefit analysis was applied in order to compare the economic costs and
benefits associated with eutrophication prevention measures at Lake Macha. The chapter
includes a particularly detailed analysis of the costs of these measures and, above all, the benefits
derived from them.
The cost-side analysis was preceded by an analysis of the total phosphorus balance (being the
chief environmental stressor) in the Lake, based on numerous yet incomplete data sources.
Measures that have had a direct effect on reducing the trophic level of the Lake waters were
identified based on this detailed analysis (contrary to the assumption, the renovation of the
wastewater treatment plant in Staré Splavy was thus not included in the costs, for instance).
The benefit analysis was based chiefly on the visitors’ preference survey conducted at Lake
Macha in 2007. The results of measures taken showed that the water quality has improved in
their wake from strongly eutrophic water with cyanobacterium colonies (so-called algal bloom),
when the Lake was useless for recreation (associated with contact with water) to a grade where
cyanobacteria are present in the Lake at an extent not observable to the common user.
Nevertheless, algae continue to be present in the water, making it relatively cloudy. This water
quality improvement corresponds to the improvement from “dirty water” to “slightly polluted
water” in the choice experiment. 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.
Subsequently, the data were cleansed of inflation and wage increases. The following conversion
of the data to the net present value and comparison of costs and benefits showed that the
measures taken are very effective measured as net social benefits (i.e., benefits minus costs of
measures).
Using the social discount rate of 5.5%, the net social benefits of the anti-phosphorus measures
were estimated at CZK 263 million, which means that the phosphorus treatment is socially
beneficial.
After this finding, the other two scenarios were defined (pessimistic and optimistic). The results
showed that in the pessimistic scenario, the project is socially non-profitable, the net loss
amounts to CZK 41.4 million; in the optimistic scenario, the net profit is a full CZK 969.1
million.
Because of these findings, a sensitivity analysis was also performed so that the critical variables
could be defined. The analysis showed that high uncertainties are mainly associated with the
benefit side, i.e., the willingness to pay for water improvement and the numbers of visitors.
This gave an incentive for the Monte Carlo analysis of the main uncertainty – the WTP. The
main question was how sensitive the NSB value is to the range of the WTP with the given
distribution. The result showed that there is a very low probability of the NSB being negative.
The question that drove the other data evaluation was, ”What is the lowest WTP that would
guarantee NSB to remain positive?” The model showed the value of CZK 34/day/visitor with the
other assumptions being equal.
To sum up, the analysis conducted in this chapter showed that the eutrophication prevention
measures at Lake Macha are highly effective using net social benefits as the main efficiency
indicator. The average WTP for the visitor used in the models (see above) was CZK
200/day/visitor, but only CZK 34/day/visitor was necessary for the NSB to remain positive.
These results could be very useful in other discussions regarding future treatment of Lake Macha
site and making policy recommendation.
Conclusions
The excessive eutrophication of rivers, lakes and other water bodies became a serious
environmental problem with broad economic consequences in the last few decades. The
eutrophication has accelerated due to massive amounts of nutrients (phosphorus and nitrogen)
from different human activities, mainly from agriculture and municipal sewage discharges. At
present, virtually every industrialised country of the world is tackling excessive eutrophication
and its negative consequences.
The consequences of water eutrophication and the quantity of affected sites indicate the
significant economic dimension of the problem. It is clear that clean water is becoming an
increasingly scarce good for both recreational and other purposes. Many epidemiological studies
world-wide deal with the consequences of water eutrophication to human health. Since the issue
has an economic dimension, significant attention is paid world-wide to studying the impacts of
water quality and water management on welfare changes. Some of the study areas were indicated
in the thesis; special attention was paid to studies dealing with impacts on recreational utility.
The literature review showed that eutrophicated reservoirs are not sporadic cases also in the
Czech Republic, but that the problem is much more extensive – approximately 80% of the water
reservoirs are excessively eutrophic. The purpose of the thesis was to contribute to the handling
of the issue, chiefly to discussing its economic aspects. In particular, it provided an economic
lookout on the recreational exploitation of water, and the effect of excessive eutrophication of
water on the (economic) preferences of holiday makers and the efficiency of the economic and
policy measures in order to reduce the eutrophication. A secondary purpose of the paper was to
verify the robustness of methods used in modelling discrete type data.
Since most of the reservoirs are quite unique cases, the economic impacts of water eutrophication
on summer recreational utility were analysed on a case study, allowing the researcher to work in
a required depth. Given the long-term tradition of summer recreation by Lake Macha, the
enormous visitation rate of the water body, and the serious ecological problems that the area
suffers from, the area was chosen for the case study of modelling demand for summer recreation
and deriving welfare changes due to water quality changes. Lake Macha is also very important to
the regional economy and is part of the ecologically valuable ecosystems of Českolipsko and a
Special Protection Area.
The following research hypotheses were formulated progressively based on the thesis objectives:
1. Water quality is the factor most significantly affecting people’s recreational utility from the
chosen site.
2. The confirmation or rejection of the above hypothesis is independent of the choice of the
discrete choice econometric model.
3. Measures taken at Lake Macha in order to improve the water quality are economically
efficient.
The choice experiment method was chosen as the research methodology suitable for answering
the first hypothesis. The cost-benefit analysis was chosen for the third hypothesis.
Based on a literature survey, several group interviews and on-site preliminary survey, the
demand determinants (attributes in choice experiment terms) and their levels were determined in
the preliminary survey. At the same time, the preliminary survey showed that there are only very
few choice experiment studies devoted to water eutrophication. Following several stages of the
preliminary survey, conducted directly on the Lake Macha beaches, the following attributes and
their levels (in brackets) were selected for the main data collection: water quality (as clean water,
slightly polluted water, and polluted water), beach overcrowding (yes, no), beach toilet and
refreshment facilities (yes, no), and entrance fee per person per day (CZK 40, 80, and 150).
The final data collection took place in two waves, its target population being visitors to the four
paid beaches surrounding the Lake. After data cleansing, the collection resulted in 331 valid
questionnaires. The data were collected using a random stratified collection. The data description
revealed no counter-intuitive findings: quite to contrary, for example the inability to discern
nuances in water quality corresponded to some other outcomes on water quality studies
conducted earlier.
The choice experiment data analysis showed that all the attributes used in the study are
significant at the 1% level of significance and the water quality is the factor that most
significantly affects people’s recreational utility. The results also showed the high value that
vacationists place on all attributes used in the choice experiment. The people’s marginal
willingness to pay for lower level of overcrowding on the beaches is CZK 18 per entrance paid.
People are willing to pay an additional CZK 100 for clear water in the Lake. The marginal
willingness to pay for dirty water is significant and negative amounting to CZK 200. If the
beaches are not equipped with refreshments and toilets, the marginal willingness to pay is
negative, amounting to CZK 74. The first hypothesis was thus confirmed, because changes in the
water quality are the most significant factor influencing the people’s utility during their stay on
the beach.
For the data analysis, various types of discrete choice models were developed (applied). The
models were compared and discussed – particularly the multinomial logit model, the nested logit
model, the probit model, and the state of art being the random parameters logit model. The
results of the data analysis showed that the model estimates are nearly the same for the first three
models and the estimates are only slightly different for the random parameters logit model
(RPL); however, the order of the attributes size did not change even in the case of the RPL
model. The confirmation or rejection of the first hypothesis is thus not dependent on the choice
of the econometric model. The second hypothesis was also confirmed.
After the first and second hypotheses were confirmed, the cost-benefit analysis could be made
because the confirmation of the first two hypotheses gave validity to the estimated benefits. On
the cost side, the correct determination of the eutrophication prevention measures was crucial.
The cost-side analysis was preceded by an analysis of the total phosphorus balance (being the
chief environmental stressor) in the Lake. Measures that have had a direct effect on reducing the
trophic level of the Lake waters were identified based on this detailed analysis. After that, the
associated flows of costs in time were estimated. Afterwards, the economic costs and benefits at
Lake Macha were compared.
The benefit analysis was based chiefly on the visitors’ preference survey conducted at Lake
Macha in 2007. The results of measures taken showed that the water quality has improved in
their wake from strongly eutrophic water with cyanobacterium colonies (so-called algal bloom),
when the Lake was useless for recreation (associated with contact with water) to a grade where
cyanobacteria are present in the Lake at an extent not observable to the common user and they do
not pose any health risk to the recreationists. Nevertheless, algae continue to be present in the
water, making it relatively cloudy. This water quality improvement corresponds to the
improvement from “dirty water” to “slightly polluted water” in the choice experiment. The most
conservative estimates from the preferred statistical model were used for aggregation, as is the
usual practice in stated preference studies.
Using the social discount rate of 5.5% for reasons being discussed in the thesis, the net social
benefits of the anti-phosphorus measures were estimated at CZK 263 million, which means that
the phosphorus treatment is socially beneficial. That means also the confirmation of the last
hypotheses. Some further steps were also taken in the cost-benefit analysis (e.g., sensitivity
analysis); however, the results suggested that the NSB are not very sensitive to the defined
assumption of the analysis. The sensitivity analysis showed that high uncertainties are associated
mainly with the benefit side, i.e., the willingness to pay for water improvement and the numbers
of visitors.
This gave an incentive for the Monte Carlo analysis of the main uncertainty – the WTP. The
main question was how sensitive the NSB value is to the range of the WTP with the given
distribution parameters. The result showed that there is a very low probability of the NSB being
negative. The results of the other optimisation model showed that the lowest WTP that would
guarantee NSB to remain positive amounts only to CZK 34/day/visitor with the other
assumptions being equal.
The results of the cost-benefit analysis could be very useful in other discussions regarding future
treatment of the Lake Macha site and making policy recommendations.
In conclusion, all the hypotheses formulated in the introduction to the present dissertation were
confirmed. The confirmation of the first hypothesis shows water eutrophication in relation to
summer recreation as a significant economic and policy issue, certainly worthy of the attention of
economic policy. The second hypothesis, or the analysis that answered it, opens room for further
discussion of the listed models. These models are not very well-known in the Czech Republic in
spite of the considerable attention paid to developing them at leading universities abroad. The
confirmation of the third hypothesis – with a great margin – points at the great efficiency of the
expended funds to reduce the eutrophication in the study area. In light of the present analysis,
therefore, more room opens up for discussing costs that could be invested in eutrophicated
reservoirs very efficiently. The cost-benefit analysis also puts the actual situation at Lake Macha
in a different light: fights break out every year concerning some CZK 3 million, required to keep
the water quality at an acceptable level for summer recreation. Room also opens up for
discussing a different arrangement of property relationships in the area so that the problem could
be resolved without interference of public authorities.
Given the limited scope of the present paper, certain additional topics offering themselves for
discussions were not treated herein. For example, the paper as a whole did not start from the
neoclassical economic paradigm; it made no broader discussion of possible applications of the
analysis results as a source of information for beach operators, who might theoretically consider
raising the entrance fee and using the revenues for improving the Lake water quality. They would
only extract a part of the Lake visitors’ increased benefits due to the improved water quality – it
could still be a Pareto-effective change, but would aim at a discussion of market
environmentalism discursus.
At the same time, the paper could have discussed the water eutrophication issue in the Czech
Republic from a different angle, such as national-level costs of reducing eutrophication, or the
ecological economics point of view – a problem of conserving valuable wetlands and aquatic
ecosystems without a purely anthropocentric view of the recreational value of water exploitation.
These additional possible research directions are suggestions for the author of the present paper
to possibly continue research in the area in future.
In the author’s opinion, the most interesting question raised by the paper is why water
eutrophication is so inadequately addressed in the Czech Republic. Is it a failure of the
environmental policy, or of its makers who do not fully realise the scope of the problem, or is it
an inappropriate arrangement of property rights failing to allow the market or the state to handle
the problem effectively, or is the problem somewhere else altogether?
The paper thus raises a number of research questions and so partially contributes to the
discussion of surface water eutrophication in the Czech Republic, as has been its broader
objective.
References
Abou-Ali, H., Carlsson, F. (2004): Evaluating the Welfare Effects of Improved Water Quality
Using the Choice Experiment Method", Working Papers in Economics no. 131, Department of
Economics, Gothenburg University.
Ackman D, Marks S, Mack P, Caldwell M, Root T, Birkhead G (1997): Swimming associated
hemorrhagic colitis due to Escherichia coli O157:H7 infection: evidence of prolonged
contamination of a fresh water lake. Epidemiology and infections, 119: 1–8.
Adamowicz, W., J. Louviere, and M. Williams (1994): Combining Revealed and Stated
Preference Methods for Valuing Environmental Amenities, Journal of Environmental Economics
and Management 26, no. 2, 271-292.
Agency for Nature Conservation and Landscape Protection of the Czech Republic official
website, cit: 12.5.2007, online:
http://www.ochranaprirody.cz/?cmd=page&type=107&lang=cs&query=M%E1chovo+jezero,
Agresti A. (2002): Categorical Data Analysis, Second Edition, Published by John Wiley & Sons,
Inc., Hoboken, New Jersey.
Ahmad, J., B. Goldar and S. Misra, "Value of Arsenic-Free Drinking Water to Rural Households
in Bangladesh", Journal of Environmental Management 74, 173-185 , 2005
ASTER (2009): Reporty odbavovacího systému Regio máchova kraje, a.s., zpracovatel ASTER
spol. s r.o., issued 28.9.2009.
Bahlaoui MA, Baleuz B, Troussellier M (1997) Dynamics of pollution-indicator and pathogenic
bacteria in high-rate oxidation wastewater treatment ponds. Water research, 31: 630–638.
Bachmann, R.W., Hoyer, M.V., Canfield, D.E. (1999): The Restoration Of Lake Apopka In
Relation to Alternative Stable States, Hydrobiologia 394, pp. 219-232.
Bateman I. J., Carson R. T., Hanemann M., Hanley N., Hett T., Jones-Lee M., Loomes G.,
Mourato S., Özdemiroğlu E., Pearce D. W., Sugden R., Swanson J. (2002): Economic Valuation
with Stated Preference Techniques: A Manual, Cheltenham: Edward Elgar, UK.
Bernes, C. (2000): Eutrophication Of Soil And Water. Swedish Environmental Protection
Agency, Swedien Environet.
Binkley, Clark S. and W. Michael Hanemann (1978): The Recreation Benefits of Water Quality
Improvement: Analysis of Day Trips in an Urban Setting", U.S. EPA, Washington, D.C., 1978.
Biro, P. (1997): Temporal variation in lake Balaton and its fish populations. Ecology of
freshwater fish, vol. 6, no. 4, pp. 196-216.
Birol, E., Karousakis, K., Koundouri, P. (2005): Using a Choice Experiment to Estimate the
Non-Use Values of Wetlands: The Case of Cheimaditida Wetland in Greece", Discussion Paper
Series no. 08.2005, Environmental Economy and Policy Research, University of Cambridge.
Blamey, R., Gordon, J., Chapman, R. (1999): Choice Modelling: Assessing the Environmental
Values of Water Supply Options, Australian Journal of Agricultural and Resource Economics,
no. 43(3): 337 – 357.
Boardman, A., Greenberg, D., Vining, A., Weimer, D. (2006): Cost Benefit Analysis: Concepts
and Practice, 3/E. Pearson Prentice Hall, Upper Saddle River, US. ISBN-13: 9780131435834.
Brouwer, R., Bateman, I.J. (2005): Temporal Stability and Transferability of Models of
Willingness to Pay for Flood Control and Wetland Conservation, Water Resources Research, vol.
41, no.3, W03017.
Carlsson, F., Martinsson, P. (2007): How Much is Too Much? An Investigation of the Effect of
the Number of Choice Sets, Context Dependence and the Choice of Bid Vectors in Choice
Experiments, Springer Science+Business, Media B.V.
Carlsson, S. (2005): Precipitation of lakes, Vattenresurs AB, Tjusta, SE-19793 Bro, Sweden, cit:
16.5.2008, online: http://www.vattenresurs.se/Precipitation.pdf.
Carson, R. T., Mitchell, R. C. (1993): The Value of Clean Water: The Public's Willingness to Pay
for Boatable, Fishable, and Swimable Quality Water, Water Resources Research, 29, 2445-54.
CNB (2010): Official website of the Czech National Bank, cit: 15.1.2010, online:
http://www.cnb.cz/cs/menova_politika/prognoza/index.html.
Cooke, G. D., Welch (1999): Effectiveness and longevity of phosphorus inactivation with alum.
J.Lake and reserv.manag. 15:3-27, pp. 4.
Cordato R.E. (2004): Toward an Austrian Theory of Environmental Economics, Quarterly
Journal of Austrian Economics, 7(1), 3-36.
ČVUT (2009): Optimalizace stavu vodních ekosystémů v povodí Robečského potoka II, České
vysoké učení technické v Praze, Fakulta stavební, katedra Hydromeliorací a krajinného
inženýrství, Praha, cit: 10.3.2010, online:
http://www.ochranaprirody.cz/res/data/187/024121.pdf.
CZSO (2010): Official website of the Czech Statistical Office, cit: 15.1.2010, online:
http://www.czso.cz/csu/redakce.nsf/i/mira_inflace.
Doksy (2007): Doksy municipality official website, cit. 12.5.2007 – 29.6.2007, online:
http://www.doksy.com.
Domencich T., McFadden D. (1975): Urban Travel Demand A Behavioural Analysis, New York
and Oxford: Elsevier.
EC (2007): Bathing water quality annual report, Bathing water results 2007 – Czech Republic,
cit: 12.09.2009, online:
http://ec.europa.eu/environment/water/water-bathing/report2008/cs_comments.pdf.
EC (2008): Structural Funds, Cohesion Fund and Instrument for Pre-Accession, European
Commission, Directorate General Regional Policy, Final Report, Submitted by TRT Trasporti e
Territorio and CSIL Centre for Industrial Studies. Cit: 12.1.2010, online:
http://www.eurosfaire.prd.fr/7pc/doc/1215583204_costs_benefits_analysis_guide_2008.pdf
Economics for the Environment Consultancy Ltd, Valuation of Benefits to England and Wales
of a Revised Bathing Water Quality Directive and Other Beach Characteristics Using the Choice
Experiment Methodology, cit. 6.11.2007, Online:
http://www.defra.gov.uk/environment/water/quality/bathing/pdf/bw_study4a.pdf
Eggert, H., Olsson, B. (2004): Heterogeneous Preferences for Marine Amenities: A Choice
Experiment Applied to Water Quality, Working Paper in Economics no. 126, Department of
Economics, Goteborg University.
Freeman, A.M. (2003): The Measurement of Environmental and Resource Values: Theory and
Methods, Resources for the Future, Washington, DC, ISBN: 1-89185-362-7.
Feenberg, D., Mills, E.S. (1980): Measuring the benefits of water pollution abatement, New
York: Academic Press, pp. 181, New York : Academic Press, ISBN 0122509501
Figueras MJ, Robertson W, Pike EB, Ashbolt NJ, Borrego JJ (2000) Sanitary inspection and
microbiological water quality. In: Bartram J, Rees G, ed. Monitoring bathing waters: a practical
guide to the design and implementation of assessments and monitoring programmes. London, E
& FN Spons, pp. 113–167. Published on behalf of the World Health Organization, Commission
of the European Communities and US Environmental Protection Agency.
Findlay, D.L., Kling, H.J., Roenicke, H., Findlay, W.J.: A Paleolimnological Study Of
Eutrophied Lake Arendsee (Germany). J. Paleolimnol. Vol. 19, No. 1, Pp. 41-54, 1998.
Garrod, G., Willis, K. (1998): Using Contingent Ranking to Estimate the Loss of Amenity Value
for Inland Waterways from Public Utility Structures, Environmental and Resource Economics
12, 241-247 ,
Garrod, G., Willis, K. (1999): Economic valuation of the environment: Methods and case studies,
Cheltenham: Edward Elgar Publishing Company.
Georgiou, S., Bateman, I.J. (2005): Revision of the EU Bathing Water Directive: economic costs
and benefity, Marine Pollution Bulletin 50, Elsevier, pp. 430–438.
Georgiou, S., I.J. Bateman, I.H. Langford and R.J. Day (2000): Coastal Bathing Water Health
Risks: Developing Means of Assessing the Adequacy of Proposals to Amend the 1976 EC
Directive, Risk Decision and Policy, Vol. 5, 49-68.
Green, C.H., Tunstall, S.M. (1991): The Evaluation of River Quality Improvements by the
Contingent Valuation Method, Applied Economics 23, 1135-1146.
Hally, J. (2002): Prací prostředky, životní prostředí a environmentální výchova, České sdružení
výrobců mýdla, čisticích a pracích prostředků, Univerzita Karlova - Pedagogická fakulta, Katedra
chemie a didaktiky chemie, Praha.
Hampl M. (2001): Trojí přístup k veřejným statkům, Finance a úvěr, 51(2), 111-125.
Hanemann M. (1984): Welfare evaluations in contingent valuation experiments with discrete
responses, American Journal of Agriculture Economics, 66(3), 332 – 341.
Håkan, E., Olsson B. (2004): Heterogeneous preferences for marine amenities: A choice
experiment applied to water quality, Working Papers in Economics, nr 126. Report.
Hanley et al., (2001): Choice Modelling Approaches: A Superior Alternative for Environmental
Valuation?, Journal of Economic Surveys, Vol. 15, No. 3, Blackwell Publishers Ltd.
Hanley N., Bell D., Alvarez-Farizo B. (2003), Valuing the benefits of coastal water quality
improvements using contingent and real behaviour’, Environmental and Resource Economics 24,
273-285.
Hanson, T.R., Hatch, L.U. (1995): Impact of Reservoir Water Level Changes on Lakefront
Property and Recreational Values, Department of Agricultural Economics, Mississippi State
University.
Hausman J.A., Wise D.A. (1978): A Conditional Probit Model for Qualitative Choice: Discrete
Decisions Recognizing Interdependence and Heterogeneous Preferences, Econometrica 46, 403426.
Hendl, J. (2005): Kvalitativní výzkum -- Základní teorie a aplikace, 2nd edition, Praha: Portál.
Hendl, J. (2006): Přehled statistických metod zpracování dat, 2nd edition, Praha: Portál.
Henscher, D. (1994): Stated preference analysis of travel choices: the state of practice.
Transportation 21, pp. 107-133.
Hensher D.A., Greene W.H. (2003): The Mixed Logit Model: The State of Practice,
Transportation 30, 133-176.
Hensher, D.A., Rose, J.M., Greene W.H. (2005): Applied Choice Analysis, Cambridge
University Press, New York.
Hosmer D. W., Lemeshow S. (2000): Applied Logistic Regression, John Wiley & Sons, New
York.
Choe, K., Whittington, D., Lauria, D.T. (1996): The Economic Benefits of Surface Water Quality
Improvements in Developing Countries: A Case Study of Davao, Philippines, Land Economics
72, no. 4, 519-537.
Integrovaný registr znečištění, Ohlašované látky - Celkový fosfor, Dopady na životní prostředí,
cit: 12.1.2010, on-line: http://www.irz.cz/latky/celkovy_fosfor.
Jensen, P.A. (2004): MS Excel Visual Basic supplement, The University of Texas in Austin,
cit:10.6.2009, online:
http://www.me.utexas.edu/~jensen/ORMM/frontpage/jensen.lib/index.html.
Kemwater ProChemie (2010): Kemwater ProChemie s.r.o. official website, sekce Obnova jezer,
cit: 20.1.2010, online: http://www.prochemie.cz/Machovo-jezero_2006.pdf.
Klouček, V., Vaverová, I. (2005): Lake restoration: Rekultivace eutrofizovaných nádrží metodou
srážení fosforu hlinitými solemi, in Vodní Hospodářství 4/2005.
Kočí, V., Burkhard, J., Maršálek, B. (2000): Eutrofizace na přelomu tisíciletí. Eutrofizace 2000,
Praha, pp. 3-13.
Kolstad, C.D. (2000): Environmental Economics. Oxford University Press, New York. ISBN 019-511954-1.
Lancaster K. (1966): A new approach to consumer theory, Journal of Political Economy, 74,
132-157.
Liberec Regional Authority (2009): Liberec Regional Authority official website, cit: 3.9. 2009,
online:
http://zivotni-prostredi.kraj-lbc.cz/page1923.
Louviere, J. (2001): Choice Experiments: an Overview of Concepts and Issues, in Bennett J.,
Blamey R., The Choice Modelling Approach to Environmental Valuation, Edward Edgar,
Cheltenham.
Louviere J., Hensher D.A., Swait J. (2000): Stated Choice Methods: Analysis and Applications in
Marketing, Transportation and Environmental Valuation, Cambridge: Cambridge University
Press.
Luce, R.D. (1959): Individual choice behaviour: a theoretical analysis, New York: Wiley.
Manski Ch. F. (1977): The structure of random utility models, Theory and Decision 8 (3), 229254.
Markandya, A., Harou, P., Bellù, L. G., Cistulli, V. (2002): Environmental Economics for
Sustainable Growth: A Handbook for Practitioners. Edward Edgar. Cheltenham.
Menesguen, A. (1999): Use Of Ecological Models In The Fight Against Eutrophication Of
French Coastal Waters, No. 24, Actes Colloq, Ifremer, pp. 31-48.
McFadden D. (1974): Conditional logit analysis of qualitative choice behaviour, in Zarembka, P.
(ed.), Frontiers in econometrics, New York: Academic Press.
McFadden D., Train K., Tye W.B. (1976): An Application of Diagnostic Tests for the
Independence from Irrelevant Alternatives Property of the Multinomial Logit Model,
Transportation Research Record 637, 39-45.
Milon, J.W., Scrogin, D. (2005): Latent Preferences and Valuation of Wetland Ecosystem
Restoration.", Ecological Economics, In Press.
MZe (2008): Zpráva o stavu vodního hospodářství České republiky v roce 2008, Odbor státní
správy, osvěty a informačních systémů ve vodním hospodářství Ministerstvo zemědělství, Odbor
ochrany vod, Ministerstvo životního prostředí, cit. 27.12.2009, online:
http://eagri.cz/public/eagri/pub/9e/9f/20/6531_4_modra_zprava_2008.pdf.
Nesměrák, I. (2006): Interní materiály Ivana Nesměráka, VUV T.G.Masaryka, Praha.
Nicholls, K.H. (1998): El Nino, Ice Cover, and Great Lakes Phosphorus: Implications for
Climate Warming, Limnology and Oceanography, Vol. 43, No. 4, pp. 715-719.
Novaczek, I. (1995): Possible roles for marine protected areas in establishing sustainable
fisheries in Canada, pages 31-36, in N. L. Shackell and J. H. M. Willison editors. Marine
protected areas and sustainable fisheries, Centre for Wildlife and Conservation Biology, Acadia
University, Wolfville, Nova Scotia, Canada.
Parsons, G.R., Helm, E.C., Bondelid, T. (2003): Measuring the Economic Benefits of Water
Quality Improvements to Recreational Uses in Six Northeastern States: An Application of the
Random Utility Maximization Model, University of Delaware Graduate College of Marine
Studies.
Pecáková I. (2007): Explained Variation Measures for Models with Categorical Responses, in:
AMSE 2007 [CD-ROM], Banská Bystrica, 177 – 182.
Phaneuf D. J., Smith V. K. (2005): Recreation Demand Models, In: Mäler, K. G. a Vincent J. R.
(eds.), Handbook of Environmental Economics, 2, Elsevier.
Public Administration official server, cit: 12.1.2010, online:
http://www.portal.gov.cz/wps/portal/_s.155/7226/_s.155/10202?docid=2312.
Rabl, A., Zwaan, B (2009): Cost–benefit analysis of climate change dynamics: uncertainties and
the value of information, Climatic Change 96, no. 3, pp. 313-333.
Rybářství Doksy (2010): Rybářství Doksy s.r.o. official website; cit: 21.1.2010, online:
http://www.rybarstvidoksy.cz.
Sanchirico, J., James, W. (2000): Dynamics of Spatial Exploitation: A Metapopulation
Approach, Discussion Papers dp-00-25-rev, Resources For the Future.
SAS Institute Inc. (1997): Getting started with the market research application, Cary, NC: SAS
Institute Inc.
Ščasný, M., Czajkowski, M., Markowska, A., Melichar, J., Škopková, H., Markiewicz, O.,
Bartczak, A. (2006): Water Quality Valuation: An empirical Evidence from CV Surveys in the
Czech Republic and Poland. Paper prepared for the the 3rd Annual Congress of Association of
Environmental and Resource Economics AERE, Kyoto, 4-7 July.
Šímová, T. (2007): Fenomén hodnoty v ekonomii přírodních zdrojů, In: Lenka Čamarová (ed.).
Ekonomie přírodních zdrojů. nakladatelství Alfa Publishing, s.r.o., společně s Liberálním
Institutem, Praha.
Škopková, H. (2007): Oceňování kvality vod: Případová studie Máchova jezera, Diploma
Theses, Charles University, Prague.
Šímová T. (2007): Fenomén hodnoty v ekonomii přírodních zdrojů, In: Lenka Čamarová (ed.).
Ekonomie přírodních zdrojů. Alfa Publishing, s.r.o. and Liberální Institut, Praha.
Slavík, J. (2007): Neoklasická ekonomie a ochrana životního prostředí. Politická ekonomie,
55(4), 526–538.
Smith, V.K., Desvousges, W.H. (1987): An empirical analysis of the economic value of risk
changes, Journal of Political Economy, 95(1), 89-114.
Soederqvist, T. (1998): Why Give Up Money For The Baltic Sea?, Environmental And Resource
Economics, Vol. 12, No. 2, pp. 249-254.
Steer Davies Gleave (2000): London underground customer priorities research, Report for
London underground.
Tagliapietra, D., Pavan, M., Wagner, C. (1998): Macrobenthic community changes related to
eutrophication in palude della rosa (venetian lagoon, italy). estuarine, coastal and shelf science
vol. 47, no. 2, pp. 217-226.
The new Bathing Water Directive of 2006 (2006/7/EC), Online:
http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2006:064:0037:01:EN:HTML.
Train K. (2003): Discrete Choice Methods with Simulation, Cambridge University Press, New
York.
Train, K., Sonnier G. (2003): Mixed Logit with Bounded Distributions of Partworths, Working
paper, Berkeley: University of California.
Thurstone, L. (1927), ’A Law of comparative judgement’, Psychological Review 34, 273-86.
Vencl, Š. (2010): Sinice - aktuální problém našich vod, cit: 12.1.2010, online:
http://www.priroda.cz/clanky.php?detail=398, 2005.
Vojáček, O. (2007): Nejvýznamnější pohledy na preference v ekonomickém myšlení, Praha
11.01.2007. In: KADEŘÁBKOVÁ, Božena (ed.). Aplikace makroekonomie ve vědeckém
zkoumání doktorandů [CD-ROM]. Praha : VŠE NF, 2007, s. 19. ISBN 978-80-245-1171-9.
Vojáček O., Melichar J. (2007): Valuing Water Quality – Choice Experiment Approach, In:
ŠAUER, Petr (ed.). Environmental Economics, Policy and International Relations. Prague :
Nakladatelství a vydavatelství litomyšlského semináře, Prague, 109–120.
Warren F. Kuhfeld, W.F., Randall D. T., Garratt, M. (1994): Efficient Experimental Design with
Marketing Research Applications, Journal of Marketing Research, Vol. 31, No. 4, pp. 545-557,
American Marketing Association, cit: 12.2.2010, online:
http://www.jstor.org/stable/3151882.
Whittington, D., Briscoe, J., Mu, X., Barron, W. (1990): Estimating the willingness to pay for
water services in developing countries, a case study of the use of contingent valuation surveys in
Southern Haiti. Economic Development and Cultural Change 38, 293-311.
Whittington, D., Mujwahuzi, M., McMahon, G., Choe, K., (1988): Willingness to pay for water
in Newala district, Tanzania, strategies for cost recovery. Water and Sanitation for Health Project
Field Report no. 246, USAID, Whashington DC.
Whittington, D., Pattanayak, S., Yang, J.C., Bal Kumar, K.C. (2002): Household demand for
improved piped water services, evidence from Kathmandu, Nepal. Water Policy 4, 531-556.
World Health Organization (2001): Bathing Water Quality and Human Health, WHO/SD
E/WSH/01.2, Protection of the Human Environment, Water, Sanitation and Health Series,
WHO, Geneva, cit: 23.5.2008, online:
http://www.crid.or.cr/digitalizacion/pdf/eng/doc14603/doc14603.pdf.
Annexes
Annex 1: Choice of attributes in preliminary survey stage
16.
How important are the following characteristics to you when choosing a holiday
area for your trip? Focus on the recreational activity that you normally do.
Please specify 5 of the characteristics that are the most important to you. Write down their
numbers in the boxes below in their order of importance. Also, specify 5 of the
characteristics that are unimportant to you.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
water temperature
water appearance
water quality
surface of the pond, reservoir or river bottom
access to the water
no litter scattered on the beach
a changing cubicle
the presence of a lifeguard
the option to rent a boat
angling opportunities
the option to take a shower
refreshment options
other recreational and sporting amenities
my family or friends come here
the costs (parking fee, entrance fee, etc.)
surface of the reservoir bottom
the surrounding scenery
another characteristic (specify): _________________________________
1. most important
1. least important
2.
2.
3.
3.
4.
4.
5.
5.
17. Please rank the below aspects of the quality of the watercourse or water body by the
importance that you personally attach to them in respect of the recreational activity that
you normally do. Please write down their numbers in their order of importance.
1.
2.
3.
4.
5.
6.
colour
smell
transparency
floating litter
algal bloom
another aspect (specify): ______________________________
Annex 2: The description of the attributes and their levels in quantitative preliminary
survey
CARD 1
Beach Overcrowding
None at all
Moderate
Overcrowding of the beach
during your stay on the beach.
Great
CARD 2
Water Quality
1 VERY CLEAN WATER
•
No algal growth
•
No algal bloom present
2 CLEAN WATER
•
Slight algal growth
•
No algal bloom present
3 SLIGHTLY POLLUTED WATER
•
Distinct algal growth
•
Beginning algal bloom
4 POLLUTED WATER
Presence of algae
and algal bloom in
the water.
•
Abundant algal growth
•
Algal bloom present
BATHING
VERY
CONVENIENT
BATHING
CONVENIENT
BATHING
LESS
CONVENIENT
BATHING
INCONVENIENT
5 VERY POLLUTED WATER
•
Substantial algal growth
•
Massive algal bloom present
BATHING BAN
CARD 3
Beach Facilities
YES
NO
Beach facilities:
- sanitary (toilets, showers, changing rooms)
- sports facilities and rentals
- refreshments
CARD 4
Medical Aid
YES
Medical aid provided by
trained medical staff.
NO
CARD 5
Entrance Fee Amount
CZK 10
CZK 50
;
Entrance fee that you pay
for one person entering the
beach for one day.
Note: Kč = CZK
CZK 100
CZK 200
Annex 3: An example of the choice card in quantitative preliminary survey
CARD 6
CHARACTERISTICS
SITE 1
SITE 2
Beach Overcrowding
GREAT
MODERATE
3 SLIGHTLY
POLLUTED
WATER
2 CLEAN
WATER
NO
NO
NO
YES
CZK 50
CZK 100
1
2
Water Quality
Beach Facilities
Medical Aid
Entrance Fee Amount
I would choose site
Note: Kč = CZK
Annex 4: Attributes selection in the quantitative survey stage
1.
When deciding, were some of the characteristics of the recreational site for you
totally unimportant?
1 yes
2 no
FILTER 12: IF THE RESPONDENT SAYS „YES“
2.
Name these characteristics, please.
______________________________
______________________________
______________________________
3.
Are there any other important characteristics of the recreational site which are
important for your decision and which were not introduced in the questionnaire?
1 yes
2 no
FILTER 13: IF THE RESPONDENT SAYS „YES“
4.
Name these characteristics, please.
______________________________
______________________________
______________________________
Annex 5: The description of the attributes and their levels in the final survey
CARD 2
CARD 3
Water Quality
Beach Overcrowding
Attribute Levels
Attribute Levels
Clean Water – Level 1
No algal growth
No cyanobacteria present
Bathing conventient
Little Overcrowding
Great Overcrowding
Prese nce
of algae and
cyanobact eria
in the water.
Overcrowding
of the beach
during your stay
on the beach.
Slightly Polluted Water
– Level 2
Perceptible algal growth
No cyanobacteria present
Bathing conventient
Polluted Water – Level 3
Abudant algal growth
Cyanobacteria present
Bathing inconventient
CARD 5
CARD 4
Beach Facilities
Entrance Fee Amount
Attribute Levels
Attribute Levels
CZK 40
YES
CZK 80
NO
Entrance fee
that you pay pe r
person per day.
WC, showers
and refreshments
on the beach
Note: Kč = CZK
CZK 150
Annex 6: The experimental design generated by SPSS for the Lake Macha case
lide1
2
1
2
1
1
1
2
1
1
Site 1
voda1 vybavenost1 vstup1 karta
1
1
2
1
1
1
2
1
3
2
2
2
2
1
1
3
1
3
3
2
1
1
2
3
3
1
2
1
2
3
4
5
6
7
8
9
Site 2
lide2 voda2 vyba2
1
1
1
3
1
3
2
1
2
3
1
2
2
2
1
1
1
2
Site 3
2
1
2
1
1
2
1
1
1
1
2
2
1
2
3
3
3
1
1
2
1
2
1
2
1
1
1
vstup2 karta
2
2
3
3
1
1
2
1
3
1
2
3
4
5
6
7
8
9
Site 4
1
1
2
1
1
1
1
2
2
3
3
1
1
2
2
1
3
2
1
2
3
4
5
6
7
8
9
1
1
1
1
2
1
2
2
1
3
1
3
1
2
2
3
1
2
1
1
2
1
1
1
1
2
2
2
3
3
1
3
2
1
2
1
1
2
3
4
5
6
7
8
9
Annex 7: The choice sets used in the final survey
Site 1
Site 2
Great Overcrowding
Little Overcrowding
Level 1
Clean Water
Level 2
Slightly Polluted
YES
CZK 80
Entrance Fee
YES
CZK 40
Entrance Fee
CARD 6
Site 1
Site 2
Little Overcrowding
Little Overcrowding
Level 1
Clean Water
Level 3
Polluted Water
YES
CZK 150
Entrance Fee
NO
CZK 80
Entrance Fee
CARD 7
Site 1
Site 2
Great Overcrowding
Little Overcrowding
Level 2
Slightly Polluted
Level 3
Polluted Water
YES
CZK 40
Entrance Fee
YES
CZK 40
Entrance Fee
CARD 8
Site 1
Site 2
Little Overcrowding
Great Overcrowding
Level 2
Slightly Polluted
Level 1
Clean Water
NO
NO
CZK 80
Entrance Fee
CZK 40
Entrance Fee
CARD 9
Site 1
Site 2
Little Overcrowding
Great Overcrowding
Level 2
Slightly Polluted
Level 2
Slightly Polluted
YES
CZK 150
Entrance Fee
YES
CZK 80
Entrance Fee
CARD 10
Site 1
Site 2
Little Overcrowding
Little Overcrowding
Level 3
Polluted Water
Level 1
Clean Water
YES
CZK 40
Entrance Fee
YES
CZK 80
Entrance Fee
CARD 11
Site 1
Site 2
Great Overcrowding
Great Overcrowding
Level 3
Polluted Water
Level 3
Polluted Water
NO
YES
CZK 150
Entrance Fee
CZK 150
Entrance Fee
CARD 12
Site 1
Site 2
Little Overcrowding
Little Overcrowding
Level 1
Clean Water
Level 1
Clean Water
NO
YES
CZK 40
Entrance Fee
CZK 150
Entrance Fee
CARD 13
Site 1
Site 2
Little Overcrowding
Little Overcrowding
Level 3
Polluted Water
Level 2
Slightly Polluted
YES
NO
CZK 80
Entrance Fee
CZK 150
Entrance Fee
CARD 14
Note: Kč = CZK
Annex 8: The final version of the questionnaire
SUMMER RECREATION – LAKE MACHA 2007
5.
Note start of interview.
Hour ________
Note questionnaire daily serial number.
Serial no. ________
Min ________
I would like to ask you a few questions concerning your recreation at Lake
Macha, what influences you while on holiday here, and how you evaluate
the Lake water quality. This survey is non-commercial and run by the
University of Economics, Prague. Completing the questionnaire will take
about 15 minutes.
All the information you tell us will remain anonymous and will be used for
research purposes only. If you don’t want to answer any question for any
reason, we will respect that. Can I start asking the questions?
PART I – INTRODUCTION
6.
Is this your first stay at Lake Macha?
1 Yes
–›› GO TO QUESTION 6
2
7.
No
–›› GO TO QUESTION 3
How many one-day trips to Lake Macha beaches have you made in
this 2007 summer season and the last 2006 summer season? Please
exclude your current visit from the estimate. Please say your
estimate for each of the beaches.
A one-day trip is a trip that lasts no longer than a day without staying
the night away from home.
NOTE FOR INTERVIEWER: SHOW CARD 1 TO THE RESPONDENT. NOTE IN TABLE 1
BELOW THE TOTAL NUMBER OF ONE-DAY TRIPS FOR EACH BEACH THAT THE RESPONDENT
HAS VISITED IN THE 2007 AND 2006 SUMMER SEASONS. IF THE RESPONDENT SAYS A
DIFFERENT BEACH, NOTE IT ON THE BLANK LINE.
NOTE FOR INTERVIEWER: IF THE RESPONDENT SAYS AN INTERVAL, ASK THEM TO SAY
JUST ONE NUMBER. IF THE RESPONDENT CANNOT RECALL THE EXACT NUMBER OF TRIPS,
ASK THEM TO SAY THEIR BEST GUESS.
TABLE 1:
Numbers of trips
1. Doksy Main Beach
2. Klůček Beach
3. Borný Beach
4. Staré Splavy Beach
5. Another beach (specify!):
2007
Summer
Season
2006
Summer
Season
8.
How many multiple-day stays to the Lake Macha area have you made
in this 2007 summer season and the last 2006 summer season?
Please exclude your current visit from the estimate.
A multiple-day stay is a trip that lasts longer than a day with staying
the night(s) away from home.
NOTE FOR INTERVIEWER: IF THE RESPONDENT SAYS AN INTERVAL, ASK THEM TO SAY
JUST ONE NUMBER. IF THE RESPONDENT CANNOT RECALL THE EXACT NUMBER OF TRIPS,
ASK THEM TO SAY THEIR BEST GUESS.
Number of stays in 2007 summer season: _________
Number of stays in 2006 summer season: _________
FILTER 1: DO QUESTION 5 BELOW ONLY WITH RESPONDENTS WHO HAVE MADE AT
LEAST ONE MULTIPLE-DAY TRIP IN THE 2007 AND 2006 SUMMER SEASONS.
9.
How many days in total have you spent on these multiple-day stays?
Please say your estimates for each beach and for the 2007 and 2006
summer seasons.
NOTE FOR INTERVIEWER: SHOW CARD 1 TO THE RESPONDENT. NOTE IN TABLE 2
BELOW THE TOTAL NUMBERS OF STAYS FOR EACH BEACH THAT THE RESPONDENT HAS
VISITED AND FOR THE 2007 AND 2006 SUMMER SEASONS. IF THE RESPONDENT SAYS A
DIFFERENT BEACH, NOTE IT ON THE BLANK LINE.
NOTE FOR INTERVIEWER: IF THE RESPONDENT SAYS AN INTERVAL, ASK THEM TO SAY
JUST ONE NUMBER. IF THE RESPONDENT CANNOT RECALL THE EXACT NUMBER OF TRIPS,
ASK THEM TO SAY THEIR BEST GUESS.
TABLE 2: Numbers of total
days on multiple-day stays
Numbers of
total days on
multiple-day
stays (2007)
Numbers of
total days on
multiple-day
stays (2006)
1. Doksy Main Beach
2. Klůček Beach
3. Borný Beach
4. Staré Splavy Beach
5. Another beach (specify!):
END OF FILTER 1
Now, we’re interested in your current stay on the Lake Macha beaches.
10.
Are you here on a multiple-day stay?
1
Yes
–›› GO TO PART II (Page 3)
2
No
–›› GO TO PART III (Page 5)
PART II – MULTIPLE-DAY STAY
FILTER 2: DO QUESTIONS 7 TO 10 BELOW ONLY WITH RESPONDENTS WHO ARE
CURRENTLY DOING A MULTIPLE-DAY STAY AT LAKE MACHA.
11.
How many nights are you going to stay away from home on this stay?
Number of nights ________
12.
Please specify the primary type of accommodation on your current
multiple-day stay.
1
hotel or guesthouse
2
own chalet or summerhouse
3
camping site
4
another (specify):
13.
Which modes of transport did you use to travel from home to your
place of accommodation?
NOTE FOR INTERVIEWER: TICK ALL OPTIONS SPECIFIED.
1
passenger car
2
train
3
bus
4
bicycle
5
another (specify):
FILTER 3: DO QUESTION 10 BELOW ONLY WITH RESPONDENTS WHO STATED PAID
MODES OF TRANSPORT ONLY (CAR, TRAIN, BUS …) IN QUESTION 9.
14.
Make an estimate of the length in kilometres of your trip using the
mode of transport from home to the place of your accommodation.
Please say your estimate for one way only.
NOTE FOR INTERVIEWER: IF THE RESPONDENT SAYS AN INTERVAL, ASK THEM TO SAY
JUST ONE NUMBER. IF THE RESPONDENT HAS DIFFICULTY SAYING THE EXACT NUMBER OF
KILOMETRES, ASK THEM TO SAY THEIR BEST GUESS.
Km ________
END OF FILTER 3
END OF FILTER 2: GO TO PART IV
PART III – ONE-DAY TRIP
FILTER 4: DO QUESTIONS 11 AND 12 BELOW ONLY WITH RESPONDENTS WHO ARE
CURRENTLY DOING A ONE-DAY TRIP TO LAKE MACHA.
15.
Which modes of transport are you using on your trip today?
NOTE FOR INTERVIEWER: TICK ALL OPTIONS SPECIFIED.
1
passenger car
2
train
3
bus
4
cycle bus
5
only cycling today
6
only walking today
7
another (specify):
FILTER 5: DO QUESTION 10 BELOW ONLY WITH RESPONDENTS WHO STATED PAID
MODES OF TRANSPORT ONLY (CAR, TRAIN, BUS …) IN QUESTION 12.
16.
Make an estimate of the distance in kilometres you are going to
travel by car, train or bus today.
NOTE FOR INTERVIEWER: IF THE RESPONDENT SAYS AN INTERVAL, ASK THEM TO SAY
JUST ONE NUMBER. IF THE RESPONDENT HAS DIFFICULTY SAYING THE EXACT NUMBER OF
KILOMETRES, ASK THEM TO SAY THEIR BEST GUESS.
Km ________
END OF FILTER 5
END OF FILTER 4: GO TO PART IV
PART IV – TRIP CHARACTERISTICS
17.
Are you exploiting Lake Macha for bathing on your current trip?
1
Yes
2
No
18.
What other body of water in the Czech Republic would you choose for
the same type of trip that you are now doing at Lake Macha? Think of
the same recreation activity that you are doing at Lake Macha today,
and the same length of stay. Please only say one body of water.
Please say the name of the body of water and the name of the district
in which the body of water is located.
Name of body of water
Name of district
_____________________
_____________________
NOTE FOR INTERVIEWER: IF THE RESPONDENT SAYS AN UNKNOWN BODY OF WATER,
ASK THEM TO SPECIFY THE REGION OR MUNICIPALITY IN OR NEAR WHICH THE BODY IS.
19.
Please say the number of persons in the group, including yourself,
with which you are staying at Lake Macha.
Total persons __________
Children (up to 18 years) ________
PART V – CHOICE OF RECREATIONAL ATTRIBUTES
NOTE FOR INTERVIEWER: READ THE FOLLOWING TEXT TO THE RESPONDENT AND GO
THROUGH THE 4 FOLLOWING CHARACTERISTICS ONE BY ONE, EXPLAINING THEIR
CONTENTS. SHOW CARDS 2 – 5 TO THE RESPONDENT ONE BY ONE.
Any body of water exploited for recreation can be described using various
characteristics. People may choose different bodies of water in the Czech
Republic for their recreation depending on:
- how the beaches on the site are overcrowded
CARD 2
- the water quality
CARD 3
- the beach facilities
CARD 4
- and the entrance fee per person per day.
CARD 5
SET 1
NOTE FOR INTERVIEWER: SHOW CARD 6 TO THE RESPONDENT.
Please imagine a situation where you are shown an information
leaflet in which these 4 characteristics are used to describe different
water bodies and reservoirs in the Czech Republic and you may
choose among them.
Which of these 2 alternatives would you choose for your beach stay
today? Would you choose Alternative 1 or Alternative 2 or would you
choose neither of them? Remember that you can only choose from
these alternatives.
1
Alternative 1
2
Alternative 2
3
stay at home
FILTER 6: IN CASE THEY SAY "STAY AT HOME" - 3
20.
Why do you say that you would rather stay at home? You can give
more reasons.
1. Polluted water
2. Too few people
3. Too many people
4. No facilities
5. Expensive
6. Another reason
(specify)
END OF FILTER 6
General
X
X
X
X
X
Site 1
1
1
1
1
1
Site 2
2
2
2
2
2
SET 2
NOTE FOR INTERVIEWER: SHOW CARD 7 TO THE RESPONDENT
21.
1
Now you have another choice between Alternative 1 or Alternative 2
or staying at home. Again, your decision concerns your choice of
recreation for your beach stay today.
Alternative 1
2
Alternative 2
3
stay at home
FILTER 7: IN CASE THEY SAY "STAY AT HOME" - 3
22.
1.
2.
3.
4.
5.
Why do you say that you would rather stay at home? You can give
more reasons.
Polluted water
Too few people
Too many people
No facilities
Expensive
General
X
X
X
X
X
Site 1
1
1
1
1
1
Site 2
2
2
2
2
2
6. Another reason
(specify)
END OF FILTER 7
SET 3
NOTE FOR INTERVIEWER: SHOW CARD 8 TO THE RESPONDENT
23.
1
Now you have another choice between Alternative 1 or Alternative 2
or staying at home. Again, your decision concerns your choice of
recreation for your beach stay today.
Alternative 1
2
Alternative 2
3
stay at home
FILTER 8: IN CASE THEY SAY "STAY AT HOME" - 3
24.
Why do you say that you would rather stay at home? You can give
more reasons.
1. Polluted water
2. Too few people
3. Too many people
4. No facilities
5. Expensive
6. Another reason
(specify)
END OF FILTER 8
General
X
X
X
X
X
Site 1
1
1
1
1
1
Site 2
2
2
2
2
2
SET 4
NOTE FOR INTERVIEWER: SHOW CARD 9 TO THE RESPONDENT
25.
1
Now you have another choice between Alternative 1 or Alternative 2
or staying at home. Again, your decision concerns your choice of
recreation for your beach stay today.
Alternative 1
2
Alternative 2
3
stay at home
FILTER 9: IN CASE THEY SAY "STAY AT HOME" - 3
26.
1.
2.
3.
4.
5.
Why do you say that you would rather stay at home? You can give
more reasons.
Polluted water
Too few people
Too many people
No facilities
Expensive
General
X
X
X
X
X
Site 1
1
1
1
1
1
Site 2
2
2
2
2
2
6. Another reason
(specify)
END OF FILTER 9
SET 5
NOTE FOR INTERVIEWER: SHOW CARD 10 TO THE RESPONDENT
27.
1
Now you have another choice between Alternative 1 or Alternative 2
or staying at home. Again, your decision concerns your choice of
recreation for your beach stay today.
Alternative 1
2
Alternative 2
3
stay at home
FILTER 10: IN CASE THEY SAY "STAY AT HOME" - 3
28.
1.
2.
3.
4.
5.
Why do you say that you would rather stay at home? You can give
more reasons.
Polluted water
Too few people
Too many people
No facilities
Expensive
6. Another reason
(specify)
END OF FILTER 10
General
X
X
X
X
X
Site 1
1
1
1
1
1
Site 2
2
2
2
2
2
SET 6
NOTE FOR INTERVIEWER: SHOW CARD 11 TO THE RESPONDENT
29.
1
Now you have another choice between Alternative 1 or Alternative 2
or staying at home. Again, your decision concerns your choice of
recreation for your beach stay today.
Alternative 1
2
Alternative 2
3
stay at home
FILTER 11: IN CASE THEY SAY "STAY AT HOME" - 3
30.
1.
2.
3.
4.
5.
Why do you say that you would rather stay at home? You can give
more reasons.
Polluted water
Too few people
Too many people
No facilities
Expensive
General
X
X
X
X
X
Site 1
1
1
1
1
1
Site 2
2
2
2
2
2
6. Another reason
(specify)
END OF FILTER 11
SET 7
NOTE FOR INTERVIEWER: SHOW CARD 12 TO THE RESPONDENT
31.
1
Now you have another choice between Alternative 1 or Alternative 2
or staying at home. Again, your decision concerns your choice of
recreation for your beach stay today.
Alternative 1
2
Alternative 2
3
stay at home
FILTER 12: IN CASE THEY SAY "STAY AT HOME" - 3
32.
1.
2.
3.
4.
5.
Why do you say that you would rather stay at home? You can give
more reasons.
Polluted water
Too few people
Too many people
No facilities
Expensive
6. Another reason
(specify)
END OF FILTER 12
General
X
X
X
X
X
Site 1
1
1
1
1
1
Site 2
2
2
2
2
2
SET 8
NOTE FOR INTERVIEWER: SHOW CARD 13 TO THE RESPONDENT
33.
1
Now you have another choice between Alternative 1 or Alternative 2
or staying at home. Again, your decision concerns your choice of
recreation for your beach stay today.
Alternative 1
2
Alternative 2
3
stay at home
FILTER 13: IN CASE THEY SAY "STAY AT HOME" - 3
34.
Why do you say that you would rather stay at home? You can give
more reasons.
1. Polluted water
2. Too few people
3. Too many people
4. No facilities
5. Expensive
6. Another reason
(specify)
General
X
X
X
X
X
Site 1
1
1
1
1
1
Site 2
2
2
2
2
2
END OF FILTER 13
SET 9
NOTE FOR INTERVIEWER: SHOW CARD 14 TO THE RESPONDENT
35.
1
Now you have another choice between Alternative 1 or Alternative 2
or staying at home. Again, your decision concerns your choice of
recreation for your beach stay today.
Alternative 1
2
Alternative 2
3
stay at home
FILTER 14: IN CASE THEY SAY "STAY AT HOME" - 3
36.
1.
2.
3.
4.
5.
Why do you say that you would rather stay at home? You can give
more reasons.
Polluted water
Too few people
Too many people
No facilities
Expensive
6. Another reason
(specify)
END OF FILTER 14
General
X
X
X
X
X
Site 1
1
1
1
1
1
Site 2
2
2
2
2
2
37.
In your opinion, what is the current water quality in Lake Macha?
Choose one of the following options.
NOTE FOR INTERVIEWER: SHOW CARD 15 TO THE RESPONDENT
1
Very clean water
2
Clean water
3
Slightly polluted water
4
Polluted water
5
Seriously polluted water
99
Cannot judge
38.
1
Have you heard of the cyanobacteria, or algal bloom, in Lake Macha?
Yes
2
No
FILTER 15: IN CASE THEY RESPOND “YES” TO QUESTION 35
39.
What do you think causes the cyanobacteria in Lake Macha?
END OF FILTER 15
40.
In your opinion, who should contribute the most to the funding for
improving water quality in Lake Macha? Choose only one option.
NOTE FOR INTERVIEWER: SHOW CARD 16 TO THE RESPONDENT
1
Holiday makers
2
Population of nearby municipalities
3
Nearby farming and industrial operations
4
The entire population of the Czech Republic
5
Beach and camp site operators
6
Another (specify):
41.
To what extent do you agree with the following on Lake Macha?
Rather
agree
Rather
disagree
Certainly
disagree
Cannot
judge
1. Lake Macha beaches are too overcrowded.
2. Toilets and showers on Lake Macha beaches are in an
excellent condition.
3. Before I go to Lake Macha beaches, I find information
on the water quality there.
Certainly
agree
NOTE FOR INTERVIEWER: SHOW CARD 17 TO THE RESPONDENT
1
2
3
4
99
1
2
3
4
99
1
2
3
4
99
PART VI – SOCIO-DEMOGRAPHIC QUESTIONS
The following questions we are going to ask you concern your person. We
are aware of the fact that they may be intimate for you. We would like to
remind you that the information you provide will be used for scientific
purposes only, not for commercial research, and that the information
acquired is anonymous. Still, tell us when you don’t feel like answering any
of the questions. We will respect that.
42.
What is your highest level of formal education attained?
1
Primary
2
Trainee (secondary without A levels)
3
Secondary with A levels, college
4
University
43.
What is your primary employment type?
NOTE FOR INTERVIEWER: SHOW CARD 18 TO THE RESPONDENT
1
Full-time employee
2
Part-time employee
3
Self-employed, entrepreneur
4
Unemployed
5
Student, trainee
6
Retired
7
Housewife/househusband
8
Maternity/parental leave
9
Working student / Working retired
44.
Please say the region and municipality where you currently live.
Region
1.
2.
3.
4.
5.
6.
7.
Prague
Central Bohemia
South Bohemia
Plzeň Region
Karlovy Vary Region
Ústí Region
Liberec Region
Municipality
45.
8. Hradec Králové Region
9. Pardubice Region
10. Vysočina Region
11. South Moravia
12. Olomouc Region
13. Zlín Region
14. Moravia-Silesia
_______________________________
How many persons are there in your household, including yourself?
Total persons ___________
Children (under 18 years) _________
46.
Which year were you born?
Year of birth _______________
NOTE FOR INTERVIEWER: IF THE RESPONDENT WON’T SAY THEIR YEAR OF BIRTH,
ESTIMATE THEIR AGE AND NOTE IN THE FIELD BELOW.
Respondent’s estimated age ______________
47.
Please choose the income bracket that best fits your net monthly
income, including any allowances and pensions.
NOTE FOR INTERVIEWER: SHOW CARD 19 TO THE RESPONDENT
1
0 – 5,500
2
5,501 – 7,000
3
7,001 – 8,500
4
8,501 – 10,500
5
10,501 – 13,000
6
13,001 – 15,500
7
15,501 – 18,000
8
18,001 – 24,000
9
24,001 – 35,000
10
35,001 and more
11
no own income
12
won’t answer
48.
Please note the end of the interview
Hour ________
Min ________
NOTE FOR INTERVIEWER: GIVE THE RESPONDENT THE SURVEY INFORMATION.
THANK YOU FOR YOUR WILLINGNESS AND CO-OPERATION.
Thank you for giving us your time and taking part in our survey. If you have any
questions, feel free to call us, write to us or send an email to the contact address
that you have received from us.
Members of the Department of Environmental Economics at the University of
Economics, Prague
PART VIII – QUESTIONS FOR THE INTERVIEWER
49.
The respondent was a:
1
man
2
woman
50.
Did the respondent take the interview seriously?
very seriously
not at all
2
1
51.
4
5
How well did the respondent understand the questions asked?
excellent
understanding
1
52.
3
not at all
2
3
4
5
How friendly was the respondent during the interview?
very friendly
not at all
2
1
53.
5
very inattentively
2
3
4
5
Was it difficult for the respondent to imagine water bodies based on
the characteristics that you showed them in Part V?
very easy
rather easy
1
2
55.
4
How attentively was the respondent listening during the interview?
very attentively
1
54.
3
neither easy
nor difficult
3
rather
difficult
4
very difficult
5
How difficult was it for the respondent to choose water bodies in the
experiment in Part V?
very easy
rather easy
1
2
neither easy
nor difficult
3
rather
difficult
4
very difficult
5
56.
Was anyone influencing the respondent during the interview?
1
Yes
2
No
57.
Cloud cover
58.
Weather
1
clear sky
1
no precipitation
2
somewhat cloudy
2
drizzle
3
cloudy
3
shower
4
overcast
4
rain
5
fog
6
storm
7
hail
59.
Name of the beach
NOTE FOR INTERVIEWER: ONLY CHOOSE ONE OPTION
1.
2.
3.
4.
60.
Doksy Main Beach
Klůček
Staré Splavy (Aquapark)
Borný
Place of interview
NOTE FOR INTERVIEWER: ONLY CHOOSE ONE OPTION
1.
2.
3.
4.
61.
restaurant, kiosk
beach
path
another (specify)
Randomness interval (respondent number in sequence)
_____________________
62.
Interviewer’s name
__________________________
63.
Interview date
Day _________
64.
Month ________
In case you have come across any other problems, please note them
briefly here:
Annex 9: Random Parameters Logit Model with heterogeneity analysis
Variable
Coefficient
Standard Error
b/St.Er.
P[|Z|>z]
Random parameters in utility functions
DIRTY
-5.484***
1.339
-4.094
0.000
EQUIP
-1.986***
0.522
-3.803
0.001
FEE
-0.031***
0.004
-7.872
0.000
Nonrandom parameters in utility functions
CROWD
-0.887***
0.130
6.814
0.000
CLEAN
1.226***
0.155
7.940
0.000
OPT_OUT -4.145***
0.300
-13.836
0.000
Heterogeneity in mean, Parameter:Variable
DIRT:SEX
0.593*
0.334
1.774
0.076
DIRT:OPA -0.562
0.350
-1.606
0.108
DIRT:KOU 0.214
0.496
0.432
0.665
DIRT:DET -0.320*
0.187
-1.717
0.086
DIRT:VEK 0.016
0.011
1.429
0.153
DIRT:PRA -0.597
0.403
-1.484
0.138
DIRT:PRI
0.000**
0.000
2.311
0.021
DIRT:DOK -0.822**
0.354
-2.325
0.020
EQUI:SEX 0.484**
0.231
2.102
0.036
EQUI:OPA -0.208
0.240
-0.865
0.387
EQUI:KOU -0.574*
0.327
-1.757
0.079
EQUI:DET 0.025
0.120
0.210
0.833
EQUI:VEK -0.002
0.008
-0.272
0.786
EQUI:PRA -0.418
0.277
-1.509
0.131
EQUI:PŘI -0.000**
0.000
-2.279
0.023
EQUI:DOK 0.087
0.220
0.395
0.693
FEE:SEX
0.000
0.001
-0.137
0.891
FEE:OPA
0.004***
0.002
2.658
0.008
FEE:KOU
0.005**
0.002
2.155
0.031
FEE:DET
-0.001
0.001
-0.968
0.333
FEE:VEK
0.000
0.000
0.609
0.543
FEE:PRA
0.002
0.002
1.299
0.194
FEE:PŘI
0.000***
0.000
4.883
0.000
FEE:DOK -0.002
0.001
-1.058
0.290
Derived standard deviations of parameter distribution
NsDIRTY
2.497***
0.960
2.602
0.009
NsEQUIP
1.224***
0.431
2.839
0.005
NsFEE
0.015***
0.002
5.926
0.000
Note:
***, **, * = Significance at 1%, 5%, 10% levels
SEX = respondent’s sex
OPA= repetitiveness of the trip at the Lake
KOU = swimming during the stay on the beach
DET = the number of children in the group staying at Lake Macha.
VEK = respondent’s age
PRA = Prague citizen
PŘI = respondent’s income
DOK = individual interviewed at the Doksy beach
Source: Own analysis