An analysis of the relative efficiency of wastewater utilities in non-metropolitan New South Wales and Victoria.

Byrnes, Joel ; Crase, Lin ; Dollery, Brian 等

1. INTRODUCTION

Several recent inquiries into Australian local government have claimed that a general degree of managerial incompetence, especially in asset management, coupled with conflict-riddled elected councils and concomitant policy deadlock can partly explain the perilous state in which the sector now finds itself (see, for instance, Allan (2006) and Dollery et al. (2007)). It follows that reform of local council governance arrangements to better reflect corporate-style managerial structures may represent a partial solution to these problems in contemporary Australian local government. First, the highly-skilled managers may be more willing to consider a career in local government if common managerial techniques applied in both the public and private sectors. Second, less scope for ongoing political interference by elected councillors in the management of local public service delivery may provide some comfort to those more accustomed to profit maximization goals. This begs the question as to the role of governance arrangements in the relative efficiency of public sector enterprises. If a corporate-style structure can be shown to lead to greater efficiency, then wholesale reform of local government service delivery arrangements may be warranted.

It is against this background that the key research questions of this paper are cast. We examine the relationship between institutional structure and the economic efficiency of urban wastewater utilities in regional New South Wales (NSW) and Victoria after having controlled for a number of exogenous factors. As we shall see in the following section, a period of reform in the governance of local government service provision in the state of Victoria presents an ideal framework in which to test our hypothesis through comparison of the relative efficiency of utilities in each state.

The paper proceeds as follows. In Section 2 some structural features of the urban wastewater sector are considered as background to the empirical investigation. Section 3 outlines the econometric technique to be employed in measuring relative efficiency, while section 4 serves to highlight the paucity of academic studies that have investigated relative efficiency in the industry. Methodological and data considerations are discussed in section 5, followed by a presentation of the results of this study in section 6. Implications for policy and concluding remarks are offered in section 7.

2. THE STRUCTURE OF URBAN WASTEWATER PROVISION IN REGIONAL AUSTRALIA

For the vast majority of the last century, the provision of urban wastewater services in both NSW and Victoria was a function of local government or alternatively water boards established by neighbouring councils. This continues to be the case in NSW, where water and wastewater services provided outside of the state capital (Sydney) and two satellite regions (the Central Coast and Hunter districts) are largely the responsibility of councils. In Victoria, widespread microeconomic reform throughout the early 1990s by the (then) Kennett state government resulted in responsibility for water and wastewater provision being transferred to regional boards, appointed by and responsible to the state government. Eighteen regional districts were established (Smith, 2004); a substantial rationalization of the sector which at one point had no less than 400 bodies with some role to play in the regulatory framework (World Bank, 2004). The standard argument based on the benefits arising from scale economies and a more business-like structure was advanced as justification for the reform (Vince, 1997).

In one sense it might be argued that this represents the main point of difference between the institutional structure of urban water and wastewater provision in the two states. While a series of local government amalgamations have since taken place in NSW (Dollery et al., 2006), reducing the number of councils with water and wastewater responsibilities, the number of utilities providing those services in NSW is still around five times greater than that in Victoria. Perhaps of most significance, the regional water authorities in Victoria are directly regulated by an independent competition watchdog (the Essential Services Commission), while councils in NSW are indirectly monitored by a state government department (Department of Water and Energy). Furthermore, while the executive of Victorian utilities is focused on running a water and wastewater business, the managers of NSW utilities can potentially be distracted by the broader concerns of local government operations and, of course, local politics.

The policy catalyst for the wide-ranging reforms in Victoria was a nationwide focus on microeconomic reform arising from the so-called 'National Competition Policy' (Sadler, 1998). A substantial portion of the reform agenda focused on the activities of Government Business Enterprises, and in particular, on increasing their economic efficiency. Urban water utilities were regulated as local monopolies in need of oversight in order to curb excess.

A separate but parallel program of reform was underway in the water policy arena, known as the Water Resources Policy (WRP), formulated by the Council of Australian Governments (CoAG). (2) Urban water issues appeared somewhat belatedly, and the intent of the WRP was to be consistent with NCP reforms in that arena. By 2004, a re-statement of the WRP was announced--the National Water Initiative (NWI). Rural water reform was the main aim of this policy. However, a relatively small section addressed urban water reform, and in particular, the performance of urban water and wastewater utilities.

Among other things, the states agreed to develop a nationally consistent framework for the benchmarking of pricing and service quality for metropolitan, non-metropolitan and rural water delivery agencies. In implementation, this has resulted in slight changes to a number of existing performance reports with the aim of bringing uniformity to the definitions of the performance measures, to enable comparisons among the states. The National Water Commission (NWC) released the first nationwide performance benchmarking reports in May 2007 (NWC, 2007a; 2007 b).

Utilities were segregated according to size (measured by the number of connected properties a utility serves). Those utilities servicing in excess of 50,000 connections were deemed 'Major Urban Utilities', while utilities responsible for between 10,000 and 50,000 connected properties were classified as Non-Major Urban Utilities. The next report will combine the two, since the small utilities will be required to report accurately on the same criteria that applied to large utilities in 2007.

A falling of the National Performance reporting framework is that it relies on partial performance indicators, expressed in absolute terms. (3) A number of authors have established the limits of this approach (see Dollery et al., 2006 for a summary), since one utility may be the benchmark on one indicator and exhibit only modest performance on another indicator. In this paper we calculate the relative efficiency (or performance) of wastewater utilities using a technique that accommodates multiple performance indicators. The following section outlines the econometric technique that was employed.

3. ECONOMETRIC TECHNIQUE

3.1 DEA as a measure of relative performance

Attempts at relative performance (or efficiency) measurement generally fall into two broad categories; Stochastic Frontier Analysis (SFA) and Data Envelopment Analysis (DEA). The concept of relative efficiency in economic analysis refers to the efficiency with which different organizations use input factors to produce an output. This allows the analyst to compare different organisations with respect to their degree of productive efficiency. Productive or technical efficiency refers to the efficacy with which a firm transforms inputs into outputs and it must be differentiated from allocative efficiency which refers to the allocation of resources between different uses.

Under the SFA approach, the parameters of a given functional form are estimated with the aim of measuring relative firm efficiency with reference to the estimated production frontier. The term 'stochastic' points to an allowance for both technical (as opposed to allocative) inefficiency (deterministic) and matters outside the control of a firm (non-deterministic) (Coelli et al., 2005).

By contrast, DEA makes no assumptions regarding the parameters of the production frontier, utilizing mathematical programming to determine the frontier as a function of the dataset itself. A hull is constructed around the data, and this is assumed to be the efficient frontier (Zhu, 2003). Firms can produce within and on the frontier, but not beyond it. In the parlance of production economics, the frontier is said to represent the feasible set of production points and equates to the observed 'best-practice' benchmark against which firms within the industry are judged.

DEA was adopted for this study since SFA would require the imposition of a number of assumptions regarding the shape of the production frontier and given the paucity of research to guide specification, it was considered prudent to employ DEA. Notwithstanding the advantages of DEA, a choice of this form carries costs. DEA is an entirely deterministic model, necessitating additional econometric steps if one wishes to account for stochastic and exogenous influences. Furthermore, incorporating the extraneous information into the DEA specification is not a particularly flexible process, requiring a number of a priori assumptions to be imposed upon the direction in which factors influence relative efficiency (Coelli et al., 2005b).

DEA calculations generally result in three interconnected measures of relative efficiency. The first is 'overall' efficiency, which can be decomposed into 'pure' efficiency and 'scale' efficiency, where scale efficiency is related to the volume of output. Assume data are obtained relating to inputs K and outputs M for a sample of N firms. For the firm these can be represented by the column vectors [x.sub.i] and [y.sub.i], respectively. The dataset consists of the input vector KxN = X and output vector MxN = Y. The following model seeks to minimize input consumption while leaving output constant.

[min.sub.[theta],[gamma]] [theta], s.t. -[y.sub.i] + Y [gamma] [greater than or equal to] 0, [theta][x.sub.i] - X [gamma] [greater than or equal to] 0, [gamma] [greater than or equal to] 0. (1)

The minimization task is achieved by [theta] while [gamma] is a Nx1 vector of constants that locates points on the frontier. Overall technical (in)efficiency is given by scores obtained in [theta], relative to [gamma]. Note that [theta] is the objective function, and operates only with respect to inputs. The linear programming problem must be solved N times, once for each firm in the sample.

Thus far it has been assumed that a given increase in inputs will result in an equi-proportionate increase in output, implying constant returns to scale. Under constant returns to scale output per unit of input remains unchanged. However, countless empirical studies have shown that certain industries benefit or suffer from variable returns to scale. Under variable returns to scale, output per unit of input either increases (i.e. economies of scale) or decreases (i.e. diseconomies of scale). To assume an industry operates under constant returns to scale, when in fact some relative efficiency could be gained through variation in scale, gives rise to the concept of scale inefficiency. DEA can be extended to allow for the calculation of 'pure' technical efficiency devoid of scale effects through the addition of a convexity constraint N1'[gamma] = 1 to provide:

[min.sub.[theta],[gamma]] [theta], s.t. -[y.sub.i] + Y [gamma] [greater than or equal to] 0, [theta][x.sub.i] - X [gamma] [greater than or equal to] 0, N1'[gamma] = 1, [gamma] [greater than or equal to]. (2)

where N1' is an N x 1 vector of ones. The constraint allows a relatively tighter envelopment frontier that is more convex than that obtained under the assumption of constant returns to scale. As a result, the efficiency scores obtained for the firms under the variable returns to scale model will be greater than or equal to those measured in the constant returns case. Measures of relative scale inefficiency are obtained by taking the ratio of overall to pure efficiency.

4. LITERATURE REVIEW

There is a paucity of relative economic efficiency studies with respect to the activities of urban wastewater utilities. Indeed, the present study would appear to be the first to examine urban wastewater utilities in Victoria, and therefore represents a genuine and timely contribution to the literature. Given the dearth of empirical evidence, we are guided by research on urban water utility efficiency, an excellent synopsis of which can be found in Coelli and Walding (2005a) and for the sake of brevity is not repeated here. Unfortunately, most studies have been in the context of the benefits and costs of public and private ownership of utilities, and as a result are not of direct relevance in this context. Furthermore, a clear pattern of evidence regarding the benefits of each has failed to emerge. However, of the extant literature three studies are worthy of closer examination: Aubert and Reynaud (2005), Woodbury and Dollery (2004) and Coelli and Walding (2005a).

Aubert and Reynaud (2005) investigated the role of regulatory oversight on the relative efficiency of water utilities in Wisconsin, USA. In sum, the authors found a significant relationship between the degree of regulatory oversight and the relative efficiency of water utilities. Those utilities required to provide extensive information to regulators were found to have higher levels of relative efficiency. Since Victorian wastewater utilities are subject to a more stringent form of economic regulation than those in NSW, the findings of efficiency gains from so-called 'hard' (as opposed to soft) regulation have important implications for the regulation of wastewater utilities in NSW and Victoria. The suggestion that there are efficiency gains attached to 'hard' regulation seems a matter well suited to empirical investigation in the current context.

There appear to be only two published studies of relative efficiency in the Australian water and wastewater sectors. Woodbury and Dollery (2004) investigated the relative efficiency of water and wastewater providers in regional NSW, finding that there was scope for general improvement in the performance of the utilities in question, indicated by an average DEA score of around 0.7 for the sample.

Coelli and Walding (2005a) studied the 18 largest urban water providers in Australia. Although this mainly involved an examination of urban water utilities in the Australian capital cities, a number of the utilities were located in regional Victoria. They found that the mean technical efficiency score of the utilities was 0.904, implying that the average utility could have reduced input consumption by 9.6 per cent without reducing output. However, the maj or conclusion was that data of much more robust quality would be required before regulatory bodies could rely upon results from efficiency studies such as theirs, at least as far as it relates to the setting of prices.

From this brief review of the extant literature it seems reasonably clear that there is a need for greater scrutiny of the efficiency of wastewater utilities in Australia. This is somewhat surprising since, as was briefly alluded to in section 2, the sector has undergone 15 years of reform.

5. METHODOLOGY AND DATA

The dataset analysed in this study consists of 14 Victorian (4) and 42 NSW (5) wastewater utilities over the period July 2000 to June 2004. (6) Utilities servicing fewer than 3,000 connections were excluded, to ensure Victorian utilities were compared against NSW utilities of a comparable size. This yielded a balanced panel of 56 observations over four years, generating 224 observations in total.

Although data relating to both labour and fixed capital were available, the input measure, Total Operating Cost, has been intentionally restricted to include only expenses related to the current operation of the wastewater business, such as maintenance of the network, treatment, wages and salaries, administration and energy consumption. Labour was excluded as an input for a number of reasons. First, the measure of labour in Victoria was aggregated across the water and wastewater businesses, while in NSW it was disaggregated. This disparity presented the unenviable task of determining how to disaggregate the Victorian labour data. Second, the data series relating to Victorian labour measures began only in 2003. Third, consultations with representatives from the urban wastewater sector in Victoria revealed that management decisions to vary the labour force were not closely related to the quantity of total wastewater treated (C. Heiner, pers. comm., 27 April, 2007).

Fixed capital was also excluded on a mixture of theoretical and pragmatic grounds. Turning first to theoretical considerations, a number of scholars have previously noted that the infrastructure related to the provision of water and wastewater services is a sunk cost, since it is difficult to conceive putting it to an alternative use (Shed, 2000). If this is so, it calls into question the inclusion of various measures of fixed capital in a DEA model since management are unlikely to seek to minimize this input. Furthermore, while additions to capital through time are likely, the opposite is not. A decline in total wastewater treated is rarely followed by the decommissioning of wastewater mains or the dismantling of pumping and treating infrastructure. Of potentially more relevance to the estimation of relative technical efficiency are current capital expenses incurred as a result of renewals activities, which is captured under operating costs.

Justification on pragmatic grounds relates to the historically poor measurement of the value of infrastructure in NSW local government, (7) made painfully clear by an independent inquiry into the financial sustainability of NSW local government, the so-called Allan report (2006). Considering the widespread lack of confidence in fixed infrastructure values, it was judged prudent to exclude this variable rather than attempt to adjust for the errors in the results. With respect to separate measures of energy and materials consumption, while the NSW data disaggregate operating costs into various classes, including administration, energy and materials, the Victorian data do not. Consequently, it was not possible to include separate input variables for materials and energy.

In order to aid comparison between years, and utilities in each state, the variable was inflated to reflect 2004 nominal values, by applying the headline consumer price index for Melbourne. The use of this less than ideal inflation factor was made necessary by data relating to Victorian wastewater utilities being inflated prior to publication, whereas data for NSW utilities were published in nominal terms.

The two outputs modelled are (1) Total Wastewater Treated and (2) Complaints per 1,000 connections. The constituent parts that form Total Wastewater Treated were similar across both states. Output quality was measured by the number of customer complaints made per 1,000 connections. This was essentially due to this data being almost universally reported, a characteristic not shared by more direct measures of quality.

It was necessary to transform the complaints variable since it was to enter the model as an output. Maximizing complaints is clearly not an objective of utility managers, and the data were modified such that maximizing the vector was akin to minimizing actual complaints. Zhu (2003: 106-7) suggested an approach to transform 'undesirable' outputs for use in DEA models, which was followed here. All data relating to utilities in NSW was sourced from the Department of Energy, Utilities and Sustainability (2005) and VicWater (2005) was the source for data relating to Victorian utilities.

Table 1 reports descriptive statistics for each variable in each of the four years. Two telling patterns emerge from an analysis of the data in this table. First, average total operating costs increased during the period, despite the variable having been adjusted for inflation. Second, average total wastewater treated fell between 2001 and 2004. Combined, this suggests a sharp increase in per unit operating costs over the period.

As mentioned earlier, we specify a Tobit regression model in which the DEA scores generated from the evaluation of equations 1 and 2 are regressed against a set of explanatory variables in an attempt to explain the determinants of relative efficiency. Table 2 outlines the suite of variables thought to influence relative efficiency, and our a priori expectations. They are grouped under the four broad themes contained in Table 2.

5.1 Returns to scale, economies of customer and production density

Although Victorian utilities recorded the proportion of sewage collected from residential customers, data limitations particular to NSW utilities forced the use of residential connections ([z.sub.1]) to the sewerage network. While it would have been preferable to include the actual quantity of tradewaste passing through the treatment plant, the proxy was expected to detect the presence of any significant relationship between relative operational efficiency and a substantial proportion of tradewaste. There was a reluctance to expect a particular sign, since the extent to which tradewaste must be treated at the treatment plant tends to vary with the particular type of industry and the level to which the waste is treated prior to being released into the sewerage network (VicWater, 2005). It is also influenced by the licensing requirements imposed by the environmental regulator. That is, not all wastewater needs to be treated to the same extent before being returned to the environment.

Lloyd (1993: 69) conveyed the additional burden felt by wastewater authorities from treating tradewaste by invoking an example from the now defunct Shepparton Water Board:

 Although the Board services a population of approximately 33,000,
 it estimates that the water and wastewater requirements of major
 food processing industries within its boundaries are such that it
 actually services the equivalent residential population of 650,000
 or 20 times the actual population.

Although it is now common practice for wastewater utilities to levy a tradewaste charge, and for specialized connections to the sewerage network to be made at the expense of the industrial customer, disproportionate tradewaste might still be expected to result in lower relative efficiency.

In this paper, following Garcia and Thomas (2001), we define production density ([z.sub.2]) as the total wastewater treated per customer, with network size and the number of customers held constant, while customer density ([z.sub.3]) is defined as the number of customers, having held the size of the network and production density constant. Our a priori expectations with relation to both are uncertain since Mays and Tung (1992) found that there are decreasing returns in the network (arising from increased customer density), yet considerable returns to scale at the treatment plant (as a result of increased production density).

A dummy variable was included to reflect utility size ([z.sub.4]). Although the specification of the variable returns to scale DEA model should have taken into account scale effects, dummy variables were included to control for the uncertainty associated with the measure of scale employed--the quantity of wastewater treated--rather than a physical measure of network size. This variable may also measure the effect of any increase in regulatory burden imposed on larger utilities. Of course, in analysing the results from the constant returns to scale DEA model, this dummy variable will likely be of crucial importance.

5.2 Treatment and pumping expenses

The major expense arising from operating a wastewater system is that relating to treatment. Accordingly, a range of variables was included to account for differences in the extent to which utilities are required to treat wastewater. The degree to which sewage is treated depends in part on where the resulting effluent is to be discharged. For instance, a utility that discharges effluent into a river that is both of considerable environmental value and is the source of raw water for a town downstream is required to 'produce' effluent of a quality close to that of the receiving environment. In contrast, effluent that is to be discharged from an ocean outfall might only require rudimentary treatment.

A dummy variable ([z.sub.5]) was included for those utilities that treat to the highest standard (tertiary treatment) while dummy variables to account for varying discharge points (land ([z.sub.6]), ocean ([z.sub.7]) and river ([z.sub.8])) were included. Since some utilities discharge to multiple points, some were assigned dummies for more than one discharge location. It is generally expected that those utilities treating to a tertiary standard will incur greater costs, resulting in a lower relative efficiency score. Consequently, it was expected that those discharging to the ocean would have the lowest treatment expenses, resulting in a positive coefficient, and those discharging to land and river would have higher treatment costs, resulting in negative signs for these variables. However, the magnitude of the coefficient was expected to be higher for those discharging to rivers.

Breaks and chokes in sewer mains are a driver of operation expenses since they must be repaired quickly to minimize spills of raw sewage (Jones and French, 1999). To account for this expense, a variable ([z.sub.9]) was included that measures the number of breaks and chokes per 100km of sewerage main. It was included because the majority of breaks and chokes are arguably beyond the direct control of managers. Such incidents usually increase during times of drought as soils shift and put pressure on pipes, and as a result of storm events which cause sewer chokes following the ingress of stormwater. Thus, a degree of uncertainty surrounds the expected sign on this coefficient.

5.3 Climatic effects

Variables to reflect rainfall were not included due to data limitations. Ideally, a variable would have been included to measure large intense rainfall events, since these tend to result in much higher quantities of stormwater being diverted to treatment plants. This rise is as a result of ingress and illegal connections to the sewerage network. Unfortunately the data were not available, and so climate variables were excluded from this analysis.

5.4 Period

The purpose of including dummy variables to represent different time periods ([z.sub.10], [z.sub.11], [z.sub.12]) is to ensure that changes in relative efficiency partially attributable to productivity change are not erroneously reflected in other variables included in the model. Given the increase in the average cost of supplying a megalitre of potable water during the period, a generally negative coefficient was expected on each of the time related dummy variables.

5.5 Institutional effects

A dummy variable to identify Victorian utilities ([z.sub.13]) was included to determine whether, as a group, Victorian wastewater providers were more or less relatively efficient than those in NSW after having controlled for the group of factors contained in Table 3. Since this represents the primary motivation for this research, we formed no a priori expectations.

Multicollinearity tests revealed no evidence of serious multicollinearity between the explanatory variables.

6. TECHNICAL EFFICIENCY RESULTS

Equations 1 and 2 were solved for each utility for each of the four years in the sample. It is important to note that direct comparisons between years are without theoretical basis, since efficiency scores are relative to the best performing utilities in each year. Descriptive statistics are reported in Table 3.

The results suggest there was considerable scope for relatively more efficient use of inputs. In the year in which average overall technical efficiency for utilities in both states was at its highest (2004), the 'average' utility could have reduced input use by 44.3 percent while leaving output unchanged. Only one utility (Gunnedah in NSW) was the benchmark in all four years in terms of overall efficiency, although Orange (also in NSW) appeared on the frontier twice. In terms of pure technical efficiency, Gunnedah was joined by the Victorian utilities Gippsland, Lower Murray and Westernport in forming the frontier in all four years. It is interesting to note that only Gippsland is from the 'Very Large' size category. With respect to scale efficiency, the results suggest a relatively high degree of scale efficiency, although utilities in NSW have a considerable advantage in this respect. Once again Gunnedah was the only benchmark utility in all four years.

It is interesting to note that there is a consistent pattern of higher relative overall technical efficiency for Victorian utilities from 2002 onward. This finding suggests that Victorian wastewater utilities, as a group, were at an advantage during the period. Of particular note, Victorian utilities were substantially more efficient in terms of relative pure technical efficiency, however this was offset by relative scale inefficiency. This result suggests the benefits of the governance arrangements in place throughout Victoria were muted by inefficiencies derived from excessive size.

6.1 Explaining Technical Efficiency Results

Three separate Tobit regression equations were estimated in order to investigate the determinants of overall, pure technical and scale efficiency. Using a technique known as 'testing down' (Kennedy, 2003), the suite of explanatory variables statistically related to each of the measures of relative efficiency were determined. In order to test the joint significance of each final model, a Wald test was conducted with the null hypothesis of joint insignificance of the variables. The results are reported in Table 4.

The results suggest that a higher proportion of residential connections is associated with higher overall and pure technical efficiency, suggesting industrial connections to the sewer network may entail relatively higher input use. The positive coefficient on the variable for production density for all three measures of relative efficiency implies some costs to utilities as a result of policies to reduce per capita indoor water consumption. However, the respective magnitudes call into question the economic significance of the results.

The results relating to the treatment and discharge variables are mixed. The sign and magnitude of the tertiary treatment co-efficient were expected. In contrast, however, the negative sign for ocean discharge is perplexing, since treatment of wastewater for disposal by this method is typically rudimentary. It may be that factors relating to the coastal location of these utilities are being captured. In a similar vein, the positive coefficient for both land and river discharge in terms of scale efficiency may reflect certain characteristics of utilities situated inland.

The result of most interest, however, relates to the dummy variable identifying Victorian utilities. Noting that the dummy variable for size was found to be insignificant in this specification, Victorian utilities were, on average, 22 percent more purely technically efficient. With respect to relative scale efficiency, Victorian utilities were found as a group to be, on average, 14 percent less scale efficient than their counterparts in NSW. This is confirmed by the seven percent advantage held by Victorian utilities in terms of overall technical efficiency. This group of results has significant policy implications and we address these in the following section.

7. CONCLUDING REIVIARKS AND POLICY IMPLICATIONS

The significance of this paper can be argued along two main fronts. First, this study represents the first analysis of the economic efficiency of regional urban wastewater utilities in NSW and Victoria. Second, to the best of our knowledge, this is the first analysis of the contribution differing governance structures make to relative (in) efficiency in the Australian water context. In combination, these two aspects of the study represent genuine contributions to the literature. Furthermore, in the context of the newly-established national performance reporting arrangements for water and wastewater utilities in Australia, the research establishes a benchmark against which future analysis of urban wastewater utilities can be measured. We noted two main policy implications from the results presented in section 6.

An unexpected finding from this study was the positive correlation between higher proportions of wastewater connections to residential customers and relative efficiency. While it is clearly not sensible to suggest utilities limit the proportion of wastewater treated from industrial customers in order to improve relative efficiency, the result should be considered by regulators and policy makers when considering the relative performance of urban wastewater utilities in regional locations. This also points to the need for councils and state governments to re-evaluate the net benefits of attracting industry to their jurisdiction and the form and quantity of incentives offered to attract their patronage.

The most important finding in the context of this paper relates to the disparity in relative efficiency scores between wastewater utilities in NSW and Victoria. Wastewater utilities in Victoria were found to be 22 percent more pure technically efficient when compared to utilities in NSW of a similar size. Why this was so cannot be deduced from this study. However, it could be hypothesized to have been a consequence of a number of related factors. First, the composition of the boards of Victorian utilities during the period was a function of relative expertise, rather than a proportional representation of the local government area each utility served. It might be argued that this contributed to a higher degree of managerial competence within Victorian utilities, due in part to the tendency for local government water utility managers in NSW to have an engineering background. Strategic decisions made by the Victorian utility boards may be less likely to be framed within an engineering paradigm, given the diversity of backgrounds of board members, diluting the propensity to 'gold-plate' infrastructure.

Second, skilled managers may be relatively more attracted to Victorian utilities due to the prospect of reporting to a board, rather than the general manager of a council, and dealing with a broader set of stakeholders, rather than simply within local government. In other words, the relatively more corporate structure may attract professionals comfortable in that environment. The implication of this assumption is that relatively more skilled employees are attracted and retained by Victorian utilities, and less so by NSW councils. The relatively poor results for NSW utilities may also suggest that the proximity of elected officials (i.e. councillors) may have resulted in some diversion of attention or resources to projects that did not constitute an efficient use of resources.

However, an interesting trade-off appears to be present. While the generally bigger utilities in Victoria appear able to attract better management expertise, giving rise to technical efficiencies, set against this is the loss of scale efficiency, insomuch as the results suggest that Victorian utilities exceed 'optimal' size. This finding adds weight to the argument that 'bigger is not better' in local public service delivery (see Dollery et al., 2007), with the obvious caveat that this result is confined to wastewater services.

These results provide support for the argument that governance arrangements are important in delivering relative efficiency gains in public service provision. More specifically, policy makers in NSW may consider reform of wastewater provision in NSW. For example, utilities with more than 10,000 connections could be required to separate from local government, following adequate compensation from the state government, to form statutory authorities owned by the state government. To mimic the Victorian structure, each authority could be governed by a board, based on relevant expertise, rather than council representation. The board would be responsible to the relevant state government minister, through a license that established the conditions by which the authority would be permitted to operate.

REFERENCES

Allan, P. (2006) Are Councils Sustainable? Final Report: Findings and Recommendations. Independent Inquiry into Local Government Inquiry (LGI), NSW Local Government and Shires Association: Sydney.

Aubert, C. and Reynaud, A. (2005) The impact of regulation on cost efficiency: an empirical analysis of Wisconsin water utilities. Journal of Productivity Analysis, 23(3), pp. 383-409.

Byrnes, J. D. (2007) Putting Water to Work: A Study of the Relative Economic Efficiency in the Urban Water and Wastewater Sectors of Regional New South Wales and Victoria, unpublished PhD thesis, University of New England Armidale.

Coelli, T. and Walding S. (2005a) Performance Measurement in the Australian Water Supply Industry. Centre for Efficiency and Productivity Analysis: St. Lucia, Queensland:

Coelli, T., Prasada Rao, D.S., O'Donnell, C.J. and Battese, G. (2005b) An Introduction to Efficiency and Productivity Analysis. Kluwer Academic Publisher: Massachusetts.

Department of Energy, Utilities and Sustainability (DEUS). (2005) 2003-04 Water Supply and Sewerage Benchmarking Report. DEUS: Sydney.

Dollery, B.E., Crase, L. and Johnson, A. (2006) Australian Local Government Economics. UNSW Press: Sydney.

Dollery, B.E., Byrnes, J.D. and Crase, L. (2007) An analysis of the new perspective on amalgamation in Australian local government. Working paper 02-2007. Centre for Local Government, University of New England: Armidale. URL: http://www.une.edu.au/clg/working-papers/03-2007.pdf.

Garcia, S. and Thomas, A. (2001) The structure of municipal water supply costs: Application to a panel of French local communities. Journal of Productivity Analysis, 16, pp. 5-29.

Heiner, C. (2007) Personal communication between Dr Joel Byrnes and Craig Heiner, North East Water, on 21 May, 2007, Albury.

Jones, M. and French, R. (1999) Local Government Engineering in Australia. Federation Press: Leichhardt.

Kennedy, P. (2003) A Guide to Econometrics (5th edition). The MIT Press: Cambridge, MA.

Lloyd, C. (1993). The delivery of sewerage and waste treatment services in Australia. In M. Johnson and S. Rix (eds), Water in Australia: Managing Economic, Environmental and Community Reform. Pluto Press: Leichhardt, Australia.

Mays, L. and Tong, Y. (1992) Hydrosystems Engineering and Management. McGraw-Hill: New York.

National Water Commission (NWC) (2007a) National Performance Report 2005-06: Major Urban Water Utilities. Water Services Association of Australia: Melbourne.

National Water Commission (NWC) (2007b) National Performance Report 2005-06: Non Major Urban Water Utilities. Water Services Association of Australia: Melbourne.

Sadler, R. (1998) The Australian experience: Managing a non-metropolitan urban water utility--paradigm shifting towards a new mindset. International Journal of Public Sector Management, 11(7), pp. 596-610.

Sheil, C. (2000) Running the risks: The rationalisation of Australia's water. Australian Journal of Public Administration, 59 (3), pp. 11-21.

Smith, D.I. (2004) Water in Australia: Resources and Management. Oxford University Press: Melbourne.

VicWater (2005) Victorian Water Review 2003-04. Victorian Water Industry Association Inc: Melbourne.

Vince, A. (1997) Amalgamations. In B. E. Dollery and N. A. Marshall (eds), Australian Local Government: Reform and Renewal. Macmillan: Melbourne.

Woodbury, K. and Dollery, B. (2004) Efficiency measurement in Australian local government: The case of New South Wales municipal water services. Review of Policy Research, 21(5), pp. 615-636.

World Bank (2004) The Institutional Economics of Water: A Cross-Country Analysis of Institutions and Performance. Edward Elgar: Cheltenham.

Zhu, J. (2003). Quantitative Models for Performance Evaluation and Benchmarking: Data Envelopment Analysis with Spreadsheets and DEA Excel Solver. Kluwer Academic Publishers: Massachusetts.

Joel Byrnes

Centre for Local Government, University of New England, Armidale, NSW 2351.

Lin Crase

School of Business, La Trobe University, Wodonga, VIC 3689.

Brian Dollery

Centre for Local Government, University of New England, Armidale, NSW 2351.

Renato Villano

School of Business, Economics and Public Policy, University of New England, Armidale, NSW 2351.

(1) Brian Dollery would like to express his gratitude to the Australian Research Council for the financial assistance offered by Discovery Grant DP0770520. The authors would like to thank anonymous referees for helpful comments on an earlier draft of the paper.

(2) COAG comprises the Prime Minister of Australia, the Premiers of the six Australian states, the Chief Ministers of the two territories and a representative of the third tier in the Australian federation, local government.

(3) The intent of the NWC is to express in relative terms in future reports (NWC, 2007a), but precisely what form 'relative' will take is unknown.

(4) The largest Victorian regional urban water authority, Barwon Water, was excluded since it was twice the size of the next largest utility.

(5) A number of NSW utilities were excluded due to data limitations. Data on the performance of water utilities is collected annually by the NSW government and published. Unfortunately, some water utilities sometimes do not complete their annual returns in sufficient detail for publication, or alternatively, some water utilities sometimes may not submit their annual returns in sufficient time for publication. We were thus obliged to omit these water utilities from our efficiency estimations. Given the small number of water utilities excluded as a proportion of the total number of utilities, we do not think that their exclusion has affected our estimates materially. Although there is potential for the exclusion of these utilities from the analysis to introduce bias, determining the extent of bias is always a difficult exercise since it is not possible to produce result based on a sample that includes the excluded utilities. A list of excluded water utilities may be found in Byrnes (2007), Appendix 1B.

(6) The data are from financial years. Henceforth, 2001 refers to July 2000-June 2001; 2002 relates to July 2001-June 2002 and so on.

(7) For a review of the problem in Australian local government data of this kind see Dollery et al. (2006).

Table 1. Descriptive Statistics of Inputs and Outputs

 Standard
Year Description Mean Deviation

2001 Total Operating Cost 3,738,612 3,501,319
 Complaints Index 135 28
 Total Wastewater Treated 4,556 5,270
2002 Total Operating Cost 4,017,957 3,761,066
 Complaints Index 134 28
 Total Wastewater Treated 4,504 5,039
2003 Total Operating Cost 4,218,759 3,937,731
 Complaints Index 76 26
 Total Wastewater Treated 4,402 4,737
2004 Total Operating Cost 4,255,662 3,838,192
 Complaints Index 93 28
 Total Wastewater Treated 4,444 4,989
56 utilities, Large (3,000 -10,000 connections) = 28
of which: Very Large (> 10,000 connections) = 28

Table 2. Variables thought to influence Relative Efficiency

Variable Code Description a priori
 expectation
Returns to Scale, Economies of Customer and Production
Density

Residential [Z.sub.l] Proportion of -
Connections connections classified
 as residential

Production Density [Z.sub.2] Kl of wastewater +
 treated per connection

Customer Density [Z.sub.3] Number of connections +
 per km of main

Very large utility [Z.sub.4] Utility serviced more -
 than 10,001
 connections

Treatment and pumping expenses
 -
Tertiary treatment [Z.sub.5] Dummy to reflect
 majority of wastewater
 treated to a tertiary
 standard

Land discharge [Z.sub.6] Dummy variable to -
 indicate discharge of
 treated effluent to
 land

Ocean discharge [Z.sub.7] Dummy variable to +
 indicate discharge of
 treated effluent to an
 ocean outfall

River discharge [Z.sub.8] Dummy variable to
 indicate discharge of -
 treated effluent to a
 river

Sewer main chokes [Z.sub.9] Number of chokes and -
and breaks main breaks per 100km
 of main

Period

2002 [Z.sub.10] Year specific dummy -
 variable: 2002

2003 [Z.sub.11] Year specific dummy -
 variable: 2003

2004 [Z.sub.12] Year specific dummy -
 variable: 2004
 Institutional effects

Victorian Utility [Z.sub.13] Dummy variable to +
 identify utilities
 located in Victoria

Source: All data was sourced from DEUS (2005) for NSW utilities and
VicWater (2005) for Victorian utilities, with the exception
of 'climate ' variables. Data under that heading was supplied by the
Bureau of Meteorology on request.

Table 3. Descriptive statistics of DEA scores

 Overall Technical Pure Technical
 Efficiency Efficiency

 2001

Statistic All NSW Vic All NSW Vic
Mean 0.487 0.483 0.501 0.569 0.526 0.698
Median 0.459 0.459 0.463 0.516 0.491 0.680
St.Dev. 0.159 0.170 0.119 0.201 0.172 0.227

 2002

Statistic All NSW Vic All NSW Vic
Mean 0.535 0.520 0.580 0.607 0.544 0.796
Median 0.511 0.489 0.560 0.546 0.498 0.823
St.Dev. 0.158 0.167 0.116 0.204 0.168 0.184

 2003

Statistic All NSW Vic All NSW Vic
Mean 0.527 0.515 0.563 0.664 0.610 0.828
Median 0.507 0.491 0.541 0.633 0.546 0.842
St.Dev. 0.167 0.184 0.095 0.209 0.194 0.162

 2004
Statistic All NSW Vic All NSW Vic
Mean 0.557 0.542 0.602 0.629 0.579 0.777
Median 0.535 0.509 0.549 0.581 0.559 0.767
St.Dev. 0.179 0.186 0.146 0.206 0.187 0.190

 Scale Technical
 Efficiency

 2001

Statistic All NSW Vic
Mean 0.879 0.918 0.760
Median 0.947 0.966 0.742
St.Dev. 0.136 0.093 0.172

 2002

Statistic All NSW Vic
Mean 0.904 0.955 0.752
Median 0.959 0.995 0.800
St.Dev. 0.124 0.059 0.143

 2003

Statistic All NSW Vic
Mean 0.808 0.846 0.694
Median 0.808 0.854 0.723
St.Dev. 0.142 0.133 0.103

 2004

Statistic All NSW Vic
Mean 0.902 0.936 0.801
Median 0.961 0.966 0.850
St.Dev. 0.137 0.095 0.185

Table 4. Explaining technical efficiency measures

Variable Description Overall Pure technical

 Coeff. Prob. Coeff. Prob.

[alpha] Constant -0.8377 0.004 -0.7418 0.030

[Z.sub.1] Residential 0.0125 0.000 0.014 0.000
 connections

[Z.sub.2] Production 0.0007 0.000 0.0004 0.027
 density

[Z.sub.5] Tertiary -0.0766 0.000 -0.1097 0.000
 treatment

[Z.sub.6] Land N/A N/A -0.0576 0.039
 discharge

[Z.sub.7] Ocean -0.0531 0.042 -0.0548 0.084
 discharge

[Z.sub.8] River N/A N/A -0.0865 0.012
 discharge

[Z.sub.10] 2002 0.0519 0.064 N/A N/A

[Z.sub.11] 2003 0.0532 0.070 0.0811 0.005

[Z.sub.12] 2004 0.0846 0.004 0.0488 0.089

[Z.sub.13] RUWA 0.0726 0.000 0.2204 0.000

e Error term 0.153 0.000 0.173 0.000

R-squared 0.165 N/A 0.306 N/A

Adjusted R-squared 0.130 N/A 0.273 N/A

Log likelihood 102.023 N/A 75.153 N/A

 Wald tests

F-statistic 407.658 0.000 304.232 0.000
Chi-square 3668.918 0.000 3042.315 0.000

Variable Scale

 Coeff. Prob.

[alpha] 0.7164 0.000

[Z.sub.1] N/A N/A

[Z.sub.2] 0.0004 0.001

[Z.sub.5] N/A N/A

[Z.sub.6] 0.0319 0.045

[Z.sub.7] N/A N/A

[Z.sub.8] 0.1103 0.000

[Z.sub.10] 0.0263 0.148

[Z.sub.11] -0.0637 0.004

[Z.sub.12] 0.0294 0.157

[Z.sub.13] -0.1465 0.000

e 0.106 0.000

R-squared 0.438 N/A

Adjusted R-squared 0.417 N/A

Log likelihood 185.908 N/A

 Wald tests

F-statistic 2618.706 0.000
Chi-square 20949.64 0.000