Production cost structure of the water

and sewage industry

An econometric cost analysis of urban water supply and
sewage treatment, with an application to a section of
Swedish communities

MADELENE MALMSTEN

Master of Science Thesis

Stockholm, Sweden 2008

 Production cost structure of the
water and sewage industry

An econometric cost analysis of urban water supply and sewage

treatment, with an application to a section of Swedish communities

Madelene Malmsten

Master of Science Thesis INDEK 2008:64

KTH Industrial Engineering and Management

Industrial Management
SE-100 44 STOCKHOLM

[2]
 Master of Science Thesis INDEK 2008:64

Production cost structure of the water

and sewage industry

An econometric cost analysis of urban water supply

and sewage treatment, with an application to a section
of Swedish communities

Madelene Malmsten

Approved Examiner Supervisor

2008-09-08 Thomas Sandberg Thomas Sandberg

Commissioner Contact person

Dimitrios F. Lekkas Dimitrios F. Lekkas

Abstract
In this thesis, I have studied how we can capture the financial characteristics of the
underlying production technology for water and wastewater systems. More broadly, I
focus on how we can get hold of the cost structure of a water and wastewater system or
total industry. After determining the (or a) cost structure of a system, the system can be
analysed in terms of e.g. efficiency, technology, growth, capacity, expenses etc. This
thesis may serve as a bridge between mathematical models on the one hand and the
water and wastewater industry on the other hand.

Foremost, the research has been established on the basis of literature studies. The
literature review is related to an empirical study of Swedish utilities. The estimation
procedure is based on a multivariate regression approach, using a Transcendental
Logarithmic cost function.

The estimated cost structure in this thesis makes it possible for one to assess a few
characteristics of the Swedish industry. The cost structure may be relevant for other
systems of the same manner as for those that have been chosen for this study. The
efficiency characteristics of the Swedish industry are expressed in terms of Economies of
production output density, Economies of customer density and Economies of scope.

This thesis urges that the fundamental statistics are missing for the Swedish case. In
order to be able to create a more throughout analysis, the quality level of information

[3]
has to be increased. It is an ongoing process for the Swedish Water & Wastewater
Association to establish the underlying statistical framework.

A few conclusions can be drawn from the study. For instance, there exists Economies of
scope between water and wastewater in the industry. There are cost advantages of a
joint production of water and wastewater. The estimations are based on a sample of
utilities (25 of them) serving 1 community. The larger utilities exhibit Diseconomies of
production output density and Diseconomies of customer density, while the smaller
utilities in the sample exhibit Economies of production output density and Economies of
customer density.

[4]
Table of Contents

Abbreviations 7

Preface 8

Acknowledgements 9

1. Chapter 1: Introduction 10
1.1. Background 10
1.1.1. New features of the water and wastewater industry 10
1.1.2. Two concepts: Actual cost estimation and cost structure
representation 12
1.2. Problem statement 13
1.3. Purposes of the study 13
1.4. Outline 13
2. Chapter 2: Methodology 15
2.1. Ontological and epistemological platforms 15
2.2. Research approach 16
3. Chapter 3: Characterization and modelling of the water and
wastewater services 18
3.1. The econometric analysis 18
3.2. The water and wastewater service system 19
3.3. Water as an input and output variable 21
3.4. Cost categories 22
3.5. Challenging features of the cost structure analysis 23
3.6. Modelling under real life conditions: a step-by-step approach 24
3.7. Identification of possible cost and scale drivers:
The explanatory variables 26
3.8. Structural-conceptual representation of the service system 29
3.8.1. Perspectives 29
3.8.2. Production growth 30
3.8.3. A few important concepts 31
3.8.4. Production patterns 34
3.9. A mathematical framework 37
3.9.1. An approach to derive cost functions 37
3.9.2. The Transcendental Logarithmic cost function 40
4. Chapter 4: Empirical study and analysis 42
4.1. Capture a representation of the total costs:
procedure and limitations 43

[5]
 4.2. Modelling the structure of variable costs, an application to a sample
of Swedish communities 44
4.2.1. The structural representation of the model 44
4.2.2. Data description 45
4.2.3. Data sorting and selection 48
4.2.4. Estimation procedure with a multivariate regression approach 50
4.2.5. Parameter estimates 52
4.2.6. Analysis: Understanding the parameter figures.
Estimating Economies of production output density,
Economies of customer density and Economies of scope 54
5. Chapter 5: Summary and conclusions 61
6. References 63

[6]
Abbreviations
OFWAT: The Office of Water Services in England

SWWA: The Swedish Water and Wastewater Association

TC: Total cost

TVC: Total variable cost

WFD (The Directive): European Union Water Framework Directive

[7]
Preface

This thesis has been conducted in cooperation with the Department of Statistics and
Actuarial-Financial Mathematics at the University of the Aegean, Karlovasi, Greece and
the Department of Industrial Engineering and Management at the Royal Institute of
Technology, Stockholm, Sweden. It constitutes the final part of a Master of Science
Degree in Industrial Engineering and Management at the Royal IT. The assignment was
essentially carried out between the 23rd of January and the 30th of June 2008.

The formation of this paper progressed during discussions between the author and the
commissioner about how to build a robust mathematical representation of the water
and wastewater services for the main purposes of 1) understanding the underlying
production technology, 2) capture the cost structure of the services, 3) estimating the
total costs of the services and 4) measuring the “behaviour” of the costs when the
service system changes in size and structure. This thesis focuses on (1), (2) and (3), since
(4) is beyond the means of this paper. It requires much information and a lot of time for
investigation.

The initial idea and desired venture was to capture the economic characteristics of the
underlying production technology for one or several Greek utilities, or for the Greek
water and wastewater industry as a whole. The analysis would thereby reflect the
municipal units of water and wastewater under the local conditions. Due to limitations
in data quantity and quality, the application could not be carried out successfully.
However, the numbered topics shown above are of interest for the global water and
wastewater industry. For that reason, the initial idea was not neglected totally. As an
alternative, the water and wastewater industry in Sweden was chosen for the empirical
study. The Swedish Water & Wastewater Association holds cost analysis as one of the
main points on its list of research priorities. The issues, viewpoints and findings
presented in this thesis may therefore contribute to earlier works in the research area of
water and wastewater services, especially for Sweden.

[8]
Acknowledgements

I would like to send my appreciation to Dimitrios F. Lekkas at the University of the

Aegean for initialising this project and for offering a lot of time to discuss the features of
the water and wastewater industry.

Additional thanks to my supervisor Thomas Sandberg at the Royal Institute of

Technology for the helpful comments in order to successfully establish this thesis.

Furthermore, I would like to express my gratitude to the Swedish Water & Wastewater
Association (Svenskt Vatten AB) and personnel for providing essential data samples
needed for the analysis.

Not to fail noticing, thanks to the two parties in Greece for being helpful and interested
in providing data, although the information was not enough to make an empirical study.

Finally, thanks to my family and friends for all the love and support.

Stockholm, June 10th 2008

Madelene Malmsten

[9]
Chapter 1
Introduction

1.1 Background

1.1.1 New features of the water and wastewater industry

During recent years, essentially during the last seven-eight years in Europe since the
establishment of the European Union Water Framework Directive (WFD), the
protection, improvement and usage of water sources have become important subjects
of discussion. This has initiated stronger quality demands from regulators on the
operations of water and wastewater establishments. The Directive was produced by the
European Commission and updates the exiting water legislation. Its key objective is to
drive the achievement of sustainable water resource management, which involves an
aim to achieve at least “good status” of all waters by the year 2015. The strong
environmental awareness spreading around the world and the proceeding
infrastructural changes due to globalization have caused the need for deeper
understandings about how water and wastewater establishments should be organised,
re-designed, re-built and expanded.

In some parts of the world, for example in China, a large percentage of the water and
wastewater facilities are old and the life time of the technology is reaching an end (City
Paper, Nanchang Municipal Government). At least parts of the total system are in need
for improvements or replacements. We can expect to find the same conditions in other
urban water and wastewater facilities around the world due to the fact that many of
them were first developed several years ago, as in the beginning of the 20th century.

The infrastructural changes may be one of the 21 st century’s biggest challenges, by

listening to concerned parties (see for example Clark R. M. et al (2002), eNewsUSA,
European Public Health Alliance among others). The supply of water to water consumers
and the treatment of wastewater should preferably be performed in an effective and
environmentally friendly manner. The new standard requirements affect the supply and
treatment plants in many ways, i.e. technologically, organisationally and economically.
The changes may well mean higher supply and treatment costs, leading to higher service
rates to customers.

[10]
There are many new forces behind the ongoing development of water supply and
wastewater treatment. Private investors have shown their interest in the industry. The
extents to which these forces are active differ from country to country. In England, the
water and wastewater facilities are privately owned and are mainly regulated by the
OFWAT authority (The Office of Water Services). In France, the individual communities
are responsible for controlling the facilities and many of the communities (75 percent)
have chosen to let private actors run the operations (Thomasson et al, 2003). Quite the
opposite holds for Greece, where most of the facilities (60-75 percent) are municipally
owned and controlled (E.Y.D.A.P, 2006). In Sweden, there are a few privately owned
facilities but most of the organisations are operated by the individual municipality
(Thomasson et al, 2003).

Communities intend to cooperate more and more across borders. Potential cost savings
can be made from the consolidations, especially for small water and wastewater
organisations that have difficulties keeping up with the new quality and investment
demands. The size of a water and wastewater system is determined by the needs and
demands from the surrounding environment. In order to reach economic and financial
efficiency, the services have to be operated within an appropriate size and structure.

Among other issues, the Directive stipulates cost analysis of the water and wastewater
services. This includes estimations of the economic impacts that water and wastewater
services have on the environment. Moreover, it includes studies of financial costs of the
water and wastewater services (Camp Dresser & McKee Ltd, 2004). Not only regulators
are interested in the analysis of costs. A matter of course is that the water and
wastewater organisations want to have information about their costs in order to control
their activities. This motivates the importance of cost analysis and cost estimation.
There are of course many other reasons why organisations are interested in estimating
the costs of supply and treatment. One of these reasons is pricing principles. The two
and only effective pricing principles are 1) prices based on the costs of water and
wastewater services and 2) prices based on market pricing (FAO, 1996).

As written earlier, the water and wastewater industry is in a changing era. The growth
creates new infrastructural patterns. How do we capture the economic and financial
impacts on the water and wastewater utilities due to the growth? How do we know in
which dimension the growth will scatter? As a first priority, are the existing water and
wastewater systems (or the industry as a whole) operating efficiently and can the
systems handle the rising changes successfully?

[11]
1.1.2 Two concepts: Actual cost estimation and cost structure
representation

To capture the cost structure of a single system (or a whole industry) is not necessarily
the same as to capture the actual costs raised by the system. The cost structure should
preferable be seen as the “behaviour” of the underlying technology. The cost structure
can be used for the purpose of measuring the “monetary behaviours” when for example
a service system changes in structure and/or in size. As will be shown in the theoretical
framework and in the empirical study, the cost structure is represented by a
parametrical expression, a so-called cost function.

Definition of cost structure: “A cost structure is the distribution of costs among the
elements of the process. The term ‘cost’ can be used generally to refer not only to
particular financial costs, but the use of resources generally. Roughly determining the
cost structure of a process involves understanding the individual steps and resources
that are consumed by the process, and characterizing how they are allocated and how
they scale with size of the task. Determining the cost structure is a useful step in any
effort to make a task more efficient, because it helps to focus attention or technology
on the places where it can be most effective” (Palo Alto Research Centre, 2008)

The cost structure of a system can be analyzed in order to assess efficiencies and
inefficiencies in the system. The information obtained from the analysis can be useful in
many other situations, e.g.

1) When planning new construction and/or replacement projects. It can be seen as

a complement to actual cost estimations.
2) When pricing the water and wastewater services.
3) When evaluating the financial situation of a system (or a total industry)

On the other hand, actual costs are more accurately estimated by using bottom-up
procedures. This means that the costs of activities are estimated separately and added
up to a total cost. Ramirez (2001) has used the bottom-up approach to estimate total
cost of a water and wastewater utility. He estimated the costs for each activity of the
water and wastewater system. Each part of the system was evaluated in terms of costs;
the investments of the constructions and the operational costs of working the system.
The model developed by Ramirez (2001) has been practically implemented and is a part
of the research programme “Sustainable Urban Water Management”, by CIT Urban
Water Management AB in Sweden. The project ended in June 2006. The model exists
now as an Excel-file in Swedish where the user can choose inputs of infrastructure,
water and wastewater treatment, recycling etc.

[12]
1.2 Problem statement
The research question is formulated as follows:

How can an economic/mathematical model be created and designed in order to 1)

assess the (or a) cost structure of the water and wastewater services and 2) measure the
“behaviour” of the costs when the service system changes in size and structure?

The question is both of a theoretical and a practical nature. For that reason, the
question will be approached and discussed from theoretical and practical points of view.
There is room for personal opinions and judgement, since the question does not impose
any absolute truths. Additional characteristics of the research question are mentioned in
Chapter 2: Methodology.

1.3 Purposes of the study

The study has several objectives. The main purposes are to examine how the
economic/mathematical model can be created and designed, to understand and
describe how we can assess the (or a) cost structure of the water and wastewater
services. This includes forming an understanding and an opinion about what variables
should be included in the model.

Another purpose of the study is to try to create a bridge between the mathematical
models presented in this study and the water and wastewater services. To my
knowledge, there is not much literature written about the relationship between the
mathematical models so often used in the econometrical literature and the theoretical
framework underlying the models.

This thesis is also a direct contribution to the Swedish research field of water and
wastewater. The empirical study applies parts of the theoretical framework on a section
of Swedish communities, each responsible for the local water and wastewater system.

1.4 Outline
This thesis is organized as follows: Chapter 1 introduces the topics of this thesis, states
the research question and purposes of the study. Chapter 2 describes the methodology
in terms of ontology, epistemology and the principles of the research approach. Chapter
3 presents theory in order to understand the industry and the following reasoning in
Chapter 4. Chapter 4 holds the empirical study, which estimates a cost structure for a
section of Swedish water and wastewater organisations. The subsequent chapter also
[13]
contains an analysis of the estimations. Chapter 5 summarizes the conclusions of the
study.

[14]
Chapter 2
Methodology

This chapter specifies the methodological choices underlying the thesis. This includes a
description of ontology, epistemology and research approach. Ontology is the
fundamental assumptions about the nature of the world and epistemology refers to the
way one chooses to collect knowledge.

2.1 Ontological and epistemological platforms

The aims of this study are, firstly, to recognize the courses of action (to understand) and,
secondly, to answer the research question (to explain). The analysing tools are
modelling and interpretation. The aims have been accomplished by looking at the
objectives partly from a positivistic point of view and partly from a hermeneutic point of
view. I use the following example (1) to illustrate how my research question could have
been formulated if I had possessed either a positivistic view or a hermeneutic view and
finally how my research question has been formulated.

Example 1: The research question

A strictly positivistic influenced question could be: “How is an economic/mathematical

model created and designed in order to 1) assess the (or a) cost structure of the water and
sewage services and 2) to measure the “behaviour” of the costs when the service system
changes in size and structure?”

On the other hand, a hermeneutic influenced question could be: “How does performer
number 1, 2, 3, etc. believe that an economic/mathematical model should be created and
designed in order to 1) assess the (or a) cost structure of the water and sewage services and
2) to measure the “behaviour” of the costs when the service system changes in size and in
structure?”

My research question (influenced by the two perspectives): “How can an

economic/mathematical model be created and designed in order to 1) assess the (or a) cost
structure of the water and sewerage services and 2) measure the “behaviour” of the costs
when the service system changes in size and structure?

[15]
The positivistic viewpoint issues an objective reality and a materialistic ontology. Thus,
a model is a reflection of the reality. This viewpoint is traditionally followed by
quantitative methods. The epistemological platform is “the experience”. In contrast, the
hermeneutic viewpoint is based on subjective opinions, a non-material ontology and the
analysis is mostly qualitative. The epistemological platform is the sense, “the reasoning”.

Due to the fact that I do not see the reality with an explicit positivistic or hermeneutic
point of view, my research question is influenced by both of the aspects. It is probably
not very unique to possess this two-folded point of view.

I do not apply a rigorous critical theory, which means that I do not seek total renovation
with my analysis. Nevertheless, I do have the ambition to make a contribution to the
research field by adding perspectives that I believe have not been addressed
(sufficiently enough) in earlier studies. The empirical applications add to this value
creation.

2.2 Research approach

I have approached the problem on a base of deductive reasoning i.e. I have taken part
of existing knowledge through literature studies and I have related the review to an
empirical study. However, I have not strictly abided by the deductive approach. Some
parts of the thesis have its roots in theory and others in practise, some parts are more
quantitative and others more qualitative. Johansson – Lindfors (1993) calls this
approach “The golden mean”, which stands for a more open relationship to one’s
choice of method. Statistical and mathematical techniques have been combined with
economic theory and methodology.

The problem statement results in both qualitative and quantitative analysis.

Considering the formulation of the question, one might simply settle with a qualitative
analysis. However, this approach does not give an answer to if the model can be
applied practically on a real-world case. If the model contains certain characteristics
that can be best understood by measuring them, it is worth doing so i.e. to
“operationalize” (Eriksson & Wiedersheim – Paul, 2006). I stand by the belief that it is
possible and necessary to quantify my qualitative statements in order to make any
useful conclusions.

To summarize my choice of method, I herewith answer the three main methodological

questions stated by Johansson – Lindfors (1993): How? Where? Why? The following
scheme includes these questions with my corresponding answers. A forth question has
been added and refers to how I have used, analysed and interpreted information.

[16]
 • How have I searched for information?
By conducting a literature study (=> “the approach”), reviewing the
findings and by speaking to some of the industry authorities in order
HOW? to take part of existing information (=> “the direct information
collection method”).

• Where have I searched for information?

Essentially, I have searched in research papers, articles, consulting
reports, journals, books and other forms of literature on the internet
and in nearby libraries. Keywords in the study are for example Water
and Wastewater Industry, Science , Technology and Management,
WHERE? Cost Modelling and Analysis. Some of the references are strictly
industry specific and others refer to a wider public. The numerical
data was collected by contacting the authorised individuals in the
industry.

• Why have I chosen this reseach approach?

The problem statement is of a complex nature and for that reason the
reliance on existing papers were a preferable choice. Most of the
needed information was seeked out in the literature, mainly by
WHY? reviewing research papers. I believe that this method brings about the
best quality of this thesis, taking into account the underlying
resources of this project.

• How have I used, analysed and interpreted information?

The numerical information have been used, analysed and interpreted
by applying statistical and econometrical methods. Some
INTER- shortcomings in the reviewed papers have been identified in order to
PRETATION try to give new theoretical contributions to this research field. The
practical applicable process goes hand in hand with the theory
developing process.

Figure 1: Summary of the choice of method. Source: Author.

[17]
Chapter 3
Characterization and modelling of the
water and wastewater services

This chapter is a theoretical framework, which is the result of a literature review on the
characteristics of water and wastewater services. It presents the emerging costs of
proving the services. The chapter intends to give the reader a brief introduction to the
field and in the same time it focuses on what needs to be understood in order to
produce a cost structure model that could be applied to a number of water and
wastewater systems. The chapter highlights a discussion about how one might choose
to represent and analyse a water and wastewater system.

3.1 The econometric analysis

Econometrics is the study of estimation and inference for economic (or financial) models
using economic (or financial) data (Hansen, 2008).

The main difference between an economic analysis and a financial analysis is that the
first one aims to improve the social well being of society in terms of income and the
second one aims to improve the financial situation of the operating entity (Asian
Development Bank, 1999). This means that the economic analysis estimates the
economic impact on the country’s economy and the financial analysis estimates the
financial impact on the operating enterprise. Financial analyses ignore to value
externalities that capture the additional social costs, such as the environmental impacts
of the enterprise’s operations (Gordon, 2005) Externalities are valued in the economic
analysis in order to sustain the health of society and environment. A core part of the
economic analysis is to identify the extent to which economic values are reflected in
pricing policies (Taylor, Stone and Webster Consultants).

I have made a financial analysis, by applying the direct cost approach. Direct costs are
the investments and operating costs incurred by the responsible agency in production of
the service (Gordon, 2005).

The analysis in this study involves an application of econometric tools on the financial
data.

[18]
3.2 The water and wastewater service system
The framework of this thesis includes the following parts of any water and wastewater
organisation:

I. Water Treatment Plant, WTP (including buildings, machinery etc.)

II. Distribution, DN (including pipes, pressure and pumping stations,

reservoirs, valves, fire hydrants and other apparatus)

III. Sewerage, SN (including sewage pipes, storm water pipes [combined or

separated], pumping stations, valves and other apparatus)

IV. Wastewater Treatment Plant, WWT (including buildings, machinery etc.)

Figure 2: A conceptual framework of the water and wastewater service, with a few descriptive variables
of the complex system. Source: Author.

[19]
The following figure (3) is an example of how to simplify the representation of a service
system, developed by Ramirez (2001).

Figure 3: System structure, Ramirez (2001) Urban Waters

[20]
3.3 Water as an input and output variable
Water can be seen as 1) a final physical good and/or 2) a service that is provided to
markets (customers) having certain demand characteristics (Kim, 1991). In this thesis,
water is regarded as a service rather than a physical good. Except this, water can be
mentioned in terms of input and output. The notions output, input and water
consumer might be confusing in the bigger context. Down follows an explanation to
how the terms are defined in this paper.

Water is an output from the production process to the water distribution system and it
is later in its impure form an input to the purification system (wastewater treatment).
The customers are the water consumers. First, water is produced and distributed
through networks to connected customers. Second, sewage water from connected
customers is collected through wastewater networks to the purification process. For a
joint water and wastewater organisation the resulting production output is therefore
the water and wastewater service. Literally, by service we mean the procedure of
supplying water and treating wastewater. Numerically, as will be shown in the
mathematical framework, volumes of supplied water and treated wastewater are the
production outputs, i.e. they are variables that represent the service.

Customers = water consumers

Water Wastewater
Treatment Water output = Sewage water Treatment
Plant water delivered input = sewage Plant
water collected
Distribution Sewerage

The Water and Wastewater services

Figure 4: Clarification of a few production terms. Source: Author.

Water delivered: estimated volume of water supplied to the boundary of each

customer’s property (OFWAT report 2005-06).

Sewage water collected: estimated volume of water returned from customer’s

properties to the wastewater network (OFWAT report 2005-06).

[21]
3.4 Cost categories
The costs of water supply and wastewater treatment can be divided into essentially
three main elements: financial, environmental and resource costs.

Financial costs: the costs of producing and distributing water, as well as the costs of
managing the wastewater treatment. The financial costs consist of

 Capital costs (incurred by purchased assets)

- Depreciation charges (based on life-expectancy of assets)
- Interest rate costs

 Operational & maintenance costs, O&M (incl. or excl. improvements,

emergency and preventive maintenance)
- Labour/Employment costs
- Electricity/Energy costs
- Material costs
- Other contracted services

Environmental costs: basically the costs of any environmental deterioration

expressed as monetary value. Environmental externalities are much more difficult to
assess then the financial statements.

Recourse costs: opportunity costs, i.e. the costs incurred by choosing one option over
an alternative option.

Other relevant cost categories that are used in economic analysis are fixed, variable,
average, marginal, unit, short- and long-run costs.

Fixed costs are costs that do not vary in line with the production.
Variable costs are costs that vary in line with the production.
Average costs of supply and treatment can be estimated by dividing total cost of
supply and treatment with the total volume of water produced and treated.

Marginal cost of water supply and treatment is calculated as the derivative of the
total cost with respect to water quantity. The marginal cost is the change in total cost
when the quantity of water changes by one unit.

Unit cost is simply a cost per unit, for example €/quantity, €/connection, €/hour,
€/kWh etc.

Depending on policies, the costs can be determined on a short- or long term basis.
Residential and non-residential water consumers are charged a price (tariff) based on
the costs per unit of consumed water and an on additional fixed cost per year (a flat
rate or fixed rate). The former reflects the short term use and the later the long term

[22]
use. The charges cover the costs of operations, maintenance and capital to some
extent and the rest is put on local authorities and financed through taxes.

The industry of providing water is very capital-intensive, second to the energy sector
(Donald, 2003). Salaries represent one of the largest groups of expenses. Expenses that
vary in line with the volume of produced water are for example expenses for energy (in
pumping) and chemicals (for treatment). Due to modern technical interventions and
improvements the costs of desalination and electricity are expected to be reduced in
the coming years.

3.5 Challenging features of the cost structure analysis

C. Barucq et al. (2006) state a number of difficulties in analyzing and comparing costs of
water and wastewater services across different cities:

1) “In the case of not having accounts separating water and wastewater services,
assumptions have to be made in order to distribute expenses between the two
services, introducing a margin of error in the reassignment of costs.“

2) “Certain investment financings by a municipality, region, State and/or European

fund could not always be clearly identified or precisely estimated”.

3) “Accounting structures vary from one capital to another.”

4) “When several players are involved in managing the services consolidating on

total costs is complex and sometimes difficult to achieve. “

5) “Differences in production factor costs among public and private operators are
a product of national features and cannot easily be interpreted in an
international system”

The British regulator OFWAT makes an important observation (report 2005-06)

concerning capital charges, namely that there exists different accounting practices
between water companies. This may complicate the procedure of making comparisons
of costs between companies and communities. In Sweden, according to SWWA
Department of Development (report 2007-13) the capital costs are in most cases
reported as an entirety cost, i.e. no separate accounting for each separate part, such as
the distribution. SWWA has introduced a fictitious/theoretical capital cost for the pipe
network based on the average age of the pipe system, its structure, climate conditions
and topography. SWWA uses the same depreciation and interest rate for every
community.

In need of financial figures, consulting reports are accessible information sources. In

these reports, many figures are approximations and not actual real values. The reports

[23]
are quite easy to get in hand, but they often include a lot of weaknesses (Kärrman, Olin
et al. 2006), i.e.

 They are only judgements of the costs, often made before a project has started.
Environmental circumstances might be missing.
 They include preliminary technical forecasts, such as amount of customer
connections.
 The O&M costs are often just presented as outlines (as a percentage of total
investments).
 They have not always taken into account the regulation requirements from
authorities.
 They are based on the individual knowledge of the planner or investigator.

There are important facts to keep in mind when determining statistical models of water
and wastewater systems, or of any other systems for that matter (John Cubbin, 2004):

A) Even with totally accurate data and models, estimates are just estimates.
B) Estimates are subject to sampling errors as long as there are limited observations.
C) There may be errors in the measurement of the dependent variable.
D) There may be variables excluded from the analysis.
E) The explanatory factors may themselves be proxies and/or subject to errors of
measurement.
F) The wrong mathematical form may have been chosen to approximate the
relationship between costs and their drivers.

Cubbin (2004) also states that simplicity is a premium when there are few observations,
as is often the case in water industry econometrics. This means that one have to have
knowledge about the industry in order to not have too many omitted variables and
determining factors.

3.6 Modelling under real life conditions: a step-by-step

approach
The British regulator OFWAT presents a practical step-by-step approach to derive
statistical cost models (Report 2005-06, Appendix 1). Noticeably, my study does not
contain such a sophisticated step-by-step procedure, because the needed resources for
that kind of endeavour are missing (i.e. excessive time and expert judgment). Some
steps are not being considered in this study.

[24]
 The OFWAT approach A modified approach for this study

Modified: Expert reviews are represented by

the studied research reports.

Performed

Modified: System structure analysis.

Performed

In practice, there
are a number of
Modified:
A) Statistical analysis to generate
iterations through
relationships between expenditure and a
steps 5, 6 and 7 as
few explanatory factors (a cost
we develop and
structure).
refine the models
B) Assessing economic measures.
C) Conclusions and judgments.

Figure 5: A step-by-step approach to derive statistical models.

Source: OFWAT (Report 2005-06, Appendix 1).

[25]
3.7 Identification of possible cost and scale drivers: the
explanatory variables
As been mentioned in the introduction, it is important for the reader to understand the
different concepts of Cost Structure Representation and Cost Estimation. With the term
“cost model” or “cost function” we mean a model that captures the technological and
structural characteristics of a system, in terms of costs. We do not aim at a model that
estimates the actual costs (capital costs and variable costs) of a system, at least not as a
first resort. The model can be used for the purpose of studying in what direction and to
what extent the costs of a business get affected by a system change. For example, by
holding a representative mathematical cost structure of a system and by knowing that
the marginal cost depends on the production quantity, we most certainly have enough
information in order to calculate numerically how the marginal cost gets affected by the
production quantity. We (simply) take the derivative of the total (annual) costs with
respect to the (annual) quantity. A well behaved cost model can also capture how other
components affect the marginal cost, e.g. how prices and operational variables
(distance, network, density etc.) affect the marginal cost.

An important step in modelling is to take a stand on critical components. Unit costs have
here been recognized as critical components, e.g. Cost/hour (Labour force), Cost/kWh
(Energy), Cost/material index (Material) and Cost/capital index (Capital). As will be
shown in the empirical study, these unit costs enter into the cost model but they are
expressed as unit prices instead: Price/hour, Price/kWh, Price/material index and
Price/capital index. Principally, we assume that the water and wastewater utilities are
price takers on the factor market. The prices are fixed and consequently the arising
costs. (Stone and Webster Consultants Ltd, 2004)

Except prices, there are other components that affect the costs for the water and
wastewater utilities: structure and length of the pipe networks, the type of treatment
plants, how closely people live to each other (Customer density), the distribution of
customers (Industrial users, Domestic users and Public users), the quality of the water
etc. It is desirable to construct a model that captures all of these components. However,
it seems reasonable to believe that it is difficult to construct such a model. The amount
of components is thousands. If we wish to characterize the water and wastewater
services by modelling, it might be necessary to separate the services in different parts
and construct models for these individual parts. This case is the normal case in Cost
Estimation. Reasonably, it is not a big deal. However, complications arise when we want
to represent the business in simply one model, which may be the case when we want to
fit a cost structure for the total industry. For example, Hanemann (1998) says that the
economic connection between water and wastewater can be neglected if the systems
are analysed separately:”The systems may involve different technological choices, but

[26]
the choices are inter-dependent: the costs of water supply are likely to affect decisions
on waste disposal, just as the costs of waste disposal are likely to affect decisions on
water intake and use.” Figure (6) is taken from OFWAT and contains the critical
components that they use for benchmarking the English water and wastewater firms.
These critical components represent expert judgments of potential cost drivers in the
industry. A few of these components have been used in the empirical study.

Generally, what variables we choose to include in the model is a judgment between

variables we want to get explained by the model and variables we have to include in the
model in order to fulfill its constraints.

[27]
 Operational Explanatory Capital Explanatory
expenditure variables included maintenance variables
in the model expenditure included in the
model

Water distribution Resident Water distribution Length of main

model population (scale) model (scale)

(log model) Small/Big mains (log unit model) Connected

(cost drivers) properties/length of
main (cost driver)

Water resources Resident Water distribution Pumping station

and treatment population (scale) model (non- capacity (scale)
model infrastructure)
Supplies from Tower and service
(linear model) boreholes (cost (log unit model) reservoir capacity
drivers)
Ratio of storage
Number of sources capacity to pumping
station capacity
Distribution input

Water power Terrain Water Billed number of

model management and customers (scale)
Power general model
(log linear model) Billed properties that
Pumping head & (log model) are non-household
Distribution input (cost driver)
(cost drivers)

Water business Billed properties Sewage treatment Total load (scale)

activities model (cost driver) model
Load treated by
(log linear model) (log unit model) sewage treatment
works

Sewerage Network Sewer length Sewage Total length of

(incl. power) model (scale) infrastructure sewers (scale)
model
(log linear model) Area of sewer Combined sewer
district (log unit cost overflows (cost
model) driver)
Resident
population

Holiday population

Large sewage Load, type of

treatment works treatment &
model effluent consents
(cost drivers)
(log linear model)

Figure 6: Potential cost drivers. Source: Author, based on OFWAT’s Appendix 1 in report 2005-06

[28]
3.8 Structural-conceptual representation of the service
system
Structural conceptions are frequently used in organisational theory. Structures can be
more or less visible, but we are anyhow often in the need to specify the underlying
structures of a process (Bakka, 2001). A structural representation of a system can
make it possible for the researcher to identify and understand its characteristics. This
may be helpful when controlling the system.

A traditional organisational analysis means that we step “inside” the entity in order to
specify its structural characteristics. Bakka (2001) has specified two common
characteristics of big organisations: complexity and formalization. An urban water and
wastewater organisation is expected to possess both of the mentioned characteristics.
The water and wastewater organisation is complex in the meaning of having many
components that interrelate to one another and it is formalized due to certain rules
and organisational plans.

3.8.1 Perspectives
When studying the characteristics and efficiency of a system we can start by analysing
it from different dimensions. For example, the water and wastewater system (or the
total industry) can be analysed vertically or horizontally. We can also analyse the
system in terms of product, function, customer or area, as well as any combination of
them all.

By using the structural perspective we can better understand the processes of the
system. A system is often analysed in the same direction as the system is broken down.
The researcher holds the responsibility to specify how the system is broken down and
in what dimension it is analysed. If this is not specified in the study or obvious by itself,
then there might be difficult to interpret and understand the analysis.

V
E Water Industry Water
Billed
R Treatment Treatment
Domestic
T Plant
In - leakages
Plant
I Public
C
A
L
HORIZONTAL

Figure 7: Vertical or horizontal perspectives of analysing a system. Source: Author.

[29]
---------------------------------------------------------------------------------------------------------------
1. Vertical analysis: Separate analysis of services, broken up into e.g. (1) water
supply and (2) wastewater treatment.
2. Horizontal analysis: Combined analysis of services, which means a joint analysis
of the water supply and wastewater treatment.
---------------------------------------------------------------------------------------------------------------
3. Product analysis: Focuses on the study of production outputs, i.e. water supplied,
water treated, and/or water charged, wastewater charged plus out (in) leakages.
4. Function analysis: A functional break down into water supply, water distribution,
sewage collection and sewage treatment.
5. Customer analysis: Analysing the distribution of customers, i.e. industrial users,
domestic users and public users.
6. Area analysis: Focuses on the characteristics of the area, i.e. formation, size,
environment etc.
---------------------------------------------------------------------------------------------------------------

The water and wastewater technology can better be understood if it is modelled as a

multiple output production process (Kim, 1995). The multiple outputs can be supplied
water and treated wastewater. This is an example of the Horizontal-product analysis. If
the outputs are regarded as supplied water and treated wastewater to customers, i.e.
water to industry, households and public uses, we focus on a customer analysis. This
analysis can be horizontal or vertical, i.e. horizontal if we regard both water and
wastewater, vertical if we regard only water or only wastewater.

One of the most essential steps in the cost analysing process is to define the
underlying functions of the system that is being studied. To my knowledge (based on
what has been found in the econometric literature) it seems to be a lack of explicit
structural representations of the systems that have been modelled. This can cause
confusions, since a water and wastewater service system is of a very complex structure
with many variables.

3.8.2 Production growth

“Growth occurs from an existing situation, with an existing infrastructure, land use and
ownership.” (Hopkins et al. 2003)

A water and wastewater service system is not a fixed state in time, it is steadily
changing. Nowadays, we can expect a metropolitan water and wastewater service
system to grow bigger and bigger in production scale. This is mainly a cause of
globalization, consolidations and growing populations.

[30]
 Example: Resemblance Stable relationships:
1) Increasing water Increasing production Customers/service area
demand per customer capacity

2) Increasing customers Demand/customer

per service area
3) Proportional Demand/customer
expansion of service
area and amount of
customers
4) Proportional Demand/customer,
expansion of Customers/community
customers and
number of local
communities

Figure 8: Examples of production growth. Source: Author.

3.8.3 A few important concepts

In order to understand the economic state of an existing service system or a total
industry, one might want to study if the systems are being operated efficiently. In that
way, one can understand if the existing systems can handle the growing changes or if
the systems are in need for re-investments. Efficiency can be measured in terms of e.g.
Economies of scale, Economies of production output density, Economies of customer
density and Economies of scope.

Economies of Scale: are achieved when it is profitable to enlarge the business i.e. to
produce more units of goods or services, increase the network size and the number of
customers. The quantity of water demanded by customers and the density of customers
remain stable. The opposite, diseconomies of scale, occurs when the production
requires more input than produced output i.e. it is not profitable to enlarge the
business. Economies of scale refer to the decrease in unit cost as the production scale
increases. The changing variables are normally said to be the production output, the
amount of customers and the network size. The network size is a geographical
representation of the service region e.g. the areal of the service region in km2 or/and
pipe line distance from the plants to the end consumers or/and distance from raw water
source or/and amount of communities.

[31]
 Decreasing variables Increasing variables

Unit costs The whole production scale: i.e.

proportional increase of production
output + amount of customers +
network size etc.

Figure 9: Economies of scale. Source: Author.

Economies of production output density: are achieved when it is profitable for the
production to increase for a given system size and a given amount of customers (Garcia
and Thomas, 2001). The only changing variable is the production output, due to the
increasing demand of water volume per customer.

Decreasing variables Increasing variables

Unit costs Production output

Figure 10: Economies of production output density. Source: Author.

Economies of customer density: are achieved when it is profitable for the production
to increase with the amount of new customers (Stone and Webster Consultants, Ltd,
2004). The system network size is held constant while the amount of water volume and
the amount of connections increase proportionally. Hence, the changing variables are
the production output and the amount of connections.

Decreasing variables Increasing variables

Unit costs Production output + amount of

customers/connections.

Figure 11: Economies of customer density. Source: Author.

Economies of scope: are achieved when a single firm can produce several goods or
services at a smaller cost than would several firms specializing in each separate
commodity (Garcia and Thomas, 2001).

[32]
 Decreasing variables Increasing variables

Unit costs Numerous/type of goods

Figure 12: Economies of scope. Source: Author.

[33]
3.8.4 Production patterns
The structural representation of a water and wastewater network can become
complicated. In order to build a cost structure model for (parts of) the Swedish industry
(in this case) we need to understand how the model should be constructed so we can
get hold of the things we want to get explained by the model. In this thesis, the model is
constructed through a system of equations (a statistical regression with each of the
firms), why it is important to control for different production patterns. Some patterns
can be difficult to control for numerically. For example, if we choose to include more
than one community in the framework, we have to control for different production
output patterns between utilities. We need to understand that outflows from one utility
might be greater than another utility. For instance, the first one might distribute water
to two communities (or more) and the second one only to one community. Different
output patterns are being shown in the following schematic examples. In order to
diminish the complications, I chose to model utilities that serve simply one community.
Yet, the determining reason for making this choice was that I did not possess
information about how many communities every utility operates. The model can be
modified later on by including how many communities every utility operates. The new
information should enter the model as one or several operational variables. The
boundary of urban service systems is thus in this thesis and at this point systems that do
not service outlying communities.

Example 2: Co-operational production output flows. I

[34]
Example 3: Co-operational production output flows. II

Example 4: Co-operational production output flows. III

[35]
Additionally, other operational variables that might be difficult to control for
numerically can be treated as dummy variables (takes values 0 or 1). For example, a
dummy variable is included in the model and takes the value 1 if the corresponding
utility distributes chemically treated water and the value 0 otherwise. Féres and
Reynaud (2006) have included dummy variables that control for licence agreements
(yes/no) and environmental regulations (yes/no). Urakami (2005) has included a dummy
for dam water and one for groundwater. Martins, Fortunato and Fernando (2006) use
one dummy for ownership. Bottasso and Conti (2003) constrain their mathematical
function by including a dummy for wastewater (yes/no). By reviewing the literature, we
can see how different authors incorporate different and various variables in their
mathematical formulas. We recognize the importance of identifying the underlying
system structures that is being studied in every case. Again, variables we choose to
include in the model is a judgment between what we want to get explained by the
model and what we have to include in the model in order to fulfill its constraints. This
choice may be more or less of one’s own free will. Information availability is a critical
issue for any industry analysis.

[36]
3.9 A mathematical framework

3.9.1 An approach to derive cost functions

Following Hanemann (1998) we can consider an establishment (in this case a water
and wastewater utility) producing water (output 1) and cleaning wastewater (output 2)
through the use of certain N inputs, e.g. capital, labour, electricity and material. The
price of input k (where k = 1, ..., N) is denoted by wk and the amount of the input k is
denoted by xk. The total cost of production and treatment for the water and
wastewater utility is then given by the function

N
TC   wk xk  w1 x1  w2 x2  ...  w N x N Equation (1)
k 1

The volume of produced and treated water/wastewater by the utility at a specific unit
of time is denoted by y. It is relevant to relate the output y to the N inputs in order to
get the production function for the utility. The literature review makes obvious that
the so- called Cobb-Douglas production function was one of the most frequently used
production functions in the applied field of water supply and treatment in the 1960s
and 1970s. The Cobb-Douglas function has however been abandoned in favour for
other production models, such as the Constant Elasticity of Substitution (CES)
production function. For the moment we hold on to the Cobb-Douglas production
function in order to understand the concept of a cost function that includes prices and
outputs rather than production quantity.

The Cobb-Douglas production function is given by the following expression,

N
y  A   x kak  Ax1a1 x 2a2  ...  x Na N A  0, a k  0, k  1,..., N Equation (2)
k 1

where A, a1, ..., aN are parameters that could be given or that should be estimated.
They depend upon the relationship between the output and the inputs.

The CES production function has a somewhat more complex parametric structure than
the Cobb-Douglas function. If we suppose that a water and wastewater utility can be
represented by a Cobb-Douglas production function and the utility wants to minimize
its total costs, a suitable problem to solve is the following (where x1, … , xN are the
optimal input levels),

[37]
 N N
min TC   w k x k subject to y  A   x kak Equation (3)
k 1 k 1

Without questioning the microeconomic proofs referred to in the notations of

Hanemann we introduce a set of functions, so-called long-run conditional input
demand functions, which give the optimal choice of inputs,

x k  g k (w1 , w 2 ,..., w N , y) k  1,..., N Equation (4)

That is, the gk function represents the demand for input xk.

This yields the total cost of production for the utility at the optimum,

N
TC ( w1 , w 2 ,..., w N , y)   w k g k ( w1 , w 2 ,..., w N , y ) Equation (5)
k 1

If we, on the other hand, are interested in the short-run total cost of production at the
optimum we fix the capital input at a level of xN, where N stands for the input N (=
capital). The other 1, …, N-1 inputs are variables (e.g. labour, electricity and material).
The total cost is denoted by TC, the total variable cost by TVC and the total fixed cost
by TFC (= wN xN).

TC ( w1 , w 2 ,..., w N , x N , y )  TVC ( w1 , w 2 ,...w N 1 , y )  w N x N 

N 1 Equation (6)
  w k g k ( w1 , w 2 ,..., w N 1 , x N , y )  w N x N
k 1

As we see in the above expression the short-run-case gives demands for inputs
conditioned on the levels of the fixed inputs, i.e. x k  g k (w1 , w 2 ,..., w N 1 , x N , y) for
k=1,…,N. In the long-run-case, the demands for inputs are conditioned on the prices of
the fixed inputs, i.e. x k  g k (w1 , w 2 ,..., w N , y) for k=1, …, N.

[38]
We can now derive the cost function TC(w1, w2, ..., wN, y) for N = 2 in the cases of a
Cobb-Douglas production function. We get the following results with the use of
Shephard's lemma1:

1 a1 a2
2
 y  a1 a2  w1  a1 a2  w 2  a1 a2
TC ( w1 , w 2 , y )   w k g k ( w1 , w 2 , y ) (a a  a 2 )     
k 1  A  a1   a2 
1 1

 y a1  a 2  a1 w 2  a1 a2
where x1  g 1 ( w1 , w 2 , y )      , x 2  g 2 ( w1 , w 2 , y )
 A  a 2 w1  Equation (7)

The same approach yields for the CES production function. The resulting cost function is
1

 
μ 1
y 1
TC(w1 ,w 2 ,y)  δ1ζ w1ζρ  δ2ζ w 2ζρ ζρ where μ  0 , ζ  0
A 1 ρ Equation (8)

where the δ are share parameters, ρ depends on the degree of substitutability of the
inputs and A and μ depend upon the units in which the outputs and inputs are
measured.2

1
For detailed information, see Andersson F, 2007, “Kostnadsminimering och Shephards Lemma”, Lecture
Notes: March 26.
2
www.minneapolisfed.org/research/prescott/macro_theory/Lectures/CESProdFn.pdf

[39]
3.9.2. The Transcendental Logarithmic cost function
Nowadays, the Transcendental Logarithmic (Translog) cost function is often used by
researchers for the purpose of representing the production technology (see for example
Reynaud, 2006, and Martins, 2006, et al.) The Translog cost function is presented in this
section.

The Translog cost function seems to be the most popular type of cost function used in
the water and wastewater industry. Stone and Webster Consultants Ltd (2004) used the
both the Translog function and a function called the generalised quadratic specification,
the last being more ad-hoc than the first. The formulas can be recognized as equally
complex in mathematical form. I use the Translog cost function in my empirical study.

Conditions: The Translog cost function is often mentioned in the literature of water
supply costs, but its practical use is not restricted to the water industry and it has been
applied within other industries such as the banking industry and the electricity industry.
The Translog cost function has the advantage of being flexible in its form i.e. it has few
restrictions on its input data. For this reason researchers have chosen the Translog cost
function in favour for other more constrained cost functions. The Cobb-Douglas cost
function restricts the inputs to be only substitutes and the CES cost function restricts the
inputs to be substitutes and/or complements. The Translog cost function relaxes these
constraints and allows the degree of complementarity and substitution to be different
between different sets of inputs. It also allows inferior inputs (see Hanemann, 1998)

Substitutes, complements, normal and inferior inputs: The meaning of inputs being
substitutes can be described as the following: two inputs i (e.g. electricity) and j (e.g.
labour) can be categorized as substitutes if an increase in the price of input i leads to an
increase of the demand for input j.

Described in the same sense, two inputs i (e.g. electricity) and j (e.g. labour) can be
categorized as complements if an increase in the price of input i leads to a decrease of
the demand for input j.

Typically, the demand of inputs increases when the production output increases.
However, there may be cases when the demand for some inputs decreases as the
production output increases. In these cases we define these inputs as inferior (second-
rated) rather than normal. For example, this usually happens when an increase of
production output (volumes of water) causes an increase in the demand for certain
inputs/input (electricity) in exchange for other inputs/input (labour). The stated
example illustrates a typical technological change when the production process gets
further mechanized and the demand for labour decreases in favour of the industrial
development.

[40]
 Type Price of input i Demand of input j Production output
(electricity) (labour)
Substitutes Increase Increase -
Complements Increase Decrease -
Normal input j - Increase Increase
Inferior input j - Decrease Increase
Table 1: Visualization of substitutes and complements, normal input and inferior input

Formula: The Translog cost function has the advantage of being flexible in its form, i.e.
it has the fewest restrictions on the input data.

2
1
ln TC(w1 ,w2 ,y)  ln β0  β y ln y   βk ln wk  δ yy ( ln y)2
k 1 2
2
1 2 2 Equation (9)
  δky ln wk ln y   δki ln wk ln wk
k 1 2 k 1 i 1

where β ' s and δ' s are the coefficien ts to be estimated and δki  δik

Notice: The exact mathematical restrictions on the Translog cost function are stated in
the empirical study.

[41]
Chapter 4
Empirical study and analysis

This empirical study includes an application of the theoretical framework on a few water
and wastewater organisations in Sweden. The chapter specifies the estimation
procedure, a structural conceptual representation that underlies the mathematical
model, a description of the data samples and the resulting analysis. A model has been
fitted for the total variable cost (TVC) of the production. If one wishes to fit a model of
the total cost (TC) of the services, the price of capital has to be included in the model.
Due to data limitations, the model was restricted to only capture the variable cost.
However, this is not a problem for the structure analysis. Either total cost or variable
cost modelling can be used to capture the cost structure of the underlying technology
(Stone and Webster Consultants, Ltd. 2004). If the variable cost function is minimized
with respect to the capital stock then it will yield the same economically relevant
information about the underlying technology as the total cost function. But of course,
the variable cost function is a representation of the variable costs in monetary terms
and nothing else. It is the concepts of Economies of scale, Economies of customer
density, Economies of production output density and Economies of scope that can be
assessed in the same manner by using either a variable cost function or a total cost
function. It is assumed that the utilities aim to minimize their costs.

TC (output, prices; Operational environment) =

= TVC (output, prices; Capital stock, Operational environment) +

+ FC (price, Capital stock)

The capital factors (= pipe length in my application) are treated as quasi-fixed inputs in
the short-run, which mean that the capital is not fully under control of the individual
organisation because of external quality investment demands (Stone and Webster
Consultants, Ltd. 2004). I have used the pipe line network of water and wastewater as a
proxy for the capital stock. The capital prices are hereby based on the network. Other
relevant proxies for the capital stock are the amount of plants, pumps, pressure stations
etc. To reduce the expressions in the cost function, I settled with only the pipe network.
However, it was more the fact that the data material was incomplete that limited the
use of variables. There were too many missing data in the samples. The simplification of
the capital stock makes it rather inappropriate to estimate Economies of scale,
Economies of production output density and Economies of customer density in the long-
run. The long-run version of the measurements requires a more accurate representation

[42]
of the capital stock. We can agree on the fact that simply the length of the pipe network
is not a genuine estimate of the total capital stock. A more truthful estimate of the
capital stock might be the book value of the utilities. Nonetheless, this choice of a proxy
can be seen as legitimate as a last resort, when we have limited amount of data. By
revising earlier relevant studies we can observe that a proxy for the capital stock is
commonly used (see for example Garcia and Thomas, 2001). Antonioli and Filippini
(2001) used the number of water wells as a simplified representation of the capital
stock. However, the non-identifiable inputs are not completely ignored. The inputs are
enclosed in a joint constant (see Parameter estimates). Furthermore, these mislaid
inputs can be separated from the constant and can be incorporated as operational
variables when more information has been collected in the future. The cost structure
can thus be interpreted accurately although the measures are valuable only in the short-
run.

4.1 Capture a representation of the total costs: procedure

and limitations
In total cost modelling, the capital price has to enter the cost function together with the
price of labour, material and electricity. The capital price for the water and wastewater
utility is usually measured as the cost of capital, comprising the interest rate on long-
term debts and the depreciation rate (Kim, 1995). However, there are often variations in
construction costs between the utilities together with different depreciation principles
between communities.

According to Tagesson (2001), we can see a clear difference between the Swedish
communities and their capital costs. Reasons for low capital costs may be:

 Buy from other community

 Large-scale production
 Topography
 Technology

Communities sometimes use different principles of calculating capital costs. This will
complicate the modelling when the capital costs have not been estimated under the
same principle for all the communities. Then, the capital price is not only a function of
technology and operating environment. It is also a function of the depreciation
principle, which is difficult to represent in the model. Tagesson (2001) states that the
different principles may lead to different prices of the water service among different
communities, during different time periods and generations (due to high life expectancy
of water facilities).

[43]
 - Depreciation principle: is the depreciation based on the historical
acquisition value or a present acquisition value?
- Depreciation time: what life-time do the assets have?
- Depreciation time distribution pattern: Linearly, progressive or
digressive?

Figure 13: Depreciation principle. Source: Author.

4.2 Modelling the structure of variable costs, an

application to a sample of Swedish communities
In this section, I have modelled the structure of the total variable costs of production,
distribution, collection and treatment for a few water and sewage organisations in
Sweden by using an econometric methodology and terminology. I have approximated
the variable cost function by using the Translog form. The availability of data has
influenced the study to simply look at the “simultaneously horizontal” process of
providing water and cleaning wastewater. This means that I have created a model for
the joint service system of water and wastewater.

It was not possible to distinguish between the two types of services and to model them
separately. If we want to study sub-activities individually, we need to have separate
accounting information i.e. separate economical, administrational, technical and
environmental figures for water supply, wastewater treatment, distribution and
sewerage.

4.2.1 The structural representation of the model

A fundamental but core part of the analysing procedure is to understand the underlying
structure of the service system. As been described earlier, there are many different
dimensions of the system that can be analysed and the researcher holds the
responsibility for trying to specify what he is studying. An over-viewing description of
the multivariate structure of the industry is informative in the big context, but equally
important in order to fully understand the system is to describe what dimensions of the
system the present study is actually focusing on. This may at least avoid conceptual
confusions due to the fact that researchers tend to focus on different parts of the
system. A good representation may also limit the risk of forgetting or overlooking some
aspects of the system. It seems to be a tendency among the econometric papers not to

[44]
clearly present the research perspectives and this may underestimate the complexity of
the system.

The following figure (14) is the structural conceptual representation of the water and
wastewater system underlying the mathematical model. The model is representative for
one community, producing water to consumers and collecting the wastewater from
consumers. I have included the amount of connections as an operating variable. I
wished to include more operational variables e.g. the capacity of the system (m3/hour)
and the distance to the users (km), but I did not possess such information. The distance
to the users demonstrates the size of the network.

Figure 14: Structural representation of a water supply and sewage treatment system attributed to one
community

4.2.2 Data description

The data used in this study comes from SWWA and contains operational data for about
200 communities, corresponding to the individual water and sewerage organisation. The
data material is extensive but there are many “gaps” (missing values) and unreasonable
values in the sample. For that reason, I had to go over the data and check for possible
errors. SWWA is gradually trying to increase the quality level of the information. It is an
ongoing but long process to establish the underlying statistical framework and to assure
that all existing organisations give their adequate contribution. Quality level 1, out of 3,

[45]
 has been reached. The results in this study should therefore be used for perspicuous
purposes only.

Table (2) presents the overall descriptive statistics for the water and sewage
organisations used in this study. Out of 200 communities, only a sample of 25
communities were selected and used in the study. The sorting procedure was made with
care and the procedure is described with further details in the following section.

The systems range from very small systems producing about 700 cubic metre of water
per day and treating about 1100 cubic metre of wastewater per day to bigger systems
producing about 50 000 cubic metre water per day and treating about 52 000 cubic
metre of wastewater per day. The specified numbers in Table (2) are annual values for
the year 2005, which was the only accessible year with relevant data. The large variation
between minimum and maximum values simply reflects the different sizes between the
systems, but it also shows that there are organizational differences. This can be
observed by looking at the different prices of labour & material and electricity. Price nr 1
is a “combined” unit price for labour and material and can be seen as a function of
produced water and treated wastewater in cubic metres. A more expected unit price for
labour is the price per total hours worked, usually obtained by dividing total wage
expenses by total hours worked. A unit price for material is recognized as difficult to
establish, due to lack of homogeneity in this input. A common way to handle this is to
construct a price index for input materials (see for example Garcia and Thomas, 2001). I
have followed this approach since I cannot separate the labour price from the material
price in a satisfactory way. Both labour and materials are thus influenced by the
production and sewage variables. The labour & material price is therefore said to be
endogenous (explained within the model).

Table 2: Sample statistics, 25 observations

Annual values, year 2005

Variable Notation Unit Mean Std. Deviation Minimum Maximum

Total Variable Cost TVC SEK 28,446,856.72 23,045,851.80 3,946,000.00 109,213,000.00
Production y1 m3 3,257,550.32 4,240,649.17 244,233.00 18,351,000.00
Sewage y2 m3 4,428,141.92 4,642,821.84 396,980.00 19,150,000.00
Price1: Labour & Material w1 SEK/m3 3.99 1.37 2.01 7.78
Price2: Electricity w2 SEK/kWh 0.75 0.11 0.44 1.00
Total network length Length Km 567.57 355.78 96.00 1482.00
Connections Connections --- 68,933.24 8,2476.08 5,145.00 340,200.00
Cost Share1: Labour & Material S1 --- 0.87 0.03 0.80 0.96
Cost Share2: Electricity S2 --- 0.13 0.03 0.04 0.20

[46]
Communities used in the study are: Bollnäs, Borlänge, Essunga, Falkenberg, Gislaved,
Hagfors, Hallstahammar, Halmstad, Herrljunga, Hjo, Härnösand, Kristinehamn, Lycksele,
Mark, Mora, Norrköping, Nybro, Olofström, Ronneby, Tierp, Tingsryd, Ulricehamn,
Uppsala, Vetlanda and Ängelholm.

Cost Data: The total (annual) variable cost is the total sum of annual electricity and
other annual operational costs for plants, distribution networks, pressure stations,
reservoirs, storm water networks and pumps. Other annual operational costs are
defined by the SWWA as the costs of labour and material (plus cost of external services)

Cost Share Data: The cost shares are calculated as the cost of each individual share-
input divided by the total variable cost.

Cost of labour & material

Cost Share1  Equation (10)
Total variable cost

Cost of electricit y
Cost Share 2  Equation (11)
Total variable cost

Price Data: The price of labour & material is obtained by dividing total annual
operational costs with the volume of water produced and wastewater treated during
the year. The price of energy is determined by dividing total annual electricity costs by
the total use of kWh.

Output Data: The output data corresponds to the annual volume of water produced and
wastewater treated. Domestic, industrial and public users constitute the main users of
water and sewage.

Capital stock Data: The capital stock data is represented by the length of the
distribution network and length of the sewerage in kilometres. This is a rather limited
representation of the capital stock.

Technical/Operational data: The technical data corresponds to the amount of total

connections to the water distribution network and the sewerage.

[47]
4.2.3 Data sorting and selection
The communities, and so the utilities, was sorted and selected with respect to the
following criteria:

 Included: utilities having complete cost information (O&M, electricity, labour &
material).
 Included: utilities having complete information about the selected operational
variables (connections, pipe length).
 Included: utilities having complete information about the volume of produced
water and treated wastewater.
 Included: utilities having similar urban customer structure (50-80 % domestic,
10-30 % industrial and 1-10 % public)

- Excluded: utilities performing services across communities and utilities letting

other communities operate all/parts of the services.
- Excluded: utilities having irrational operational estimates, e.g. 0 percent out (in)
leakage.
- Excluded: utilities having unreasonable input prices, i.e. extremely high or low
prices.

As a first requirement, the utilities had to possess complete cost information, taking the
cost categories into consideration. Second, the operational variables were selected
based on the total distribution of operational variables. Two of the most comprehensive
operational variable samples were the amount of connections and the length of pipe
lines. I settled on the opinion to not include more than these two operational variables
in the model, due to defective data. More variables would result in templates rather
than true estimates. With respect to the limited amount of information, it seemed most
reasonable to keep the model simple but as realistic as possible. The selection of
operational variables involves a judgement about how important the variables are for
the model’s accuracy and what we want the model to tell us.

In order to control for different urban environments, the municipalities were selected
based on the similarity of structure. The 25 selected utilities operate in similar
community structures, bearing upon the distribution of households, industry and public
services.

Utilities that operate across communities were excluded. The collected statistical
information did not clarify how many communities were served by the individual
districts. It was not possible to set apart if the utilities in the sample were serving 2, 3, 4
or 5 (etc.) communities. It was only possible to spot if the utilities were serving 1

[48]
community. For this reason, only utilities serving 1 community was grouped and
selected. Other ways of grouping and selecting municipalities could be based on:

- Community size (amount of connections, size of the production, area size etc.)
- Geographical conditions
- Operational logistics (distance to pole municipality, quality of infrastructure etc.)

I am interested to capture how the cost structure gets affected by the size of the
system. No size grouping was performed. We will eliminate the size effect if all of the
observations in the selected sample have the same or very similar size.

The amount of observations does not allow us to include too many variables, because
then the system will be undetermined. We need to have at least one condition
attributed to every parameter.

[49]
4.2.4 Estimation procedure with a multivariate regression
approach3
The following expression follows a Translog approximation form. It represents a cost
structure for the utilities chosen in this study (thus a cost structure for a part of the
Swedish water and wastewater industry):

ln TVC  Constant  a1 ln w1  a 2 ln w 2  b1 ln y1  b2 ln y 2 
 c ln Length  d ln Connections  0.5a11 ln w1 ln w1  a12 ln w1 ln w 2
 0.5a 22 ln w 2 ln w 2  0.5b11 ln y1 ln y1  b12 ln y1 ln y 2  0.5b22 ln y 2 ln y 2 
 0.5cc ln Length ln Length  cd ln Length ln Connections 
 0.5dd ln Connection ln Connections  e11 ln w1 ln y1  e12 ln w1 ln y 2 
 e 21 ln w 2 ln y1  e 22 ln w 2 ln y 2  f1c ln w1 ln Length  f 2 c ln w 2 ln Length 
 g1d ln w1 ln Connections  g 2d ln w 2 ln Connections 
 h1c ln y1 ln Length  h1d ln y1 ln Connections 
 h2c ln y 2 ln Length  h2d ln y 2 ln Connections
Equation (12)

where w1 and w2 are the prices for labour & material and electricity, respectively. Y1 and
y2 are the produced water and treated wastewater, respectively. The Length and
Connections variables are the included operational variables in the model. The Constant
is one of the parameter estimates and it encloses other omitted variables, which for all
the reasons could not be caught in separate variables.

The explicit parameter estimates are a1, a2, b1, b2, c, d, a11, a12, a22, b11, b12, b22, cc, cd,
dd, e11, e12, e21, e22, f1c, f2c, g1d, g2d, h1c, h1d, h2c, h2d.

The above expression, together with the so-called Cost Share equations S1 and S2 forms
a multivariate regression system. The Cost Shares are the first derivative of ln(TVC) with
respect to the individual prices w1 and w2. Thus, we get two Cost Share equations and
the system of equations to be estimated is:

ln(TVC i )  lnTVC ( yi , wi , CS i , Z i )  εTCVi , where i = 1, ..., 25.

S j,i  S j,i ( yi , wi , CS i , Z i )  ε Sharej,i
, where j = 1, 2 and i =1, ..., 25.

3
The computer program used for this section was MATLAB 7.5.

[50]
εTVC i and εSharej,i are, respectively, the residuals associated with the cost equation i and
the cost share equation j for utility i.

The system of equations is estimated by Ordinary Least Squares (OLS). The numerical
values of the parameter estimates are presented in the next section. In order to
estimate the regression system above, all the variables are normalized by their sample
means. The Translog cost function needs a reference point (Garcia and Thomas, 2001).
This means that the first-order parameters are most accurately representative for the
sample mean of the total selection of 25 utilities. For example, that is to say that if the
price for electricity (w2) increases by 1 percent, the total variable cost will increase by a1
percent for the “average” utility (i.e. an utility with mean[TVC], mean[y1], mean[y2],
mean[w1], mean[w2] etc). In addition, it can be confirmed that the mean of the
estimated ln(TVC) given by equation (12) equals the mean of actual ln(TVC). Figure (15)
illustrates this statement.

There are certain restrictions on the model: Homogeneous of degree one in factor
prices, concave in factor prices and symmetry in certain parameters (a12 = a21, b12=b21).
The Cost Shares are not allowed to be negative. The homogeneity in input prices is
fulfilled when the following parameter restrictions hold:

a1  a 2  1
a11  a 21  a12  a 22  0
e11  e 21  e12  e 22  0
f 1c  f 2c  g1d  g 2 d  0

[51]
4.2.5 Parameter estimates

Table 3: Estimated parameters

No. Parameter Corresponding variable Estimates

OLS (White) Std. Error
1 Constant Constant 17.3584 0.0132
2 a1 ln w1 0.8848 0.0048
3 a2 ln w2 0.1156 0.0049
4 b1 ln y1 0.5116 0.0976
5 b2 ln y2 0.4240 0.1013
6 c ln Length (-0.0099) 0.0504
7 d ln Connections 0.0086 0.0691
8 a11 ln w1*ln w1 0.1087 0.0130
9 2*a12=2*a21 ln w1*ln w2 (-0.1016) 0.0121
10 a22 ln w2*ln w2 0.0900 0.0259
11 b11 ln y1*ln y1 0.0727 0.8252
12 2*b12=2*b21 ln y1*ln y2 (-0.0744) 0.4158
13 b22 ln y2*ln y2 0.6191 0.3223
14 cc ln Length*ln Length 0.0789 0.1904
15 cd ln Length*ln Connections 0.1360 0.2494
16 dd ln Connections*ln Connections (-0.1984) 0.0840
17 e11 ln w1*ln y1 0.0423 0.0177
18 e12 ln w1*ln y2 (-0.0010) 0.0165
19 e21 ln w2*ln y1 (-0.0392) 0.0182
20 e22 ln w2*lny2 0.0008 0.0175
21 f1c ln w1*ln Length (-0.0297) 0.0136
22 f2c ln w2*ln Length 0.0276 0.0138
23 g1d ln w1*ln Connectons (-0.0026) 0.0080
24 g2d ln w2*ln Connections 0.0019 0.0082
25 h1c ln y1*ln Length (-0.5378) 0.3635
26 h1d ln y1*ln Connections 0.4901 0.4864
27 h2c ln y2*ln Length 0.3225 0.3565
28 h2d ln y2*lnConnections (-0.6311) 0.5951

29 R-squared Cost equation 0.998

30 R-squared Cost Share1 (Labour & Material) 0.689
31 R-squared Cost Share2 (Electricity) 0.690

32 N Observations 25

[52]
 19

18.5

18

17.5

Figure 15: Estimated ln(TVC) and actual ln(TVC)

17

[53]
16.5

Real ln(Total Variable Cost) Normalized

16

15.5

15
15 15.5 16 16.5 17 17.5 18 18.5 19
Estimated ln(Total Variable Cost) Normalized
4.2.6 Analysis: Understanding the parameter figures. Estimating
Economies of production output density, Economies of
customer density and Economies of scope
In this section, I implement the theoretical concepts of structural characterization,
system analysis, critical factors and efficiency for the Swedish sample of utilities.
Subsequently, I present and put into practice the mathematical measurements of
Economies of production output density, customer density and scope. The meaning of
Economies of scale is presented on a theoretical stage. The economies are first
evaluated at the sample mean of variables, i.e. the final estimates capture the state of
an “average” utility. Second, the economies are evaluated for “small” utilities [Total
production: 6.4∙105-5.0∙106 m3] and “large” utilities [Total production: 5.1∙106-3.8∙107
m3] in the sample. This section is consequently the final session of the empirical study.

Measure Decreasing return Constant return to Increasing return to

to size size size

Economies of <1 =1 >1

production output
density*

Economies of <1 =1 >1

customer density*

Economies of scale * <1 =1 >1

Measure Joint production Separate

production

Economies of <0 >0

scope**

Table 4: Definition of efficiency measures. Sources: * Garcia and Thomas (2001), ** Hajargasht, Coelli
and Rao (2006).

[54]
Horizontal-product analysis of Economies of production output density and Economies
of customer density:

Economies of production output density in the short-run (for the joint production of
water and wastewater) =
1
  ln TVC  ln TVC 
   
  ln y1  ln y 2 

 (b1  b11 ln y 1  b12 ln y *2  e11 ln w *1  e 21 ln w *2  h1c ln Length *  h1d ln Connections * 

*

 b2  b22 ln y *2  b12 ln y *1  e12 ln w *1  e 22 ln w *2  h´2 c ln Length *  h2 d ln Connections * ) 1 

 (0.5116  0.0727 ln y *1  0.5  0.0744 ln y *2  0.0423 ln w *1  0.0392 ln w *2 
 0.5378 ln Length *  0.4901ln Connections * 
 0.4240  0.6191 ln y *2  0.5  0.0744 ln y *1  0.0010 ln w *1  0.0008 ln w *2 
 0.3225 ln Length *  0.6311ln Connections * ) 1
 (0.9616  0.0355 ln y *1  0.5819 ln y *2  0.0413 ln w *1  0.0384 ln w *2
 0.2153 ln Length *  0.1410 ln Connections * ) 1 
 {for the " average" utility}  0.5116  0.4240   1.06
1

Equation (13)

Notation in the formula (*): Normalized variables.

The expression shows us that the Economies of production output density for the
Swedish utilities “overall” are dependent mostly on the treatment of wastewater, the
capital stock (represented by Length) and the amount of connections. Above all,
wastewater is a critical factor among the utilities, especially for utilities with large
production of treated wastewater. Wastewater has a large effect on the total variable
cost of production. This may be a cause of having expensive treatment procedures
or/and too much in-leakage (over flow). In-leakage is water that searches its way into
the sewage network although it does not belong in the system, for example rainwater,
groundwater or surface water that forces themselves into wells and pipes. The amount
of in-leakage water in the system can vary between 0 and several hundred percent of
the normal water flow (Weglert T, 2005)

On the other hand, the effect of wastewater for the average firm is not that critical as it
is for larger production. A force of direction, if the volume of water supplied to
customers increases by 1 percent (all other things unchanged) then the total variable
cost increases by approximately 0.51 percent for the average firm. If the volume of
wastewater increases by 1 percent (all other things unchanged) then the total variable
cost increases by approximately 0.42 percent. Conclusively, water and wastewater are
about equally affecting the total variable cost for the average utility. For the industry “as

[55]
a whole” we can conclude that especially wastewater has a large impact on the variable
cost.

The price of electricity has a higher effect (it is a sensitivity factor) on the total variable
cost when the amount of wastewater increases than when the amount of water
increases. On the other hand, the price of labour & material has a higher effect on the
total variable cost when the amount of water increases than when the amount of
wastewater increases.

The estimated Economies of production output density are 1.06 for the average utility.
This leads us to conclude that the average utility exhibits Economies of production
output density (returns to size are increasing, because Economies of production output
density > 1). It would be profitable to produce more units of goods with the existing
network and the amount of existing customers (i.e. to have an increasing demand from
the existing users to produce water and to clean wastewater). On the contrary, it would
not be profitable for the large utilities to produce more units of goods with their existing
network and their amount of existing customers.

Measure “Small” utilities in The “average” “Large” utilities in

the sample utility the sample

Economies of = 1.34 = 1.06 = 0.86

production output
Marked increasing Increasing return Decreasing return
density
return to size to size (roughly to size
constant return to
size)

Table 5: Estimated Economies of production output density for different utility sizes. Source: Author.

Economies of customer density in the short run (for the joint production of water and
wastewater) =
1
  ln TVC  ln TVC  ln TVC 
 0.5116  0.4240  0.0086  1.03
1
   
  ln y1  ln y 2 Connections 

Equation (14)

A 1 percent increase of the amount of connections (all other things unchanged) results
in approximately a 0.0086 percent increase of the total variable cost for the average
utility. A proportionally increase of 1 percent respectively of the volume of water, the

[56]
volume of wastewater and the amount of connections brings about an 0.97 percent
increase of the total variable cost for the average utility.

Measure “Small” utilities in The “average” “Large” utilities in

the sample utility the sample

Economies of = 1.04 = 1.03 = 0.87

customer density
Increasing return to Increasing return Decreasing return
size (roughly to size (roughly to size
constant return to constant return to
size) size)

Table 6: Estimated Economies of customer density for different utility sizes. Source: Author

The estimated economies of customer density are 1.03 for the average utility. This leads
us to conclude that the average utility exhibits economies of customer density (returns
to size are increasing, because economies of customer density > 1). It would be
profitable for the average utility and for the small utilities to serve more customers (thus
to also produce more units of goods) in their existing network. On the contrary, it would
not be profitable for the large utilities to serve more customers with their existing
network. The larger utilities does not benefit from customer growth when having their
existing network.

Horizontal-product analysis of Economies of Scale:

A theoretical measure of Economies of scale in the short run (Equation 15) and in the
long run (Equation 16):
1
  ln TVC  ln TVC  ln TVC  ln TVC 
    
  ln y  ln y  ln Connections  ln Network size 
1 2
Equation (15)

 ln TVC
1
 ln Capital stock
 ln TVC  ln TVC  ln TVC  ln TVC
  
 ln y1  ln y 2  ln Connections  ln Network size Equation (16)

Network size stated in the previous formulas is a geographical representation/measure

of the service region. The size of a network can be measured in e.g.

[57]
 a) Km (the length of the total pipe network)
b) Km2 (the area of the service region)
c) Communities (the amount of communities that form a service region)

How we choose to represent the network size in the formulas depends on how the
business is enlarged. Here is a short example. By the definition of Economies of scale the
density of customers has to remain unchanged once the production has varied in size. If
the network size is represented only by (a) then the proportionally varying variables are
the supplied water, the treated wastewater, the amount of customers and the length of
the total pipe network. Thus, the resulting measure of customer density
(Customers/Km) remains unchanged. However, we do not control for any other
variations by this measurement. Since we have not included a representation for the
area, the measure of customers per areal (Customers/Km2) may have increased,
remained the same or decreased. In the same manner, since we have not included a
representation for the amount of communities, the measurement of customers per
community (Customers/community) may have increased, remained the same or
decreased. In order to control for an area variation (e.g. an enlargement), we include
one more expression in our equation. In order to control for community consolidations,
we include yet one more expression in our equation. All of the possible size variations
should preferably be included in the mathematical formula in order for the desired
customer density to remain unchanged once the production has varied in size (e.g.
Customers/[Km & Km2 & community]). Here is a general formula for the short-run case
with more than one network size representation:
1
  ln TVC  ln TVC  ln TVC  ln TVC  ln TVC  ln TVC 
    () () ()...
  ln y1  ln y 2  ln Connections  ln (a )  ln (b)  ln (c ) 

Equation (17)

I did not possess any information about the area (e.g. in km2) of the service regions or
the amount of served communities, which meant that I could not tag on the proper
measurement of Economies of scale. The furthest to reach at this point are the
measurements of Economies of production output density and Economies of customer
density.

 2 log( TVC )  0.0744

Economies of scope (water and wastewater) =  b12  0
 log Y1 log Y2 2

Equation (18)

There are Economies of scope between water and wastewater in the Swedish industry,
i.e. it appears to be cost advantages with the joint production of water and wastewater.

[58]
An autonomous example; Horizontal-customer analysis of Economies of production
output density and Economies of scope:

This is simply a short example of a horizontal-customer analysis, in comparison to the

horizontal-production analysis. The parameter estimates are based on estimation
procedures like the previous ones. The estimates in this subdivision should be taken
more slightly, because they rely on an approximation procedure that is not that
rigorous. This is mainly due to missing values. This small section serves as an illustration
to how the industry analysis can be approach from another point of view.

Economies of production output density in the short run (for the joint production of
water and wastewater to domestic users and industry users) =
1
  ln TVC  ln TVC 
  0.5428  0.4225  1.04
1

  ln y 
 ln yindustry 
 domestic
Equation (19)

For the simplicity to calculate, I have excluded public users. It is a small part of the total
production. It seems logical that the measurement of Economies of production output
density with the customer perspective gives almost the same result as the analysis with
the product perspective. They represent the same thing, using two different
perspectives. The result of 1.04 is a little less than the previous 1.06, which may be the
result of not including the public users. We can conclude that domestic users have a
large effect on the total variable cost of production. A 1 percent increase of the total
water and wastewater demand from domestic users (all other things unchanged) brings
about a 0.54 percent increase of the total variable cost. A 1 percent increase of the total
water and wastewater demand from industry users (all other thing unchanged) brings
about a 0.42 percent increase of the total variable cost.

 2 log( TVC ) 0.6725

Economies of scope (domestic and industry) =  0
 log YDometic log YIndustry 2

Equation (20)

The estimate of Economies of scope is greater than 0 which indicates that it is not
profitable for the average utility to serve both domestic and industry users. It seems to
be more profitable to specialize in different production units. This contradicts a study
made by Kim (1995) that found cost advantages of operating with a combination of
residential and residential outputs for a cross-section of 60 utilities for the year 1973 in
the United States. If we exclude the possibility of estimation errors, the individual
industry analysis can be seen as case-specific, depending on the underlying production

[59]
technology and environment. The analysis can be spit up into different product,
customer or function groups.

[60]
Chapter 5
Summary and conclusions

In this thesis we have seen how an economic/mathematical model for the water and
wastewater industry can be constructed by using a multivariate regression approach.
The accurate mathematical formula (including explanatory variables) we choose to
apply on the statistical data is a choice based on the desired venture. A second
(alternative) approach is the stochastic frontier analysis. While the regression method
generates an expression of the average cost, the stochastic frontier method determines
how close each utility is to the expenditure achieved by the best of the industry. The
second approach is often used for the purpose of benchmarking (OFWAT, 2005).

Each estimated parameter in the mathematical formula describes the relationship

between the total (variable) cost and the corresponding system input. The parameters
represent a cost structure for a few Swedish utilities in the industry. The results from
the empirical study could be of interest for authorities in the Swedish water and
wastewater industry (especially).

In order to properly build a mathematical model of a single system (or total industry) we
need to form an understanding about the underlying production technology. A critical
step is to clarify which structural-conceptual representation is underlying the
mathematical model. This step can be approach from a dimensional perspective.

The Translog cost function is better interpreted as a “cost structure function” for the
existing industry rather than a “cost estimation function” for new constructions. The
Translog cost function has been normalized in order to be easily interpreted for an
average utility. The efficiency measures have been derived from the variable cost
formula and have been calculated for different sizes of the utilities (“small”, average and
“large”). We can also see that the coefficient of the capital stock has a negative sign,
which goes along the lines of cost theory (Filippini, 2007).

It is important to be critical to one’s own choice of method. Thus, the estimations need
to be refined with more sophisticated econometrical methods because the parameter
restrictions stated for w1 and w1 is not perfectly fulfilled (the parameters a1+a2 does not
summarize perfectly to 1). Moreover, the data used for this study is cross-sectional data
(i.e. data-sets that do not reflect difference in time). More interesting would be the use
of panel-data (i.e. data-sets that reflects changes in time).

[61]
Can the estimates really be verified in reality? The estimated cost structure is an
approximation of the actual state of the industry. The efficiency measures can be
estimated for every utility in the study, although the measures are most accurately
interpreted for the average utility. The average approach is a common way of
representing the conditions of an industry. However, the average utility is a fictive utility
based on actual values of the existing industry and cannot be identified independently.
As an alternative to the average approach, we can use the median approach.

This study points out the need for high quality methods of collecting information and
presenting statistics about the industry. It is vital in order to make any further analysis.

[62]
References

Andersson F, 2007, ”Kostnadsminimering och Shephards Lemma”, Lecture Notes: March

26

Bakka F.J. et al, 2001, “Organisationsteori: Struktur – kultur – processer”, Malmö; Liber
AB

Barucq C. et al, 2006, ”Analysis of Drinking Water and Wastewater Services in Eight
European Capitals: the Sustainable Development Perspective”, Le Vivaldi; BIPE

Bottasso A. and Conti M, 2003, “Cost inefficiency in the English and Welsh Water
Industy: An Heteroskedastic Stochastic Cost Frontier Approach”, Università di Genova;
DIEM.

Camp Dresser & McKee Ltd, 2004, “Economic Analysis of Water Use in Ireland, Final
Report”, Department of the Environment, Heritage and Local Government

Clark R. M. et al, 2002, “Cost Models for Water Supply Distribution Systems”, Journal of
Water Resources Planning and Management, Volume 128, Issue 5, pp. 312-321.

Cubbin J. et al, 2004, “Assessing Ofwat’s Efficiency Economics, A report for Water UK,
Mars 2004”, City University, UK

Donald E. et al, 2003, “Managing urban water supply”, Springer; ISBN: 1402017200

Eriksson L. T. and Wiedersheim-Paul F, 2006, “Att utreda forska och rapportera”,

Malmö; Liber AB

E.Y.D.A.P (Athens water and sewerage company), 2006, Financial statements, Galatsi

FAO Corporate Document Repository, 1996, “Water sector policy review and strategy
formulation. A general framework”, Natural Resources Management and Environment
Department

Féres J. and Reynaud A, 2006, “Does formal or informal environmental regulation have
an impact on costs? An empirical analysis of the Brazilian manufacturing industry”, São
Paulo, Brazil

Filippini M. et al, 2007, “Cost efficiency and economies of scale of Slovenian water
distribution utilities: an application of panel data stochastic frontier methods”,
University of Lugano, Switzerland.

[63]
Filippini M. et al, 2005, “Economies of Scale in the Swiss Hydropower sector”, CEPE
Working Paper No. 44, Zurich

Garcia S. and Thomas A, 2001, “The structure of Municipal Water Supply Costs:
Application to a Panel of French Local Communities”, Université des Sciences Sociales,
Toulouse. Journal of Productivity Analysis, 16, 5-29; Kluwer Academic Publishers, Boston

Gordon D, 2005, “Incorporating environmental costs into an economic analysis of water

supply planning: a case study of Israel”, Simon Frazer University

Hajargasht G, 2006, “A Dual Measure of Economies of Scope”, School of Economics,

University of Queensland, Australia.

Hanemann W. M, 1998 “Determinants of urban water use”. Pp. 31-75 Chapter 2 in

Urban Water Demand Management and Planning by Baumann D, Boland J. and
Hanemann W. M, eds. New York; McGraw-Hill.

Hansen B. E, 2008, “Econometrics”, University of Wisconsin.

Hopkins, Levis D, 2003, “Economies of Scale in Wastewater Treatment and Planning for
urban growth”, Department of Urban and regional planning

Johansson-Lindfors M. B, 1993, “Att utveckla kunskap”, Lund; Studentlitteratur

Kim Y, 1995, ”Marginal Cost and Second-Best Pricing for Water Services”, Department of
Economics, Western Kentucky University, U.S.A.

Kärrman E. et al, 2006, ”Kostnader för VA i omvandlingsområden”, APM 6:2006, ISSN

1402-134X/RTN 200207-0253

Lee J, 1998, ”Guidelines for the economic analysis of water supply project”, Economics
and Development Resource Center, Asian Development Bank, Philippines

Martins R. et al, 2006, “Cost Structure of the Portuguese Water Industry: a cubic cost
function application”, Faculdade de Economia da Universidade de Coimbra, Portugal

Nanchang Municipal Government, 2000, “Urban Water Supply in the 21st Century”, City
Paper

OFWAT, report 2005-06, “Water and sewerage service unit costs and relative
efficiency”; Econometric models-opex_2005-06_Annex1.pdf

Ramirez J. I, 2001, “Kostnadsmodell Urban Water”, Forskningsprogrammet Urban

Water, Chalmers Tekniska Högskola, Sweden

Stahre P. et al, report 2007-13, ”Värdering av vatten- och avloppsledningsnät”, Svenskt

Vatten Utveckling; SWWA.

[64]
Stone and Webster Consultants Ltd, 2004, “Investigation into evidence for economies of
scale in the water and sewerage industry in England and Wales”, Final report, London

Taylor A (Stone and Webster Consultants),”Economic Analysis of the EU Water

Framework Directive”, andrew.taylor@shawgrp.com

Tagesson T, 2001, “Redovisning och beräkning av kapitalkostnader i kommunala va-

verksamheter, Department of Business Studies, Kristianstad University College.

Thomasson A, 2003, ”Uthålliga VA-system – internationella erfarenheter av

organisationsformer och drivkrafter i en VA-sektor i utveckling”, Rapport 135
Forskningsprogrammet ”Sustainable Urban Water Management”, Chalmers Tekniska
Högskola, Sweden.

Urakami T, 2005, “Identifying scale economies for different types of water supply and
organizations in Japan”, School of Business Administration, Kinki University, Japan

Weglert T, 2005, “Vatten och avlopp för gruppbebyggelse och enskilt boende”, ICA
bokförlag

Internet sources:

The Federal Reserve Bank of Minneapolis:

http://www.minneapolis.org/research/prescott/macro_teory/Lectures/CESProdFn.pdf
Visited: 2008-03-02

The European Public Health Alliance, “How will water shape the 21 st century?”:
http://www.epha.org/a/694. Visited: 2008-05-26

eNewsUSA, ”Water supply challenges for the 21st century”:

http://enewsusa.blogspot.com/search/label/water. Visited: 2008-05-14

Palo Alto Research Centre: http://www.parc.com. Visited: 2008-04-16

Files:

SWWA: VASS Drift 2005, Nivå 1, Svarsdata och nyckeltal.xls

Ramirez et al. (Urban Water): BilagaB_Kalkylmodell.xls

[65]