Alongside this specific, albeit incomplete, research into the economics of current structures of governance, some effort has been made to estimate the full range of net benefits that might be expected to be derived from reform of the federal system. Numerous government publications, consultancy reports, popular media analysts and non-economics academic papers quote various estimates made by Drummond as to the costs of duplication and coordination inherent within the current federal structure (for example, Access Economics 2004; House of Representatives Standing Committee 2003; Parliament of Victoria 1998). These estimates vary from $10-$40 billion dollars per annum, depending on which calculation is chosen for quotation. The most often quoted figure, however, is the $20+ billion claimed to be able to be saved from moving to a two tier system that abolishes the States (Drummond 2002).
By contrast, a study commissioned by state and territory governments, which focuses only on the direct and indirect overlap costs associated with special purpose payments, argues that the cost savings are more like $1 billion (ACIL Economics and Policy Pty Ltd 1996; Tredbeck and Cutbush 1996). A more recent study using Commonwealth Grants Commission data produces a similar order of magnitude estimate, namely $0.8m, for the cost of overlap and duplication between the Commonwealth and state levels (Garnaut and Fitzgerald 2002; Allen Consulting Group 2006). Yet another recent report estimates these costs to be $8.92 billion (Access Economics 2006).
These are vastly different estimates, suggesting the need for a careful examination of the underlying methodologies in order that we might be able to move towards development of a framework capable of resolving these differences and in so doing get a better handle on the real order of magnitude of potential savings (and/or costs) from various reform options. This is especially important given flaws inherent in the Drummond and Access Economics (2006) approaches, and the narrow focus of the alternative estimates. Some of these flaws have already been commented on by others (e.g. Twomey and Withers 2007).
Drummond’s (2001, 2002) approach commences with the assumption that government expenditure associated with any jurisdiction can be divided into a fixed cost component and a single variable cost component. The single variable cost component is calculated as a function of the aggregate population level in the jurisdiction; while the fixed cost component is assumed to be associated with overheads or administration of the level of government in question, and hence able to be avoided or eliminated if this level of government is phased out. Since the variable cost component is solely a function of population levels, it is argued that such costs will remain unaltered in magnitude should responsibility for that population shift to another jurisdiction. The sizes of these two components are estimated by fitting a simple linear regression model to pooled time series/cross-section data relating to three consecutive years (1998-99 to 2000-01) and eight jurisdictions (states and territories) – that is a total of 24 observations. The derived regression equation is:
E = 1.5883 + 6.6152 P
where E = total public sector expenditure ($ billion in 2000-01 dollars), and P = population (in millions).
The fixed cost estimate is $1.5883 billion and the variable (or marginal) cost estimate is $6,615 per head of population. It is this fixed cost estimate which it is argued to be saved if a particular jurisdiction is merged with another. Abolition of seven of the eight States (with the remaining one absorbing the responsibilities of those abolished and in effect administered by the Commonwealth) is argued to yield cost savings of seven times the fixed cost estimate – namely $11.12 billion (= 7 x 1.5883). Drummond also assumes that individual state deviations around the estimated regression line were accounted for by higher or lower variable costs per head of population and hence able to be eliminated through horizontal amalgamation. An algorithm was developed for achieving these mergers in ascending population size order. Application of this algorithm resulted in an almost doubling of the estimated cost savings from abolition of the States, namely $20.22 billion.
A number of serious criticisms can be made with respect to this aspect of the Drummond approach:
24 observations is too small a number of observations to support the findings reported, especially given the comparative lack of variability in the total level of government expenditure in any given jurisdiction over successive years and the use of pooled time series and cross sectional data – a longer time series of observations is available, even though the recent shift from cash to accrual accounting within government finance statistics may generate the need for care to be exercised in splicing relevant time series together;
The assumption of a single variable cost component based on population levels ignores a large volume of available data relating to determinants of inter-jurisdictional variations in costs. For example, federal and state grants commissions have long recognised in their horizontal equalisation grants that aggregate population size is only one relevant factor that needs to be taken into account when seeking to understand inter-jurisdictional expenditure requirements. Other factors of relevance include population composition, population dispersion, and geography including climate and terrain. Ignoring these other factors leads to an overestimation of the fixed cost component. In technical terms the regression equation is inappropriately specified due to omitted variables, such that at the very least there is a need to move to a multiple regression equation that includes other variable cost terms;
In Drummond’s estimation the effects of all these other variable cost factors are being inappropriately attributed to the fixed costs component, leading to its overestimation. Given the importance of this fixed cost component to the remaining steps in the Drummond approach, this represents a serious problem;
While Drummond reports the coefficients related to his fixed and variable cost components, he does not report the associated standard deviations or t test statistics, such that the statistical significance of his parameter estimates related to these two components cannot be determined. In addition he does not report results of any regression diagnostic tests conducted – yet previous modelling experience in this area suggests that the equation as estimated has problems that require it to be re-specified and re-estimated in a different format. In particular, as other critics (Access Economics 2004; Parliament of Victoria 1998) have pointed out, it is more usual to estimate models relating to government expenditure in per capita terms because of serial correlation between total expenditure and population;
The assumption that functions carried out by a state government could be transferred to a national body without altering either the fixed cost or variable cost component is simply unrealistic. In particular it ignores the coordination costs that could be expected to result from attempts to manage population-oriented services such as education, health and policing from a single centralised location in a nation as large as Australia with its vast areas of sparsely dispersed populations. It also ignores research relating to economies/diseconomies of scale in provision of the services for which our state governments are currently responsible – which for many functions have resulted in regionalisation of service provision rather than the adoption of a one size fits all model within many states;
The implied assumption that population is the only determinant of variable or marginal costs of service provision suggests, for example, that transfer of responsibility for providing state level services to a particular area within Queensland would not involve any increase in per capita costs if this provision was administered from say Perth rather than Brisbane. This is clearly unrealistic.
By using pooled time series/cross-section estimation the assumption is being made that all state governments host equal fixed and variable costs – so that amalgamation of any two states is argued to liberate cost savings equal to an identical quantum of fixed costs. This assumption too is unrealistic. For example it ignores the geography and demographic composition of the various states, as well as differences in the level and nature of the services that they currently provide in each major area of responsibility which would inevitably need to be bought into alignment following a merger. More significantly this assumption would be able to be dispensed with, if the regression exercise was to be redone with the underlying data collected over a longer time period;
The assumption that individual state deviations around the regression line were accounted for by higher or lower fixed costs and not higher or lower variable costs could also be readily tested via use of a more extended time series of expenditure data and the introduction of a series of dummy variables for the different states.
Drummond (2001, 2002) explored cost savings associated with change options other than abolition of the States, for example a new states option and an option that eliminated local rather than state governments were developed. However, the cost savings estimates derived were based on the same regression methodology underlying the abolition of the States option, and so most of the criticisms outlined above continue to apply. Some additional unrealistic assumptions were introduced when these options were explored. In particular:
the new states option assumed that data and associated parameter estimates derived from Tasmania would be representative of any new state likely to be introduced. This is patently not realistic given the geography, economy and demography of Tasmania compared with potential mainland new states (such as the New England region of NSW and the Northern regions of Queensland, as mentioned by Wiltshire, this volume).
the elimination of local governments option assumed that the Australian Capital Territory (ACT) could be regarded as representative of all other ‘reformulated states’ – on the basis that the ACT government performs a mixture of both state and local government options. Again this is simply not realistic given the geography, economy and demography of the ACT compared with the rest of Australia. The ACT, for example, has the vast majority of its population employed in the public sector on incomes significantly higher than the national average, it has only one major urban centre which was centrally planned and administered from its inception, and it covers a very small geographic area. The costs of local government service provision under these circumstances are vastly different to those in either major metropolitan areas or remote rural regions elsewhere in Australia.
Despite the problems inherent in the Drummond (2001, 2002) analysis, he did attempt to estimate the full range of net benefits that might be expected to be derived from reform of the current federal system. Accordingly, this work has provided a significant departure point for ongoing work and significant stimulation to further debate and more targeted research. In addition, most alternative estimates also suffer their own problems. Most of the alternative estimates that have been published focus on only a narrow subset of potential savings (e.g. from reducing the Commonwealth Government’s reliance on specific purpose payments as a form of intergovernmental grants) and/or ignore the role of local governments in the federation.
The most comprehensive alternative analysis to date, by Access Economics (2006), focuses on a large range of possible categories of cost savings that could be argued as able to be derived from reform. These categories were classified into two groups: (a) costs associated with spending-related efficiencies (including cost of administering intergovernmental grants and costs of coordination across different levels of government when overlap of functions exists) estimated at $5.1 billion, and (b) costs associated with taxation-related efficiencies (including the use of inefficient taxes by the States) estimated at $3.8 billion.
However, in the Access Economics (2006) analysis, the dollar values placed on these various categories of costs have not been derived from detailed economic or statistical analysis but rather from a set of assumptions or educated guesses. For example 10% of the dollar amount associated with special purpose payments to the States is argued to be associated with states spending above the efficient level due to lack of adequate coordination and/or oversight by the Commonwealth. No justification is given for why 10% represented a reasonable assumption. Similarly it was assumed that 3% of the expenditure by states was on activities that represented ‘overlap and duplication’ with the Commonwealth. In arriving at this 3% figure, no specific areas of overlap or duplication were identified or costed, nor was any insight provided as to why 3% was a reasonable figure to select from the 99 other possibilities between one and 100. In terms of cost shifting, the area of pharmaceuticals and public hospitals was singled out, and state governments accused of shifting costs on to the Commonwealth in the order of 3% of Commonwealth expenditure in this area. Cost shifting in other areas or from the Commonwealth to states were ignored; and even in the restricted area of pharmaceuticals and public hospitals, no reasoning was provided to justify the selection of 3% rather than 2% or 5% or some other percentage. In addition, the focus of the Access Economics (2006) report was purely on inefficiencies argued to be associated with current federal-state relations – local government and various regional organisational arrangements discussed earlier in this chapter are not mentioned.
As a result, there remains considerable scope for further research in this area.