Author’s Accepted Manuscript

Estimating marginal abatement costs for industrial

water pollution in Colombo

Asha Gunawardena, Atakelty Hailu, Ben White,

Ram Pandit

www.elsevier.com/locate/envdev

PII: S2211-4645(16)30089-6
DOI: http://dx.doi.org/10.1016/j.envdev.2016.11.001
Reference: ENVDEV313
To appear in: Environmental Development
Received date: 20 April 2016
Revised date: 3 November 2016
Accepted date: 3 November 2016
Cite this article as: Asha Gunawardena, Atakelty Hailu, Ben White and Ram
Pandit, Estimating marginal abatement costs for industrial water pollution in
C o l o m b o , Environmental Development,
http://dx.doi.org/10.1016/j.envdev.2016.11.001
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
Estimating marginal abatement costs for industrial water pollution in

Colombo

Asha Gunawardena1,2*, Atakelty Hailu1, Ben White1, Ram Pandit1

1
School of Agricultural and Resource Economics, University of Western Australia
2
Institute of Policy Studies, Sri Lanka

[email protected]

[email protected]

*
Corresponding author. Asha Gunawardena

Abstract

This paper estimates the cost of effluent discharge regulations for firms located in the lower

Kelani River catchment in Sri Lanka. The river provides water for many economic purposes

including drinking water to the capital city, Colombo and a variety of ecosystem services.

Employing multi-input and multi-output translog production technology, we estimate

shadow prices of effluents and technical efficiency of firms belonging to eight industries.

With the given level of compliance, we extend our analysis to compute total abatement cost

for firms under different policy scenarios, related to simultaneous reduction in

concentration of three water pollutants including current regulatory standards. The evidence

of poor compliance and wide variations in firm and industry specific shadow prices (marginal

abatement costs) provide a strong case for a comprehensive redesign of environmental

policy to control water pollution by industries in the heavily industrialized river catchment.
1. Introduction

Degradation of water quality of major urban rivers due to point and non-point sources has

become a serious challenge in many developing countries (Schaffner et al., 2009). This is

mainly due to rapid urbanization accompanied by population and industrial expansion in

these countries where development of water purification infrastructure facilities and policy

has lagged behind industrialization (Biswas and Tortajada, 2009; Qin et al., 2011). While the

cost of environmental asset degradation is significant to society in these countries, the full

cost of environmental externalities has not been accounted for because the value of

ecosystem services is not readily measured or accounted for in national or regional accounts.

This study focuses on water polluting industries in the lower Kelani River catchment. The

river is the main source of drinking water for the city of Colombo. Apart from provision of

drinking water, the river is a source of hydropower generation, industrial and irrigation

water, and is used for washing, bathing, fishing and recreation. Due to population growth

and industrialization, water quality in the lower river has deteriorated rapidly over the last

20 years. Therefore, managing river water quality has become a critical issue due to the cost

of maintaining drinking water standards and the costs of deteriorating ecosystem services.

Recent studies report that the river is of poor quality endangering the aquatic life and

degrading the ecology of the estuary ecosystem (Herath and Amaresekera, 2007). The

urgent need to improve river water quality has created a strong interest among policy

makers to experiment with different policy instruments in addition to the existing regulatory

standards.

The use of market based instruments to control environmental pollution has been strongly

encouraged by economists despite inconclusive evidence in developing and emerging

2
countries where institutional, financial, political and human resource limitations restrict the

development of such instruments (Blackman, 2009; Blackman and Harrington, 2000;

Kathuria, 2006; Rahman and Ancev, 2014). These instruments create economic incentives for

industries to reduce pollution by imposing implicit or explicit price on emissions. Well-

designed and implemented market based instruments have the ability to provide the overall

least cost means of achieving desirable levels of emission reductions (Stavins, 2003) by

equalizing the marginal or incremental abatement cost across polluters (Baumol and Oates,

1988; Tietenberg, 2006). In many cases, the marginal cost of abatement among firms varies

due to size, industrial category, location, price and quality of inputs and differences in

abatement technology. The potential cost savings achieved by implementing a market based

instruments tends to be high where marginal abatement costs are heterogeneous across

firms (Newell and Stavins, 2003).

However, information on abatement costs is not readily available to policy makers due to

the absence of markets and observable prices for pollutants. There are two main economic

approaches that can be used to estimate the cost of abatement of undesirable outputs: cost

function approach and distance function approach. The estimation of firm level efficiency

and shadow prices using distance functions has been the more widely employed approach

(Hailu and Veeman, 2000; Lee et al., 2002; Lee, 2005). Distance functions can be used to

characterize environmental production technology in a multi-input and multi-output setting

including undesirable outputs where information on regulations and input prices is not

required. Duality theory (Shephard, 1953, 1970), which provides a fundamental link

between traditional measures of productivity and distance function productivity measures,

can be used to calculate shadow prices. For example, the cost function is dual to the input

distance function and the revenue function is dual to the output distance function. The

3
calculated shadow prices with respect to a particular pollutant can be defined as the

marginal cost of abatement (MAC) for this pollutant as it represents the additional cost

incurred to reduce the pollution by one unit. The MAC provides useful information at a firm

level, linking current emission levels to the cost of reducing emissions. At a policy level,

these values can be used as an important tool to determine the economically efficient levels

of pollution reduction to maximize societal welfare (Vijay et al., 2010).

Estimation of shadow prices using distance functions has followed three main approaches:

non-parametric or Data Envelopment Analysis (DEA) approaches; parametric functions

estimated using linear programming (PLP) approaches; and the parametric Stochastic

Frontier Analysis (SFA) approaches (Zhou et al., 2014). The DEA is a frontier analysis

technique (Cooper et al., 2011) which constructs a piece-wise production boundary using the

observed input and output data points (Du et al., 2015). The advantage of DEA is that it is

not necessary to specify a functional form for the underlying production technology (Zhou et

al., 2014). However, it does not guarantee the differentiability of the estimated distance

function which is important in estimating shadow prices. This is because DEA derivatives are

not defined at the vertices of the piece-wise linear frontier.

In contrast, parametric distance function methods have become popular in the literature

due to their differentiability which is an essential feature for estimating shadow prices

(Zhang and Choi, 2014). PLP methods allow simple imposition of important constraints in

the frontier estimation. The only shortcoming of this estimation approach is that it ignores

statistical noise and attributes all deviations from the frontier to inefficiency. On the other

hand, the stochastic frontier approach allows stochastic treatment of these deviations by

decomposing into an inefficiency term and a random disturbance term that accounts for

measurement errors and random noise. In addition, SFA allows for hypothesis testing in

4
relation to parameter values (Kuosmanen and Kortelainen, 2012). However, the SFA

approach limits the researchers’ ability to apply prior monotonicity restrictions on the

parametric distance function (Rezek and Campbell, 2007).

Shephard distance function (and radial) and directional distance function have been used in

the literature. Both can be specified as flexible forms – radial as translog and directional as

generalised quadratic, allowing for the global imposition of linear homogeneity and

translation properties, respectively. Many recent studies have tended to use the directional

distance function that can allow for increases of desirable output(s) while contracting the

undesirable output(s). Nevertheless, the shadow price estimates vary depending on the

directional vector which is used to expand or contract the input and output set. Therefore,

the choice of an appropriate directional vector plays a key role in shadow price estimations

(Zhou et al., 2014). However, there is no consensus in the literature regarding the choice of

directional vector. Further, the use of a direction that implies a radical change in input or

output mix is akin to assumption of a structural change that is more consistent with long-run

rather than short-run possibilities. Therefore, the use of shadow prices based on a radial

distance functions can be more appropriate as these functions maintain quantity mixes at

observed levels (Ma and Hailu, 2016).

A few studies have calculated shadow prices using distance functions in developing country

contexts (Dutta and Narayanan, 2011; Mandal, 2010; Murty et al., 2007; Murty and Kumar,

2003; Murty et al., 2006; Van Ha et al., 2008; Xu et al., 2010) . To our knowledge this is the

first study carried out in Sri Lanka to estimate shadow prices of water pollutants in

manufacturing industries. In this study, we calculate firm and industry specific shadow prices

for Biochemical Oxygen Demand (Simonovic and Fahmy), Chemical Oxygen Demand (COD)

and Total Suspended Solids (TSS) using input distance functions estimated by PLP (Coelli et

5
al., 2013; Hailu and Veeman, 2000). Given the importance of cost information on policy

decisions, we extended our analysis to simulate total abatement cost for different policy

scenarios based on simultaneous reduction of pollutant concentration to different levels

including those consistent with the current regulatory standards. Our choice of input

distance functions in this study is based on the following three reasons: (1) in the presence

of undesirable outputs, an input-based efficiency measure is easy to interpret as it

represents the proportional change in inputs with both desirable and undesirable outputs

held constant (Hailu and Veeman, 2000; Murty et al., 2006); (2) input based efficiency is

more appropriate as most firms have more control over inputs than outputs; (3) shadow

price estimates using input distance function do not require the choice of a direction vector

as is the case of directional distance functions.

The paper is organized as follows. Section 2 presents the current policy context and Section

3 explains the theoretical concepts behind the methodology we used. Section 4 describes

the data and Section 5 reports and discusses the empirical results. Section 6 concludes the

paper.

1 Policy Context

Sri Lanka uses command and control regulatory measures administered by the Central

Environmental Authority (CEA) to control industrial pollution. The key regulatory measures

adopted by CEA are Environmental Protection Licencing (EPL) and concentration standards 1.

The CEA issues Environment Protection Licenses (EPLs) to firms, a mandatory provision to

start a new business in Sri Lanka. These licenses are renewed for existing businesses,

1
The national concentration standards for water pollutants discharge into inland water bodies are given as 30mg/l for BOD,
250 mg/l for COD and 50 mg/l for TSS.
6
annually for high polluting industries2 and triennially for medium to low polluting industries,

after checks and verification on whether the wastewater quality meets the existing effluent

standards. There are no river specific policies for managing water quality, except the ban on

establishing new pollution intensive industries upstream of the water extraction point in the

Kelani River.

Empirical evidence on water quality evidence in Sri Lanka suggests that the current approach

to pollution regulation in rivers is ineffective (AECEN, 2006; Vasantha, 2008). Firstly,

emission standards are based on discharge concentrations which do not restrict total

pollution loads. Secondly, EPLs provide no incentive to reduce pollution by industries with

varying emission levels as all industries under the same pollution category are charged a

uniform fee irrespective of their pollution levels. Thirdly, CEA has limited regulatory and

weak enforcement powers. The number of cases handled by the legal unit of the CEA was

about 252 in 2012 where 64 were new cases related to industrial pollution (CEA, 2012).

Fourth, budgetary constraints have resulted in limited resources within CEA. The lack of a

well-managed information system is also a hindrance to effective monitoring and

enforcement. In addition, the lack of public pressure on polluting industries because of

limited public awareness and ineffective public complaint processes has resulted in poor

compliance.

Recently, the CEA has been exploring new options3 such as market-based instruments for

effective control of industrial wastewater pollution (Vasantha, 2008). In 2007, CEA proposed

a Wastewater Discharge Fee (WDF) program in Sri Lanka. However, implementing such a

program presents a number of challenges due to overlapping legal and institutional

2
Industries have been categorised by the CEA based on their level of pollution under three main categories; A-high polluter,
B-medium level polluters and C-small scale polluters
3
CEA also initiated a program in 2011 to increase the voluntary compliance named as the national Green Award Scheme.
This program recognizes and publicizes private and public sector institutions that operate in environmental friendly manner.
7
functions and lack of procedures to design and collect fees. In addition, a lack of technology

to measure pollution levels at the firm level and the absence of systematic up-to-date

industrial pollution database make implementation of this program difficult. In many cases,

it is difficult for government agencies such as CEA to set appropriate fees on industrial

emissions due to unavailability of empirical information. This study is the first to address and

provide much needed information on abatement costs which policy makers could potentially

use to develop more effective and efficient policies. This is done in the context of

production efficiency frontiers allowing us to also to generate useful information about the

performance of water polluting industries in Colombo, Sri Lanka

2 Methodology

Shephard (1953) was the first to introduce distance functions. The functions can be

employed to describe multi-input and multi-output production technology in order to

estimate technical efficiency and productivity measures without resorting to specific

behavioural assumptions such as profit maximization and cost minimization (Coelli et al.,

2005). There are two types of distance functions: output distance functions and input

distance functions. An output distance function characterizes production technology by

considering maximum proportional expansion of the firm’s output vector for a given set of

inputs. An input distance function represents the production technology by looking at the

maximal proportional contraction of the input vector for a given output vector.

In this paper, we use input distance function which provides meaningful and explicit

measure of production efficiency as it considers proportional savings of inputs (costs) while

keeping both desirable and undesirable outputs constant. In the case of output efficiency

measures, efficiency is defined in terms of proportional expansion of both desirable and

undesirable outputs but the net welfare gain from such an expansion cannot be determined.
8
The net welfare gain or loss depends on the difference between benefits gained from the

expansion of desirable output and the damage caused by simultaneous expansion of

undesirable outputs. Therefore, interpretation of radial output efficiency changes is

ambiguous in the presence of undesirable outputs (Hailu and Veeman, 2000; Murty et al.,

2006).

3.1. The technology set and input distance function

The production technology of each water polluting firm can be described using input

sets, ( ), representing the set of all input vectors that produce output vector

with the output vector consisting of both desirable and undesirable outputs (for

example water pollutants). The input distance function can be defined against the input

requirement set as follows:

( ) { ( ⁄ ) ( )} (1)

That is the input distance function indicates the maximum amount by which an input vector

can be deflated or contracted given the output vector and the production technology.

The input distance function is linearly homogenous, non-decreasing and concave in and

non-increasing and quasi-concave in (Coelli et al., 2005). The value of the distance

function is equal to 1 (if is located on the inner boundary of the frontier) or greater than 1

if is able to produce . In other words, the distance function provides a complete

representation of the production technology.

9
 ( ) ( ) (2)

3.2. Derivation of Shadow prices

There are different ways to abate pollutants such as having an end-pipe treatment plant,

changing the production process to make it more cleaner or changing input mix. However,

all these differences are captured in a vector of inputs and a vector of outputs irrespective of

differences. And the focus of this type of modelling (producer analysis) is at the firm level

and the trade-offs between inputs and outputs implied by the sample data. Therefore, it

does not specify particular abatement methods or rule out any.

We employed input distance function to calculate shadow prices of pollutants following

(Hailu and Veeman, 2000). Shadow prices of pollutants can be derived from the cost

function (3) and (4) using the behavioural assumption of cost minimization and the duality

between cost function and input distance function. A formula for shadow price can be

derived using envelope theorem on the first order conditions of the cost minimization

problem defined against a distance function representation of the technology as shown in

Eq. (5) and (6) below.

( ) { ( )} (3)

( ) { ( ) } (4)

( ) ( ) ( ) (5)

( ) ( ) ( ) (6)

10
where is the Lagrangian multiplier and equals the value of optimized cost function,

allowing us to derive (7) from (Färe and Primont). Using (6), the ratio of shadow prices of

outputs can be written as:

( )

( ) (7)

This ratio reflects the trade-off between two outputs in the production technology. For

example, if i is an undesirable output and j is desirable output, the ratio represents the

number of units of desirable output j that would be forgone to reduce the emission of one

unit of pollutant j is the producer shadow price for i th pollutant.

If we assume (as is commonly done) that the market price for the desirable output equals its

shadow price (Färe et al., 1993; Färe et al., 2005; Hailu and Veeman, 2000; Shephard, 1970),

then the shadow price for undesirable output can be witten as:

( )

( ) (8)

The shadow price of undesirable output is positive as the input distance function is non-

decreasing in pollutant outputs and the derivatives have opposite signs. We use Eq. (8) to

calculate shadow prices for three water pollutants (BOD, COD and TSS).

11
 3.3. Estimation of parametric input distance function

The input distance function is homogenous in inputs. The flexible functional form that

allows us to improve this property globally is the translog form. We estimate the translog

parametric input distance function frontier with linear programming techniques.

( ) ∑ ∑

( ) ∑∑

( ) ∑ ∑

∑ ∑ (9)

Aigner and Chu (1968) were the first to use mathematical programming techniques to

estimate parameters of production function. This method minimizes the sum of deviations

of the values of the function from the unknown frontier that is being estimated subject to

monotonicity and homogeneity restrictions as specified in Hailu and Veeman (2000). For this

study, the ability to impose inequality constraints is very important as we need to treat

desirable and undesirable outputs asymmetrically in the specification of technology (Färe et

al., 1993; Hailu and Veeman, 2000).

The objective function of the linear program is to choose a set of parameter estimates that

minimize the sum of deviations of log values of the input distance function from zero. We

impose monotonicity, homogeneity and symmetry conditions as constraints. Also we impose

the constraint that the estimated input distance value should be equal or greater than one

so that the data points in the same are feasible for the estimated technology. The derivative

properties of the input distance function with respect to desirable and undesirable outputs

12
are specified to be different. As Input distance function measures maximum proportional

reduction of inputs while outputs are held constant, the function should be non-decreasing

in inputs (Eq. 10) and non-increasing in desirable outputs (Eq. 11). While desirable outputs

can be freely disposable, reduction of undesirable outputs consumes resources and,

therefore, can be achieved only at the expense of desirable outputs or increase of inputs.

Therefore, input distance function should be non-decreasing in undesirable outputs (Eq. 12).

The implementation of these monotonicity conditions can be done through the derivative

signs imposed on the estimation of parameters.

( ) n=1,..4 (10)

( ) m=1 (11)

( ) m=2,..4 (12)

The input distance function like the output (and other) distance functions is a full

representation of the production technology (Färe and Primont, 1995)-. As a full

representation of the technology, it reflects the trade-offs between inputs and outputs

(desirable and undesirable). Therefore, whether some components (outputs) are changing or

not, the function will reflect what any bundle of inputs and outputs are in terms of potential

input savings. For example, it is possible for the efficiency value to increase when inputs are

reduced even when undesirable outputs stay the same.

To implement the estimation we use APEAR, an R based package for productivity and

efficiency analysis.

13
 3 Data

The data used in this paper come from a survey of water polluting industries that was

conducted in 2013 in Sri Lanka. A representative sample of firms was selected from the

database of industries available from the Central Environmental Authority (CEA). The sample

consists of high polluters4 and medium level of polluters located within 1 km of the river (see

Figure 1). There were 324 water polluting individual firms in total and our sample comprised

74 of them. We interviewed production and administrative managers of 74 firms to collect

information on inputs, desirable outputs and other required firm specific data for the year

2012.

Fig 1: Spatial distribution of water polluting firms within 1 km of the lower Kelani

4
The CEA categorized firms into three categories : high, medium and low polluting firms based on the size, and nature of
the pollutants
14
In addition to desirable outputs, these firms produce wastewater (i.e. undesirable output) in

their industrial processes. The wastewater is discharged into the Kelani River either treated

or untreated. The quantity of wastewater discharged by each firm was recorded during the

survey. However, the quality of the wastewater (concentration of BOD, COD and TSS in mg/l)

discharged by each firm was not available for the entire sample. Water quality data were

available with the CEA only for 25 firms. Therefore, wastewater samples were collected,

from the rest of the firms by visiting them (without prearranged appointments) with an

environmental officer of the CEA and an officer from a private laboratory. Information on

undesirable outputs was estimated based on water samples collected from the firms at the

survey time. Therefore, we assumed the concentration levels were not changing throughout

the year5.

The majority of firms (66 per cent) in our sample are high polluters. Although having an

environmental protection licence (EPL) is a mandatory requirement for these industries, only

53 per cent of the firms currently have licences. Little more than half of the firms have end-

of-the-pipe treatment plants and only 29 per cent of firms are found to be in compliance6

with the existing concentration standards for all three pollutants. Even though firms

classified as high polluters are monitored annually by the CEA, only 63 per cent of them have

treatment plants. In this category, only 53 per cent currently own licenses with an even

smaller proportion (49 per cent) complying with existing concentration standards. Among

medium level polluters which accounts for 33 per cent of the sample and are monitored

once every three years, only 52 per cent possess licences, 40 per cent own treatment

facilities and 12 per cent comply with current standards.

5
The concentration levels can be changed due to changes in treatment technology or changes in the production process.
However, we believe the probability of such changes within a year is not great although the same cannot be said about the
wastewater volume over the course of a year.
6
Having a license or a treatment plant does not assure that a firm comply with regulations. We checked the compliance
using concertation of pollutants in the wastewater samples.
15
In productivity measurements, the production technology is defined as the transformation

of a vector of inputs into a vector of outputs (Coelli et al., 2005). Depending on the empirical

exercise, input and output vectors can be defined at different levels: micro level and macro

level. Micro-level studies can be confined to a particular industry such as electricity (Coelli et

al., 2013) , paper and pulp (Hailu and Veeman, 2000) or ceramic (Picazo-Tadeo and Garc¡a-

Reche, 2007). However, there can be macro level studies that examine whole industrial

sector with different types of industries (Murty and Kumar, 2003, Murty and Kumar, 2002), a

country with several provinces (Zhang et al., 2008) and cross-country studies (Kocher et al.,

2006). It is customary or common practice in the Literature on growth accounting or

productivity trends to aggregate all data across industries into a national data set (Solow,

1957) . In our analysis, we aggregated input and output data in monetary terms for the

whole industrial sector as our interest is on investigating the shadow prices for water

polluting industries located in the river catchment.

For the estimation, we aggregated inputs into four categories: raw materials, labour, energy

use, and service and operational costs. We also aggregated outputs into four: one desirable

output (total sales) and three undesirable outputs (pollution loads of BOD, COD and TSS).

Total sales were aggregated in US$ and pollution loads were reported in kilograms. Only

some firms had end-of-pipe treatment facilities to for wastewater. For these firms, we

recorded the cost of annualized capital and operating cost under treatment costs (Table 1).

16
 Table 1 : Descriptive statistics of output and input variables (000s)

Variable Unit Mean Standard deviation

1 Total sales US$ 4700.36 17900.00
2 BOD load Kg 3.78 15.67
3 COD load Kg 6.74 26.87
4 TSS load Kg 3.33 19.39
5 Raw materials US$ 1936.74 11700.00
6 Labour US$ 96.21 267.83
7 Energy use US$ 65.48 284.48
8 Services and operations US$ 42.74 204.44
9 Treatment cost US$ 3.02 6.09

Table 2 shows the cost share of each input by industry. Raw materials account for the

biggest cost share in all industries followed by cost of labour, power and service and

maintenance and the cost of waste water treatment. The share of waste treatment is

comparatively small, 0 to 4.4%; however, information on treatment cost is available only for

the firms that have end-of-pipe emission treatment facilities.

Table 2: Mean cost share of each input by industry

Services Waste
Number Raw Energy and water
Industry of Firms material Labour use operations treatment
Chemical 6 0.607 0.235 0.092 0.065 0.001
Food 12 0.666 0.147 0.133 0.051 0.002
Beverages 5 0.509 0.242 0.123 0.092 0.035
Livestock farms 10 0.674 0.232 0.018 0.032 0.044
Vehicle service 25 0.469 0.375 0.038 0.083 0.034
Textile and leather 8 0.463 0.351 0.090 0.069 0.027
Mineral 5 0.773 0.109 0.093 0.021 0.004
Waste recycling 3 0.434 0.329 0.045 0.192 0.000
Total 74 0.561 0.276 0.070 0.069 0.024

17
 4 Empirical results

In this section, we report shadow price estimates, abatement cost under different policy

scenarios and the technical efficiency scores obtained from PLP (parameter estimates are

shown in Table 3).

Table 3: Parametric estimates of input distance function for water polluting firms

Variable Parameter Values Variable Parameter Values

lny1 1 -0.7923 lnx2x2 β13 -0.0209
lnx1 2 0.4261 lnx3x2 β14 -0.0098
lnx2 3 0.1110 lnx4x2 β23 0.0739
lnx3 4 0.0550 lnb1x2 β24 0.0001
lnx4 β1 0.4078 lnb2x2 β34 0.0000
lnb1 β2 0.0002 lnb3x2 γ11 0.0000
lnb2 β3 0.0014 lnx3x3 γ12 0.0192
lnb3 β4 0.0002 lnx4x3 γ13 -0.0017
lny1y1 11 -0.0932 lnb1x3 γ14 -0.0002
lny1x1 22 -0.0106 lnb2x3 γ21 -0.0005
lny1x2 33 0.0336 lnb3x3 γ22 0.0000
lny1x3 44 0.0135 lnx4x4 γ23 -0.0685
lny1x4 12 -0.0365 lnb1x4 γ24 0.0003
lny1b1 13 0.0000 lnb2x4 γ31 0.0002
lny1b2 14 -0.0003 lnb3x4 γ32 -0.0001
lny1b3 23 -0.0001 lnb1b1 γ33 -0.0001
lnx1x1 24 0.0545 lnb2b1 γ34 0.0001
lnx2x1 34 -0.0431 lnb3b1 γ41 0.0000
lnx3x1 β11 -0.0077 lnb2b2 γ42 -0.0004
lnx4x1 β22 -0.0037 lnb3b2 γ43 0.0002
lnb1x1 β33 -0.0001 lnb3b3 γ44 -0.0002
lnb2x1 β44 0.0004 Intercept 0 2.4205
lnb3x1 β12 0.0000
Y1 : Total Sales (US$) b1 : BOD load (kg)
x1: Raw materials (US$) b2 : COD load (kg)
x2 : Cost of labour (US$) b3 : COD load (kg)
x3 : Cost of energy (US$)
x4 : Cost of services and operations (US$)

18
 5.1 Shadow prices

We computed shadow prices for the undesirable outputs (BOD, COD and TSS) using the

parameters (Table 4) of the input distance function estimated by PLP. This was done using

the shadow prices ratios as illustrated in the Eq. (8). The shadow price values are based on

the marginal rate of transformation between undesirable and desirable outputs. Therefore,

these values can be interpreted as marginal cost for pollution abatement for industries. The

mean and median shadow prices for each industry (see Table 4) were calculated using the

firm level shadow price values.

Table 4: Shadow prices of BOD, COD and TSS (US$/ kg)

Industry Number BOD COD TSS

of firms Median Mean Median Mean Median Mean
Chemical 6 16.5 43.26 6.22 15. 6.61 10.60
Food 12 1.83 13.20 2.02 6.1 2.02 12.73
7 74
Beverages 5 0.01 60.11 0.01 65. 0.00 0.0011
3
Livestock farms 10 0.82 40.37 5.72 11. 0.49 1.55
22
Vehicle service 25 2.51 12.85 5.68 12. 7.46 19.61
13
Textile and leather 8 3.86 26.65 2.52 5.6 6.42 13.83
52
Mineral 5 18.4 42.30 7.24 24. 7.13 19.31
6
Waste recycling 3 0.11 5.76 0.01 1.4 1.03 25.73
95
Total 74 3.02 25.48 4.41 14. 3.24 13.59
9
76
On average, the cost of abatement of a kilogram of undesirable outputs is found to be US$

25.48 for BOD, US$ 14.76 for COD and US$ 13.39 for TSS (Table 4). The shadow prices show

a wide variation across firms: US$ 0 to 325.5 for BOD, US$ 0 to 251.8 for COD and US$ 0 to

168.37 for TSS.

Figure 2 depicts the distribution of shadow prices for the three pollutants. The graph shows

that shadow prices for three pollutants are within the range of 0-20 for majority of firms.
19
 .06
Kernel density

.04
.02
0

0 20 40 60 80 100
Shadow prices

SP of BOD SP of COD SP of TSS

Fig 2: Kernel distribution of shadow prices

We undertook further an empirical analysis to understand the main reasons for variations in

shadow prices of BOD (Table 5)7. The variation can be explained by the fact that our sample

firms comes from different industries8. For example, compared to vehicle servicing firms, the

shadow prices for beverages and livestock firms are significantly higher. The variation of

shadow prices of firms within the same industry is due to scale of operations9 and the

compliance to environmental regulations even though all firms are operating under the

same regulatory emission standards (Table 5). Firms with bigger pollution loads tend to

have lower shadow prices. However, whether a firm categorised by the CEA as a higher

polluter was not found to be statistically significant in explaining the pollution load

7
The mean shadow prices of firms were regressed on the firm characteristics
8
Having different industries has an influence on variation of shadow prices; however this is not the only factor that
influences the variations. Since the shadow prices are estimated as ratios, small change in denominator can result bigger
values. This is a common issue on shadow price estimates.
9
We used water consumption by firms as a proxy for the scale of operation for water polluting firms.
20
variations. We undertook similar analyses for COD and TSS and found the same set of

variables significantly influencing the shadow prices.

Table 5: Empirical analysis of factors affecting shadow prices

Dependant variable: Shadow price of BOD (USD/kg) Adjusted R2 =0.2524

Variable Coefficient Std. Err. P>t

High polluters (A) -4.083 15.824 0.797
Water consumption (in 1000 m3) -0.009** 0.003 0.002
Industrial category
Chemical 18.097 25.184 0.475
Food 25.371 19.746 0.204
Beverages 105.770** 33.373 0.002
Livestock 47.466* 20.405 0.023
Mineral 11.773 26.859 0.663
Textile 30.970 23.142 0.186
Waste recycling 24.668 33.580 0.465
Compliance with BOD standard 72.993*** 15.621 0.000
Constant -15.685 14.610 0.287

The Table 6 compares the differences in shadow prices of pollutants for compliant and non-

compliant firms. Firms that are already complying with standards have very high shadow

prices for all three water pollutants. This means that those who are in compliance already

operate in the upper segment of their MAC, because they have already abated, hence the

higher shadow prices. The mean shadow price for firms that do not adhere to existing

standards is US$ 6.40, US$ 7.32 and US$ 9.75 for BOD, COD and TSS, respectively (Table 6).

Table 6: Shadow prices by compliance to the existing concentration standards

Compliant firms Non-compliant firms

BOD COD TSS BOD COD TSS
Number of
firms (n) 27 40 38 47 34 36
Mean 58.70 21.10 17.25 6.40 7.32 9.75
shadow
price
(US$/ kg) (16.25)* (6.85) (4.55) (2.26) (2.00) (2.71)
21
* Standard deviations are reported in brackets

5.2 Abatement cost under different policy scenarios

The results presented above provide information on abatement costs at the margin, given

existing compliance patterns. It is also possible to use the estimated distance function to

derive abatement cost curves under different levels of pollution. The cost curves can be

generated for individual pollutants or a simultaneous reduction of pollutants. For each firm,

we simulated the marginal abatement cost of simultaneous pollutant reductions under the

assumption that the firm’s efficiency level would remain the same10. Given the fact that the

current regulation on effluent discharge to the water bodies based on concentration

standards, we simulated total cost for firms to meet a range of concentration standards.

The scenarios (1-9) shown in Table 7 are based on different concentration levels of three

main pollutants: BOD, COD and TSS. Scenario 9 shows the current regulatory effluent

standards while scenario 1 shows the combination of the highest concentration levels for

the three water pollutants. Then we simulated the cost of simultaneous reduction of the

three pollutants from current levels to the levels indicated in the row for each in each

scenario. We report the corresponding abatement as a percentage of effluent loads and

also the abatement costs as percentage of total production cost and of total revenue in the

Table 7.

Table 7: Abatement cost simulations for policy scenarios based on simultaneous reduction of

three pollutants.

10
We assumed that the efficiency of firms would be remained in the same level in the short-run.
22
 Abatement as a % of Abatement
Concentration levels
Scenarios effluent loads cost
(mg/l)
As a % of As a % of
BOD COD TSS BOD COD TSS
total total
production revenue
1 200 500 250 72 23 5 0.06 0.03
2 180 450 200 73 25 5 0.07 0.03
3 150 400 180 75 27 7
5 0.08 0.03
4 130 380 150 76 28 9
6 0.08 0.04
5 110 350 130 77 29 9
6 0.09 0.04
6 90 320 110 78 30 1
6 0.10 0.05
7 70 300 90 79 31 2
6 0.11 0.05
8 50 270 70 80 32 3
6 0.13 0.06
9 30* 250 50* 81 33 4
6 0.15 0.07
5 water bodies
*Indicates the current effluent standards for discharging waste water to inland
* 7

The cost estimates on overall industrial pollution treatment is not available in Sri Lanka. In

India, for example the cost of water pollution treatment would account for 2.5 per cent of

the industrial GDP (Kumar and Murty, 2011). A study on rubber industry conducted in Sri

Lanka using data from 2003-2005 found that the pollution tax required to bring the average

firm to compliance would be 8.7 percent of average annual turnover (Edirisinghe, 2014).

Compared with these figures it is apparent that the overall cost for non-complying firms (in

our sample) to meet the current regulatory standard is very low (0.15 per cent of total

production cost and 0.07 percent of the industrial revenue). However, given the

heterogeneity of firms and MAC, especially small firms and firms with high MAC cost would

pay higher cost than the average values suggest.

The total abatement cost is comparatively high in the case of current regulatory effluent

concentration standards, 0.06 per cent of the total production cost and 0.15 per cent of the

total revenues for the firms (see scenario 9 in the Table 7). Firms could save if they are

allowed to first meet less stringent concentration standards. For example, lifting

concentration standards from scenario 9 to 1, would reduce the costs significantly (Figure 3).

23
The policy scenario 1 in Table 7 shows the pre-treatment standards of some common central

waste treatment plants with considerably lower costs compared with existing emission

standards. As the industries face higher individual costs in treating waste to very low

concentrations, there would be a potential for cost saving if firms treat their waste to pre-

treatment standards and direct the discharges to a common treatment facility. . However,

such treatment facilities are only available in industrial parks.

Total abatement cost (Million US$)
0.22
0.18
0.14
0.10

1 2 3 4 5 6 7 8 9
Scenarios

Fig 3: Total abatement cost and the policy scenarios 11

5.3 Technical efficiency

The measures on technical (input) efficiency using input distance function frontiers are

summarized by industry in Table 8. In general, technical efficiency estimates are low for all

water polluting industries in the sample, indicating that there is untapped potential for

efficiency improvement. The mean technical efficiency is 35 per cent implying that there is a

11
The scenarios are based on concentration levels of pollutants and corresponding pollution loads, as reported in Table 2.7
24
substantial room for cost savings by reducing inputs while keeping the outputs constant.

Similar studies that estimated technical efficiency scores for industries with undesirable

outputs, especially in developing countries, report similarly low efficiency values. For

example, the technical efficiency score for paper recycling units in Vietnam varied from 0.37

to 1.00 (Van Ha et al., 2008) while the technical efficiency scores for Chines industrial

provinces ranged from 0.16 to 1.00 (Zhang et al., 2008).

Table 8: Input efficiency based on parametric input distance functions

Industry No of Firms Efficiency

Mean Median
Chemical 6 0.2671 0.1275
Food 12 0.2885 0.1589
Beverages 5 0.3106 0.1863
Livestock farms 10 0.4442 0.2968
Vehicle service 25 0.3938 0.3431
Textile and leather 8 0.1719 0.1322
Mineral 5 0.3438 0.2164
Waste recycling 3 0.6343 0.8259
Total 74 0.3500 0.2366

We carried out an empirical analysis of technical scores with industry categories, pollution

category and degree of compliance to current regulation on concentration standards (Table

9). Compared to vehicle services, food and textile industries found to be significantly less

efficient. The higher the degree of compliance, the lower the efficiency of firms; however,

this effect appear very small or negligible.

Table 9: Empirical analysis of factors affecting efficiency

Dependant variable: Firm efficiency Adjusted R2 =0.0557

Variable Coefficient Std. Err. P>t
High polluters (A) 0.015 0.095 0.872
Industrial category

25
 Chemical -0.124 0.138 0.372
Food -0.325** 0.145 0.029
Beverages -0.080 0.148 0.589
Livestock -0.227 0.175 0.199
Mineral -0.005 0.145 0.973
Textile -0.326** 0.141 0.024
Waste recycling 0.007 0.216 0.975
Degree of compliance to standards -0.004 * 0.002 0.050
Constant 0.725 *** 0.162 0.000

5 Conclusions and policy implications

Controlling water pollution in inland water bodies is one of the key challenges in developing

countries given the dependence on surface water for drinking, industrial use, irrigation and

recreational uses. Although continuous deterioration of water bodies is a common issue in

many of these countries, environmental regulations are not linked to the surface water

quality targets. In this study, we examined existing regulations on industrial water pollution,

its cost implications and incentives for compliance by industries.

First, we investigated the cost of pollution for a representative sample of firms belonging to

eight types of water polluting industries located in the Kelani river catchment. These

industries operate under regulatory standards on emission concentrations but only 29 per

cent of them are compliant with the standards. Using parametric input distance functions,

we estimated industry and firm specific shadow prices (marginal costs) of water pollutants.

Our results reveal a wide variation in shadow prices among firms and also firms within the

same industry. The variations of firm-specific shadow prices are due to differences in scale of

operation. The compliance with existing standards also contributes to the differences.

Shadow price estimates for all three pollutants (BOD, COD and TSS) are significantly higher

for the compliant firms compared to non-complaint firms.

26
Second, we simulated the potential total abatement cost for the firms under different policy

scenarios for simultaneous reduction of concentrations s including current regulation on

three water pollutants. The overall abatement cost to bring non-complying firms to

compliance is not very high considered as a percentage of total firm production cost and

revenues. As marginal cost increases with lower concentration levels, firms with

comparatively lower effluent concentrations face higher marginal costs. Therefore, having

uniform effluent concentrations standards across all firms may not yield optimal results in

terms of minimizing cost. Hence, the cost heterogeneity among firms makes a strong case

for market based instruments such as effluent discharge tax that equalize marginal

abatement cost among all polluting firms and provide least cost solution while achieving

pollution reduction targets.

Third, we examine the technical efficiency of firms and our findings show that average

efficiency is 35 per cent. This suggests there is a room for substantial savings in inputs (and

costs) without sacrificing outputs. We also found that the firm efficiency is negatively related

to the degree of compliance to current regulation; implying that there is no incentives for

firms to comply.

The evidence on poor compliance and wide variations in shadow prices (MAC) makes a

strong case for a new design of comprehensive environmental policy to control industrial

pollution as an alternative to existing command and control regulations. The shadow price

estimates can be used as guidelines to design market based policy instruments such as

emission-based taxes or tradeable permits that would cap the level of pollution released into

the river. The case for a serious consideration of alternative approaches is made stronger by

the evidence of weak enforcement of current regulations. Therefore, setting appropriate

27
economic instruments would provide incentives for firms to control emissions in socially

optimum ways without imposing a greater burden on complying firms.

Acknowledgements
We acknowledge the financial support of the Department of Education and Training,

Government of Australia for conducting this research and the South Asian Network for

Development and Environmental Economics (SANDEE) and its sponsors for field work. We

also acknowledge support from officers of the Central Environmental Authority for providing

information and support in data collection in Sri Lanka.

References

AECEN, 2006. Environmental Compliance and Enforcement in Sri Lanka: Rapid Assessment.
Asian Environmental Compliance and Enforcement Network, USAID, Sri Lanka
Aigner, D.J., Chu, S.F., 1968. On Estimating the Industry Production Function. The American
Economic Review 58, 826-839.
Baumol, W.J., Oates, W.E., 1988. The Theory of Environmental Policy. Cambridge University
Press, UK.
Biswas, A., Tortajada, C., 2009. Changing Global Water Management Landscape. In: Biswas,
A., Tortajada, C., Izquierdo, R. (Eds.), Water Management in 2020 and Beyond. Springer
Berlin Heidelberg, pp. 1-34.
Blackman, A., 2009. Colombia's discharge fee program: Incentives for polluters or
regulators? Journal of Environmental Management 90, 101-119.
Blackman, A., Harrington, W., 2000. The Use of Economic Incentives in Developing Countries:
Lessons from International Experience with Industrial Air Pollution. The Journal of
Environment & Development 9, 5-44.
CEA, 2012. Annual Report. Central Environmetal Authority ,Ministry of Environment, Sri
Lanka.
Coelli, T.J., Gautier, A., Perelman, S., Saplacan-Pop, R., 2013. Estimating the cost of improving
quality in electricity distribution: A parametric distance function approach. Energy Policy 53,
287-297.
Coelli, T.J., Rao, D.S.P., O'Donnell, C.J., Battese, G.E., 2005. An introduction to efficiency and
productivity analysis. Springer Science and Business Media, USA.
Cooper, W.W., Seiford, L.M., Zhu, J., 2011. Data Envelopment Analysis: History, Models, and
Interpretations: Handbook on Data Envelopment Analysis. Springer US.
Du, L., Hanley, A., Wei, C., 2015. Estimating the Marginal Abatement Cost Curve of CO2
Emissions in China: Provincial Panel Data Analysis. Energy Economics 48, 217-229.
Edirisinghe, J.C., 2014. Taxing the Pollution: A Case for Reducing the Environmental Impacts
of Rubber Production in Sri Lanka. Journal of South Asian Development 9, 71-90.

28
Färe, R., Grosskopf, S., Lovell, C.A.K., Yaisawarng, S., 1993. Derivation of Shadow Prices for
Undesirable Outputs: A Distance Function Approach. Review of Economics and Statistics 75,
374-380.
Färe, R., Grosskopf, S., Noh, D.-W., Weber, W., 2005. Characteristics of a polluting
technology: theory and practice. Journal of Econometrics 126, 469-492.
Färe, R., Primont, D., 1995. Multi-Output Production and Duality: Theory and Applications.
Kluwer Academic Publishers, Boston.
Hailu, A., Veeman, T.S., 2000. Environmentally Sensitive Productivity Analysis of the
Canadian Pulp and Paper Industry, 1959-1994: An Input Distance Function Approach. Journal
of Environmental Economics and Management 40, 251-274.
Herath, G., Amaresekera, T., 2007. Assessment of Urban and Industrial Pollution on Water
Quality: Kelani River Sri Lanka. Southeast Asian Water Environment 2 2, 91-98.
Kathuria, V., 2006. Controlling water pollution in developing and transition countries—
lessons from three successful cases. Journal of Environmental Management 78, 405-426.
Kumar, S., Murty, M.N., 2011. Water Pollution in India: An Economic Appraisal. India
Infrastructure Report 2011,Water: Policy and Performance for Sustainable Development.,
DFC and Oxford University Press, New Delhi.
Kuosmanen, T., Kortelainen, M., 2012. Stochastic non-smooth envelopment of data: semi-
parametric frontier estimation subject to shape constraints. Journal of Productivity Analysis
38, 11-28.
Lee, M., 2005. The shadow price of substitutable sulfur in the US electric power plant: A
distance function approach. Journal of Environmental Management 77, 104-110.
Ma, C., Hailu, A., 2016. The Marginal Abatement Cost of Carbon Emissions in China. The
Energy Journal 37.
Mandal, S.K., 2010. Do undesirable output and environmental regulation matter in energy
efficiency analysis? Evidence from Indian Cement Industry. Energy Policy 38, 6076-6083.
Murty, M., Kumar, S., Dhavala, K., 2007. Measuring environmental efficiency of industry: a
case study of thermal power generation in India. Environmental and Resource Economics 38,
31-50.
Murty, M.N., Kumar, S., 2003. Win–win opportunities and environmental regulation: testing
of porter hypothesis for Indian manufacturing industries. Journal of Environmental
Management 67, 139-144.
Murty, M.N., Kumar, S., Paul, M., 2006. Environmental regulation, productive efficiency and
cost of pollution abatement: a case study of the sugar industry in India. Journal of
Environmental Management 79, 1-9.
Newell, R.G., Stavins, R.N., 2003. Cost Heterogeneity and the Potential Savings from Market-
Based Policies. Journal of Regulatory Economics 23, 43-59.
Qin, H.-P., Su, Q., Khu, S.-T., 2011. An integrated model for water management in a rapidly
urbanizing catchment. Environmental Modelling and Software 26, 1502-1514.
Rahman, M.A., Ancev, T., 2014. Economic analysis of alternative pollution abatement
policies: the case of Buriganga River, Bangladesh. Interdisciplinary Environmental Review 15,
66-87.
Rezek, J.P., Campbell, R.C., 2007. Cost estimates for multiple pollutants: A maximum entropy
approach. Energy Economics 29, 503-519.
Schaffner, M., Bader, H.-P., Scheidegger, R., 2009. Modeling the contribution of point
sources and non-point sources to Thachin River water pollution. Science of The Total
Environment 407, 4902-4915.

29
Simonovic, S.P., Fahmy, H., 1999. A new modeling approach for water resources policy
analysis. Water Resour Res 35, 295-304.
Stavins, R.N., Karl-Göran, M., Jeffrey, R.V., 2003. Chapter 9 - Experience with Market-Based
Environmental Policy Instruments. Handbook of Environmental Economics. Elsevier, pp. 355-
435.

Keypoints
 Abatement costs for industries in Kelani river catchment were estimated using
distance functions.
 Costs for different policy scenarios based on effluent concentration standards were
simulated
 Variation of marginal abatement costs makes a strong case for use of market based
instruments

30