Waste Management 39 (2015) 277–286

Contents lists available at ScienceDirect

Waste Management journal homepage: www.elsevier.com/locate/wasman

Review

Is bigger better? An empirical analysis of waste management in New South Wales Pedro Carvalho a,b,1, Rui Cunha Marques b,⇑, Brian Dollery c,d,2 a

LAMEMO Laboratory, Department of Civil Engineering, Federal University of Rio de Janeiro, COPPE/UFRJ, Av. Pedro Calmon – Ilha do Fundão, 21941-596 Rio de Janeiro, Brazil CESUR – Center for Urban and Regional Systems, Instituto Superior Técnico, University of Lisboa, Av. Rovisco Pais, 1049-001 Lisbon, Portugal c CLG – Centre for Local Government, University of New England, Armidale, NSW, Australia d Faculty of Economics, Yokohama National University, Japan b

a r t i c l e

i n f o

Article history: Received 25 July 2014 Accepted 20 January 2015 Available online 18 February 2015 Keywords: Economies of size Economies of output density Economies of scope Waste management Recycling New South Wales

a b s t r a c t Across the world, rising demand for municipal solid waste services has seen an ongoing increase in the costs of providing these services. Moreover, municipal waste services have typically been provided through natural or legal monopolies, where few incentives exist to reduce costs. It is thus vital to examine empirically the cost structure of these services in order to develop effective public policies which can make these services more cost efficient. Accordingly, this paper considers economies of size and economies of output density in the municipal waste collection sector in the New South Wales (NSW) local government system in an effort to identify the optimal size of utilities from the perspective of cost efficiency. Our results show that – as presently constituted – NSW municipal waste services are not efficient in terms of costs, thereby demonstrating that ‘bigger is not better.’ The optimal size of waste utilities is estimated to fall in the range 12,000–20,000 inhabitants. However, significant economies of output density for unsorted (residual) municipal waste collection and recycling waste collection were found, which means it is advantageous to increase the amount of waste collected, but maintaining constant the number of customers and the intervention area. Ó 2015 Elsevier Ltd. All rights reserved.

Contents 1. 2. 3. 4. 5. 6.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Municipal solid waste sector in New South Wales Methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sample and model specification . . . . . . . . . . . . . . . Empirical results and policy implications . . . . . . . . Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

1. Introduction Rising living standards in many regions of the world has led to a growing demanding for municipal solid waste (MSW) services. ⇑ Corresponding author. Tel.: +351 218418305. E-mail addresses: [email protected] (P. Carvalho), rui.marques@tecnico. ulisboa.pt (R.C. Marques), [email protected] (B. Dollery). 1 Tel.: +55 2139387393. 2 Tel.: +61 0266732500. http://dx.doi.org/10.1016/j.wasman.2015.01.024 0956-053X/Ó 2015 Elsevier Ltd. All rights reserved.

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

277 278 279 280 281 284 285 285

This has led to the allocation of ever greater resources to collect and treat this waste which has put into question the long-run sustainability of these services (Santibañez-Aguilar et al., 2013; Broitman et al., 2012). At the same time, environmental concerns have stimulated the growing reuse and recycling of MSW, thereby sparking a further rise in the costs of MSW collection and treatment services. To make matters worse, these services have usually been provided in the form of natural or legal monopolies (Callan and Thomas, 2001), which has mitigated the reduction of costs associated with MSW services as a whole. Taken together, this

278

P. Carvalho et al. / Waste Management 39 (2015) 277–286

makes the empirical analysis of the cost structure of these services, including economies of size (ESize) and economies of scope, vital to the formulation of effective public policies to make these services more cost efficient. ESize are a measure that reflects the cost reaction due to a proportional increase in all outputs as well as other variables which characterize the dimension of the outputs (like the number of customers and the size of the intervention area, in the case of the waste sector). This measure is greater than one when the unit (average) cost of production falls to a proportional increase in the amount of outputs, number of customers and area size, which means that in these circumstances it is advantageous to increase, in this case, the intervention area of MSW services supply. Otherwise, if ESize are smaller than one, it means that it is more advantageous to decrease the area of MSW services supply, and it is said that diseconomies of size occur. In the literature ESize are often called ‘‘economies of scale’’ erroneously. Although these two concepts are related, they only coincide under certain circumstances. While ESize refer to how costs respond to increases in outputs, economies of scale refer to how outputs respond to increases in all inputs together (Chambers, 1988). In addition to ESize, in industries such as the waste collection sector, it is also possible to investigate economies of output density (EOD). EOD is a quantity that (unlike ESize) measures the reaction of costs relative to the increase in the only one of the outputs (or several outputs), maintaining constant the number of customers and the sized of the intervention area. And finally economies of scope are related to the fact of whether or not the joint production of different goods or services is more advantageous. When there are economies of scope it means that joint production is more advantageous, otherwise it is said that there are diseconomies of scope. With growing concerns about the sustainability of waste services, a number of studies have focused on their cost structure. Early work began to emerge in the 1960s in the U.S. For example, Hirsch (1965) examined the cost structure of the refuse collection services provided by 24 municipalities in the St. Louis City – County area, corresponding to a population of about one million residents. He found no significant evidence of the existence of economies of scale. In a similar vein, Hall and Jones (1973) investigated 22 local communities of Texas and found economies of scale for a community size up to 9600 inhabitants. Stevens (1978) evaluated the refuse collection services of firms that operated in 340 North American cities: constant returns to scale were found for towns with populations greater than 50,000 people. Folz (1999) studied the cost structure of municipal recycling in American cities with less than 50,000 inhabitants and found economies of scale. Callan and Thomas (2001) analyzed the costs of the waste services in 110 Massachusetts cities and towns. They found EOD for the provision of recycling services and EOD and constant returns to scale for refuse collection and disposal services, as well as economies of scope in joint recycling and disposal services. In Europe most empirical research originated somewhat later. For instance, Antonioli and Filippini (2002) studied 30 Italian refuse collection firms which served municipalities and found EOD and economies of scale for most output levels, but diseconomies of scale for large firms. More recently, Abrate et al. (2012) examined economies of scale and economies of scope in 529 Italian municipalities. They found constant returns to scale in the refuse collection for populations of about 42,500 inhabitants, but overall diseconomies of scale in municipalities with greater than 100,000 inhabitants, as well as economies of scope in the joint provision of disposal and recycling services. In Spain, a study carried out by Bel and Fageda (2010) found economies of scale for Galician municipalities with less than 50,000 inhabitants. Simões et al. (2010) analyzed 29 Portuguese utilities that provided urban waste disposal services and identified an optimal size for utilities at around 300,000 inhabitants. More recently, Carvalho and

Marques (2014) analyzed the selective collection and recycling services in Portugal. They found significant EOD in recycling (especially for glass and paper), where optimal size corresponded to about 400/550,000 inhabitants, and where smaller utilities displayed economies of size and larger utilities diseconomies of size. Given this body of empirical literature, it seems reasonable to conclude that modest optimal sizes of around 10,000–50,000 inhabitants exist for collection services, while disposal services show higher optimal sizes. The bulk of studies also report considerable EOD for collection services, especially for small utilities, as well as economies of scope in the joint provision of disposal and recycling services. In this paper we seek to empirically assess economies of size (ESize), EOD and economies of scope in the waste collection sector in New South Wales (NSW) local government and identify the optimal size of utilities in terms of lower average unit costs. To this end, we applied the Stochastic Frontier Analysis methodology (SFA) to estimate the cost function of waste collection services. Our paper attempts to augment an existing embryonic Australian empirical literature on waste collection and disposal services (see, for instance, Worthington and Dollery, 2001; Woodbury et al., 2003) by seeking – for the first time – to examine the impact of waste market structure in Australia. The paper is divided into five main parts. Section 2 describes the MSW sector in NSW local government. Section 3 outlines the empirical methodology applied in the paper. Section 4 presents the sample and the model employed, whereas Section 5 discusses the empirical findings of the study. The paper ends with some brief concluding remarks in Section 6.

2. Municipal solid waste sector in New South Wales In NSW local authorities bear statutory responsibility for the provision of domestic (municipal) waste management services (as regulated by the NSW Local Government Act 1993), including waste collection, recycling and landfill disposal. This Act enables local councils to levy a charge for the provision of these services. Although domestic recycling services are not mandatory for municipalities, its environmental benefits, together with community expectations, as well as the existence of a waste and environment levy, make domestic recycling services a priority for local governments (DECCW, 2010). The diversion target from landfilling to other process options imposed by the Department of Environment, Climate Change and Water (DECCW) for municipal waste was set at 66% to 2014. Municipal waste includes household garbage, recycling and organics using kerbside and clean-up collections and drop-off facilities and other materials collected in municipal parks and gardens, street sweepings, waste from public places and from council operations. The regular waste stream corresponds to about 59% of mixed residual waste, 26% sorted for recycling (paper and cardboard, glass, plastic and metal) and 15% sorted organics (food and garden waste) (DLG, 2013). The waste collection service in NSW is provided by 151 councils. In practice, the process of waste collection and disposal by local authorities typically means that household municipal waste is collected by commercial contractors using automated vehicles once a week. Households must place municipal waste in prescribed wheeled containers – termed ‘wheelie-bins’ – and place these at the side of the road, together with two boxes for recyclable material: one for glass, metal, plastic and other material and the other box for paper material. Some local authorities are more prescriptive than others regarding the nature of recyclable material. Every second week, a separate wheelie-bin must be used for organic waste, sometimes referred to as ‘garden garbage’. Should households wish to dispose of additional municipal waste or recyclable

279

P. Carvalho et al. / Waste Management 39 (2015) 277–286

of customer density (ECD). In addition, these nonparametric methodologies have other disadvantages compared with parametric ones, namely, they are deterministic methods and are very sensitive to outliers and extreme data and also suffer from the problem of the ‘‘curse of dimensionality’’, particularly the traditional full frontier nonparametric methods, such as DEA (Daraio and Simar, 2007). The cost functions domestic recycling services estimation from SFA consists in evaluating the following function (Kumbhakar and Lovell, 2000):

Ei ¼ Cðyi ; wi ; bÞ  expðv i Þ=CEi

Fig. 1. Municipal waste framework in NSW.

material, then they can take this waste directly to the municipal garbage processes site directly and pay a small fee. In NSW there are also 321 landfills, 287 recycling facilities and 70 composting facilities. In addition, several composting facilities exist, although incineration has not been employed yet. While waste generation has fallen in recent years, waste generation in per capita terms continues to be substantial, exceeding 2 kg per person per day. However, the waste diversion rate from landfilling is one of the highest in the world. Fig. 1 presents a synoptic illustration of the municipal waste framework in NSW. There is a high and standardized level of service since it is regulated by law and strictly complied with. Although waste operators in NSW are generally considered to be reasonably efficient, doubts exist as to whether their cost structure is efficient. In particular, it is worthwhile considering whether the amalgamation of at least some of the 151 operators in NSW would be desirable from a cost efficiency perspective. In addition, it is important to determine empirically if the waste recycling and collection are complementary functions and if potential economies of scope may exist. To this end, Section 3 sets out our empirical strategy to analyze in detail the market structure of the NSW waste sector.

3. Methodology Among the various parametric methodologies employed in the empirical literature, Frontier Analysis (SFA) has been one of the most appealing, particularly in efficiency analysis, because of its comparative strengths. This methodology was initially proposed by Aigner et al. (1977) and Meeusen and van den Broeck (1977) for the purpose of estimating the efficiencies of Decision Making Units (DMU) and their corresponding efficient frontiers. Perhaps for this reason, SFA has chiefly been applied in this area and only sometimes in cost function estimation, since in many situations the cost functions estimated using this methodology do not satisfy the regularity conditions imposed by the economic theory, and mathematically it is difficult to impose a priori these regularity conditions. As a result, the SFA methodology is often disregarded in cost function estimation. However, methods have been proposed in the literature to overcome this problem, such as Henningsen and Henning (2009). In the present study, the regularity conditions were not met only for a small percentage of DMUs, so it was not necessary to impose them. In the literature nonparametric methodologies (such as Data Envelopment Analysis – DEA) are also applied in the efficiency estimation as well as in the search for economies of scale and economies of scope. However, these methodologies do not enable to search for other kind of economies, such as EOD and economies

ð1Þ

where Ei is the expenditure incurred by utility i, C(yi, wi; b) is the deterministic cost frontier, vi represents the effects of random shocks on utility i and CEi is the cost efficiency of utility i defined as CEi = exp(ui), yi and wi are the outputs produced and input prices respectively and b the technology parameters to be estimated. It is assumed that ui is a non-negative random independent variable and identically distributed (i.i.d.) with a specific distribution function and vi also a random variable i.i.d., two-sided normally distributed with mean zero and a standard deviation Nð0; r2m Þ and independent of ui . Given some fragilities of some SFA models, several other models have been proposed in the literature (see, for example, Greene, 2008, for an overview of existing models). For example, Pitt and Lee (1981) and Schmidt and Sickles (1984) proposed models which assume that the cost inefficiency of each producer is constant over time. However, such models turn out to be misspecified since they do not capture changes that may occur in the efficiency of producers over time, either due to internal or external changes. Accordingly, other models have subsequently been proposed which enable the cost inefficiency of each producer to vary over time (see, for example, Battese and Coelli, 1992, 1995). The initial first models are termed Time-invariant inefficiency models and the latter Time-varying inefficiency models. A further problem addressed in the literature lies in the component which reflects whether the inefficiency (ui) is uncorrelated with the element representing the effects of random shocks (vi) and with the regressors. The approaches that impose the condition that ui are uncorrelated with the vit and with the regressors are named random-effects models (see, for instance, Pitt and Lee, 1981; Battese and Coelli, 1995) and the approaches which relax these assumptions are called Fixed-effects models (e.g., Schmidt and Sickles, 1984). More recently, the fact that traditional models are not able to distinguish between the unobserved heterogeneity (which is outside the control of producers but can nonetheless affect their costs significantly) from inefficiency of producers themselves has also been raised. Greene (2005a) proposed two new models to overcome this problem: the ‘true’ fixed-effects and ‘true’ random-effects models. In the present context, several cost functions using different models were estimated for the waste collection sector in NSW and from these cost functions ESize and economies of density were estimated. In addition to ESize, in some industries, such as the waste collection sector, where, in addition to the main outputs, there are other variables which characterize the dimension of the outputs (like the number of customers and the size of the intervention area), EOD and ECD can also to be estimated. ESize, in the waste collection sector, measures the reaction of the costs of production relative to a proportional increase of the amount of outputs (y), number of customers (cu) and intervention area size (area) simultaneously (Caves et al., 1984). ESize is defined as: 1

ESize ¼ ð@ ln C=@ ln y þ @ ln C=@ ln cu þ @ ln C=@ ln areaÞ

ð2Þ

280

P. Carvalho et al. / Waste Management 39 (2015) 277–286

Table 1 Statistics of the variables considered.

Minimum 1st Quartile Median 3rd Quartile Maximum Average Standard deviation

Total domestic waste collection costs (excluding tipping costs) (A$)

Domestic waste collected (tons)

Recycling collected (tons)

Population (inh)

Area size (Km2)

Staff price (A$/employee)

Other costs price (A$/kg)

562 142,211 601,796 2,641,848 18,922,543 1,830,259 2,497,927

155 1683 6082 13,631 145,000 11,663 15,444

1.562 352 2136 8965 77,687 6423 9938

1389 6594 20,255 57,884 283,458 42,539 52,925

5.8 185.6 2432.5 4344.6 53510.8 3837.8 6570.2

18,622.23 40,904.58 46,432.93 52,404.24 77,649.67 47,116.46 8707.79

0.03 1.00 1.57 3.03 4377.68 25.66 260.34

When ESize > 1, it is said that the utility exhibits ESize, and indicates that a proportional increase in the amount of outputs, number of customers and area size corresponds to a reduction in the average cost of production, which means that the utility should increase the output produced, number of customers and area size to become more cost-effective. By contrast, Otherwise, if ESize < 1, the utility displays diseconomies of size, and it becomes more advantageous to reduce its production levels. Thus, the optimal size of utilities that is, the corresponding size to the lower (unit) average costs is reached when ESize = 1. In contrast, EOD measures the reaction of costs relative to the increase in the main output, maintaining the number of customers and the size of the intervention area constant, and are defined as: 1

EOD ¼ ð@ ln C=@ ln yÞ

ð3Þ

ECD allows us to assess the behavior of costs due to a proportional increase in both the output and number of customers, maintaining the size of the area constant:

ECD ¼ ð@ ln C=@ ln y þ @ ln C=@ ln cuÞ

1

ð4Þ

Accordingly, as in the case of ESize, there are EOD and ECD when these economies are greater than one; otherwise there are diseconomies. In addition, in this paper the existence of possible economies of scope in the joint provision of unsorted municipal waste and recycling waste collection services are also evaluated. These were estimated bearing in mind the sufficient condition for the existence of economies of scope provided by economic theory, which holds that there are economies of scope when the cost function exhibits a weak cost complementarity between the two outputs: in this case between the amount of household unsorted municipal waste collected (U) and the amount of recycling waste collected (R). Put differently, if the following condition is met:

@ 2 C=@U@R < 0

ð5Þ

That is, for a translog equation, as is the case:

@ 2 C=@U@R < 0 () @ 2 ln C=@ ln U@ ln R þ ð@ ln C=@ ln U  @ ln C=@ ln RÞ < 0 ð6Þ

costs (TC) in the provision of MSW service were considered as the dependent variable. Total costs include all the costs involved in collection (including the transportation costs to the end facilities), except for tipping costs (or gate fees) (in thousands of A$). Two outputs and two input prices were considered. The outputs are the amount of household waste (unsorted) collected (U) (tons) and the amount of recycling waste collected (R) (tons). However, to make possible to estimate economies of density and ESize the following variables were also included in the models: Population (Pop) (number of inhabitants) and the size of the intervention area (Area) (in km2). The inputs prices considered were the price of staff (SP) and other price (OP). The price of staff was estimated dividing the employee costs by the number of equivalent full time staff (A$/ employee) and the other price was estimated dividing the remaining costs by the amount of household waste collected (unsorted and recycling) (A$/ton). In Table 1 a summary statistics of these variables is presented. It is known that many other factors (variables) may have influence on the cost of waste services, such as the collection frequency, kind of collection (kerbside collection or drop-off systems) and others. Unfortunately, accurate official data does not exist on these additional factors. While it would have been preferable to include some of these variables in our empirical analysis, the absence of the requisite data necessarily meant that we were unable to do so. Nonetheless, given the high degree of homogeneity in the process of municipal waste collection and disposal across New South Wales local councils, and the fact that our analysis captures the key variables, we believe that the variables considered in this study already provide a reasonable perspective on the cost of provision of waste services in New South Wales. Nevertheless, if more information about other factors (variables) that might influence the costs is available, it must also be considered in future studies. In the same line, it would be very interesting to integrate the environment impacts in the study by carrying out a life cycle assessment (Feo and Malvano, 2012). The translog was the functional form chosen to represent the cost functions, due to its flexibility, ease of estimation and interpretation of results. However, because of multicollinearity and heteroscedasticity problems, some terms of the complete translog function were excluded, resulting in the following restricted translog cost function:

lnðTC=POÞ ¼ a þ bU ln U þ bR ln R þ bPop ln Pop þ bArea ln Area 4. Sample and model specification

1 1 2 2 þ bPop2 ðln PopÞ þ bArea2 ðln AreaÞ 2 2

The sample comprises data on the waste collection service in 184 local utilities in NSW (there are councils with more than one utility). The data sample was observed over a period of six years, between 2000–01 and 2005–06 and was collected from the NSW Division of Local Government. In order to calculate EOD, ECD and ESize in MSW service in the NSW cost functions were estimated. For this purpose, the total

þ bU

Pop

ln U  ln Pop þ bU

þ bR

Pop

ln R  ln Pop þ bR

þ bPop

Area

1 þ cPW 2

Area Area

ln U  ln Area

ln R  ln Area

ln Pop  ln Area þ cPW

PO2

PO

lnðSP=OPÞ  lnðSP=OPÞ

lnðSP=OPÞ ð7Þ

281

P. Carvalho et al. / Waste Management 39 (2015) 277–286

Model I Pooled Model

Model II Fixed-effects model (Schmidt and Sickles, 1984)

Model III Random-effects model (Pitt and Lee, 1981)

Fig. 2. Efficiencies estimated from Models I–III as function to the amount of municipal waste collected.

In cost function estimation, the linear homogeneity in input prices was imposed and the remaining regularity conditions – namely non-negativity of the monotonicity of input prices, marginal costs and the concavity of input prices – were checked after cost functions had been estimated. In order to obtain the most robust results possible, several different models for cost functions were estimated and the results compared. All models were estimated using the parametric SFA methodology. 5. Empirical results and policy implications In order to draw reliable inferences, five different models for the cost function were estimated. In the first model, the structure of panel data was not considered (Model I – Pooled model); that is, it was assumed that the observations are all independent of each other, ignoring the fact that some observations correspond to the same entity but were observed at different points in time (i.e., different years). Two models where the structure of panel data and the inefficiency of the utilities are constant throughout the study period were assumed (Time-invariant inefficiency models): Model II (Fixed-effects – Schmidt and Sickles, 1984) and the Model III (random-effects – Pitt and Lee, 1981). Finally, two other models were employed, where, in addition to considering panel data, it was also assumed that the inefficiency of the utilities underwent variation throughout the study period (Time-varying inefficiency models): Model IV (True fixed-effects – Greene, 2005a) and Model V (True random-effects – Greene, 2005a). Table A.1 in Appendix A lists the coefficients of the frontier cost function for the various models. Given that Model I (pooled model) does not take into account the structure of panel data, it is expected that this model might

provide biased results. In fact, as we can see from Fig. 2, Model I generates very high efficiencies, implying that it is not the most appropriate model. Nevertheless, with respect to EOD, ECD and ESize, Model I provides plausible results, considering the circumstances of the sector. The results from Model I show high and statistically significant EOD for both unsorted municipal waste and recycling waste collection services, and only ECD and ESize for smallest utilities (1st quartile) (Tables 2 and 3). Regarding Models II (Fixed-effects – Schmidt and Sickles, 1984) and III (random-effects – Pitt and Lee, 1981), despite taking into account the structure of panel data (in contrast to Model I – pooled model), these models contain various imperfections. One of their main shortcomings is the fact that they tend not to distinguish inefficiency from the unobserved heterogeneity usually present in the real-world waste sectors. That is particularly true in the fixed-effects models, because in these models time-invariant regressors cannot be included and, for this reason, unobserved heterogeneity is absorbed by the inefficiency component, resulting in overestimated inefficiency estimates (Greene, 2005b). This is true in the present paper, where the fixed-effects model provides low efficiencies (Figs. 2 and 3). The random-effects models allow for the inclusion of time-invariant regressors. However, this is achieved at the cost of imposing the hypothesis that the inefficiency component is independent from (and therefore uncorrelated with) the regressors. By applying the Hausman test statistic it was demonstrated that in the present study the random-effects model is preferred when compared with the fixed-effects model (chi2(13) = 12.17; Prob > chi2 = 0.5142) and, therefore, in this paper the inefficiency of the utilities is not correlated with regressors. It is evident that the random-effects models are those which provide more plausible and statistically significant estimates of

282

P. Carvalho et al. / Waste Management 39 (2015) 277–286

Table 2 Estimated degrees of ESize and density estimates and cost complementarities between unsorted municipal waste and recycling waste collection services.

Quartile

Economies of output density (domestic waste collected)

1st Median 3rd

Economies of output density (recycling collected)

1st Median 3rd

Economies of customer density

1st Median 3rd

Economies of size

1st Median 3rd

Cost complementarity between domestic waste and recycling collected (A$/ton2)

1st Median 3rd

Time-invariant inefficiency models

Time-varying inefficiency models

Model I

Fixed-effects model (Schmidt and Sickles, 1984) Model II

Random-effects model (Pitt and Lee, 1981) Model III

True fixed-effects model (Greene, 2005a) Model IV

True random-effects model (Greene, 2005a) Model V

11.516 (10.750) 6.874+++ (2.111) 6.007++ (2.410)

28.142 (44.842) 32.138 (22.678) 17.483 (12.171)

13.045+ (7.596) 41.488 (32.354) 17.445++ (7.480)

26.145 (57.128) 3.844+++ (0.805) 2.71+++ (0.616)

15.542 (19.654) 4.929+++ (1.182) 3.712+++ (1.000)

16.921+++ (4.029) 36.748+++ (12.272) 50.018 (36.905)

439.826 (2401.916) 215.31 (352.181) 108.651 (146.594)

27.538+ (13.973) 1411.004 (19869.390) 63.115 (67.360)

49.968 (31.076) 226.868 (475.720) 52.814 (42.322)

39.137++ (19.294) 927.532 (7964.140) 84.416 (105.652)

1.188++ (0.088) 0.862+++ (0.021) 0.75+++ (0.033)

2.081 (1.963) 1.778 (1.289) 1.73 (1.859)

1.622++ (0.260) 0.899+++ (0.036) 0.713+++ (0.047)

8.689 (51.584) 1.281 (0.843) 1.018 (0.997)

1.152 (0.121) 0.861+++ (0.031) 0.762+++ (0.053)

1.558+++ (0.205) 0.879+++ (0.033) 0.717+++ (0.045)

3.007 (4.778) 3.778 (7.058) 2.609 (5.561)

2.636+ (0.909) 0.865++ (0.057) 0.636+++ (0.058)

3.301 (10.387) 1.968 (2.570) 1.112 (1.569)

1.493+ (0.280) 0.884++ (0.054) 0.737+++ (0.076)

0.0055 (0.0085) 0.0043 (0.0072) 0.0037 (0.0072)

0.00052 (0.00905) 0.00057 (0.0089) 0.00095 (0.0085)

0.0025 (0.0070) 0.0054 (0.0067) 0.0044 (0.0067)

0.0129** (0.0057) 0.0146** (0.0057) 0.0188** (0.0074)

0.0070 (0.0059) 0.0084 (0.0057) 0.0112* (0.0064)

Pooled model

In brackets are the standard errors. Statistically significant at the 1% level. + Statistically significant different from one at 10% level. ++ Statistically significant different from one at 5% level. +++ Statistically significant different from one at 1% level. * Statistically significant at the 10% level. ** Statistically significant at the 5% level.

⁄⁄⁄

Table 3 Dimension of utilities by percentiles. Percentile

10th 20th 25th (1st quartile) 30th 40th 50th (median) 60th 70th 75th (3rd quartile) 80th 90th

Domestic waste collected (tons)

Recycling collected (tons)

Population (inh)

Area size (Km2)

Staff price (A$/ employee)

Other costs price (A$/kg)

Total domestic waste collection costs (excluding tipping costs) (’000 A$)

577.893 1319.400 1683.000

3.619 209.400 352.315

3550 4989 6594

22 91 186

36,390 39,688 40,905

0.735 0.892 1.001

70.57 114.97 142.21

2358.636 3718.278 6082.000

500.000 949.800 2136.270

7603 12,013 20,255

335 1309 2433

42,100 44,281 46,433

1.082 1.287 1.566

186.37 300.97 601.80

9028.952 11769.760 13631.000

3717.136 7656.724 8964.710

32,327 46,493 57,884

3230 3730 4345

48,281 51,060 52,404

2.000 2.541 3.028

1219.95 1961.87 2641.85

18210.306 31879.566

10604.392 16743.716

63,725 137,181

4836 8071

53,902 58,623

3.585 5.675

3452.26 5892.68

economies when compared to the degrees of economies obtained by fixed-effects models (Table 2). However, these two models (Models II and III) require a large number of observations for each DMU (corresponding to several time instants) to achieve accurate

inefficiency estimates (Battese and Coelli, 1988), which is not true in this paper since each entity only has six observations that correspond to the study period (i.e., six years). Furthermore, they assume that inefficiency does not vary over time, not allowing to

283

P. Carvalho et al. / Waste Management 39 (2015) 277–286

Model IV True fixed-effects model (Greene, 2005a)

Model V True random-effects (Greene, 2005a)

Fig. 3. Efficiencies estimated from Models IV and V as function to the amount of municipal waste collected.

analyze the evolution of efficiency levels of organizations throughout the study period. Thus, taking into account these shortcomings from Models II and III, they most likely provide skewed results. In order to obtain the most robust results possible, two other models were applied: Model IV (True fixed-effects – Greene, 2005a) and Model V (True random-effects – Greene, 2005a), which enable us to distinguish unobserved heterogeneity from the inefficiency of the utilities. In addition, these models also enable us to evaluate the evolution of efficiency levels of organizations over time. To distinguish unobserved heterogeneity from inefficiency, these models use an additional error component, which makes it possible to separate the inefficiency from the parameter that reflects the unobserved heterogeneity, thus allowing us to obtain coefficients less affected by heterogeneity and consequently models with less bias. As we have noted earlier, since in the present context case the inefficiency of the utilities does not seem to be related to the regressors, the Model V (True random-effects model) provides the least biased results. This can be confirmed by the magnitude of efficiencies and degrees of economies estimated from this model, which seem quite plausible (Fig. 3 and Table 2). The Model IV (True fixed-effects model) provides very low efficiency levels, which implies, as expected, that the efficiencies are possibly skewed, probably due to the ‘‘incidental parameter’’ problem (Hsiao, 2003) which is a consequence of the number of observations per entity (only six) being low. Furthermore, with respect to the economies estimated from this model, it appears that these are neither probable nor statistically significant, and must thus be discarded. By contrast, Model V (True random-effects model) provides more plausible results. Regarding the economies estimated, this model provides plausible and statistically significant degrees of economies bearing in mind the market structure. In addition, the economies estimated from this model were similar to the degrees of economies obtained by Model I (pooled) and Model III (random-effects model – Pitt and Lee, 1981), which may also indicate that this model provides consistent results (Table 2). From this model, it is also possible to infer that the sector is characterized by an average cost inefficiency level of about 15% (Fig. 3). In light of these considerations, it is assumed that Model V (True random-effects model – Greene, 2005a) provides the least biased results. From the results obtained by Model V, the waste collection service sector in NSW presents high EOD in unsorted municipal waste collection. However, this result was only demonstrated for median and large utilities (Table 2), implying utilities which collect more than 3700 tons/year of unsorted municipal waste (utilities with dimension above the 40th percentile), corresponding to more than 12,000 inhabitants (Fig. 4 and Table 3). Only for these utilities are EOD for unsorted municipal waste collection statistically

Fig. 4. Estimated degrees of EOD for unsorted municipal waste collection service (top) and for recycling waste collection service (below) as function of the utilities size and corresponding confidence intervals (shaded) to a 95% confidence interval.

significant. For utilities providing waste collection services to less than 12,000 inhabitants, the hypothesis of constant returns to output density for unsorted municipal waste collection cannot be rejected. However, bearing in mind the sector configuration, it is very likely that high EOD are also found for unsorted municipal waste collection service, not only for the large utilities but also for the small utilities. This means that there are advantages in terms of cost for all utilities to collect larger amounts of unsorted municipal waste. Concerning ECD, according to the results, the hypothesis of constant returns to customer density for small utilities which provide municipal waste collection services to less than 12,000 inhabitants (40th percentile) (Fig. 5 and Table 3) cannot be rejected, and there are diseconomies ECD for larger utilities. This means that for the utilities providing the waste services to more than 12,000 inhabitants there are no advantages to increasing proportionally the quantities of unsorted and recycled waste and the number of customers within their current intervention area. Regarding ESize, our results demonstrate that in the municipal waste collection service in NSW there are diseconomies of size for median and large utilities, corresponding to utilities providing the

284

P. Carvalho et al. / Waste Management 39 (2015) 277–286

Fig. 5. Estimated degrees of ECD as function of the utilities size and corresponding confidence intervals (shaded) to a 95% confidence interval.

Fig. 7. Cost complementarity between the outputs: amount of household unsorted municipal waste collected and the amount of recycling waste collected as function of the size utilities and corresponding confidence intervals (shaded) to a 95% and 90%.

between these two outputs – has negative average values, they are not statistically significant different from zero at a 95% confidence level. However, for a 90% confidence level, condition (6) takes statistically significant negative values but only for utilities larger than those corresponding to the 60th percentile size (Fig. 7 and Table 3); that is, only for utilities that collecting more than about 9000 tons/year of unsorted municipal waste, or providing the collection service to more than 32,300 inhabitants. Accordingly, it is argued that there are economies of scope in the joint provision of these two services, especially for utilities providing collection services to more than 32,300 inhabitants.

6. Conclusions Fig. 6. Estimated degrees of ESize as function of the size utilities and corresponding confidence intervals (shaded) to a 95% confidence interval.

municipal waste collection service over 20,255 inhabitants or collect more than about 6100 tons/year of unsorted municipal waste (Fig. 6 and Table 3). This means that these utilities should reduce their intervention area to become more cost efficient. According to our results, the optimal size of utilities corresponds roughly to dimensions corresponding to provide the municipal waste collection service to about 12,000 inhabitants (40th percentile), since it is to this dimension that ESize is closest to unity (Fig. 6). This implies that the utilities in the sector should move towards this size to become more cost efficient. In other words, utilities providing municipal waste collection services to less (more) than 12,000 inhabitants should proportionally increase (decrease) the amount of unsorted, recycling waste collected, customers and their intervention area to become more cost efficient. Finally, we also sought the presence of economies of scope in the joint provision (by the same entity) of unsorted municipal waste and recycling waste collection services. This was accomplished by estimating the cost complementarity between the outputs: the amount of household unsorted municipal waste collected and the amount of recycling waste collected. It is known that a sufficient condition for the existence of economies of scope is for the cost function to provide evidence of a weak cost complementarity between each pair of outputs; that is, if cost function (C) meets the condition (5). Although there is some evidence of the existence of economies of scope in the joint provision of unsorted municipal waste and recycling waste collection services, our results do not demonstrate this at a 95% confidence level. This is because whereas condition (6) – which represents the existence of cost complementarity

Several conclusions flow from the empirical analysis conducted in this paper with significant public policy implications for structural change to the waste collection sector in NSW. In the first place, we have established that in general ‘bigger is not better’ in waste collection. Put differently, it is not efficient in terms of costs that the large waste collection utilities increase their size in NSW, or that they increase their intervention area through mergers or other forms of structural reform. Indeed, the optimal size of these utilities was estimated to be roughly the median size of the existing utilities in the sector, corresponding to utilities which provide municipal waste collection services to populations in the range of 12,000–20,000 inhabitants. For utilities servicing larger populations, we found diseconomies of size, indicating that these utilities should proportionally decrease the amount of unsorted, recycling waste collected, customers and their intervention area to become more cost efficient. Secondly, our results showed high EOD for unsorted municipal waste collection, especially for the median and large utilities (i.e., utilities which collect waste from more than 12,000 inhabitants). This means that there are economic advantages in terms of costs for these utilities to collect larger amounts of unsorted municipal waste. However, with respect to recycling waste collection, EOD was only apparent for small utilities providing services to less than 6600 inhabitants, thereby not rejecting the hypothesis of constant returns to EOD for larger utilities. Finally, some evidence of the existence of economies of scope in the joint provision of unsorted municipal waste and recycling waste collection services was established. However, the presence of this kind of economies was demonstrated only for larger utilities which provide collection s to more than 32,300 inhabitants (for a 90% confidence level). This suggests that a possible increase in the size of median utilities, through mergers, could result in gains due to economies of scope, but at the cost of a loss in ESize.

285

P. Carvalho et al. / Waste Management 39 (2015) 277–286 Table A.1 Estimation results of the frontier cost function models. Time-invariant inefficiency models

Constant bU bR bPop bArea bPop2 bArea2 bU_Pop bU_Area bR_Pop bR_Area bPop_Area

cSP_OP cSP_OP2 Vsigma bU bPop Constant

rU rV

Model I Pooled model

Model II Fixed-effects model (Schmidt and Sickles, 1984)

Model III Random-effects model (Pitt and Lee, 1981)

13.223*** (0.196) 0.145*** (0.045) 0.027*** (0.009) 0.988*** (0.043) 0.023 (0.024) 0.111* (0.066) 0.042*** (0.011) 0.0098 (0.026) 0.019 (0.024) 4.1e05 (0.006) 0.012*** (0.004) 0.061** (0.027) 0.816*** (0.048) 0.024 (0.015)

13.406*** (0.399) 0.031 (0.022) 0.005 (0.008) 0.536 (0.414) 0.827** (0.363) 0.018 (0.291) 0.278* (0.143) 0.014 (0.016) 0.02 (0.018) 0.004 (0.005) 0.001 (0.004) 0.012 (0.118) 0.972*** (0.04) 0.004 (0.013)

12.584*** (0.074) 0.024 (0.019) 0.001 (0.01) 1.088*** (0.046) 0.043 (0.043) 0.226*** (0.064) 0.068*** (0.018) 0.013 (0.011) 0.033*** (0.012) 0.011 (0.007) 0.009* (0.005) 0.07*** (0.027) 1.025*** (0.047) 0.001 (0.014)

0.263*** (0.033) 0.483*** (0.049) 1.27*** (0.048) 0.051 (6.707) 0.578*** (0.009)

0.941 0.447

k Log-likelihood * ** ***

Time-varying inefficiency models

751.9893

0.678*** (0.065) 0.473*** (0.013) 1.433*** (0.062) 737.1598

Model IV True fixed-effects model (Greene, 2005a,b)

Model V True random-effects (Greene, 2005a,b)

0.26*** (0.054) 0.004 (0.009) 0.525 (0.508) 0.273 (0.448) 0.031 (0.447) 0.04 (0.163) 0.074** (0.033) 0.054** (0.024) 0.011* (0.006) 0.005 (0.003) 0.166 (0.2) 0.797*** (0.052) 0.014 (0.013)

13.048*** (0.074) 0.203*** (0.049) 0.001 (0.009) 0.958*** (0.056) 0.031 (0.038) 0.055 (0.079) 0.038** (0.017) 0.044 (0.028) 0.035 (0.023) 0.009 (0.006) 0.005 (0.003) 0.03 (0.032) 0.837*** (0.049) 0.003 (0.012)

0.548*** (0.074) 1.392*** (0.202) 2.991*** (0.248) 0.184*** (0.021) 0.369*** (0.035)

0.43*** (0.054) 1.028*** (0.142) 2.269*** (0.145) 0.209*** (0.064) 0.421*** (0.38)

289.7215

572.5059

Statistically significant at the 10% level. Statistically significant at the 5% level. Statistically significant at the 1% level.

Appendix A See Table A.1. References Abrate, G., Erbetta, F., Fraquelli, G., Vannoni, D., 2012. The costs of disposal and recycling: an application to Italian municipal solid waste services. Reg. Stud., 1– 14 Aigner, D.J., Lovell, C.A.K., Schmidt, P., 1977. Formulation and estimation of stochastic frontier production function models. J. Econom. 6, 21–37. Antonioli, B., Filippini, M., 2002. Optimal size in the waste collection sector. Rev. Ind. Organ. 20, 239–252. Battese, G.E., Coelli, T.J., 1988. Prediction of firm-level technical efficiencies with a generalized frontier production function and panel data. J. Econom. 38, 387– 399. Battese, G.E., Coelli, T.J., 1992. Frontier production functions, technical efficiency and panel data: with application to paddy farmers in India. J. Prod. Anal. 3, 153– 169. Battese, G.E., Coelli, T.J., 1995. A model for technical inefficiency effects in a stochastic frontier production function for panel data. Empir. Econm. 20, 325– 332.

Bel, G., Fageda, X., 2010. Empirical analysis of solid management waste costs: some evidence from Galicia, Spain. Resour. Conserv. Recy. 54, 187–193. Broitman, D., Ayalona, O., Kanb, I., 2012. One size fits all? An assessment tool for solid waste management at local and national levels. Waste Manage. 32, 12201979-1988. Callan, S., Thomas, J., 2001. Economies of scale and scope: a cost analysis of municipal solid waste services. Land Econ. 77, 548–560. Carvalho, P., Marques, R., 2014. Economies of size and density in municipal solid waste recycling in Portugal. Waste Manage. 34, 12–20. Caves, D.W., Christensen, L.R., Tretheway, J.A., 1984. Economies of density versus economies of scale: why trunk and local service airline costs differ. Rand. J. Econ. 15, 471–489. Chambers, R.G., 1988. Applied Production Analysis. Cambridge University Press, Cambridge. Daraio, C., Simar, L., 2007. Advanced Robust and Nonparametric Methods in Efficiency Analysis. Methodology and Applications. Springer, New York. DECCW, 2010. Review of Waste Strategy and Policy in New South Wales. Department of Environment, Climate Change and Water, Sydney. DLG, 2013. Comparative Information on NSW Local Government: Measuring Local Government Performance 2011/12. Premier & Cabinet, Department of Local Government. Feo, G., Malvano, C., 2012. Technical, economic and environmental analysis of a MSW separate kerbside collection system applied to small communities. Waste Manage. 32, 1760–1774.

286

P. Carvalho et al. / Waste Management 39 (2015) 277–286

Folz, D., 1999. Municipal recycling performance: a public sector environmental success story. Public Admin. Rev. 59, 336–345. Greene, W., 2005a. Reconsidering heterogeneity in panel data estimators of the stochastic frontier model. J. Econom. 126, 269–303. Greene, W., 2005b. Fixed and random effects in stochastic frontier models. J. Prod. Anal. 23, 7–32. Greene, W.H., 2008. The econometric approach to efficiency analysis. In: Fried, H.O., Lovell, K.C.A., Schmidt, S.S. (Eds.), The Measurement of Productive Efficiency and Productivity Growth. Oxford University Press, New York. Hall, J., Jones, L., 1973. Costs of solid waste management in rural Texas communities. South J. Agric. Econ. 5, 115–119. Henningsen, A., Henning, C., 2009. Imposing regional monotonicity on translog stochastic production frontiers with a simple three-step procedure. J. Prod. Anal. 32, 217–229. Hirsch, W., 1965. Cost functions of an urban government service: refuse collection. Rev. Econ. Stat. 47, 87–92. Hsiao, C., 2003. Analysis of Panel Data, second ed. Cambridge University Press. Kumbhakar, S.C., Lovell, C.A.K., 2000. Stochastic Frontier Analysis. Cambridge University Press, Cambridge.

Meeusen, W., van den Broeck, J., 1977. Efficiency estimation from Cobb–Douglas production functions with composed error. Int. Econ. Rev. 18, 435–444. Pitt, M., Lee, L., 1981. The measurement and sources of technical inefficiency in the Indonesian weaving industry. J. Dev. Econ. 9, 43–64. Santibañez-Aguilar, J.E., Ponce-Ortega, J.M., González-Campos, J.B., Serna-González, M., El-Halwagi, M.M., 2013. Optimal planning for the sustainable utilization of municipal solid waste. Waste Manage. 33, 2607–2622. Schmidt, P., Sickles, R., 1984. Production frontiers and panel data. J. Bus. Econ. Stat. 2, 367–374. Simões, P., De Witte, K., Marques, R., 2010. Regulatory structures and operational environment in the Portuguese waste sector. Waste Manag. 30, 1130–1137. Stevens, B., 1978. Scale, market structure and cost of refuse collection. Rev. Econ. Stat. 60, 438–448. Woodbury, K., Dollery, B., Prasada, R., 2003. Is local government efficiency measurement in Australia adequate? An analysis of the evidence. Public Product. Manag. Rev. 27, 77–91. Worthington, A., Dollery, B., 2001. Measuring efficiency in local government: an analysis of New South Wales municipalities’ domestic waste management function. Policy Stud. J. 29 (2), 232–249.