Introduction1
In the uk, the privatization of utility industries has led to the development of regulatory regimes to prevent monopoly abuse. there are two main methods of economic regulation, namely cost of service, alternatively known as rate of return regulation, and price-cap regulation. In very broad terms, whereas cost of service regulation sets prices based on operating expenditure (oPeX) and capital expenditure (caPeX) figures with an allowance for a normal profit, price-cap regulation places a ceiling on prices or revenues. as is now recog-nized, price-cap regulation provides incentives for management to economize on oPeX and caPeX. as a result, all privatized utilities in the uk have been subject to price-cap regulation.
an approach to determining the revenue required to meet efficient costs would be to base future revenue on the costs achieved by the firm in previous years, adjusted for any known exceptional items or expected cost changes. the prob-lem with this approach, which is essentially identical to the method adopted under cost of service regulation, is that incentives for cost efficiency are reduced.
Where management expects that cost increases can be passed through to con-sumers, there is little incentive to reduce them. moreover, if future oPeX and caPeX funding is based on the firm’s forecasts of its own costs, there will be an incentive for management to pad these costs.
the objective of incentive based utility regulation is, therefore, to establish an incentive compatible set of cost comparisons that can be used to determine an efficient firm’s revenue needs. benchmarking or ‘yardstick competition’ as it is sometimes described involves comparing the performance of a regulated firm with some comparator. In a seminal paper, Shleifer argued that a regulated firm’s revenue needs should be assessed by looking at costs in comparable firms or industries (Shleifer, 1985). benchmarking costs reduces the effects of the company’s own costs on prices; in extremis, if a firm’s costs have no effect on its own revenues, the incentives for management to reduce costs will be maxi-mized. this chapter therefore considers the arguments for yardstick competition and highlight difficulties in achieving efficient and effective benchmarking, fo-cusing on the uk’s experience of implementing price-cap regimes in
telecommunications, gas, electricity, and water and sewerage.2 the chapter be-gins by reviewing the case for benchmarking.
The case for benchmarking
benchmarking or yardstick competition (we use the terms interchangeably3) can provide regulators with information about efficient oPeX and caPeX re-quirements, thereby reducing the regulated firm’s informational rents.
benchmarking is now used extensively: for example in costa rica for transport tariff setting, in telecommunications regulation in Hungary (rossi and ruzzier, 2000, p. 82), in dutch electricity and telecommunications (dte, 1999), and for electricity regulation in norway and new South Wales, australia. It is also a component of some price-cap setting in the uk.
the price-cap regime for privatized utilities was first developed in the uk in the early to mid-1980s. under rPI-X price cap regulation, a factor, X, is intro-duced, which effectively increases or reduces real prices relative to the retail price index (rPI) over the price-cap period (often five years). However, in littlechild’s 1983 proposals to the government for the regulation of british telecom, which was the first privatized utility in the uk, there was almost no reference to how price caps should be set.4 at privatization, bt’s initial X factor was set behind closed doors in discussion between the government, its bankers and city investors, so as to ensure a successful share flotation. littlechild stated in his proposals that ‘X is a number to be negotiated’ (littlechild, 1983, para.
13.17) and he appears to have expected competition to develop quickly in tele-coms. therefore, littlechild (1983) provides virtually no guidance as to how price caps should be reset after the completion of a regulatory period. However given that effective competition did not quickly develop in telecoms, and the privatization of other network industries with varying degrees of monopoly power necessitated the development of more permanent price-cap regimes, the attention of regulators quickly turned to the process of setting appropriate price caps for the industries (Saal, 2003).
arguably, regulatory regimes should set out to mimic the discipline imposed by a competitive market. In a highly competitive market, prices are external to the costs of any individual firm and the prices facing any firm in the industry change at the same rate as the growth in the industry’s, not the firm’s, unit costs.
broadly, there are two approaches to setting a price cap. the first is ‘bottom up’, under which the regulator starts by taking the firm’s forecast operating ex-penditures, asset base, depreciation rate and capital exex-penditures, and computes the necessary revenue requirement taking into account demand changes. the alternative is a ‘top down’ approach, where a revenue ceiling is imposed (based on forecasts of the firm’s costs – as similar to the bottom up approach – or through some form of benchmarking performance) and both oPeX and caPeX requirements are managed within this cap. a pure price-cap regime
approxi-mates the latter. In the uk the price-cap regime is not pure and is based on a combination of both ‘bottom up’ and ‘top down’ approaches. the process gener-ally involves a truncated future cash flow model including estimates of oPeX and caPeX over the life time of the price cap (Vass, 1999) and an expected rate of return set by reference to the estimated cost to the firm of raising capital.5
achieving reasonably correct cost estimates is demanding in terms of the in-formation that the regulator requires. In particular, the regulator needs information about the firm’s efficient levels of oPeX and caPeX, but regulated firms will have an incentive not to search out these costs given that cost mini-mization requires management effort. In 1985 Shleifer set out to tackle the inherent information asymmetry in economic regulation by arguing that the firm’s revenue requirement to cover its oPeX and caPeX should not be based on its own costs. Instead it should be based on benchmarked costs or a relative efficiency measurement. as he comments at the outset of his paper, the objective should be to find ‘some relatively simple benchmark, other than the firm’s present or past performance, against which to evaluate the firm’s potential’
(Shleifer, 1985, p. 319). Shleifer’s solution was to identify ‘comparable firms to infer a firm’s attainable cost level’ (p. 312). Provided that the regulator has two or more firms under its jurisdiction, then yardstick competition can over-come information asymmetry. each regulated firm, i, is assigned a ‘shadow firm’. the shadow firm becomes the benchmark for setting the revenue require-ment. the efficient outcome for each firm is to select cost, ci = c*, where c* is the cost consistent with the minimum feasible cost for any given output. as long as the information provided by the benchmarking exercise leads the regulator to impose c*, the regulated firm sets price (P) according to efficient costs:
Pi = ci = c*.
this is best achieved when ‘a firm’s choice of ci has no effect on the price it gets’ (Shleifer, 1985, p. 322).
Shleifer recognized, however, that certain requirements were necessary for this approach to regulation to work satisfactorily. In particular, he was at pains to emphasize (1985, p. 323):
It is essential for the regulator to commit to not paying attention to firms’ complaints and to be prepared to let the firms go bankrupt if they choose inefficient cost levels.
unless the regulator can credibly threaten to make inefficient firms lose money … cost reduction cannot be enforced.6
Shleifer was also aware that this form of yardstick competition required a
‘shadow firm’ comparable in terms of its cost structure. If the firms scrutinized for benchmarking purposes by the regulator were heterogeneous in terms of
efficient cost structures, the result would be unreliable. the result would also be unreliable if the firms faced different demand functions, although for simplic-ity Shleifer assumes a common demand function for much of his analysis. to overcome the lack of a perfect ‘shadow firm’, Shleifer recognized that multivari-ate regression models would need to be developed to reflect characteristics that could account for cost differences between firms that are not within the control of management (e.g. topography, customer density, regional wage costs, etc.).
Shleifer was aware that the use of multivariate regression as the basis for yardstick competition is demanding in terms of modelling. If some significant environmental or ‘exogenous’ variables are omitted from the model, and these impact on any particular firm’s costs differently from the way they impact on the costs of other firms, the outcome will not be optimal due to omitted variable bias. It is also difficult for the regulator to know what variables are truly exoge-nous to the firm and which are at least partially under its control. However, Shleifer also notes that applying more and more detailed models can lead to additional complexity that outweighs any useful economic gain, thereby antici-pating the potential for oPeX and caPeX modelling to employ more and more regulatory resources for less and less marginal benefit. thus he warns against the marginal cost of the exercise overwhelming its marginal benefit in terms of more accurate price setting. Finally, Shleifer makes reference to the dangers of
‘collusive manipulation’ of yardstick competition by participating firms (1985, p. 327). Yardstick competition requires managerial independence across the firms used as comparators; otherwise the results will be biased.7 Shleifer’s (1985) paper has made an important contribution to the economics of regulation literature. today yardstick competition is widely used by regulators internation-ally and has led to various methods of modelling relative efficiency. the four main approaches to comparative performance measurement are as follows:
1. Productivity indices, and particularly nonparametric index approaches to productivity measurement such as labour productivity and total factor pro-ductivity (tFP) indicators based on, for example, törnqvist indices.8 2. Stochastic analysis of production and cost functions,9 such as translog cost
functions, which are a second order approximation to any arbitrary func-tional form. the standard olS (ordinary least squares) econometric method assumes that the residual term is normally distributed around zero. an al-ternative is to use colS (corrected ordinary least squares) where the regression equation is estimated using olS and then, for cost functions, shifted to the efficient frontier by adjusting the constant term by subtracting the value of the largest positive residual. as a result, the function passes through the most efficient firm and bounds the other firms. an extension where there are multiple inputs and outputs is to estimate distance functions (Shephard, 1970).10 more recently, regulators have been experimenting with
stochastic frontier analysis (SFa). this is based on the assumption that part of the error term results from inefficiency.11 Stochastic methods allow for diagnostic tests that are lacking in mathematical methods (discussed next) and therefore allow an analysis of the probability that the results are inac-curate.12 However, a problem with stochastic approaches is that a functional structure is imposed on both the data and the error term.
3. Mathematical modelling, especially the use of dea (data envelopment analysis). bogetoft (1994, 1997) has demonstrated that yardstick competi-tion based on dea may be optimal in regulatory environments with technological uncertainty.13 but being non-stochastic, arguably dea is less satisfactory for dealing with the uncertainties that surround the data that enter into benchmarking exercises.14 also, unlike in econometric analysis standard statistical tests cannot be readily applied to test for the significance of the variables included in the model. norway has experimented with the use of dea (Weyman-Jones, 2003), as has the netherlands electricity regu-lator (dte, 2000). but in new South Wales the reguregu-lator, IPart, switched from using dea to a more basic engineering-based approach to benchmark-ing because of concerns about the reliability of dea results. In the dutch electricity sector, dea was later abandoned.
4. Engineering models. an alternative approach is to calculate theoretical production functions based on engineering data. Farrell (1957) argued against theoretical functions in favour of empirical efficient frontiers based on best-observed practice because of the difficulty of estimating theoretical frontiers accurately: ‘the theoretical function is likely to be wildly optimis-tic’ (1957, p. 255). nevertheless, some utility regulators in latin america rely on theoretical functions (Fischer and Serra, 2000).15 although not widely used in the uk, engineering analysis has entered into the regulatory process through commissioned work on issues such as telecoms intercon-nection, network reliability and serviceability, and efficient capital project costs, for instance in the water industry. Process analysis, under which consultants compare how firms do common tasks, including how many staff they use, has some similarities to this approach to performance measure-ment, though it may not involve the use of sophisticated engineering models.
In practice, these different performance measures can be used on their own, but can also be used as part of a more complex process for setting prices and profits based on a range of yardsticks. bauer et al. (1998) have proposed a set of con-sistency conditions when different measures are adopted by regulators, if they are to be relied on, namely that the different methods used should provide con-sistent efficiency levels and rankings and identification of best and worst performers, and also be consistent in their results over time. coelli and Perelman
Figure 6.1 Comparisons of efficiency benchmarking
olS regression
cost frontier
Frontier shift oPeX
output ya
ca a
b
(1999) have suggested combining the results from alternative modelling exer-cises by using the geometric means of the performance scores for each data point in order to reduce potential bias. However, regardless of the precise pro-cess adopted, the modelling of efficient costs is unlikely to produce an outcome that completely replaces ‘negotiation’ in price-cap setting.16 Indeed, as the uk’s experience demonstrates, yardstick measures are best seen as a tool for regula-tors to boost their negotiating position vis-à-vis regulated companies.
Figure 6.1 illustrates benchmarking methods by focusing on oPeX as an example and assuming, for simplicity, a linear frontier. olS regression analysis establishes an estimated average relationship between oPeX and output in the industry. Subsequently, this average cost function is shifted down (to b) using colS techniques, which effectively assume that the firm with the largest nega-tive olS residual is on the efficient frontier.17 the inefficiency of all firms not on the frontier is then measured as the distance between their actual oPeX and that which is predicted by the frontier. this distance can be interpreted as a measure of the potential efficiency ‘catch-up’ that an inefficient firm could achieve if it became efficient. Figure 6.1 also shows the potential effect of a
‘frontier shift’, perhaps caused by positive technological change or experience curve effects. a frontier shift reflects the impact of technological change and changes in working practices on efficient best practice performance across the industry, over time.
an important distinction must be drawn between ‘catch-up’ and ‘frontier shift’ effects when setting prices with benchmarking techniques. catch-up or what is sometimes referred to as the ‘stretch’ factor refers to the efficiency gap that needs to be closed between the firm and the efficient frontier. For example, the uS Fcc (Federal communications commission) approved a price-indexing plan for at&t in 1988 with a stretch factor of 0.5 per cent, equivalent to 20 per cent of the company’s estimated tFP growth. Similar catch-up factors have been introduced for the interstate services of local exchange telecom carriers in the uS and for north american electricity utilities. they are also applied by uk regulators, as discussed below. by contrast, frontier shift represents an estimate of the actual or potential productivity improvement achieved by a firm that is already on the frontier, and is therefore already efficient. the actual decomposi-tion of efficiency change and technical change is typically derived using colS, dea or SFa techniques, but could alternatively be estimated using engineering models.
While any shortfall from the estimated frontier is usually interpreted as inef-ficiency, there is always the possibility that the estimated frontier and hence the resulting efficiency scores are biased because of measurement errors or model misspecification. For this reason, regulators may be reluctant to put complete reliance on frontier estimates. moreover, leaving aside the theoretical extreme of ‘perfect competition’, expecting all firms to be on the efficient frontier is in-consistent with the usual operation of competitive markets, whose outcomes some economists see economic regulation as attempting to emulate. In real-life competitive markets the average performer, not the best performer, can determine prices and firms with above average performance then earn super-normal returns.
If regulators were to determine their view of the expected performance in the industry based on the most efficient firm, then this would penalize firms earning average returns. these firms would not necessarily be penalized in a normal competitive market accommodating a number of suppliers with different costs.
bernstein and Sappington (1999) demonstrate that when modelling the aver-age X factor, which includes the total frontier shift and efficiency gains that a regulated industry can be expected to achieve over the regulatory period, the relative change of tFP in the regulated industry (TFPr) compared to the com-petitive economy as a whole (TFPc) becomes important. also relevant are changes and differences in input prices in the economy (Wc) relative to those in the regulated industry (Wr). thus they demonstrate that the X factor can be ex-pressed as the difference between the total factor productivity growth potential in the regulated sector and that in the remainder of the competitive economy, plus the difference in the rate of growth of input prices in the two sectors:
X = ∆(TFPr – TFPc) + ∆(Wc – Wr)
However, bernstein and Sappington (1999) also emphasize that this approach to setting X depends on a number of assumptions, including that the regulated sector is small in comparison to the economy as a whole, otherwise its perform-ance will materially affect economy-wide tFP and input prices. nevertheless, average tFP growth in the economy often serves as a starting point for deter-mining X factors. For instance, in uS telecommunications the average rate of productivity growth is used as a benchmark (dte, 1999; Jamasb and Pollitt, 2001).
The UK experience
In the uk, benchmarking performance is now part of resetting regulatory price caps and is used to help determine appropriate X factors (Jamasb and Pollitt, 2001).18 over time there appears to have been movement towards a more com-mon approach to benchmarking across the regulatory offices, see below, no doubt as the result of increasing experience, monopolies and mergers commis-sion (now competition commiscommis-sion) investigations,19 demonstration effects and government policy that favours more consistency in regulation across the dif-ferent regulatory bodies (dtI, 1998, para. 80). nevertheless, some important differences remain. the regulatory offices undertake their own productivity and cost analyses and from time to time use research produced by different outside consultancy firms. expert judgement also plays a part in efficiency assessment by the regulatory offices and the experts come from diverse backgrounds includ-ing industry and academia. there is, therefore, some heterogeneity in the approaches adopted by uk regulatory offices when setting price caps.
In the uk, benchmarking performance is now part of resetting regulatory price caps and is used to help determine appropriate X factors (Jamasb and Pollitt, 2001).18 over time there appears to have been movement towards a more com-mon approach to benchmarking across the regulatory offices, see below, no doubt as the result of increasing experience, monopolies and mergers commis-sion (now competition commiscommis-sion) investigations,19 demonstration effects and government policy that favours more consistency in regulation across the dif-ferent regulatory bodies (dtI, 1998, para. 80). nevertheless, some important differences remain. the regulatory offices undertake their own productivity and cost analyses and from time to time use research produced by different outside consultancy firms. expert judgement also plays a part in efficiency assessment by the regulatory offices and the experts come from diverse backgrounds includ-ing industry and academia. there is, therefore, some heterogeneity in the approaches adopted by uk regulatory offices when setting price caps.