This paper sets out the consequences for cost reduction of three efficiency target scenarios which, broadly speaking, cover the range of targets that my office believes are plausible.

In setting efficiency targets or 'X factors' for individual companies, I shall be mindful of the need to maintain incentives. The X factors will be higher for the relatively inefficient companies than for the efficient ones.

The companies are, generally speaking, outperforming the efficiency assumptions made at the 1994 Periodic Review. Although price limits tightened significantly then, water companies have not experienced the rigors of competitive markets nor the difficulty of securing finance in a non-regulated environment. These points suggest scope exists for further improvement, perhaps at a faster rate than in the past.

Customers want to see continuing investment to improve services and the environment without seeing bills rise. The range of efficiency savings documented in this report would be sufficient to finance substantial capital enhancement investment without any need for prices to increase.

The range of possible savings is inevitably very wide at this stage of the price review. To enable me to consider company responses in preparation for the publication of Prospects for prices in October 1998, companies are invited to respond to this paper by the end of June. I hope companies, and others, will comment on — and challenge — the results.

I hope companies will tell us, and their customers, how they are judging their current levels of efficiency and the scope they see for further progress. In particular, I should be interested in whether they have carried out the benchmarking studies which so many firms in competitive markets have found so helpful — and, indeed, necessary for survival. Indeed some of them may be able to make international comparisons from their own experience in the world market. I should be interested to learn what companies have done to raise awareness of what is possible through benchmarking and the extent to which they have set objective performance standards for key activities to match or surpass the best in the world.

If the work emanating from companies is incomplete Ofwat may commission its own work in this area.

I C R BYATT

Director General of Water Services

2.THE EFFICIENCY FRAMEWORK FOR THE 1999 PERIODIC REVIEW

2.1 Estimating the scope for future efficiency to assume in price limits — X factors

The Director General of Water Services (the Director) identified the key constituents of K in the recent paper Setting price limits for water and sewerage services — the framework and business planning process for the 1999 Periodic Review (February 1998). The price limit year by year was seen as a function of past outperformance (P_o), future efficiency gains (X), quality standards (Q), and enhancements to the security of supply (V) and service levels (S). The Director's conclusions on the framework and approach to estimating future efficiency — or X factors — were set down in the paper.

This paper builds on these foundations by describing the work carried out by Ofwat to date to assess the scope for future efficiency savings at both industry and company level. The paper includes the early results from the analysis. The work is set in the context of the approach taken to similar issues during the 1994 Periodic Review.

Past outperformance (P_o) will be dealt with in Prospects for prices, which is due to be published in October 1998.

2.2 General approach to assessing the scope of the X factor

The assessment of the X factor has two key elements:

This paper outlines the Director's approach to both elements (see 2.4 and 2.5). The approach is tailored to the three main components, operating expenditure, capital maintenance expenditure and capital enhancement expenditure. It is usual to add together the latter two components under the term capital expenditure (capex) but, for the purposes of price setting, they are considered separately.

As in the 1994 Review, the Director intends to consider separately the scope for efficiency improvements in operating expenditure, capital maintenance expenditure and capital enhancement expenditure. He will test his conclusions by looking closely at the interactions between the expenditure areas. An alternative approach to efficiency assessments, based on total costs, has been rejected following a detailed study.

The approach to opex will be similar to that used in 1994, updated in the light of the recent performance of companies. As set down in Setting price limits, variations in service performance between companies will be dealt with separately (see A proposed approach to assessing overall service to customers — a technical paper (March 1998)). In 1994, levels of service performance was one of the non-econometric factors taken into account in deriving the relative opex efficiency of companies.

The approach to capital maintenance has evolved since the first Periodic Review. In 1994, capital maintenance expenditure on underground assets was reviewed using a historical serviceability approach. Capital maintenance expenditure on water and sewerage surface assets was subject to an upper limit of £20 per property per annum. Company specific efficiency improvements were set for both capital maintenance expenditure and capital enhancement expenditure by an examination of comparative capital works' unit costs (using the cost base (see 2.6) and a separate comparative analysis of the costs of meeting the Urban Waste Water Treatment Directive). In addition, a general capital productivity assumption was applied to all companies.

For the 1999 Periodic Review, the historical serviceability analysis has been extended to cover surface assets. Statistical models similar to those for opex have also been developed for the comparative analysis of historic capital maintenance expenditure. These models will help identify high and low cost companies and so guide judgements on the scope for improvements in efficiency. The cost base exercise will be repeated to inform judgements on the scope for efficiency gains in both capital maintenance and capital enhancement expenditure (see Chapter 5).

The approach to capital enhancement expenditure will be similar to that used in 1994. Companies will submit their estimates for the cost of achieving enhanced performance required by new legislation. These projections will be largely based on companies' current data on capital unit costs. The cost base will then be used to test whether there is scope for improvements in capital unit costs in individual companies when compared to their peers. A general capital productivity assumption will also be applied to all capital enhancement expenditure.

2.3 The components of total expenditure and trends since privatisation

The total expenditure by the industry, in 1997–98 prices, since 1989–90 for both the water and sewerage services is shown in Table 1.

Table 1: Total expenditure by water companies in the period 1989–90 to 1996–97

(£ billion at 1997–98 prices)
Service area	Operating expenditure		Capital maintenance		Capital enhancement
	£bn	% of total	£bn	% of total	£bn	% of total
Total water service	13.0	60	5.0	51	7.3	50
Total sewerage service	8.6	40	4.8	49	7.3	50
Total both services	21.6	100%	9.9	100%	14.6	100%

Note: Numbers may not add up because of rounding.

Figure 1 shows the composition of operating costs and capital maintenance costs. The large slice relating to 'other operating expenditure' expenditure includes costs associated with customer services, associated companies and doubtful debts.

Figure 1: Proportions of operating and capital maintenance expenditure by purpose — 1996–97

The total actual expenditure shown in Table 1 is generally less than was assumed in the price limits set in 1989 by the Secretaries of State and in 1994 by the Director. For example, in 1989, operating expenditure, including quality improvements, was expected to increase by 20% by 1994, whereas it only increased by 2.5%. In 1994, an efficiency improvement of 10% by 1999 was assumed. Companies have reported operating expenditure reductions, including new expenditure on water quality and other improvements, of around 12% since the 1994 price review. These reductions may, on latest estimates, reach between 15% and 20% by 1999. The reduction is, however, less — 6% since the price review — when new expenditure to meet quality obligations is added. It is not yet clear whether the performance of the water companies on operating expenditure is as good as that achieved by the other utilities that were privatised in the 1980s. For example, the regional electricity companies have cut their distribution operating expenditure by around 20% between 1992–93 and 1996–97. Further work is planned to compare the performance of the water companies with other sectors of industry. 2.4 The overall scope for efficiency savings In 1994, Ofwat concluded that the overall scope for improvements in opex that should be included in price limits was 2% per annum for the period 1995–96 to 1999–2000 and 1% per annum thereafter. For the period 1995–96 to 1999–2000 the improvement was split into two elements; a general assumption (1% per annum) and company specific improvements depending on their relative efficiency. For both capital maintenance and capital enhancements expenditure, Ofwat assumed capital productivity improvements at a rate of 1% a year, in addition to company specific adjustments derived from the Cost Base as described in MD 127 (Capital unit costs in the water industry — the 1994 Periodic Review cost base) (March 1997). In total, these adjustments corresponded to a 2–2.5% per annum reduction in capital unit costs. For the 1999 Periodic Review, Ofwat will be commissioning consultants to review the scope for overall improvements in both opex and capex. The trends in other utilities, particularly those that have been subject to growing direct competition, would suggest that continuing scope for improvements should be expected in the water industry. Indeed, some companies that were assessed as below average efficiency in 1994 had reduced total operating expenditure by more than 20% by 1996–97. Companies also cite improvements in capital efficiency of the order of 15%. This suggests that, particularly for inefficient companies, the scope for improvements during the next price limit period could be much larger than those assumed for the period 1995–96 to 1999–2000. This implies that the combined effect of the general efficiency assumptions and the catch-up arising from the comparative efficiency analyses within the industry for both opex and capex could be well in excess of 2.5% and possibly up to 4% or 5% a year for the next five years. The scope for improvements will not be uniform across companies. Those companies with historically high levels of opex and capital maintenance might be expected to account for a larger proportion of the cost reductions expected of the industry as a whole. Therefore, comparative assessments of companies' operating and capital maintenance costs will be a vital tool for use in the 1999 Periodic Review. The work to date in assessing comparative levels of expenditure is summarised in the next section. 2.5 The comparative expenditure of the companies Rankings of company comparative opex were set down in the 1996–97 Report on water and sewerage service operating costs and efficiency (December 1997). Further work on the econometric models has changed the bandings for some companies. Since the 1994 Review, considerable effort has been devoted to the development of comparative capital maintenance models, similar to the approach developed for comparative operating expenditure efficiency in the early 1990s. The comparative capital maintenance work has built on these foundations, taking on board the learning points. The analysis has been reviewed by Professor Mark Stewart (University of Warwick) but has yet to be subject to public challenge. Ofwat follows a step by step approach to the statistical assessment of company performance (see Appendix 1). As part of this process, econometric equations are derived that describe the relationships between expenditure, outputs, company size and a very limited number of statistically significant explanatory factors. These relationships are called econometric models. The current econometric models for both opex and capital maintenance expenditure are summarised in Appendices 2 and 3. Ofwat has decided to publish these now to ensure that all interested parties have an opportunity to understand and contribute to challenging and refining the work. It is for higher cost companies to explain, justify and quantify why they should not be classed as less efficient than their peers. Only at the end of the process of challenge and after assessments of company specific factors, will a final relative efficiency ranking be derived for use at the Periodic Review. The results of the models for the water and sewerage services are set down in Chapter 4. Some companies have a much better cost performance than suggested by the models — these are banded as 'A' companies. Other companies' cost performance is not as good as the models suggest it should be and their actual expenditure is well above that which is predicted by the models — they are banded as 'E' companies. Table 2 maps out a matrix pattern showing clearly that some companies have lower than expected expenditure in both areas — the A/A companies, while others have higher than expected expenditure in both areas — the E/E companies. It also draws attention to those companies with low expenditure in one area but high expenditure in the other. The relative efficiency targets for such companies are considered further in Chapter 5. Table 2: Matrix pattern for opex and capital maintenance analysis   Table 2: Matrix pattern for opex and capital maintenance analysis Operating expenditure analysis A High/low High/low Low Low Low B High/low As expected As expected Low Low C High As expected As expected As expected Low D High High As expected As expected High/low E High High High High/low High/low   E D C B A Capital maintenance expenditure analysis Tables 3 and 4 show for the water and sewerage services the relative cost positions of the water and sewerage companies. Table 3: Preliminary results from the econometric models — water service Table 4: Preliminary results from the econometric models — sewerage service 2.6 The cost base In the 1994 Review, comparative capital unit costs were used to assess the relative procurement efficiency for both capital maintenance and capital enhancement expenditure in an exercise known as the cost base. The cost base exercise will be repeated for the 1999 Review with an initial submission from companies due in June 1998. The cost base will contain standard costs that are relevant to both capital maintenance and capital enhancement expenditure. The standard costs have been selected to reflect probable capital expenditure programmes in the period 2000–05 and specified so that improvements in capital efficiency since 1992–93 can be identified. In addition to the cost base, for capital maintenance expenditure there will be a related assessment of comparative efficiency arising from the econometric models. The cost base will assess the relative efficiency in procurement and implementation of capital projects. The econometric models will provide an overall indication of comparative capital maintenance efficiency in carrying out the right capital maintenance activity at the right cost on the right assets at the right time. 2.7 Implications of comparative efficiency rankings for company specific elements of X factors In 1994, company specific assumptions of opex efficiency improvements were such that 25–35% of the identified difference between the least efficient company and the leading representative company's performance were removed over the first five years (ie by 1999–2000). The efficiency improvements for capital maintenance and capital enhancement expenditure, derived from the cost base, represented the immediate elimination of about a quarter of the gap between company reported costs and the lower quartile yardsticks. The size of the company specific catch-up that will be assumed in 1999 will depend on judgements on a number of factors. These include: the relative ranking in the final efficiency tables; the robustness of efficiency analysis; the chosen benchmark performance; how much of the gap should be removed; whether this should be reduced because of opex/capex performance relationships; and how the catch-up should be phased over the price limit period. These decisions are crucial to maintaining the correct balance of incentives. In particular, too lax assumptions will not challenge the companies to innovate. Successful innovation by the companies is in the long-term interests of customers. The final efficiency ranking will be influenced by the companies' performance in 1997–98 and 1998–99, reported in the July Return for 1998 and 1999. This ranking will also reflect the Director's judgements on any company specific factors that affect the outcome of the models. Judgements on the robustness of the ranking will vary from service to service and between opex and capex. Decisions on the most appropriate benchmarks have yet to be taken. In 1994 the leading comparator, ie the representative company with the best performance, was used for opex and the lower quartile of the industry was used for the cost base. The current preferred approach is to use the leading comparator as the benchmark, provided that this represents a reasonable proportion of the industry, and then phase the catch-up towards this performance over the price limit period. Conventional practice for yardstick competition would imply using the average performance as the benchmark, with the catch-up to this performance being assumed in year one. The Director consulted on these issues in The business planning process, customer consultation and information requirements for the 1999 Periodic Review (July 1997). There was a mixed response, some respondents suggesting the catch-up in two years, others proposing five years (as in the 1994 Review) with tougher targets in the early years. In Setting price limits (February 1998) Ofwat inclined towards a phasing of the catch-up over the five years, but with greater progress achieved in the early years. 2.8 Possible expenditure savings through improving efficiency Cost savings through projections of improved efficiency are built into current price limits. Chapter 3 sets out the assumptions made at the 1994 Periodic Review. Broadly, price limits assumed 2% per year improvements in operating efficiency, comprising 1% per year overall improvement and the remainder achieved by less efficient companies catching up. For the most efficient companies, the 1% per year overall improvement was reduced to 0.5% per year. On capital expenditure, the efficiency improvements were set at a level equivalent to an average of 1.5% per year for improvements in procurement efficiency and an additional 1% per year target for technological improvement. These assumptions reduced expenditure requirements by around £0.8 billion for opex and £0.4 billion for capital maintenance over the period 1995–2000, compared with static levels of efficiency. In the event, companies are likely to achieve more, possibly of the order of £2 billion for opex. Expenditure on capital maintenance is likely to be less than was assumed at the last review by as much as £0.5 billion. The extra savings will be passed on to customers through the Po adjustment. At the 1999 Periodic Review, the combination of overall targets and company specific catching-up could generate significant further savings in expenditure requirements. As indicated above, the size of the savings are dependent on a number of judgements yet to be made. These judgements will be influenced by developments since 1994. In particular, a number of companies have reported large reductions in opex — some have achieved more than 20% reductions in total opex in four years. On capex, some companies report that they are achieving efficiencies of 15% or more. Indeed, a joint government/industry review of procurement and contractual arrangements in the UK construction industry — Constructing the team (Sir Michael Latham, 1995) — identified the potential for reductions of up to 30% in construction costs. These developments point to targets that are tougher than those assumed in 1994, particularly for the less efficient companies. In Table 6, three efficiency scenarios have been developed to illustrate the possible scale of expenditure reductions over the next price limit period. In Chapter 5, there is a year by year breakdown of these scenarios. At this stage of the Periodic Review it is desirable to look at a wide variety of options. These have been grouped together in three scenarios. These are still subject to considerable uncertainty. For example the correct balance between industry target and catch-up is uncertain. The scenarios can be described as follows: All three scenarios are at least as tough as those targets set at the 1994 price review. Scenario 1 — Steady pressure. Scope for efficiency improvements similar to that assumed in 1994 for the period 1995–96 to 1999–2000. This assumes that studies will show that the water industry has moved forward at a similar pace to improvements in the economy as a whole and that there remains potential for further savings. This implies about 2–2½% per year overall improvement for opex and capital maintenance split broadly equally between a general efficiency assessment and the company specific efficiency catch-ups. It is anticipated that for the most efficient companies a slightly lower general efficiency assessment might be applied. Scenario 2 — Challenging. Scope for efficiency improvements greater than assumed in 1994. This assumes that studies will show that the water industry has lagged behind productivity improvements elsewhere in the economy. This scenario is supported by the gains made in the other utilities that have been exposed to more direct market competition and the claims made by some companies about savings promised during the current price limits. It implies overall improvement of around 3–3½% per year for opex and capital maintenance. Scenario 3 — Aggressive. Assumes companies are far from exhausting the scope for efficiency improvements. This assumes that studies will show that the water industry has lagged far behind productivity improvements elsewhere in the economy. Benchmarking studies supporting this scenario would be expected to show the continued presence of old practices that have not been subject to the rigours of direct market competition. Overall the judgement would be that there is an accelerating efficiency gap between a cosseted monopoly service and external comparator industries. This gap would need to be addressed by a more aggressive approach to the X factor for the period 2000–05. It would imply improvements of 4–5% per year for opex and capital maintenance. To carry out this analysis at this stage of the Review, a number of very broad assumptions have had to be made. In particular, the scale of expenditure reductions resulting from less efficient companies catching up depends on both the size of the gap between the most and least efficient, and on how many companies are in each efficiency band. The results of the scenario analysis are set down in Chapter 5. The three scenarios for each of the expenditure areas are summarised in Table 6 (on page 18). These numbers need to be treated as illustrative and with extreme caution. They indicate the possible range of efficiency savings which might be achieved by the 1999 Periodic Review. However there is still a lot of work to be done by Ofwat and the companies to refine this analysis. These efficiency scenarios have been developed on a set of assumptions in each of the three expenditure areas, operating costs, capital maintenance and capital enhancement. The assumptions for each area are: The scope for general efficiency, that is for all companies steady pressure 1% challenging 1½% aggressive 2% These assumptions may be softened for leading companies. The additional scope for company specific adjustments dependent on assessment of each company's efficiency compared with that of the most efficient companies. Each scenario assumes different amounts of catch-up as shown in Table 5: Table 5: Catch-up assumptions for the three scenarios (%)   Opex Capital maintenance Capital enhancement Steady pressure 50 30 50 Challenging 60 40 60 Aggressive 70 50 70 The timing of the delivery of company specific improvements can be either flat over the five year period or concentrated in the early years (front loaded). In the steady pressure scenario, the improvements are assumed to be flat, while in the challenging and aggressive scenarios they are front loaded. In 1994, company specific targets to lessen the efficiency gap were equivalent to 1% per year reduction in opex, for the industry as a whole. The corresponding annual reductions in the three scenarios are around 1½%, 2% and 2½% respectively, although these figures might be different if the distribution of large companies is more towards one end of the efficiency range than the other. In scenarios 2 and 3, the targets would close the efficiency gap faster than in scenario 1 – they are front loaded rather than flat. In 1994 flat targets were set for opex. For capital maintenance, the £400 million reduction illustrated in scenario 1 in Table 6 is a similar level of efficiency to that assumed in current price limits. As with opex, scenarios 2 and 3 for capital maintenance assume that a greater portion of the efficiency gap between companies can be caught up and that this catch-up occurs earlier in the quinquennium. This leads to combined targets for both general and company catch-up equivalent to 3% per annum and 4% per annum respectively. For capital enhancement, scenario 1 also contains similar assumptions to those in current price limits which would result in a 7% saving over the period 2000–05. Scenarios 2 and 3 assume 60% and 70% catch-up respectively, leading to 11% and 13% savings. Table 6: Illustrative impact of different efficiency scenarios on reductions in expenditure requirements 2.9 Using X to finance Q, S and V   Illustrative description of the efficiency scenarios Illustrative impact on expenditure requirements Expenditure category Scenario General Efficiency improvement Elimination of gap to leading comparators % Profile Total reductions over the period 2000–05 £m or % Reduction in 2004–05 £m or % Operating expenditure Scenario 1 – Steady pressure 1% 50% Flat £900 £300 Scenario 2 – Challenging 1.5% 60% Front loaded £1,300 £400 Scenario 3 – Aggressive 2% 70% Front loaded £1,700 £500 Capital maintenance expenditure Scenario 1 – Steady pressure 1% 30% Flat £400 £150 Scenario 2 — Challenging 1.5% 40% Front loaded £700 £200 Scenario 3 — Aggressive 2% 50% Front loaded £900 £250 Capital enhancement expenditure Scenario 1 – Steady pressure 1% 50% Flat 7% saving 12% saving Scenario 2 — Challenging 1.5% 60% Front loaded 11% saving 16% saving Scenario 3 — Aggressive 2% 70% Front loaded 13% saving 20% saving The Director has indicated that improvements in water company efficiency following the Po adjustment would be used to mitigate the upward pressures on customers' bills arising from further quality improvements (Q), other service improvements (S) and ensuring continuing security of supplies (V). Ofwat will be addressing these trade-offs, at both an industry and company level, in Prospects for prices (October 1998). 2.10 Key dates in taking forward efficiency issues This paper sets down the Director's proposed approach to efficiency issues. For the Prospects for prices paper (October 1998) the Director will make assumptions on appropriate ranges for efficiency savings at the next review. These assumptions will take account of: responses to this paper; the outcome of research into the scope for general efficiency; and company results for the year 1997–98. Following publication of Prospects for prices, companies will have an opportunity to discuss their relative efficiency with Ofwat before submitting their views on the scope for general efficiency and future savings within their own companies as part of the Draft Business Plan. Finally, the Director will take judgements based on company submissions and 1998–99 results to product his draft determinations. Companies and other interested parties will have an opportunity to comment on these before they are finalised in November 1999. A more detailed timetable is set out in Appendix 7.

3. APPROACH TO EFFICIENCY AT THE 1994 PERIODIC REVIEW 3.1 Overview In 1994 the Director considered separately the issues of operating and capital efficiencies. For each he made an assessment of the general scope for savings across the industry and an assessment of the different relative efficiencies of companies within the industry. This section contains an outline of the key elements of the approach the Director took, showing how it achieved a balance of incentives. The strategy was subject to detailed scrutiny by the Monopolies and Mergers Commission (MMC) in the South West Water and Portsmouth Water referrals on the Director's determinations. The MMC endorsed the strategy used by Ofwat. The current framework has evolved from this base. 3.2 How the overall scope for efficiency savings was assessed The Director's assessments of the overall scope for savings across the companies formed the basis for annual efficiency targets that were built into price limits. This gave companies incentives to outperform targets, but companies could not raise prices through failure to meet these targets. 3.2.1 Operating expenditure efficiency The Director assessed the overall scope for efficiency savings on operating expenditure in the water industry as 2% per annum for 1995–2000 and 1% per annum for the subsequent five years. The Director commissioned a report on general efficiency trends from Bosworth, Stoneman & Roe. They calculated that the reduction in real unit operating cost in productive and manufacturing industries over the period 1979 to 1989 was around 1–1.25% a year. Quantification was difficult, but the evidence was consistent with a judgement that real reductions in unit operating costs in the wider economy were 1% a year greater than measured, after taking into account known biases in the data and unaccounted for quality improvements. This would lead to a central estimate for manufacturing industry over the period 1980–90 of 2–2.25% per annum. The Director noted that quality enhancements in the water industry were accounted for separately. Companies reported to the Director the extra spending they considered necessary to meet new quality obligations and this was scrutinised by the Director separately (see 3.2.2 and 3.3). Other industries would not make the same distinction. The limited evidence available suggested that this distinction between the water industry and others might be significant. It was therefore assumed that the overall scope for efficiency savings in opex would be 2% and that all companies would be able to achieve at least a 1% efficiency saving. Appendix 5 provides a summary of the Bosworth, Stoneman & Roe report. A full copy of this report has been placed in the Ofwat Library. 3.2.2 Capital efficiencies across the industry An assessment was made of an industry-wide target for continuing capital productivity efficiency arising from technological progress, following two separate studies of the scope for widescale introduction of new innovations. A conservative judgement was made, owing to uncertainties in the timing and take up of newer and emerging technologies, and it was assumed, when price limits were set, that there would be scope for continuing reductions of 1% per year in capital expenditure. As a result, companies had incentives for early adoption of lower cost technologies and practices. The 1% per annum target was applied to both capital maintenance and capital enhancement expenditure, in addition to the reduction implied by comparative analysis of capital unit costs in the cost base exercise. The additional adjustments arising from the cost base ranged between 0-7.5% in total over the five year period. This resulted in the expectation that there would be an average reduction in total capital costs across the industry of 2-2.5% per annum. 3.3 Assessing relative efficiency between companies Within the overall framework of incentives to improve efficiency, Ofwat recognised that the scope for improvement was not the same for all companies, because of differences in starting levels of efficiency. To take this into account in price limits, Ofwat assessed relative efficiency for each company, for operating expenditure and for capital expenditure. 3.3.1 Assessing relative operating expenditure efficiency Companies' relative operating expenditure efficiency was assessed using a combination of econometric techniques, as described in Appendix 1, and adjustments for non-econometric effects, such as significant differences in levels of service or particular circumstances of single companies. Econometric models were developed for the water service as a whole, resources and treatment, distribution and business activities. Each company was given an efficiency ranking for each activity and an overall assessment was reached. For the sewerage service, econometric models were developed for operation of the sewer network and for large sewage treatment works. Assessments of the comparative costs of small sewage treatment works and sludge treatment used unit costs but took into account works size, load and treatment levels. For sludge disposal, unit costs were used, taking into account the various disposal methods. Business activity performance was assessed using costs per property billed. The operating expenditures predicted by each sub-service model were summed for each company. The average of these sums with the results of the overall service models was calculated — it was considered that both approaches had equal validity. These averaged predicted expenditures were compared with actual reported expenditures, less local authority rates, third party costs and National Rivers Authority charges, none of which were modelled. Companies were ranked on percentage differences. The ranked companies were each allocated a provisional expenditure band. The bands were determined so that companies with similar reported expenditure relative to predicted expenditure were allocated to the same band. Converting from provisional expenditure bands to final efficiency bands required individual company circumstances to be taken into account, either quantitatively, to accommodate a specific cost, or qualitatively, based on an assessment of quality of service, trends in expenditure or other specific factors. It was recognised that the differences between predicted and actual expenditure, even after adjustment for specific factors, did not translate directly to differences in efficiency. Some of the differences were the result of errors in the data used to model expenditure and especially in the models themselves, as simple models could not take all factors into account. The approach adopted was to set company specific efficiency targets that would move individual company expenditure towards those of the best performers, over a five year period. The amount of movement was taken to be around 25% – 35% of the difference in predicted cost. In addition to the company specific targets, all companies had a target of 1% per year for reducing costs, resulting from the overall scope for efficiency as discussed earlier. For the best performers, only the 1% overall target was set but, for those with the highest expenditure, the combined target was a 3.5% expenditure reduction per year. 3.3.2 Assessing relative capital maintenance efficiency During the Review Ofwat developed a range of analytical tools to interpret the information submitted by companies in their Strategic Business Plans. These were used to form a view of the levels of capital maintenance expenditure that would need to be required by companies over the ten year period 1995–2005. The main tool used to assess comparative efficiency for both capital maintenance expenditure and enhancements was the comparative capital unit cost approach — the cost base. This was developed so that the Director could compare capital expenditure across the industry and identify those companies that appeared to be more efficient at procuring capital assets than others. Companies with higher capital unit costs were considered to have more scope for savings in their expenditure projections than companies with lower capital unit costs.   4.RECENT RESULTS IN COMPARATIVE EXPENDITURE 4.1 Expenditure trends since the 1994 Periodic Review The majority of companies have, so far, been successful in reducing operating expenditure by a significant amount. Total opex decreased by 6% between 1992–93 and 1996–97, in spite of the new costs of operating additional water and sewage treatment plants to meet European Union directives. Without these new costs, companies report that expenditure would have decreased by 12% in the same period. Ofwat published comparisons of 1992–93 and 1996–97 operating expenditure in the 1996–97 Report on water and sewerage company operating costs and efficiency (December 1997). The trends in capital maintenance expenditure are less clear. Large variations in capital maintenance expenditure can occur on an annual basis, as it becomes necessary to renovate or replace major assets. In recent years, reported capital maintenance expenditure has generally been in line with the amounts assumed in price limits. However, a number of companies are reporting savings in unit costs of up to 10%. This would suggest that the reductions in capital unit costs that have been claimed have been offset by higher overall levels of capital maintenance activity. For capital enhancement expenditure, comparisons can be made between the actual expenditure of the companies and the amount assumed in price limits. The main component of the capital enhancement programme for both water and sewerage services is expenditure to meet new quality standards. Companies are currently reporting that they have made considerable savings in capital enhancement expenditure compared with the amounts that were assumed in price limits. At the end of 1996–97, there was a considerable cumulative difference between the expenditure that was assumed in price limits and actual expenditure by the companies. However, companies are reporting that a substantial part of this difference will be spent later in the quinquennium and that the overall difference on the quality programme by the end of the five years to 2000 is likely to be of the order of 15%. The sources of these efficiency savings in capital expenditure vary between companies. A large proportion of the savings are reported by companies to be due to uptake of new construction techniques. Examples of this have been: more sophisticated measurement of flows at sewage treatment works, use of Granular Activated Carbon sandwich techniques for water treatment, greater use of no-dig techniques for mains and sewer renewal and general better management of contracts and resources. In addition, companies with relatively high capital unit costs, as measured by the cost base, may have approached the 1994 yardsticks at a faster rate than was assumed when price limits were set. The industry cost base submission in June 1998 will enable Ofwat to measure the progress made by individual companies in reducing their capital unit costs during the current price review period. 4.2 Results of the econometric analysis 4.2.1 Water and sewerage services There are important differences between the water and sewerage services that affect how comparative performance is assessed. For the water service, it is generally possible to derive robust econometric models for explaining expenditure, because there are 28 separate companies, a sufficient number of comparators. In contrast, there are only ten separate sewerage companies. For these, Ofwat uses disaggregated data, for example for several sub-areas of each company or for individual large treatment works. With disaggregated data it is not usually possible to assess the role of economies of scale. However, the range of sizes of the sewerage companies is much less than that of the water companies. It is important that expenditure is allocated reliably between the water and sewerage services because misallocations can affect assessments of economies of scale which, in turn, could affect how efficiency is assessed for the smaller water only companies. 4.2.2 Operating expenditure econometric models The econometric models used at the 1994 Periodic Review were published in the 1993–94 Report on the cost of water delivered and sewage collected (December 1994). The models resulted from extensive research and consultation with the industry. Some of the early research was published as research papers in 1993 and 1994 (copies available in the Ofwat Library). For the water service, the econometric models have been re-estimated each year since the 1994 Review. The form of the resources and treatment and business activities models have not changed, only the coefficients have altered. The water distribution model has been reformulated (see Appendix 2). An econometric model for power expenditure has replaced the previous unit cost approach. This was found necessary because of apparent economies of scale on power expenditure. The combined sub-service models are now judged to be more robust than the overall water service models. This is mainly because expenditure is now allocated more reliably than in 1992–93. It is intended to use an overall model only as a check on the results of the sub-service models. Data was collected for sewerage areas and individual large sewage treatment works for 1996–97, the first such data collection since 1992–93. The sewer network model and large treatment works model have been revised and improved. For other parts of the sewerage service a unit cost approach is retained, using similar calculations to those used in the 1994 Periodic Review. Ofwat is aware of the potential for new explanatory factors to become material, for example with the increase in domestic metering and new requirements for water treatment, and will be examining 1997–98 data for evidence of this. Some factors are suspected but prove difficult or impossible to measure consistently, for example the split between urban and rural areas within companies. Where companies can demonstrate that a particular explanatory factor is a material cost driver across the industry and can be measured consistently, Ofwat will consider its use, if its inclusion leads to a better model on engineering and statistical grounds. 4.2.3 Capital maintenance expenditure econometric models Companies must renew their assets to ensure continuity of services for current and future customers. Ideally, an asset should only be renewed when the present value of additional operating expenditure required to maintain serviceability becomes greater than the capital cost of renewal. To achieve the optimum level of capital maintenance, companies must carry out the right capital maintenance activity at the right cost on the right assets at the right time. The approach to capital maintenance at the 1999 Periodic Review starts with serviceability to customers. Before making any assessment of comparative efficiency based on historic expenditure, Ofwat will assess whether the trend in serviceability to customers has deteriorated. Declining serviceability may be indicative that the company has failed to carry out sufficient capital maintenance — its activity may be too low, ineffective, or a combination of both. The company will be challenged to explain its position. Stable trends in serviceability will be indicative that recent capital maintenance activity has been adequate. Companies will also be submitting an update of their asset inventory in August 1998. This will enable the changes in the condition and operational performance of their assets from 1992–93 to 1997–98 to be identified. Account of capital enhancement expenditure during this period will need to be taken so that the state of the underlying asset base can be assessed. Where there is evidence that assets have deteriorated companies will be challenged to explain why this is so. This may indicate that recent levels of capital maintenance expenditure have not been sufficient. Unless serviceability is deteriorating, current levels of capital maintenance expenditure can be taken to be a reasonable reflection of the requirement for capital maintenance activity over the period of the next price limits. There is a wide range of unit capital maintenance costs across the industry, suggesting that some companies are more effective at maintaining serviceability to customers than others. This range of unit costs may be attributable to variations in the effectiveness with which companies target their capital maintenance expenditure leading to inappropriately high levels of activity. Companies' capital maintenance strategy and unit costs have been assessed using the same statistical techniques as those used in the approach to operating expenditure. In common with the opex models, the capital maintenance econometric models inform judgements about how far individual companies are from the best in the industry. An important difference between the capital maintenance and opex econometrics is that opex models take annual data for both expenditure and explanatory factors. The capital maintenance models currently take an average of the capital maintenance expenditure over the four year period 1993–94 to 1996–97, and explanatory factors that relate to 1992–93. This is in order to take account of annual variations in capital maintenance expenditure and ensures that the explanatory factors are not affected by modelled expenditure. The capital maintenance econometric models developed from company data are set down in Appendix 3. This data has been derived from the Capital maintenance return (November 1997). A copy of the datasets used to develop the econometric models presented in this report is available from the Ofwat library. In common with the opex models, the capital maintenance models have also been developed using the principles outlined in Appendix 1. 4.2.4 Robustness of the econometric models The econometric models for both opex and capital maintenance predict a level of cost for a particular company based on the performance of the other companies in the industry. These predicted costs can be compared with companies' actual costs and companies can then be grouped into cost variation bands according to the size of the difference between actual and predicted costs. Some of the difference may be the result of shortcomings in the models as well as comparative cost performance. Whilst accepting this limitation, the econometric modelling of reported company expenditure, their outputs and statistically significant explanatory factors, reveals wide ranges of cost performance. High cost companies will need to be able to demonstrate why their costs are higher than other companies in similar situations. Table 7 summarises the results of the econometrics using a consistent banding convention linked to the gap between reported expenditure and the central estimate of predicted expenditure from the relevant model. Bandings for both operating expenditure and capital maintenance expenditure are shown in the table. The banding convention is: Band A: well below predicted expenditure (less than 85% of C) Band B: below predicted expenditure (85–95% of C) Band C: around predicted expenditure (within 5% of modelled expenditure) Band D: above predicted expenditure (105–115% of C) Band E: well above predicted expenditure (more than 115% of C) The results presented compare predicted expenditure with reported expenditure. They do not take account of legitimate, company specific factors that may affect expenditure. The bands, therefore, are not definitive statements of the relative efficiency of individual companies. It should be noted that opex accounts for between 50% and 80% of the combined total of opex and capital maintenance. Appendix 4 sets out the results for each of the econometric models developed for the water and sewerage services. These are presented in matrix form, showing comparisons between operating expenditure bandings and capital maintenance expenditure bandings. Table 7: Results of the econometric modelling summarised for the water and sewerage services   Water service Sewerage service Company Operating expenditure banding Capital maintenance expenditure banding Operating expenditure banding Capital maintenance expenditure banding Water and sewerage companies Anglian C E C A Dwr Cymru E D D A North West B C C D Northumbrian C D C C Severn Trent C E C E South West D C E E Southern D A E C Thames C E C E Wessex B E B E Yorkshire C E B A Water only companies Bournemouth & West Hampshire B E     Bristol E B     Cambridge B A     Chester C C     Essex & Suffolk E B     Folkestone & Dover D D     Hartlepool D E     Mid Kent C D     Mid Southern B B     North Surrey D A     Portsmouth A E     South East E E     South Staffordshire C C     Sutton & East Surrey C E     Tendring Hundred E B     Three Valleys C C     Wrexham A B     York A A     4.3 Limitations of the analysis The analysis and results in this report are preliminary. Much more analysis remains to be done, to take into account, for example: suggestions from companies; additional data, for 1997–98, which companies will report in July 1998; and company specific factors not covered by the models. 4.4 Total efficiency and interactions between operating expenditure and capital maintenance expenditure As shown in the matrices in Chapter 2 and Appendix 4, there are a number of cases where companies appear efficient on opex and inefficient on capital maintenance, or vice versa. This raises the wider question of whether it is appropriate to consider different categories of expenditure separately and whether the system of analysis sets up perverse incentives to companies. For example, companies may have the scope to make investments that reduce operating costs. Such investments may be classified as capital maintenance. Some companies might adopt strategies where they spend extra on capital maintenance in order to reduce operating expenditure. An active policy of leakage detection and repair is an example of this. Overlap also occurs because of accounting practices which affect how costs, such as leakage control, are allocated between operating expenditure and capital maintenance. It is essential that the overall framework for judging comparative efficiency maintains the incentive to undertake cost saving investments as these become viable in net present value terms. It is also important that different practices in allocation do not affect the overall adjustments resulting from the structure proposed above. It is, therefore, considered appropriate to assess efficiency on operating expenditure and capital maintenance together. This approach helps to ensure that companies are not unduly penalised, or rewarded, because of the way they choose to minimise expenditure, or the way in which they allocate expenditure. In practice, it means that a company's band for operating efficiency could be adjusted if it is an outlier on capital maintenance efficiency and vice versa. Such adjustments would take into account the actual expenditure involved. An alternative to this framework would be an aggregate approach to efficiency involving total efficiency measures. Companies and others commented to the Director in response to his paper The proposed framework and approach to the 1999 Periodic Review (June 1997) that total efficiency, or at least the interaction between operating and capital efficiencies, needed to be examined. There were concerns about the incentives provided to companies by treating them separately. The Director considered the issue of total efficiency in 1996 to see if the difficulties with this method could be overcome. Bosworth, Stoneman & Thanassoulis conducted a study into the feasibility of assessing total efficiency through production functions. This concluded that capital expenditure should be considered in assessing efficiency, but that defining appropriate inputs and outputs for the industry was problematic and that defining data requirements and collecting such data was very difficult. There are a considerable number of factors beyond management control, for example geographic areas and the inherited capital stock, that would need to be adjusted for. Although the research concluded that some data envelopment analysis might be possible, there are practical difficulties in collecting sufficient data of good quality to an appropriate degree of detail. This leads the Director to conclude that separate efficiency comparisons using operating expenditure and capital maintenance expenditure are likely to lead to more robust conclusions than analysis undermined by patchy and unreliable data. He will, therefore, assess these separately but base his judgements on the total of opex and capital maintenance analyses. Furthermore, having considered these comments and the research, the Director considers that the best way to identify any interaction of capital maintenance and operating efficiencies is to assess comparative efficiency in both areas using similar econometric techniques. The relationship between the two assessments will then be reviewed and an allowance for any significant interaction, where it can be identified, would be made when future efficiency targets are set. A summary of the study is given in Appendix 6. A copy of the full report is available in the Ofwat Library.

5. IMPLICATIONS OF RELATIVE EFFICIENCY RANKINGS AND THE SCOPE FOR EFFICIENCY IMPROVEMENTS 5.1 Implications of the work for the X factor Ofwat has looked at the possible implications for expenditure of different judgements on the robustness of the relative efficiency rankings and the overall scope for efficiency improvements by the industry as a guide to understanding the impact of the X factor. This section describes these simulations. 5.2 Assessing the scope for future savings in operating expenditure and capital maintenance expenditure together As the Director stated in Setting price limits (February 1998), he intends to treat efficiency savings in operating expenditure and capital maintenance expenditure cumulatively. For those companies who are equally efficient or inefficient in both areas their operating expenditure and capital maintenance can be treated independently. Section 2.8 set down three scenarios which varied in terms of general improvement and company specific efficiency assumptions. The general improvement assumptions vary between 1% (steady pressure scenario) and 2% (aggressive scenario) while the company specific assumptions vary in both the degree and timing of the catch-up to the leading comparators. 5.2.1 Relationship between company specific efficiency targets for operating costs and capital maintenance This section considers the interaction between company specific factors for operating expenditure and capital maintenance and how the Director intends to take account of this at the Periodic Review. Where companies are efficient in one area and inefficient in the other it is presumed that this is a result of a conscious decision to increase spending in their inefficient area in order to be efficient in the other. In other words Ofwat recognises that the separate measures of efficiency in the two areas may be too harsh. For this reason it is proposed to give such companies less tough efficiency targets for their inefficient area than would otherwise be the case. That is, they are not to be penalised for decisions which make them leading comparators in one area at the expense of apparent inefficiency elsewhere. Taking scenario 2 as an illustration, companies would be expected to move 60% towards the benchmark company in operating expenditure and 40% towards the benchmark company in capital maintenance. Table 8 illustrates the percentage changes over the five year period which might be expected. Similar tables can be derived for other scenarios. Note that for those companies falling in the top left or bottom right hand corners, the targets for their inefficient areas are relatively softer than for companies falling elsewhere in the matrix. This has not made a significant difference to the industry wide figures presented in this report. In addition to the percentage changes in Table 8, all companies will also be set a general efficiency target for both opex and capex as set out in Table 6. 5.2.2 Timing of catch-up to leading companies The degree of catch-up to the leading comparators for opex ranges between 50% (steady pressure scenario) and 70% (aggressive scenario) and between 30% (steady pressure scenario) and 50% (aggressive scenario) for capital maintenance. The timing of the catch-up for the company specific assumptions can affect the total savings expected. Two timing options have been adopted, one assumes the improvements to be achieved evenly over the period while the more challenging option is for the improvements to occur early in the period. Table 9 below sets out the timing assumptions for the more challenging front loaded scenario. Table 9: The front loaded option Illustrative apportionment of catch-up through price limit period Year 2000–01 2001–02 2002–03 2003–04 2004–05 Proportion of catch-up 30% 30% 20% 10% 10% The potential for savings range from £430 million for capital maintenance and £920 million for opex under scenario 1 (steady pressure) to £850 million for capital maintenance and £1,660 million under the aggressive assumptions in scenario 3. 5.2.3 Detailed profiles of efficiency savings in operating costs and capital maintenance Tables 10, 11 and 12 illustrate the break down of the efficiency savings that would be available using the steady pressure, challenging and aggressive efficiency assumptions as set out in Table 6. It should be remembered that these scenarios are only illustrative at this stage. The Director will consider a wide range of options before setting efficiency targets. He may decide to go further in some areas than others. Table 10: Illustrative impact of 'X' factors on industry expenditure for the price limit period Scenario 1 – Steady pressure (Assuming a normal distribution of companies across the relative efficiency scale (by turnover)) Water service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 1 Opex savings — company specific assumptions 20 50 70 90 110 340 2 Opex savings — general improvement assumption (1% / year) 10 30 40 60 70 210 Opex savings 40 70 110 150 180 550 3 Capex (maintenance) savings — company specific assumptions 10 10 20 30 30 100 4 Capex (maintenance) savings — general improvement assumption (1% / year) 10 10 20 30 40 110 Capex (maintenance) savings 10 30 40 60 70 210 Total water service savings 50 100 150 200 250 760 Sewerage service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 1 Opex savings — company specific assumptions 20 30 50 60 80 230 2 Opex savings — general improvement assumptions (1% / year) 10 20 30 40 50 140 Opex savings 30 50 70 100 120 370 3 Capex (maintenance) savings — company specific assumptions 10 10 20 30 40 110 4 Capex (maintenance) savings — general improvement assumptions (1% / year) 10 20 20 30 40 110 Capex (maintenance) savings 10 30 40 60 70 220 Total sewerage service savings 40 80 120 160 200 590 Total savings — both services 90 180 270 360 450 1,360 Note: numbers may not add exactly because of rounding. Table 11: Illustrative impact of 'X' factors on industry expenditure for the price limit period Scenario 2 – Challenging (Assuming a normal distribution of companies across the relative efficiency scale (by turnover)) Water service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 1 Opex savings — company specific assumptions 40 80 110 120 140 490 2 Opex savings — general improvement assumption (1½% / year) 20 40 60 80 100 320 Opex savings 60 120 170 210 240 800 3 Capex (maintenance) savings — company specific assumptions 10 30 40 40 50 170 4 Capex (maintenance) savings — general improvement assumption (1½% / year) 10 20 30 40 50 160 Capex (maintenance) savings 20 50 70 80 100 330 Total water service savings 90 170 240 290 340 1,130 Sewerage service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 1 Opex savings — company specific assumptions 30 60 70 80 90 330 2 Opex savings — general improvement assumptions (1½% / year) 10 30 40 60 70 220 Opex savings 40 80 120 140 160 550 3 Capex (maintenance) savings — company specific assumptions 10 30 40 40 50 170 4 Capex (maintenance) savings — general improvement assumptions (1½% / year) 10 20 30 40 60 170 Capex (maintenance) savings 30 50 70 90 100 340 Total sewerage service savings 70 140 190 230 270 880 Total savings — both services 150 310 430 520 600 2,020 Note: numbers may not add exactly because of rounding. Table 12: Illustrative impact of "X" factors on industry expenditure for the price limit period Scenario 3 – Aggressive (Assuming a normal distribution of companies across the relative efficiency scale (by turnover)) Water service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 1 Opex savings — company specific assumptions 50 90 130 140 160 570 2 Opex savings — general improvement assumption~(2% / year) 30 60 80 110 140 420 Opex savings 80 150 210 250 300 990 3 Capex (maintenance) savings — company specific assumptions 20 30 50 50 60 210 4 Capex (maintenance) savings — general improvement assumption (2% / year) 10 30 40 60 70 220 Capex (maintenance) savings 30 60 90 110 130 420 Total water service savings 110 220 300 360 420 1,410 Sewerage service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 1 Opex savings — company specific assumptions 30 60 90 100 110 390 2 Opex savings — general improvement assumptions (2% / year) 20 40 60 80 90 290 Opex savings 50 100 140 170 200 670 3 Capex (maintenance) savings — company specific assumptions 20 40 50 50 60 210 4 Capex (maintenance) savings — general improvement assumptions (2% / year) 20 30 40 60 70 220 Capex (maintenance) savings 30 70 90 110 130 430 Total sewerage service savings 80 170 240 280 330 1,110 Total savings — both services 190 380 540 650 760 2,520 Note: numbers may not add exactly because of rounding. 5.3 The effect of efficiency adjustments to capital enhancement expenditure In addition to efficiency improvements in operating and capital maintenance expenditure, efficiencies are also expected in capital enhancement expenditure. These will be assessed using comparative capital unit costs in the cost base exercise. Expectations of continuing capital productivity will also apply, following results of further work. Tables 13–15 set down the effect of these two adjustments on an illustrative enhancements programme of £5 billion over the five years (comprising £2 billion water and £3 billion sewerage) for the three scenarios in Table 6, Section 2.8. The capital enhancement scenarios in the tables assume, as in 1994, that capital unit costs of an average company are 15% higher than the lower quartile yardsticks and that companies will be expected to close some of this gap over the five year period. The three scenarios differ in the size of the differential that companies are expected to be closed (between 50 and 70%) as well as the timing of the delivery of these savings. The scenarios also include the effects of assumptions for general improvements in efficiency of between 1 and 2%. The potential for savings on a projected capital enhancement programme of £5 billion range between £370 million under scenario 1 to £650 million under the assumptions in scenario 3. Table 13: Illustrative impact of "X" factors on enhancement expenditure for the price limit period Scenario 1— Steady pressure Water service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 3 Capex (enhancement) savings — company specific assumptions 5 10 20 20 30 90 4 Capex (enhancement) savings — general improvement assumption (1% / year) 5 10 10 20 20 60 Capex (enhancement) savings: water 10 20 30 40 50 150 Sewerage service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 3 Capex (enhancement) savings —company specific assumptions 10 20 30 40 50 140 4 Capex (enhancement) savings — general improvement assumptions (1% / year) 5 10 20 20 30 90 Capex (enhancement) savings: sewerage 15 30 40 60 70 220 Total savings — both services 30 50 70 100 120 370 Note: numbers may not add exactly because of rounding. Table 14: Illustrative impact of "X" factors on enhancement expenditure for the price limit period Scenario 2 - Challenging Water service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 3 Capex (enhancement) savings — company specific assumptions 10 20 30 30 40 130 4 Capex (enhancement) savings — general improvement assumption (1.5% / year) 5 10 20 20 30 80 Capex (enhancement) savings: water 20 30 50 50 60 210 Sewerage service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 3 Capex (enhancement) savings —company specific assumptions 20 30 40 50 50 190 4 Capex (enhancement) savings — general improvement assumptions (1.5% / year) 10 20 30 30 40 130 Capex (enhancement) savings: sewerage 30 50 70 80 90 320 Total savings — both services 40 80 110 140 160 530 Note: numbers may not add exactly because of rounding. Table 15: Illustrative impact of "X" factors on enhancement expenditure for the price limit period Scenario 3 — Aggressive Water service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 3 Capex (enhancement) savings — company specific assumptions 10 20 30 40 40 150 4 Capex (enhancement) savings — general improvement assumption (2% / year) 10 20 20 30 40 110 Capex (enhancement) savings: water 20 40 60 70 80 260 Sewerage service savings year by year 2000–01 £m 2001–02 £m 2002–03 £m 2003–04 £m 2004–05 £m Five year total £m 3 Capex (enhancement) savings —company specific assumptions 20 40 50 60 60 230 4 Capex (enhancement) savings — general improvement assumptions (2% / year) 10 20 30 40 50 160 Capex (enhancement) savings: sewerage 30 60 80 100 120 390 Total savings — both services 50 100 140 170 200 650 Note: numbers may not add exactly because of rounding. Appendix 1: ECONOMETRIC APPROACH 1.1 Choice of statistical method Ofwat uses a number of tools to compare the relative efficiency of the water and sewerage companies. Direct comparisons of unit costs across the industry are simple and straightforward. They can be very useful but do not take account of differences in operating environment and service performance. These differences may explain why some companies should have higher or lower expenditure than others. Statistical techniques, such as multiple regression, provide a means to assess the impact of different operating environments. These statistical techniques are called econometrics. Ofwat and Professor Stewart (University of Warwick) developed econometric models for the 1994 Periodic Review. The econometric models can take into account factors that describe the size and operating environment of different companies. These models require a larger amount of data than simple unit cost comparisons and consistency between companies, to ensure comparisons are fair. There are some factors which are difficult to quantify in terms of expenditure or value to customers, such as the levels of service provided by a company. There are other factors which are company specific or affect the ability of a company to achieve efficiency savings. These are not easy to incorporate into an econometric model. Nevertheless, these factors are important and can be taken into account by making adjustments to the results of unit cost analysis or econometric models. In the past, Ofwat has made such adjustments in producing relative efficiency assessments. It is important to bear in mind that no method of assessing efficiency is perfect. An element of judgement will always be involved in arriving at the final answer, regardless of the technique used. There are a number of different approaches to generating econometric models. The two main approaches are stochastic frontier models and regression analysis. The former determines how close each company is to the expenditure achieved by the best in the industry, which provides a benchmark for comparisons. The latter generates the average cost in the industry. Companies with expenditure above or below the average are taken to be less efficient or more efficient respectively. Ofwat explored both stochastic frontier and regression approaches to assessing operating cost efficiency in a number of research papers published for the 1994 Periodic Review. Ofwat concluded that the regression approach was the most appropriate. This is because stochastic frontier models rely on a number of assumptions about the form of the relationship between expenditure and explanatory factors. These assumptions may not hold for the information collected from the water companies. The regression approach does not require such strong assumptions. However, it is assumed that once the explanatory factors are taken into account, the difference between a company's actual expenditure and that predicted by the model, the 'residual', provides a measure of its efficiency. There are alternatives to these two approaches, such as panel data analysis and data envelopment analysis (DEA). They also have limitations. Panel data analysis, for example, uses time series data and so requires assumptions about how expenditure or efficiency varies over time. In 1994 Ofwat used DEA to confirm the results of the operating cost regression models. These alternative analyses provide a useful challenge to the results of the preferred method, regression analysis. It is important to use explanatory factors that are outside a company's immediate control in the econometric models. Econometric models must not rely on cost drivers that include measures of current inefficiency. It is possible to develop statistically valid models through analysis of the data, which do not reflect the way in which the industry operates. It is therefore important that the models make sense, in the identification of cost drivers and explanatory factors, to the industry as a whole. 1.2 Deriving robust statistical models Deriving satisfactory statistical models requires a systematic approach. The step by step approach adopted is set down in Table 1.1. Table 1.1: Step by step approach used to derive the statistical models 1 Expert review of potential cost drivers 2 Data collection and its validation 3 Identification of atypical expenditure and exceptional items 4 Produce revised data for statistical analysis 5 Generate plausible conceptual models to limit statistical analysis 6 Statistical analysis to generate robust relationships between expenditure and explanatory factors 7 Expert review of the statistical models 8 Preliminary assessments of relative cost 9 Review company specific factors to assess validity and impact on preliminary judgements 10 Review results of parallel analysis and assess impact on preliminary judgements 11 Finalise judgements on relative efficiency In practice, there are a number of iterations through steps 5, 6 and 7 as models are developed and refined. 1.3 Economic approach to operating costs and capital maintenance The Ofwat approach to efficiency assessment using econometric tools was applied only to operating costs at the 1994 Review. The approach has been extended to cover capital maintenance costs for the 1999 Review. Ofwat has developed the capital maintenance models over the last two years in a joint research project with a number of companies and advised by Professor Stewart (University of Warwick). Preliminary analysis using data already available to Ofwat was carried out in 1995. This analysis was then taken forward into an extended pilot study involving data collection and discussion of cost drivers with seven companies (Anglian Water, Thames Water, Severn Trent Water, Wessex Water, Dwr Cymru (Welsh Water), Essex & Suffolk Water and North Surrey Water) during 1996. Data was collected from all companies in the November 1997 Capital Maintenance Returns (Periodic Review 1999 Information Requirement C). This data was then reviewed for consistency and used to develop the capital maintenance econometric models included in Appendix 3 of this paper. The models will be updated using data collected in the 1998 July Return and in the light of comments received from interested parties, and the results of this analysis will be included in Prospects for prices in October 1998.

Appendix 2: OPERATING EXPENDITURE MODELS 2.1 Water service The four models developed to date are summarised in Table 2.1. Table 2.1: Current water service models Sub-service Model type Explanatory variables Water distribution Log linear with constant returns to scale Distribution input less total leakage, number of connected properties Water resources and treatment Linear model for unit cost Distribution input, summary treatment measure Water power Log linear Distribution input, average pumping head Water business activities Log linear Number of billed properties 2.1.1 Resources and treatment model The summary treatment measure used in this model is a single measure of the expenditure, excluding power, for the operation and maintenance of treatment works. It is derived by applying a unit cost for works of different types and size to the proportion of distribution input in each category. Economies of scale at the level of individual works are incorporated in the unit costs. A linear rather than logarithmic form of model is considered appropriate. Distribution input is used as the scale factor, rather than water delivered, as this represents the volume of water that is treated. This approach avoids penalising those companies with justifiably above average levels of leakage, but may unfairly advantage those companies whose leakage levels are above their economic level. The current relative operating efficiency model is given in Table 2.2. Table 2.2: Water service: resources and treatment model Water service Resources and treatment expenditure Data: JR97 as issued December 1997 Modelled cost: resources and treatment functional expenditure less power expenditure, less Environment Agency service charges (£ million), divided by distribution input Explanatory variables: Coefficient Standard error Constant 4.70 3.00 Summary treatment measure (unscaled) 3.87 0.67 Form of model: Resources and treatment expenditure less Environment Agency charges and power = (4.70 + 3.87 x summary treatment measure (unscaled)) x distribution input/1000 Statistical indicators: No of observations (obs): 25 R2 0.59 This model is similar to that used at the last Periodic Review. Ofwat proposed changes to improve the precision of the summary treatment measure and consulted with the industry in December 1997. The proposals were generally welcomed and have been put in place for JR98. Ofwat will test the significance of a new category of treatment, taking in the most complex, expensive processes, to see if it is a cost driver. In fitting the model, it was necessary to remove Cambridge Water, Essex & Suffolk Water and Tendring Hundred Water because they were outliers, having unusually low or high expenditure. This may reflect company specific circumstances not covered in the model, company accounting policy in allocating expenditure, real differences in efficiency or a combination of these factors. Companies have been banded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 2.3. Table 2.3: Resources and treatment expenditure results Band Company Much lower than predicted expenditure North West, Thames, Bournemouth & West Hampshire, Cambridge, Chester, Mid Southern, Portsmouth, South Staffordshire Lower than predicted expenditure Anglian, Northumbrian, Southern, Mid Kent, York As predicted expenditure Three Valleys, Wrexham Higher than predicted expenditure Folkestone & Dover, Hartlepool, North Surrey, Sutton & East Surrey Much higher than predicted expenditure Dwr Cymru, Severn Trent, South West, Wessex, Yorkshire, Bristol, Essex & Suffolk, South East, Tendring Hundred The expenditure bands are as defined in Section 4.2.4 of this report. 2.1.2 Distribution model The distribution model used at the last Periodic Review was no longer statistically valid when tested on 1996–97 data. It used water delivered, mains length and proportion of water delivered to measured non-households as explanatory factors. There are, however, several potential cost drivers that could influence distribution expenditure and many of these are being reviewed in the light of 1996–97 data. Attention is also being paid to which measures of volume are most appropriate (distribution input, water delivered, water delivered less supplies to large users, etc). An interim model has been developed. It uses distribution input minus total leakage as a volume measure to derive unit expenditure. The explanatory factor is total connected properties divided by this volume. This model does not relate expenditure to company assets, such as mains length. It appears from initial analysis that total storage capacity may be as good a measure as mains length, but that neither are able to explain the variation in unit expenditure. No factors have so far been found that account for the high unit expenditure in some companies. The current relative operating efficiency model is given in Table 2.4. Table 2.4: Water service: distribution model Water service Distribution expenditure Data: JR97 as issued December 1997 Modelled cost: Log to base e of (distribution functional expenditure excluding power expenditure (£million), divided by distribution input less total leakage (megalitres per day (Ml/d)) Explanatory variables: Coefficient Standard error Constant -3.74 0.23 Log to base e of (total connected properties (000s) divided by distribution input less total leakage (Ml/d)) 0.68 0.36 Form of model: Log to base e of (distribution functional expenditure excluding power expenditure/(distribution input – total leakage)) = -3.74 + log to base e of (total number of connected properties (000s)/(distribution input – total leakage)) x 0.68. Statistical indicators: No of obs.: 28 R2 0.12 Companies have been banded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 2.5. Table 2.5: Distribution model results Band Company Much lower than predicted expenditure North West, Wessex, Chester, Mid Kent, North Surrey, Wrexham, York Lower than predicted expenditure Severn Trent, Essex & Suffolk, Mid Southern As predicted expenditure Anglian, South West, South East, South Staffordshire, Three Valleys Higher than predicted expenditure Northumbrian, Yorkshire, Bournemouth & West Hampshire, Bristol, Portsmouth, Sutton & East Surrey Much higher than predicted expenditure Dwr Cymru, Southern, Thames, Cambridge, Folkestone & Dover, Hartlepool, Tendring Hundred 2.1.3 Water service power model Many companies have made savings on power expenditure and it was considered that it might be useful to look at this cost separately. Another reason is that the allocation of power expenditure between distribution and resources and treatment can be problematic. For most companies, power expenditure is almost entirely for pumping, although there are exceptions to this as, for example, some water treatment processes are energy intensive. In theory, the energy used in pumping should be a linear function of volume pumped multiplied by the dynamic head. Dynamic head incorporates the extra energy needed to overcome friction, in addition to static head. However, the coefficient on distribution input multiplied by head implies significant economies of scale. This may result from companies with higher power consumption having scope to negotiate lower cost contracts for electricity. The current relative operating efficiency model is given in Table 2.6. Table 2.6: Water service: power expenditure model Water service Power expenditure Data: JR97 as issued December 1997 Modelled cost: Log to base e of power expenditure (£ million) Explanatory variables: Coefficient Standard error Constant -8.64 0.18 Distribution input (Ml/d) x average pumping head (m) 0.92 0.02 Form of model:Log to base e of power expenditure = -8.64 + (Log to base e of distribution input x average pumping head) x 0.92 Statistical indicators: No of obs.: 28 R2 0.99 The high value of R2 arises from modelling the log of total expenditure, to incorporate economies of scale. Had unit expenditure been used, the value would be considerably lower. Companies have been banded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 2.7. Table 2.7: Power model results Band Company Much lower than predicted expenditure North West, Hartlepool, Portsmouth Lower than predicted expenditure Northumbrian, Yorkshire, Bournemouth & West Hampshire, South Staffordshire, Sutton and East Surrey, Three Valleys, Wrexham As predicted expenditure Southern, Thames, Bristol, Essex & Suffolk, Folkestone & Dover, Mid Kent, North Surrey, Tendring Hundred, York Higher than predicted expenditure Dwr Cymru, South West, Wessex, Mid Southern, South East Much higher than predicted expenditure Anglian, Severn Trent, Cambridge, Chester 2.1.4 Business activities model Business activities in this context include the costs of customer services, scientific services, regulation and doubtful debts. Local authority rates are considered separately. The model is very similar to that derived on 1992–93 and subsequent data, with number of billed properties as apparently the only significant cost driver. There appear to be economies of scale, such that a company with twice the billed properties of another would be expected to have expenditure per property of 95% of the smaller company. On 1996–97 and earlier data, there appear to be no other sensible cost drivers than the number of properties billed. However, the proportion of billed properties that are metered could become a significant driver if increased domestic metering leads to material increases in operating expenditure on customer services. Ofwat will check for this in the JR98 data, and look at other classes of property (eg household, non-household). The current relative operating efficiency model is given in Table 2.8. Table 2.8: Water service: business activities model Water service Business activities expenditure Data: JR97 as issued December 1997 Modelled cost: Log to base e of business activities expenditure less local authority rates (£m) Explanatory variables: Coefficient Standard error Constant -3.93 0.28 Log to base e of number of billed properties (000s) 0.92 0.05 Form of model:Log to base e of business activities expenditure = -3.93 + log to base e of number of billed properties x 0.92 Statistical indicators: No of obs.: 28 R2 0.94 Companies have been banded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 2.9. Table 2.9: Business activities results Band Company Much lower than predicted expenditure South West, Wessex, Yorkshire, Bournemouth & West Hampshire, Essex & Suffolk, Portsmouth, Tendring Hundred, York Lower than predicted expenditure Northumbrian, Cambridge, Sutton and East Surrey, Wrexham As predicted expenditure Anglian, Severn Trent, Mid Southern, Three Valleys Higher than predicted expenditure   Much higher than predicted expenditure Dwr Cymru, North West, Southern, Thames, Bristol, Chester, Folkestone & Dover, Hartlepool, Mid Kent, North Surrey, South East, South Staffordshire 2.2 Sewerage service The six models developed so far are summarised in Table 2.10. Table 2.10: Current sewerage service models Sub-service Model type Explanatory variables Sewerage network Log linear Sewer length, critical sewer length, area, resident population, holiday population Sewerage network: power Log linear Volume of sewage collected, average pumping head Large sewage treatment works Log linear Total load, use of biological treatment, use of activated sludge, tight effluent consent for suspended solids and BOD, own sludge expenditure included, sludge centre expenditure included Small sewage treatment works Unit cost Works size, works type, load Sludge treatment and disposal Unit cost Weights of dry solids, disposal route Business activities Unit cost Billed properties The individual sewerage sub service models and results are shown below. The expenditure bands are not efficiency bands as some of the company specific factors are not taken into account. 2.2.1 Sewer network model Data collected in 1996-97 was used to revise the coefficients of the 1994 published model. The results were incorporated in the 1996-97 Report on water and sewerage operating costs and efficiency (December 1997). Following further work, the model has been substantially revised. It now takes into account critical sewer length and holiday population, which were not in the previous model. The factors relating to pumping station capacity and coastline length are no longer in the model, as both were found not to have a significant effect on costs, given the factors listed. Other potential explanatory factors are under review, and companies' views are welcome. Pumping expenditure is modelled separately from the volume of sewage collected multiplied by pumping head. Other explanatory factors, including the capacity and number of pumping stations and the proportion of stormwater pumping stations, are being reviewed. The current relative operating efficiency models are given in Tables 2.11 and 2.12. Table 2.11: Sewerage service: sewer network expenditure Sewerage service Sewer network expenditure Data: JR97 as issued December 1997 Modelled cost: Log to base e of sewerage network functional expenditure (£million), less Environment Agency charges, less pumping expenditure, per kilometer of sewer, for each sewerage area. Explanatory variables: Coefficient Standard error Log to base e of area of sewer district per kilometer of sewer 0.10 0.03 Log to base e of residential population per kilometer of sewer 0.17 0.20 Proportion of sewer length classed as critical 1.44 0.46 Holiday population divided by resident population 1.77 0.48 Constant -7.47 0.43 Form of model: Log to base e of sewerage area functional expenditure (less pumping expenditure and Environment Agency charges) per kilometer of sewer = -0.56 + log to base e of (area of sewer district/ sewer length) x 0.10 + log to base e of (residential population/sewer length) x 0.17 + (critical sewer length/total sewer length) x 1.44 + (holiday population/residential population) x 1.77 Statistical indicators: No of obs.: 64 R2 0.53 Table 2.12: Sewerage service: sewer network model power cost Sewerage service Sewer network expenditure Data: JR97 as issued December 1997 Modelled cost: Log to base e of sewerage power expenditure(£m) for each sewerage area. Explanatory variables: Coefficient Standard error Constant -6.52 1.18 Log to base e of (volume of sewage collected multiplied by average pumping head) 0.69 0.16 Form of model: Log to base e of sewerage area power expenditure = 0.39 + log to base e of (volume of sewage collected x average pumping head) x 0.69 Statistical indicators: No of obs.: 64 R2 0.22 Companies have been banded by comparing their actual expenditure level with that predicted by the summed models. The results of this analysis are shown in Table 2.13. Table 2.13: Sewer network expenditure results Band Company Much lower than predicted expenditure North West, Wessex Lower than predicted expenditure Yorkshire As predicted expenditure Anglian, Severn Trent Higher than predicted expenditure Northumbrian, South West Much higher than predicted expenditure Dwr Cymru, Southern, Thames 2.2.2 Large sewage treatment works model Data collected in 1996-97 was used to revise the coefficients of the 1994 published model. The results were incorporated in the 1996-97 Report on water and sewerage operating costs and efficiency (December 1997). As with the sewerage network model, this model has been substantially revised since, and companies' suggestions for further improvements are welcome. The variables relating to trade effluent, distance to next works and some combinations of treatment process and consent condition that were used in the 1994 published model are no longer statistically significant. The current model separates treatment process from consent conditions, as these variables are significant when looked at individually and might be expected to act independently. The current relative operating efficiency model is given in Table 2.14. Table 2.14: Sewerage service: large works model Sewerage service Costs of sewage treatment at large works Data: JR97 as issued December 1997, with corrections (to be issued) Modelled cost: Log to base e of functional expenditure on sewage treatment at large works (£million) Explanatory variables: Coefficient Standard error Constant -8.99 0.26 Log to base e of total load1 0.80 0.03 Biological treatment used 0.23 0.10 Activated sludge used 0.54 0.09 Tight effluent consent for both suspended solids and BOD2 0.20 0.05 Costs include own sludge treatment 0.18 0.06 Costs include sludge centre expenditure 0.33 0.07 Form of model: Log to base e of functional expenditure on sewage treatment at large works = -8.99 + (log to base e of total load) x 0.80 + 0.23 if biological treatment used + 0.54 if activated sludge used + 0.2 if tight effluent consent for both suspended solids and BOD + 0.18 if costs include own sludge costs + 0.33 if costs include sludge centre expenditure. Statistical indicators: No of obs.: 335 R2 0.80 1Total load in this example is estimated as population equivalent x 120. 2Tight effluent consent is defined as 30 mg/litre or less suspended solids and 20 mg/litre or less BOD. Companies have been banded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 2.15. Table 2.15: Large works results Band Company Much lower than predicted expenditure   Lower than predicted expenditure Anglian As predicted expenditure North West, Severn Trent, Yorkshire Higher than predicted expenditure Southern Much higher than predicted expenditure Dwr Cymru, Northumbrian, South West, Thames, Wessex 2.2.3 Small works sewage treatment model A relatively straightforward analysis was carried out on summary data provided by companies for smaller works. This data included direct expenditure and loads for all works within a matrix of size bands and treatment levels. For completeness, all sizes of sea outfalls were included within this analysis. Functional expenditure was derived by removal of any sludge treatment expenditure included within direct expenditure from works above 600kg BOD5/day and allocating general and support expenditure pro rata to direct expenditure and consistently with the regulatory accounts. The approach adopted was to calculate an industry average unit cost for each works' size band and treatment level, and to combine these with each company's treated load figures to derive company predicted expenditure for each category of works. Comparison of the sum of this expenditure for each company with the sum of its actual expenditure provided an indication of each company's efficiency in this area. Companies have been graded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 2.16. Table 2.16: Small works results Band Company Much lower than predicted expenditure North West, Northumbrian, Severn Trent Lower than predicted expenditure Southern, Wessex As predicted expenditure South West Higher than predicted expenditure Dwr Cymru, Yorkshire Much higher than predicted expenditure Anglian, Thames 2.2.4 Sludge treatment and disposal model Sludge treatment and disposal costs were analysed using a similar approach to that for small sewage treatment works. Data was available on the direct expenditure and weights of dry solids for a number of disposal routes for each company. Total costs for each company were derived by adding general and support expenditure pro rata to direct expenditure. Industry average unit costs for each disposal route were calculated by dividing the total industry expenditure by the total weight of dry solids treated and disposed of by that route. These were then used to derive a predicted expenditure for each company for comparison with its actual cost. Large differences between actual and predicted expenditure for some companies probably have more to do with the difficulties of allocating expenditure between sewage treatment and sludge treatment than efficiency. However, it is important that these results are included in the overall assessment to correct for the effect of these different expenditure allocations on the sewage treatment models. Companies have been banded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 2.17. Table 2.17 : Sludge treatment and disposal model Band Company Much lower than predicted expenditure Anglian, Thames Lower than predicted expenditure Dwr Cymru, Wessex, Yorkshire As predicted expenditure North West, Northumbrian Higher than predicted expenditure Severn Trent Much higher than predicted expenditure South West, Southern 2.2.5 Business activities model Business activities for the sewerage service included the costs of customer services, scientific services and regulation. The approach adopted was to compare the actual expenditure per property billed for sewerage with the industry average unit cost. This approach is similar to that adopted for the water service, with the exception that no allowance for economies of scale is included, as the size range of sewerage companies is less than that for water companies. Differences between actual and predicted expenditure probably reflect the different allocation of expenditure by each company rather than efficiency. However, for the reasons quoted above, it is important that these costs are included in any overall assessment of the sewerage service. Companies have been banded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 2.18. Table 2.18: Business activities model results Band Company Much lower than predicted expenditure Thames, Wessex Lower than predicted expenditure Yorkshire As predicted expenditure Northumbrian, Severn Trent Higher than predicted expenditure Anglian, Dwr Cymru, South West Much higher than predicted expenditure North West, Southern   Appendix 3: CAPITAL MAINTENANCE EXPENDITURE MODELS 3.1 Water service The four sub-service models are summarised in Table 3.1. Table 3.1: Initial water service models Sub-service Model type Explanatory variables Water resources and treatment Log linear with constant returns to scale Distribution input, distribution input from surface water works, distribution input from works larger than 25Ml/d capacity Water distribution infrastructure Log linear with constant returns to scale Length of main, proportion of mains by length <150mm in diameter, proportion of communication pipes that are lead Water distribution non-infrastructure Log linear with constant returns to scale Pumping station capacity, service reservoir and water tower capacity divided by pumping station capacity, number of meters divided by distribution input Water management and general Log linear with constant returns to scale Total billed properties, proportion of properties that are non-household The banding of companies for each of the sub-services is summarised below. The results from each model are then combined to produce a capital maintenance expenditure assessment for the water service in Table 6 in Section 2.8. 3.1.1 Resources and treatment model Capital maintenance expenditure on water resources and treatment increases uniformly with company size, ie there are constant returns to scale. A unit cost model was consequently developed with distribution input the surrogate for company size. Additional explanatory factors in the model included water treatment works size and whether works treat surface or ground water. Large water treatment works are shown to be less expensive to maintain per megalitre per day (Ml/d) of output compared with smaller works, while works that draw water from rivers incur greater capital maintenance expenditure per Ml/d of water treated than works that draw water from ground water sources. Table 3.2 Water service: resource and treatment model Water service Resource & treatment expenditure Data: Capital Maintenance Return (CMR)1993–97 Modelled cost: Log to the base e annual average resources and treatment functional expenditure (£000s), divided by distribution input (Ml/d) Explanatory variables: Coefficient Standard error Constant 2.13 0.22 Proportion of distribution input from works >25Ml/d capacity -1.47 0.43 Proportion of distribution input from surface water works 1.60 0.37 Form of Model: Log to the base e of (average annual resources and treatment functional expenditure / distribution input) = 2.13 – 1.47 x (proportion of total distribution input from works >25Ml/d capacity) + 1.60 x (proportion of total distribution input from surface water works) Statistical indicators: No of obs.: 27 R2 = 0.44 The water resources and treatment model has been derived from an examination of potential explanatory factors, submitted in the Capital Maintenance Return (CMR), representing treatment works type and size, pumping station number and capacity, dams and aqueducts, water treated, distribution input, connected properties, asset condition and asset life characteristics. In fitting the model, it was necessary to remove Cambridge Water as an outlier due to unusually low expenditure. This may reflect company specific circumstances not covered in the model, company accounting policy in allocating expenditure, real differences in efficiency or a combination of these factors. Companies have been banded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 3.3. Table 3.3: Resource and treatment results Band Company Much lower than predicted expenditure North West, Northumbrian, Bristol, Cambridge, Essex & Suffolk, Hartlepool, Mid Southern, North Surrey, South Staffordshire, Tendring Hundred, Three Valleys, Wrexham Lower than predicted expenditure   As predicted expenditure Southern, Mid Kent, Anglian Higher than predicted expenditure Portsmouth Much higher than predicted expenditure Dwr Cymru, Severn Trent, South West, Thames, Wessex, Yorkshire, Bournemouth & West Hampshire, Chester, Folkestone & Dover, Sutton & East Surrey, South East, York. 3.1.2 Distribution infrastructure model Capital maintenance expenditure on water distribution infrastructure increases uniformly with company size, ie there are constant returns to scale. A unit cost model was consequently developed with length of main the surrogate for company size. Additional explanatory factors include the proportion of mains that are small in diameter and communication pipe material. Capital maintenance expenditure per length of main rises with both the proportion of small diameter mains and the proportion of communication pipes that are lead. The current relative capital maintenance efficiency model is given in Table 3.4. Table 3.4: Water service: distribution infrastructure model Water service Distribution infrastructure expenditure Data: CMR 1993–97 Modelled cost: Log to the base e annual average water distribution infrastructure functional expenditure (£000s), divided by length of main (km) Explanatory variables: Coefficient Standard error Constant -4.06 1.12 Proportion of mains that are < 150mm diameter 4.20 1.40 Proportion of communication pipes that are lead 0.77 0.35 Form of Model: Log to the base e (average annual distribution infrastructure expenditure/length of main) = -4.06 + 4.20 x (proportion of mains that are <150mm diameter) + 0.77 x (proportion of communication pipes that are lead) Statistical indicators: No of obs.: 28 R2 = 0.32 The water distribution infrastructure model has been derived from an examination of potential explanatory factors, submitted in the CMR, that include mains' condition, burst rate, connected properties, length of main, mains diameter and communication pipe material. Companies have been banded by comparing their actual expenditure level with that predicted by the model. The results of this analysis are shown in Table 3.5. Table 3.5: Distribution infrastructure model results Band Company Much lower than predicted expenditure South West, Bournemouth & West Hampshire, Chester, Essex & Suffolk, Folkestone & Dover, North Surrey, York Lower than predicted expenditure Thames As predicted expenditure Anglian, Northumbrian, Southern, Yorkshire, Mid Kent, South Staffordshire, Sutton & East Surrey Higher than predicted expenditure Mid Southern, Wrexham Much higher than predicted expenditure Dwr Cymru, North West, Severn Trent, Wessex, Bristol, Cambridge, Hartlepool, Portsmouth, South East, Tendring Hundred, Three Valleys 3.1.3 Distribution non-infrastructure model Capital maintenance expenditure on water distribution non-infrastructure increases with pumping station capacity and water storage capacity. It was considered appropriate to impose constant returns to scale relative to pumping station capacity. An additional explanatory factor is the relative number of meters to distribution input that is positively correlated to unit expenditure. The current capital maintenance efficiency model is given in Table 3.6. Table 3.6: Water service: distribution non-infrastructure expenditure Water service Distribution non-infrastructure expenditure Data: CMR 1993–97 Modelled cost: Log to the base e annual average water distribution non-infrastructure functional expenditure (£000s), divided by pumping station capacity (kw). Explanatory variables: Coefficient Standard error Constant -3.55 1.04 ln service reservoir and water tower storage capacity divided by pumping station capacity 0.44 0.13 ln number of meters divided by distribution input 0.66 0.24 Form of Model: Log to the base e (average annual distribution non-infrastructure functional expenditure /pumping station capacity) = -3.55 + 0.44 x ln (water storage capacity/pumping station capacity) + 0.66 x ln (number of meters/distribution input) Statistical indicators: No of obs.: 27 R2 = 0.35 The water distribution non-infrastructure model has been derived from an examination of potential explanatory factors, submitted in the CMR, representing water storage, pumping assets, distribution input, connected properties, meter numbers, asset condition and asset life characteristics. In fitting the model, it was necessary to remove North Surrey Water as an outlier due to unusually low expenditure. This may reflect company specific circumstances not covered in the model, company accounting policy in allocating expenditure, real differences in efficiency or a combination of these factors. Companies have been banded by comparing their actual expenditure level with that predicted by the preferred model. The results of this analysis are shown in Table 3.7. Table 3.7: Distribution non-infrastructure model results Band Company Much lower than predicted expenditure Dwr Cymru, North West, Southern, Severn Trent, Bournemouth & West Hampshire, Bristol, Cambridge, Hartlepool, North Surrey, Portsmouth, South East, Three Valleys, Wrexham, York Lower than predicted expenditure Chester As predicted expenditure Tendring Hundred Higher than Predicted expenditure Anglian, Mid Southern Much higher than predicted expenditure Northumbrian, South West, Thames, Wessex, Yorkshire, Essex & Suffolk, Folkestone & Dover, Mid Kent, Sutton & East Surrey, South Staffordshire 3.1.4 Management and general model Capital maintenance expenditure on water management and general increases with company size. The appropriate scale variable used as a surrogate for company size was the number of billed properties. Significant dis-economies of scale were evident but were not considered appropriate and thus a unit cost form of the model was adopted. In addition to billed properties, the proportion of properties that are non-household is also correlated to unit expenditure. The current capital maintenance efficiency model is given in Table 3.8. Table 3.8: Water service management and general model Water service Management and general expenditure Data: CMR Modelled cost: Log to the base e annual average management and general expenditure (£000s), divided by billed properties (000s) Explanatory variables: Coefficient Standard error Constant 0.88 0.49 Proportion of properties that are non-household 11.26 6.22 Form of Model: Log to the base e (average annual water management and general expenditure / number of billed properties) = 0.88 + 11.26 x (proportion of properties that are non-household) Statistical indicators: No of obs.: 28 R2 = 0.11 Companies have been banded by comparing their actual expenditure level with that predicted by the preferred model. The results of this analysis are shown in Table 3.9. Table 3.9: Management and general results Band Company Much lower than predicted expenditure Dwr Cymru, Southern, Cambridge, Folkestone & Dover, Mid Southern, North Surrey, Portsmouth, Tendring Hundred, York Lower than predicted expenditure Wessex, Bristol, South East, Wrexham As predicted expenditure South Staffordshire Higher than predicted expenditure Essex & Suffolk Much higher than predicted expenditure Anglian, North West, Northumbrian, Severn Trent, South West, Thames, Yorkshire, Bournemouth & West Hampshire, Chester, Hartlepool, Mid Kent, Sutton & East Surrey, Three Valleys 3.2 Sewerage service The five models developed so far are summarised in Table 3.10. Table 3.10: Current sewerage service models Sub-service Model type Explanatory variables Sewerage infrastructure Log linear with constant returns to scale Sewer length, number of combined sewer overflows per km of sewer Sewerage non-infrastructure Unit cost Total capacity of sewage pumping stations Sewage treatment Log linear with constant returns to scale Total load received at treatment works, proportion of load treated by a tertiary process, proportion of load treated by works of size 600 kgbod5/d or greater Sludge treatment and disposal Unit cost Total weight of dry solids Sewerage management and general Unit cost Total number of billed properties The banding of companies for each of the sub-services are summarised below. The results from each model are then combined to produce a capital maintenance expenditure assessment for the sewerage service in Table 3.11. 3.2.1 Sewerage infrastructure model Capital maintenance expenditure on sewerage infrastructure increases uniformly with company size, ie constant returns to scale. A unit cost form of the model was consequently developed with total length of sewer the surrogate for company size. Unit expenditure is also related to the number of combined sewer overflows per length of sewer. The current capital maintenance efficiency model is given in Table 3.11. Table 3.11: Sewerage service: sewerage infrastructure model Sewerage service Sewerage infrastructure Data: CMR 1993–97 Modelled cost: Log to base e of average annual sewerage infrastructure functional expenditure (£million), divided by the total length of sewer. Explanatory variables: Coefficient Standard error Constant -6.25 0.20 Log to the base e of the number of combined sewer overflows divided by the total length of sewer. 0.42 0.06 Form of Model: Log to the base e of (average annual sewerage infrastructure expenditure/total sewer length) = - 6.248 + 0.421 x log to the base e (number of combined sewer overflows/total sewer length) Statistical indicators: No of obs.: 61 R2 = 0.47 The sewerage infrastructure model was derived from an examination of potential explanatory factors representing sewer type (critical, S24, brick and masonry) and condition, connected properties, sewer collapses and number of CSOs. Companies have been banded by comparing their actual expenditure level with that predicted by the preferred model. The results of this analysis are shown in Table 3.12. Table 3.12: Sewerage infrastructure model results Band Company Much lower than predicted expenditure Wessex, Yorkshire Lower than predicted expenditure Anglian, Dwr Cymru, Southern As predicted expenditure   Higher than predicted expenditure South West Much higher than predicted expenditure North West, Northumbrian, Severn Trent, Thames 3.2.2 Sewerage non-infrastructure model A unit cost approach resulted from modelling sewerage non-infrastructure data. Each company's average annual expenditure divided by the total capacity of pumping stations was compared with the weighted average industry cost. Comparison of this expenditure with the average provided an indication of each company's efficiency in this area. The results of this analysis are shown in Table 3.13. Table 3.13: Sewerage non-infrastructure model results Band Company Much lower than predicted expenditure Anglian, Dwr Cymru, North West, Northumbrian, Severn Trent, Yorkshire Lower than predicted expenditure   As predicted expenditure   Higher than predicted expenditure   Much higher than predicted expenditure South West, Southern, Thames, Wessex The sewerage non-infrastructure model was derived from an examination of potential explanatory factors representing the number of domestic properties connected to the sewerage system, total and average capacity of all pumping stations and in each size band, and asset condition. 3.2.3 Sewage treatment model The data indicates that there are economies of scale for capital maintenance expenditure on sewage treatment, using total load received at sewage treatment works as a surrogate for scale. However, experience and the modelling results from water resources and treatment indicate that constant returns to scale should apply to such a scale variable. Constant returns to scale have therefore been imposed. Total load received at treatment works is the scale variable used as a surrogate for company size. Expenditure is also related to the complexity of treatment process and the size of treatment works. Companies with load treated by a tertiary process incur greater maintenance costs, while those with load dealt with in works of size 600 kg BOD5/d or greater (size bands five and six) will incur lower maintenance costs. The current capital maintenance efficiency model is given in Table 3.14. Table 3.14: Sewerage service: sewage treatment model Sewerage service Sewage treatment Data: CMR 1993–97 Modelled cost: Log to base e of average annual sewage treatment functional expenditure (£million), divided by the total load received at sewage treatment works. Explanatory variables: Coefficient Standard error Constant -7.77 0.39 Proportion of total load received at works that is treated in works of size 600kgBOD5/d or greater. -2.87 0.45 Proportion of total load received at works that is treated by a tertiary process. 1.12 0.34 Form of Model: Log to the base e (average annual sewage treatment expenditure/total load) = - 7.77 + 1.12 x (proportion of load treated in works of size 600kgbod5/d or greater) – 2.87 x (proportion of total load that is treated in a tertiary process) Statistical indicators: No of obs.: 587 R2 = 0.49 The sewage treatment model was derived from an examination of potential explanatory factors representing load received at different works, average load, volume of load, equivalent population served, connected properties, size of treatment works, complexity of treatment works and volume of waste water and trade effluent. Yorkshire area three was excluded from the modelling as it appeared to be an outlier with unusually low expenditure Companies have been banded by comparing their actual expenditure level with that predicted by the preferred model. The results of this analysis are shown in Table 3.15. Table 3.15: Sewage treatment model results Band Company Much lower than predicted expenditure Anglian, Dwr Cymru, North West, Yorkshire Lower than predicted expenditure   As predicted expenditure Northumbrian Higher than predicted expenditure   Much higher than predicted expenditure Severn Trent, South West, Southern, Thames, Wessex 3.2.4 Sludge treatment and disposal model A unit cost approach was used for the sludge treatment and disposal model. Each company's average annual expenditure divided by the total weight of dry solids disposed of was compared with the weighted average industry cost. Comparison of this expenditure with the average provided an indication of each company's efficiency in this area. The results of this analysis are shown in Table 3.16. Table 3.16: Sludge treatment and disposal model results Band Company Much lower than predicted expenditure Dwr Cymru, Anglian, North West, Northumbrian, Southern, Yorkshire Lower than predicted expenditure   As predicted expenditure   Higher than predicted expenditure   Much higher than predicted expenditure Severn Trent, South West, Thames, Wessex The sludge treatment and disposal model was derived from an examination of potential explanatory factors representing the number of sewage treatment works that treat only their own sludge, the number of sewage treatment works that are sludge centres, the number of sludge centres not situated at sewage treatment works and the amount of sludge disposed of to each route. 3.2.5 Sewerage management and general model A unit cost approach was adopted for sewerage management and general expenditure, comparing each company's average annual expenditure per billed property with the weighted average industry cost. Comparison of this expenditure provided an indication of each company's efficiency in this area. The results of this analysis are shown in Table 3.17. Table 3.17: Sewerage management and general model results Band Company Much lower than predicted expenditure Dwr Cymru, Northumbrian, Southern, Thames, Yorkshire Lower than predicted expenditure Wessex As predicted expenditure   Higher than predicted expenditure Severn Trent Much higher than predicted expenditure Anglian, North West, South West No other form of model development was possible. Only ten sets of data were collected from the industry, as it is inappropriate to subdivide into regions the centralised management and general function. Insufficient information was available to enable sewerage management and general data to be combined with water management and general data for regression analysis.

Appendix 4: SUMMARY MATRICES The following tables compare expenditure bands for operating expenditure and capital maintenance expenditure across each sub-service for the water and sewerage services. Tables 4.2, 4.3 and 4.4 are for the water sub-services and Tables 4.5, 4.6, 4.7 & 4.8 are for the sewerage sub-services. Tables 4 and 5 of the report compare the overall position. A consistent banding convention is used throughout and is based on the difference between reported expenditure and the central estimate of predicted expenditure from the relevant model. The banding convention in the tables is: Band A: well below predicted expenditure (less than 85% of C) Band B: below predicted expenditure (85–95% of C) Band C: around predicted expenditure (within 5% of modelled expenditure) Band D: above predicted expenditure (105–115% of C) Band E: well above predicted expenditure (more than 115% of C) Table 4.1 illustrates how models have been combined in the preparation of the matrices in this appendix. In some instances two models have been developed for a particular asset type for operating expenditure, whereas only one has been developed for capital maintenance expenditure and vice versa. For operating expenditure on power, there is no similar model for capital maintenance, so these costs are not compared at the sub-service level. However, water power costs are included in the water service matrix found in Table 3 of this paper. Table 4.1: Combining the results from the models for opex and capital maintenance Operating expenditure models Capital maintenance expenditure models 4.2 Water resources and treatment Water resources and treatment 4.3 Water distribution Water distribution infrastructure Water distribution non-infrastructure 4.4 Water business activities Water management and general 4.5 Sewer network Sewerage infrastructure Sewerage non-infrastructure 4.6 Sewage treatment large works Sewage treatment small works Sewage treatment 4.7 Sludge treatment and disposal Sludge treatment and disposal 4.8 Sewerage business activities Sewerage management and general As set out in Chapter 4 of this paper, the results presented here only compare the expenditure predicted by the model with company actual expenditure and do not take account of legitimate, company specific factors. However, the tables do allow operating expenditure and capital maintenance expenditure performance to be compared. It might be expected that a company would generally perform equally as well in operating expenditure as it would in capital maintenance expenditure. However, a company may have a strategy of investing heavily in technological solutions with automated processes resulting in operating expenditure savings. Such a company may appear to perform better in operating expenditure relative to capital maintenance expenditure. Accounting practices could explain why some companies do not have the same performance assessment for both types of cost. Table 4.2: Water resources and treatment Table 4.3: Water distribution Table 4.4: Water business activities and management and general Table 4.5: Sewer network Opex expenditure banding A North West Wessex B Yorkshire C Severn Trent Anglian D South West Northumbrian E Thames Southern Dwr Cymru   E D C B A Capital maintenance expenditure banding Table 4.6: Sewage treatment Opex expenditure banding A B Severn Trent Northumbrian North West C Yorkshire D South West Southern Wessex Anglian Dwr Cymru E Thames   E D C B A Capital maintenance expenditure banding Table 4.7: Sludge treatment and disposal Opex expenditure banding A Thames Anglian B Wessex Dwr Cymru Yorkshire C North West Northumbrian D Severn Trent E South West Southern   E D C B A Capital maintenance banding Table 4.8: Business activities and management and general Opex expenditure banding A Wessex Thames B Yorkshire C Severn Trent Northumbrian D Anglian South West Dwr Cymru E North West Southern   E D C B A Capital maintenance expenditure banding   Appendix 5: SCOPE FOR GENERAL EFFICIENCY Summary of the final report on Water and Sewerage Industries: General Efficiency and the Potential for Improvement By Professor Derek Bosworth University of Manchester Institute of Science and Technology Professor Paul Stoneman Warwick Business School Joanne Roe Warwick Research Institute The aim of this Report is to identify the magnitude and causes of underlying movements in Real Unit Operating Costs (RUOC) and productivity, both in the UK as a whole and in relevant comparator industries, based on both evidence from the published literature and on original research, in order to draw conclusions as to potential developments in real unit operating costs in the water and sewerage industry. The emphasis of the existing literature is upon labour productivity, which is only one element determining costs, nevertheless, it is still possible to make some inferences about potential changes in real unit operating costs. The main features of labour productivity changes are that it: Is procyclical; Exhibits long term differences (the post–War period is one of relatively rapid movement); Shows medium term trends – the 1970s exhibits slower rates of improvement than the 1960s and 1980s; And is faster in manufacturing than services. The review of the literature relating to manufacturing, for example, suggests estimates of labour productivity growth of approximately 4% during the 1960s, 2-3% in the 1970s and 4-5% in the 1980s. Total factor productivity measures reveal that a significant proportion of the growth in labour productivity can be attributed to the growth in raw material and intermediate inputs, as opposed to the growth in capital. One study suggests that somewhere in the order of 75% of labour productivity growth can be attributed to the substitution of material and intermediate inputs for labour. This substitution effect can be traced to the cost effectiveness of sub-contracting and 'buying-in', as opposed to producing in-house. The published estimates tend to underestimate the rate of improvement in the quality of output, producing a downward bias to the measured rate of productivity growth. Unaccounted improvements in the quality of inputs impart an opposite bias. What this implies for water and sewerage industry comparisons, however, depends on the relative quality changes in inputs and outputs vis à vis comparator industries. The key area of interest is in the rate of change in real unit operating costs, which is where the literature is weakest. The review of the literature suggests a range of estimates from about – 1.0% to – 2.4% per annum (using the implicit GDP deflator) over the 1980s for manufacturing as a whole. These findings are most sensitive to estimates of the rate of growth of material and intermediate inputs, the area in which published results are least precise. The literature has found support for a range of influences on labour productivity growth, including: Capital/labour substitution: which tends to occur through new investment in expanding industries and through scrapping old plant in declining industries. The literature suggests that, at various times (1970s) and in various countries (Sweden), public sector investment has contributed more to the growth of labour productivity than private sector investment. Intermediate input/labour substitution which raised gross output per unit of labour (by definition), but also raises value added per employee, as inefficient in-house activities are substituted by more efficient sub-contracted, 'bought-in' activities. Privatisation of previously public enterprises, which, it is argued, raised labour productivity and perhaps raises of the trend rate of growth of labour productivity, although some authors suggest that this is more the result of liberalisation and competition than privatisation per se. Economies of scale, which were thought to be significant in the 1960s (a 1% increase in all inputs produced about a 1.1% increase in output), but were exhausted in the 1970s. Reductions in unit sizes in the 1980s were not thought likely to lead to significant diseconomies. Learning effects, which are extremely important for some industries (the modal value is that a doubling of output leads to a 20% reduction in costs). Regulation is often quoted as a potential influence on productivity improvement, perhaps accounting for about half the residual (ie that which cannot be explained by observed changes in factor inputs). The one up-to-date published forecast suggests that, across all sectors, the 1990s will be very similar to the 1980s (the 1990s overall productivity performance may be 0.2% per annum higher). However, manufacturing productivity growth is forecast to be slightly down (by about 0.2% points on the 1980s), as is primary and utilities (0.4% lower). The choice of comparator industries was based on key performance ratios and key features of the water and sewerage industry, such as: natural monopoly, large-scale operations, capital intensive operations, long term investment, key processes (abstraction, cleansing and distribution), external and seasonal factors, quality variations and regulation, guaranteed demand, metering, periodic invoicing, bad debts, organisational control, privatisation, employment structure and technological change. From these factors a list of approximately 25 potential comparator (3-digit) industries have been isolated, including public sector industries (ie electricity, gas, post, telecommunications), and more traditional private sector industries (ie brewing, distilling, chemicals, pharmaceuticals, road haulage), which has been reduced to 11 actual comparators plus All Manufacturing. Of the various comparator industries it is considered that the five most appropriate are the Extraction of Sand and Gravel, Extraction of Miscellaneous Materials, Pharmaceuticals, Brewing and Malting and Miscellaneous Manufacturing Industries. We find that the average performance of these five industries closely matches estimates of the performance of the manufacturing sector as a whole. On the key issue of movements over time in RUOC, we find that in manufacturing as a whole, using unadjusted Census of Production Data, RUOC fell at a rate of 0.74% per annum between 1979 and 1990. The average for the five comparators was 1.1% per annum. Comparable estimates for the water industry have been calculated using Census of Production Data. However, this has involved a number of major data problems. In particular the report discusses problems surrounding: the definition of the industry and changes in that definition with privatisation; the coverage of undertakings in the industry; the measurement of output; the measurement of output prices; and the measurement of input prices. The estimates for the water industry that have been produced are thus subject to a number of health warnings. The estimates do suggest however that compared to manufacturing and the other comparator industries (on average), between 1979 and 1990, that prices rose more quickly, and that real unit material costs rose that than fell. However real unit labour costs fell faster. Real unit operating costs are estimated as rising in the water industry as opposed to falling in manufacturing as a whole and in the comparators on average. Overall it is argued that in the water industry RUOC has been increasing faster than in manufacturing in general. A discussion of quality issues leads us to suggest that the measured difference ought to be increased if the comparisons are to be made on a constant quality basis. Consideration of the methods used for constructing estimates of the growth of RUOC in Manufacturing leads to a discussion of the biases involved in using unadjusted Census of Production data, and the implications of using different indicators for the price of intermediate inputs, materials and fuels. It is argued that taking account of such biases and also taking account of unmeasured quality change, that a 'best' estimate of the rate of growth of RUOC in manufacturing between 1979 and 1990 (using the implicit GDP deflator) is in the range from –2.0% to –2.25% per annum. Alternative methods are considered for forecasting changes in RUOC in the water and sewerage industry over the next decade. It is argued that the best indicator will reflect the rates of growth of RULC [Real Unit Labour Costs] and RUMC [Real Unit Material Costs] in manufacturing as a whole over the next decade but weighted by water industry shares in total operating costs. It is further argued that the forecasting exercise in the Report suggests that manufacturing performance in the next decade will be very similar to that seen in the 1980s and thus the 1980s performance figures may be used for forecasting. Ignoring quality issues, and using unadjusted Census data for estimating the growth of RULC and RUMC in manufacturing, suggests a best forecast to be a rate of growth of RUOC in the water industry of –0.7% per annum. However, given the problems of bias resulting from the use of unadjusted Census data to calculate RUMC and RULC in manufacturing, this figure might well be increased by a further 0.5% per annum, ie to –1.2% per annum. Introducing a figure to allow for unaccounted quality change in manufacturing output, the estimate may be further increased to –2.2% per annum. This approach thus suggests that in the absence of changes in output quality, RUOC in the water industry over the next decade might grow at around –2% per annum. This is a considerable turnaround from a measured rate of increase of 1.5% per annum seen between 1979 and 1990 although this figure itself is subject to a margin of error.   Appendix 6: TOTAL EFFICIENCY The measurement of comparative total efficiency in the sewerage and water industry: An exploratory study By Professor Derek Bosworth University of Manchester Institute of Science and Technology Professor Paul Stoneman and Dr Emmanuel Thanassoulis Warwick Business School EXECUTIVE SUMMARY 1. The objective of this report is to assess the validity and practicality of undertaking comparative efficiency assessments of companies in the water and sewerage industry encompassing all inputs, operational and capital (as opposed to operational inputs alone) and to make recommendations upon the content and costing of a full study. 2. Capital costs represent a considerable proportion of total costs and differ significantly across companies. Thus an 'overall'measure of efficiency should be preferred to measures relating solely to operating costs. 3. The following efficiency concepts are defined and explained: scale efficiency; technical efficiency; allocative efficiency; social efficiency; dynamic efficiency; other measures. The regulated non-competitive nature of this industry considerably complicates the interpretation of efficiency measures that involve prices in their calculation. 4. In principle, efficiency analysis can be undertaken at the line of business level and/or at the company level. However there are particularly severe problems in the allocation of costs to lines of business. 5. There are significant problems in the water industry in defining the appropriate set of inputs and outputs and measuring them. A multi dimensional measure of output based upon physical indicators is to be preferred. 6. Inputs may be defined in a disaggregated manner but the greater the degree of disaggregation the greater are the data requirements. Alternatively inputs may be aggregated into a total cost measure. A measure of total costs may be obtained from company accounts or separate reporting to Ofwat. 7. Capital inputs are particularly difficult to measure. A single measure of capital costs equivalent to the economists' concept of 'normal' profits can be defined. To operationalise the suggested approach it will be necessary to have data on measures of company capital assets calculated on a net replacement cost basis. 8. Management can only influence the efficiency of a company with respect to those factors that are under its control within the period of analysis. Inter company comparisons aimed to meet the objectives of Ofwat should exclude the influence of fixed factors when measuring comparative efficiency (or, synonymously, if such factors cannot be excluded, consider such factors as valid reasons for differences in measured efficiency). 9. Factors beyond management control in addition to geographic location might include the inherited (ie pre-privatisation) capital stock. The problems associated with this can be best treated by proceeding to measure efficiency as if the capital stock was not inherited but actually under management control, but to allow that revealed patterns of efficiency differences can be attributed, in the short run, to the nature of individual company inheritances. In the long run, when the capital stock is under management control, the nature of the inheritance would no longer be a valid rationale for measured efficiency differences. 10. The feasibility of three different approaches to measuring efficiency are studied: the use of accounting data; stochastic, frontier and deterministic production function techniques; and Data Envelopment Analysis (DEA). 11. The direct use of accounting data is found not to be a practical route, primarily because such data is unable to generate a useful measure of output. Accounting data can however be used for the measurement of costs. 12. The use of (stochastic, deterministic and frontier) production function techniques is more promising and can yield estimates of both technical and (input) allocative efficiency. The analysis suggests that the cost function version of this approach is to be preferred. To undertake such analysis a panel data set over at least a five year period would be required. This is a strict data requirement and cannot currently be met. 13. The use of DEA techniques is practical and feasible and its use is illustrated. The advantage of DEA is that the technique allocates to each unit of observation (subject to minor restrictions) input and output weights that maximise the measure of efficiency for that unit of observation. In the use of DEA there are still questions to be resolved relating to the appropriate measures of inputs and outputs to be used in the analysis. 14. The objectives of the study have been met. The recommendations are; i. The analysis of comparative efficiency in the water and sewerage industry should encompass total efficiency ie both capital and operating inputs. ii. Both production function and DEA approaches are feasible and useful methods for approaching comparative total efficiency measurement in the water and sewerage industry. iii. The data requirements of the production function approach are more demanding than those of the DEA approach. iv. Given the current data situation and the potential outputs from the alternative approaches, the DEA approach is to be preferred. v. There is still much work to be undertaken if the DEA approach is to be fully operationalised especially as regards the definition and measurement of appropriate inputs and outputs. 15. A full DEA analysis of comparative efficiency at the company level will cost approximately £60,000. An extension to line of business analysis would cost in the region of a further £15,000. Appendix 7: 1999 PERIODIC REVIEW TIMETABLE Outline efficiency framework timetable. Item Brief description Timetable 1 Publication of Assessing the scope for future improvements in water company efficiency — A technical paper Consultation on the technical paper April 1998 May to June 1998 2 Release of data used in capital maintenance analysis (from company Capital Maintenance Return — PR99C) and opex modelling data sets May 1998 3 Updating of relationships in light of company performance in 1997–98 (JR98 numbers) July to September 1998 4 Relative efficiency and scope for future savings identified in Prospects for prices October 1998 5 Feedback to companies on current assessments of relative efficiency (operating costs, capital maintenance and capital unit costs (Cost Base)) October to December 1998 6 Publication of 1997-98 Report on water and sewerage operating costs and efficiency December 1998 7 Efficiency issues discussed at the formal meetings between the Director and companies January & February 1999 8 Company submissions on their relative efficiency and scope for future savings as part of the draft Business Plans April 1999 9 Updating of relationships in light of company performance in 1998-99 (JR99 numbers) June & July 1999 10 Assessing company submissions and judgements necessary for the draft determinations April to July 1999 11 Company representations on the draft determinations September & October 1999