Measuring efficiency in local government: An analysis of New South Wales municipalities'...

Public sector reform has become an established dimension of policymaking in many developed countries, including Australia. Although the ongoing program of public sector reform in Australia has focused mainly on the Commonwealth government and some state governments, especially Victoria, it is now being applied to local government. Key aspects of this process have been administrative reforms (compulsory competitive tendering and contracting-out), structural reforms (local council consolidations), legislative reforms (fiscal transparency and accountability), and workplace reform (labor market deregulation). Another part of this process has been the collection of new ideas associated with what has come to be known as the "New Public Management." Central ingredients in this movement have been the notion of explicit standards and measures of performance in the public sector, the greater emphasis on outputs rather than inputs, the shift to greater competition in the public sector, an emphasis on private-sector styles of management practice (i.e., "letting managers manage"), and a stress on greater discipline and parsimony in resource use (Hood, 1991). Finally, there is a greater awareness on the behalf of the Commonwealth government of the desirability of promoting efficiency through the system of intergovernmental financial assistance. In common with the other pressures for greater efficiency and effectiveness in local public service provision, this process can be used for accurate and meaningful measures of local government efficiency for the purposes of comparative performance assessment and process benchmarking.

This article is centrally concerned with the evaluation of technical and scale efficiency in New South Wales (NSW) local governments using the nonparametric approach to efficiency measurement. We examine technical and scale efficiency for a single function of Australian local government: namely, domestic waste management services. The article itself is divided into four main parts. The first section outlines the nonparametric approach to efficiency measurement for local public services and provides the formulation of the model employed. The second section provides the specification of inputs and outputs for domestic waste management services, both discretionary and nondiscretionary. The results obtained from this analysis are discussed in the third section. The paper ends with some brief concluding remarks.

Model Formulation

The method used to measure efficiency at the local level is based upon data envelopment analysis (DEA), a mathematical programming approach to frontier estimation pioneered in Charnes, Cooper, and Rhodes (1978), extended in Banker, Charnes, and Cooper (1984), and outlined in Fare, Grosskopf, and Lovell (1994). There are several advantages of the DEA approach in evaluating the efficiency of government service providers (Worthington & Dollery, 2000). These include inter alia its ability to handle the multiple inputs and outputs characteristic of public sector production, especially where it is difficult or impossible to assign prices to many of these factors, and its capacity to incorporate differences in operating environments beyond management control, particularly for purposes of comparative performance assessment and process benchmarking (Steering Committee for the Review of Commonwealth/State Service Provision, 1997). Measuring efficiency in this manner is consistent with both the literature associated with the efficiency analysis of government service providers in general, such as Ganley and Cubbin (1992), Kittelsen and Forsund (1992), Mensah and Li (1993), and Carrington, Puthucheary, Rose, and Yaisawarng (2000), and with the majority of past empirical approaches to efficiency measurement in the local public sector, notably Charnes, Cooper, and Li (1989), Cook, Roll, and Kazakov (1990), Grosskopf and Yaisawarng (1990), Deller (1992), Vanden Eeckaut, Tulkens, and Jamar (1993), and De Borger and Kerstens (1996a).

Figure 1 illustrates the derivation of the efficiency measures found in DEA in the single-input (x), single-output (y) case. As shown, these envelopment surfaces may be either linear, as in the constant returns to scale (CRS) case, or convex, as with variable returns to scale (VRS). The CRS and VRS cases are detailed: The CRS surface is the straight line OICM, and the VRS surface is GABCDEF. For ease of exposition, the interior (or inefficient) councils are represented by point K. The efficiency of any interior point (such as K) is indicated intuitively by the distance between the envelope and itself. In the case of an input orientation, focus falls on maximal movement toward the frontier through the proportional reduction of inputs. For example, using an input orientation and the council depicted by point K, the measure of technical efficiency will be given by hi/hk in the CRS case and by hj/hk in the VRS case. A measure of scale efficiency is provided by the ratio hi/hj. Using an output orientation, the technical efficiency of point K would be given as nk/nm in the CRS case and nk/nl in the VRS case, and the scale efficiency would be provided by nl/nm. Finally, for a council on the envelope surface, as denoted by C, the technical efficiency ratio would be qc/qc for technical efficiency under both VRS and CRS with an input orientation (a value of unity), and the scale efficiency measure in this case would also be qc/qc.

The specific extension of DEA to the multiple-input, multiple-output case was first introduced by Charnes et al., (1978) and extended in Seiford and Thrall (1990). Consider N local councils each producing M different outputs using K different inputs. The envelopment form of the input-orientated DEA linear programming problem is specified as follows:

Specification of Inputs and Outputs

The variables used to provide efficiency measures using the nonparametric methodology are outlined in Table 1. Following Smith and Mayston (1987), Valdmanis (1992), Kooreman (1994), Thanassoulis and Dunstan (1994), and Thanassoulis, Boussofiane, and Dyson (1996), a single function is employed to evaluate DEA as a tool of efficiency analysis in government service provision. The activity selected in the current study is the provision of domestic waste management and recycling services by NSW ending 31 December 1993 (the first year in which statements were prepared under AAS27 Financial Reporting by Local Government) and is obtained from the NSW Department of Local Government (NSWDLG), the NSW Local Government Grants Commission (NSWLGGC), and the Australian Bureau of Statistics (ABS). Descriptive statistics are also provided in Table 1.

The model used to conceptualize local council behavior is a traditional production-based approach. Table 1 details the inputs (both discretionary and nondiscretionary) and outputs for the provision of domestic waste management and recycling services in NSW local government councils. The provision of these services generally is classified as a "community-related" function. This function also usually is acknowledged as a core service of local government, especially since the provision of waste services usually involves a significant proportion of councils' total resources (New South Wales Department of Local Government, 1998). Within the context of NSW local governments' responsibilities, waste is recognized as being composed of four components: (a) domestic waste, (b) council operational waste, (c) commercial and industrial waste, and (d) construction and demolition waste (Independent Pricing and Regulatory Tribunal of New South Wales, 1997, p. 90). While local councils have an important role in managing all four waste streams, they have a primary responsibility in providing what is referred to as the domestic waste management service (DWMS).

An important consideration is that all waste activities in NSW are now subject to the Waste Minimization and Management Act 1995. The underlying principles of the Act are: (a) a 60% reduction in waste disposal by the end of the year 2000 (per capita reduction on 1990 disposal rates); and (b) the establishment of a waste management hierarchy of the following order: (1) avoidance, (2) reuse, (3) recycling and reprocessing, and (4) disposal. The Act also provides that waste services should be coordinated in nominated waste management regions, that councils should adopt efficient waste management practices and policies, and that councils also should operate in accordance with the principles of ecologically sustainable development.1

Two problems immediately arise when calculating the efficiency of DWMS for local governments. First, one problem that potentially may arise here is that waste management services is one of the most frequently "contracted-out" services in the Australian local public sector. However, the shift to accrual accounting and the adoption of a common accounting standard in the form of AAS27 Financial Reporting by Local Government has ensured that all current and capital costs are recognized within the reporting period, whether provided "inhouse" or purchased via contract (the Independent Pricing and Regulatory Tribunal of New South Wales [IPART], 1997 report suggests that where waste services are not contracted out, labor, capital [equipment utilized], overheads, and other costs would add additional dimensions to council performance). Second, whereas all or nearly all local councils in NSW operate waste collection services, only those councils covered by the Waste Recycling and Processing Service NSW (WRPS) have information collected on recyclable material collected and disposal costs. The total sample of 173 NSW local governments accordingly is reduced to 103 individual councils.

A large number of factors are thought to have an impact on the efficiency of waste collection. In common with other local government functions, these may be grouped broadly as: (a) characteristics of the existing service (such as frequency of service); (b) the community's service requirements (including the manner of collection); (c) limitations on the service posed by the environment (such as complexities posed by population density and topography and the influence of garden area, family size, household income, and restaurant usage); (d) council's utilization of various productive factors (including the degree of automation); and (e) other factors (including the extent of green space, and street sweeping and litter bin services) (IPART, 1997). However, the recent IPART (1997, p. 90) inquiry has identified a number of conflicts that make the measurement of efficiency in DWMS particularly problematic.

One example is that there may be a degree of conflict between strictly efficient performance and compliance with the Waste Minimization and Management Act if the cheapest method of waste management is disposal to landfill, yet the Act seeks to minimize disposal to landfill. Another example is associated with councils' recycling efforts and involves ownership of recyclable material. The PART (1997, p. 90) inquiry notes that where a council maintains ownership, any proceeds from the sale of recycled material will offset costs to some degree. Alternatively, where ownership is transferred to a collection contractor, the proceeds should be considered in deriving the cost of the recycling service. Unfortunately, there is no data set available reflecting all factors relevant to calculating DWMS efficiency at the present time.

In terms of nondiscretionary inputs, eight categories are employed. These are: the number of properties receiving DWMS (x1); the occupancy rate (x2) (council population divided by the number of serviced properties); urban density (x3) (urban population divided by the urban residential area); population distribution (x4) (the sum of population centers greater than 200 multiplied by their distance from council headquarters divided by the number of urban properties); and an index of waste disposal costs (x5) (based on the standardized tonnage of garbage collected, the cartage distance to the receiving depot, and the receiving charge at that depot). Once again these measures are identical to those employed by the NSWLGGC to calculate expenditure disability factors in DWMS (see New South Wales Local Government Grants Commission, 1994). The occupancy measure recognizes the variation in DWMS expenditures required for households with a higher than average occupancy rate, the urban density measure indicates the constraints placed on operating machinery in densely populated areas, while the measure of population distribution indicates costs associated with travel and duplication of services in local government areas (LGAs) where population is widely dispersed. As an example, narrow streets (associated with high urban density) may reduce the ability to use large, specialized equipment. Similarly, the extent of on-street parking may reduce the ability to use some automated collection equipment and accordingly increase manual labor requirements. According to the NSWLGGC (1994, p. 55) methodology for calculating standardized unit expenditure for residential garbage services, the largest marginal input requirement for a 1% increase in the contextual variable is for the occupancy rate, followed by disposal costs, and finally, urban density and population distribution.

A comparable study of U.K. local authorities by Domberger, Meadowcroft, and Thompson (1986) used similar variables to add additional dimensions to DWMS efficiency. In their cost function approach, Domberger et al. (1986) employed frequency of collection, density of population units, and distance to disposal points. In common with the present study, Domberger et al. (1986, p. 74) used the number of units serviced rather than population, arguing that "population served seemed less appropriate on a priori grounds (the number of pick-up points is likely to be a more important determinant of costs than the number of people served by the collection service) and this was confirmed by our analysis". However, in contrast with the present study, Domberger et al. (1986, p. 75) argued that "the density of units is likely to have a negative effect on total cost; the proximity of pick-up points and shorter walking distances in areas of high density would suggest that costs should be lower in these areas."

Of these nondiscretionary inputs, one of the most important is the index of waste disposal costs. Given that most Australian garbage is disposed of in landfill sites near or beyond the urban fringe, the cost of transport will vary slightly with the distance of a local council from the landfill site. This may result in some geographic differences in the level of disposal costs (Neutze, 1997, p. 174). However, a more significant contributor to differences in the cost of disposal is the charges at the landfill site. Ideally, these would include the value of the site used for landfill, the environmental impact of these operations, and a scarcity rent associated with the exhaustible nature of these sites. It is also possible that this measure would provide some indication of the propensity of a council's ratepayers to engage in illegal dumping. All other things being equal, higher charges for dumping domestic waste, and the greater the distance to a collection site, the more likely illegal dumping will occur. A commensurate increase in the cost of surveillance by the council could also be expected (Neutze, 1997).

As with the contextual inputs, problems arise when obtaining reliable data on discretionary DWMS inputs and outputs for local councils. The principal difficulty is that the available data usually are not disaggregated sufficiently for the purposes of the analysis. For example, total costs for labor and capital could be listed as separate items, and variables identifying whether the service is provided "in-house" or by "contract," and the degree of automation also could be used. Moreover, there also is considerable diversity among the waste management practices of councils, which in turn influences the specification of outputs. For instance, in 1992 (the latest year for which these figures were collected) of the 72% of councils that offered DWMS, 72% provided "big bins" (240-liter bins, sometimes referred to as "wheelie" bins), 18% provided "normal/other bins" (55liter or any other than "big" bins), and 38% both "big" and "normal/other" bins (New South Wales Department of Local Government, 1993, p. 19).

Similarly, the recycling services offered by councils vary considerably, a condition that may have a dramatic influence on the rate of recycling. For example, the average rate of recycling in urban metropolitan councils was 23.09%, compared to 11.35% in urban fringe councils, 11.43% in urban rural councils, 11.46% in rural agricultural councils, and 10.42% in rural councils with significant growth. As discussed, one reason for this may be differences in the recycling services offered. For instance, of the 23% of councils offering a recycling service, 78% were collected weekly, 10% twice monthly and 2% monthly, (New South Wales Department of Local Government, 1993, p. 19). In ideal circumstances, the vector of discretionary outputs would also include collection quantities, the frequency of garbage service, and place of pick-up (streetfront or within the residence) (Independent Pricing and Regulatory Tribunal of New South Wales, 1997, p. 92). Reliable data on these variables are not available.

Accordingly, the discretionary input employed in DWMS is total collection cost (x6), whereas the three measures of discretionary outputs are: the amount of garbage collected, in kilograms (y1); the amount of recyclables collected (y2) (also in kilograms); and the implied recyclable rate (y3) (recyclable material as a proportion of total garbage collection). Although the specification of these variables is not ideal, especially that concerning outputs, it does effectively serve two purposes.

First, to some extent the collection of garbage is exogenously imposed upon a council by legal requirements. Increasing the volume of garbage collected thereby tends to provide some indication of the council's success in deterring illegal dumping by providing timely and effective collection services, and accordingly maintaining the quality of the environment (Neutze, 1997). Second, the distinction between "recyclable" and "nonrecyclable" domestic waste highlights efforts by councils to constrain the high costs associated with landfill site or incineration, and promote local environmental objectives. Moreover, the absence of a charging system for household garbage that relates to volume has meant that the primary means of limiting the demand for garbage collection in recent years has been education. Neutze (1997, p. 95) has argued that this is an appropriate method for discouraging the excessive use of public disposal facilities since it: [T]akes advantage of the interest of individuals in protecting the natural environment and emphasizes a range of options including composting organic wastes and recycling paper, some plastics, glass and metal cans. In addition, recycling has been encouraged by the free provision of containers for, and free collection of, recyclable materials, and free or subsidized provision of compost containers.

A similar argument has been advanced by Domberger et al. (1986) when the amount of waste paper reclaimed was used as an output in a study of U.K. DWMS cost efficiency. The implied recyclable rate therefore indicates efforts the council has made to promote the recycling of domestic waste, both in the provision of separate collection services and promotion of these services within the community.

The final set of variables (z1 - z5) detailed in Table 1 relates to the Australian Classification of Local Government (ACLG) categories, which in turn are based upon objective geographic/demographic criteria. It is argued that other considerations still may have an influence on a council's efforts to attain an efficient outcome, even after the vector of nondiscretionary inputs is taken into account. For example, in waste management services there may be additional complexities relating to the distance to waste disposal facilities or proximity of this facility to residential areas. If the vector of dummy variables in either of these cases proves to be an insignificant influence on efficient outcomes, then local governments across NSW should be able to be compared solely on the basis of the input/output vector and individual disability factors. Alternatively, evidence of a systematic relationship between one or more ACLG categories may focus the search for excluded disability factors, or analysis of managerial conditions unique to that local government classification.

Empirical Results

The results of the analysis of technical and scale efficiency using local governments' waste management and recycling function is presented in Table 2. The nondiscretionary inputs posited to exert an influence on performance include the number of properties receiving the service, population density, and occupancy rate. The discretionary input is total collection expenditure, while the discretionary outputs are the total tonnage of garbage and recyclable material collected and the implied rate of recycling. This particular model includes nondiscretionary inputs in the efficiency calculations themselves; however, an alternative for future work would be to leave the nondiscretionary variables out of the DEA model and examine them in more detail in a second-stage regression.

As indicated, of the 103 councils examined, 42 (41 %) are judged as being purely technically efficient, while 37 (36%) are scale efficient. The results for pure technical efficiency indicate that, on average, inputs could be reduced to 67.12% of the current level based upon observable best-practice, while the results for scale efficiency suggest that productivity losses due to scale effects account for 15.47% of inputs. However, more councils are either scale efficient or nearly so, with 75% of councils having an efficiency score greater than 97.34%. On the other hand, 50% of councils are less than 75% purely technically efficient when compared to best-practice.

The results for waste management and recycling services indicate that the larger portion of overall technical efficiency is the result of purely technical inefficiency, rather than scale effects. That scale inefficiency that does exist is largely the result of operating at a smaller than optimal scale (53 councils subject to increasing returns-to-scale) as against scale diseconomies. Banker's (1996) tests of returns-to-scale reject the null hypothesis of constant returns-to-scale, and we may conclude that the provision of waste management and recycling services is subject to variable returns-to-scale.

The distribution of waste management and recycling efficiency across the narrowest definition of ACLG categories is presented in Table 3. It should be emphasized that the sample of 103 councils used in this analysis comprises only 59% of all NSW local governments, and relates only to those councils covered by the Waste Recycling and Processing Service NSW (WRPS).

There is also significant variation in the average level of technical and scale efficiency (in brackets respectively) across the broader ACLG categories; urban developed (UD) (0.5757/0.9223), urban fringe (UF) (0.5718/0.9067), urban regional (UR) ((0.6756/0.8950), rural significant growth (RSG) (0.2492/0.2492), and rural agricultural (RA) (0.4951/0.7645). Combined with Table 3 is the suggestion that urban developed councils are generally less efficient, either purely or nearly so, compared to urban regional councils, with regard to both technical and scale efficiencies. Further, scale efficiencies are generally higher in urban fringe councils and lower in rural councils with significant growth, and technical efficiency is highest in urban rural councils and lowest in rural councils with significant growth.

However, these results are not supported on the basis of the statistical tests detailed in Table 4. The Welch test indicates that the distribution of pure technical efficiency varies from the overall population for urban regional and rural significant growth councils, whereas the Mann-Whitney test provides support on this basis only for urban regional councils. On the other hand, Banker's (1996) asymptotic tests for both an assumption of exponential and half-normal distributions support the hypothesis that urban fringe councils are less scale and purely technically efficient than urban regional councils, which in turn are less efficient on average than urban developed councils. Examples of purely technically efficient councils are spread across a number of categories. Examples include Gunnedah, Scone and Tamworth (RA), Manly and North Sydney (UD), and Penrith (UF). However, scale efficient councils tend to be concentrated in the larger urban and regional developed categories. These include Blacktown, Mosman, and Bankstown in the former, and Newcastle and Lake Macquarie in the latter.

The components of overall efficiency are examined using efficiency scores and total slacks (radial and nonradial) in Tables 5 and 6, respectively. The average level of slack across all geographic categories (as a percentage of the observed amount) is 56.7% for recyclables, 32.8% for expenditure, and 13.9% for garbage. All other things being equal, urban developed councils have greater slacks in all three outputs (i.e., garbage, recyclables, and the recycling rate), and the level of input (i.e., collection expenditure). These results hold even after the vector of nondiscretionary inputs is taken into account, most of which is the result of congestion factors rather than municipal size or geographic location. This would suggest that the impact of congestion, the inability to operate machinery, and difficulties in waste disposal in metropolitan areas are significant influences on a council's ability to attain efficient outcomes. Moreover, it is only in the urban developed category that significant slacks in all discretionary inputs and outputs exist. Both urban regional and rural agricultural councils have substantial slacks in recyclables and the recycling rate, but both are relatively productive in collecting garbage within the constraints imposed by their respective local government areas.

In terms of expenditure slack, urban regional and urban fringe councils tend to have higher expenditure slacks. The results indicate that the emphasis on improving productive performance in urban fringe councils should fall on reducing inputs, whereas urban regional and rural agricultural councils need to place more attention on promoting recycling and increasing the rate of recycling. Although the output weights used in DEA are derived from the sample itself, it would be possible to restrict weights in order to recognize the efforts by councils to promote recycling. Unfortunately, information of this type is not available for Australian local government.

However, the alternative logistic regression approach presented in Table 6 indicates that both urban developed and rural agricultural councils are generally less technically and scale efficient. A reason for this discrepancy would appear to be that while many urban developed and rural agricultural councils are not purely efficient in either respect, their relative efficiency scores, on average, are relatively high. The results in this section highlight the benefits of using a number of different approaches to interpret efficiency variation across groups of interest. Put differently, simple descriptive analysis, or an emphasis on the numbers of efficient councils alone, is likely to result in misleading inferences.

Concluding Remarks

The first section of this article, focusing on technical and scale efficiency in local government, examined cross-sectional technical and scale efficiency at the municipal level using the mathematical programming approach to efficiency measurement. The approach selected directly incorporates the effect of nondiscretionary environmental factors on efficiency indices, and thereby allows the comparison of efficiency of public sector entities with different operating environments. The results indicate that technical and scale efficiency varies significantly across individual councils at the local level. The results also suggest that it is possible to construct a uniform framework for measuring efficiency in local public services, provided allowance is made for the nondiscretionary environmental or contextual factors that affect the production correspondence relating inputs to outputs. However, even after allowing for differences in councils' operating environments, variations in efficiency remain, and these may be related to several imposed conditions.

The second section of the article focused on the individual components that determine efficiency in local governments' waste management and recycling function. All other things being equal, urban developed councils have greater input slacks in expenditure, while regional and rural councils have greater output slacks in recycling programs. A number of promising areas for further research are highlighted by these results in particular. This includes using surveys of ratepayers/citizens to ascertain a jurisdiction's subjective preferences for local public services, and incorporating these into efficiency analyses. A further area is to utilize a more disaggregated data set to identify more specific sources of inefficiency in local public services. These additional variables may include information relating to the extent of contracting-out, the type and frequency of service delivery, and the degree of mechanization.2