Aim of the study

(1)

ONGOING EFFICIENCY IN NEW ME THOD DECISIONS FOR DUTCH ELECTRICITY AND GAS NETWORK

OPERATORS

DUTCH OFFICE O F ENERGY REGU LATION

NOVEMBER 2012 FINAL REPORT

Submitted by:

Cambridge Economic Policy Associates Ltd

(2)

Executive summary ... 1

Glossary ... 9

1. Introduction ... 10

1.1. Aim of the study ... 10

1.2. Previous studies ... 10

1.3. Overview of approach ... 14

1.4. Structure of report... 14

2. Concepts and definitions ... 16

2.1. Introduction ... 16

2.2. Productivity ... 16

2.3. Illustration of frontier shift and catch-up ... 17

2.4. Capital maintenance ... 18

2.5. Network growth ... 18

2.6. Liberalisation versus privatisation ... 19

2.7. Gross frontier shift versus net frontier shift ... 19

2.8. Real price effects ... 20

3. Approaches to setting ongoing efficiency targets or efficient cost targets ... 21

3.1. Common methodologies ... 21

3.2. Regulatory precedents... 24

4. NMa’s regulatory approach ... 27

4.2. Electricity and gas distribution ... 27

4.3. Electricity transmission (TenneT) ... 30

4.4. Gas transmission (GTS) ... 31

4.5. Summary ... 31

5. Applicable approaches to setting productivity/ or efficient cost targets ... 33

5.2. Approaches ... 33

5.3. Sector appropriateness... 35

5.4. Transmission operators ... 37

6. Empirical analysis of applicable approaches ... 39

6.2. Total Factor Productivity (TFP) ... 39

6.3. Output price indices... 47

6.4. RUTC ... 49

6.5. RUOE ... 54

(3)

6.6. Labour, energy, materials and services (LEMS) cost and productivity measure ... 55

6.7. Academic/ consultant studies ... 56

6.8. Summary of empirical analysis ... 57

7. Conclusions ... 60

7.1. Distribution networks ... 60

7.2. Gas transmission ... 63

7.3. Electricity transmission ... 64

7.4. Summary ... 65

Annex A – Theoretical definition of efficiency... 66

Annex B – Liberalisation versus privatisation ... 68

Annex C – Regulator case studies ... 70

Annex D – TFP sensitivity analysis ... 78

Annex E – Output price indices sensitivity analysis ... 79

Annex F – LEMS ... 81

Annex G – RUOE ... 83

Annex H – Summaries of Academic Studies ... 89

Annex I – Methodologies ... 95

Bibliography ... 97

Important notice

This report has been commissioned by NMa. However, the views expressed are those of CEPA alone. CEPA accept no liability for use of this report or any information contained therein by any third party. © All rights reserved by Cambridge Economic Policy Associates Ltd.

(4)

1

EXECUTIVE SUMMARY

This report has been prepared by Cambridge Economic Policy Associates (CEPA) for the Dutch Competition Authority (NMa). This report provides advice on setting an efficient cost frontier shift for the four energy sectors that NMa regulates – electricity distribution and gas distribution (DNOs), electricity transmission (TenneT), and gas transmission (GTS).

CEPA’s advice on possible efficient cost frontier shifts is to be applicable for the regulatory period commencing in 2014. NMa is still developing many aspects of the next regime and a method for setting ongoing efficiency targets and/or changes in the level of inputs will help inform their development of other areas of the regime. It will be for NMa to decide on how to use the estimates presented in this report.

Cost frontier shift

Cost frontier shift is also sometimes referred to in the regulatory literature as ongoing efficiency.

This encompasses productivity gains resulting from technical change over time and also takes account of changes in input prices over time. It does not include changes in productivity resulting from changes in the relative efficiency of firms within the regulated industry (sometimes referred to as catch-up efficiency).

Methodology

We have approached this study by, firstly, assessing the regimes that the NMa operates across the four sectors and, secondly, assessing the applicability of different empirical measures against this. We started the second stage by considering the eight methods that have been used regularly by regulators in other jurisdictions, consultants and academics. These are listed in Table E.1 below. We differentiated these measures by the level that they cover, i.e., total coverage or partial, and whether it is a productivity measure (only covering changes in volume of outputs and inputs) or a cost measure (taking into account productivity and change in input costs, including cost of capital, above that of inflation). If a sector/firm’s input prices moved in line with inflation then the productivity and cost measure would provide the same result.

Indirect comparisons use data/measures from other sectors, for example, assessment of historical productivity growth or cost trends for selected sectors of the Dutch economy, or other regulated industries. International comparisons use data from regulated companies in the same sector in other countries/jurisdictions. While within sector comparisons use historical productivity growth or cost trends from within the regulated industry.

We have focused on using indirect comparisons for assessing the scope for cost frontier shift where possible. This reduces the risk of any ‘gaming’ by regulated companies when within sector data is used and allows comparisons to sectors with a large number of privatised companies facing a high level of competition. This approach has the downside of not exactly covering the activities carried out by the network operators, however this can be mitigated by using productivity/ cost measures from sectors carrying out similar types of activities.

(5)

2 Table E.1: Methods for measuring productivity/ cost change

Method Proposed comparison type Cost measure or

productivity measure Total factor productivity/cost measures

Total Factor Productivity (TFP) Indirect comparisons Productivity Output price indices Indirect comparisons Cost measure

Malmquist TFP indices Within/ international

comparisons

Productivity

Real Unit Total Cost (RUTC) Within comparisons Cost measure Partial factor productivity/cost measures

Labour productivity Indirect comparisons Productivity Real Unit Operating Expenditure

(RUOE)

Within/ indirect comparisons Cost measure

LEMS cost measure Indirect comparisons Cost measure

LEMS productivity Indirect comparisons Productivity

Using the data available to us we produced empirical results for:

· TFP;

· Output price indices;

· RUTC (distribution networks only);

· RUOE (using data from the UK and Netherlands);

· LEMS cost measures; and

· LEMS productivity.

For all measures apart from RUTC and RUOE we have used the EU KLEMS dataset information for the Dutch economy. The data originates from Statistics Netherlands. We have also selected sectors which may undertake similar activities to the network operators. In addition to calculating our own estimates we have undertaken a review of recent academic and consultant reports which set out empirical results for the productivity/cost trends in energy transmission or distribution networks.

In any empirical analysis measurement issues can exist, these may be due to inaccurate data or the inappropriate use of the data. While we have attempted to identify measurement issues these cannot always be ruled out and, indeed, when using estimates from other studies and regulators it is not clear to what extent their analysis may have been impacted by measurement issues.

Therefore, we advocate using multiple sources in order to triangulate a range for setting the scope for the efficient cost frontier.

Our preferred approach is to use total cost measures where possible, as these cover all costs and take into account the change in input prices as well. The TFP results provide useful information,

(6)

3 particularly if it were considered that growth in input prices were different in the comparator sectors to the regulated sectors (and hence output prices in comparators sectors will be less appropriate as a proxy for the movements in efficient costs). Output price measures can be applied directly to total costs (although there is an exception to this discussed below) as it reflects all costs required to produce an output. The only difference in using output price indices or productivity measures in setting an efficient cost frontier is that if productivity measures were used, further consideration would need to be given to whether an adjustment for input prices is required. Our analysis of Dutch output prices and TFP suggest that historically there was little difference in input price growth in the selected sectors and the Dutch economy indicating that an adjustment for input price differentials is not be required. This assumption holds only on the basis that the movements in the input prices in the selected selectors are representative of the movements in the input prices faced by the network operators.

NMa use a concept of total costs (opex plus depreciation plus the cost of capital) to set a notional income target, and hence price path, for the companies. We note that depreciation and cost of capital are not in themselves subject to efficiency improvements; however expenditure on new capital can lead to efficiency improvements which would in turn lead to lower growth in future depreciation and cost of capital requirements (than if efficiency gains were not achieved).

In the following sections we have set out the percentage changes so that a positive number reflects a productivity improvement or a cost reduction.

Distribution networks

NMa’s approach to setting catch-up efficiency for the DNOs and its use of a single adjustment factor for setting a notional income target means that consideration needs to be given as to whether additional factors in addition to an ongoing efficiency are required to set an efficient cost frontier target. For instance, NMa’s current approach of allowing a fixed depreciation amount for around 30 years from liberalisation plus depreciation on any new capex including replacement expenditure means that depreciation is continually growing.¹ This is offset to some degree by the RAB also falling, however our analysis of the DNOs’ costs indicated that depreciation was growing at a much greater rate than the RAB.

NMa’s current approach of using RUTC captures historical movements in the DNOs’

depreciation. This means that it is not comparable with the productivity or cost measures from indirect comparisons or other studies as similar depreciation arrangements will not be in place.

We are cautious of using RUTC measures to set efficient cost targets as they rely on within industry data which is based on companies not facing the strong incentives that privatised companies face and thus we have focused on measures using indirect comparisons. Also, setting an explicit cost efficiency target, rather than one which is not separated from the ongoing fixed depreciation charges, provides greater transparency on what incentives have been placed on the companies.

1 While we have referred to 30 years as being the depreciation timeframe, we note that the time period for the depreciation varies across the network operators, from 19 to 32 years for electricity distribution operators and 28 to 35 for gas distribution operators.

(7)

4 Table E.2 below provides the estimates for the different empirical methods we have calculated, the results from other academic studies and the regulatory decisions from other jurisdictions.

Table E.2: Productivity growth and cost trend estimates (total expenditure/ costs)

Measure Range (% per

annum)

Frontier-shift only

Cost measure or productivity measure Estimated

TFP (GO) – selected sectors 0.5% Yes Productivity measure

Output price indices (GO) relative to CPI – selected sectors

0.5% Yes Cost measure

RUTC (2003-2010) – Gas distribution

2.2% Yes** Cost measure

RUTC (2003-2010) – Electricity distribution

0.6-1.5% Yes** Cost measure

Other studies/ regulatory precedents Gas distribution

Other studies 1.9-2.9%* No (some studies’

estimates may include catch-up efficiency)

Productivity measure

Regulatory precedents 0.8-1.5% Yes Productivity measure

Electricity distribution

Other studies 2.1-3.6% No (some studies’

estimates may include catch-up efficiency)

Mix

* We have excluded the partial productivity measures from this range.

** Depends on the assumption that there was not a significant amount of catch-up during this period.

Other studies on distribution networks’ efficient cost frontier change suggest higher rates of change may be achievable than that indicated by the output price indices – although some of these measures include catch-up as well as frontier shift. We note that some of these studies may not take into account input price growth as is the case with the output price indices. Evidence from other jurisdictions indicates that input prices (namely engineering wages and materials) have historically increased at a slightly higher rate than CPI, although this was not the case in our analysis of the selected Dutch sectors. This would mean that combining the estimates in Table E.2 with a positive input price differential (e.g. input price inflation is greater than economy wide inflation) would lead to a lower efficient cost frontier.

(8)

5 We consider that as the change in output price indices, an average of 0.5% for the selected industries, is lower than the estimates produced in other studies and set by regulators in other jurisdictions it provides a good indication of a lower bound for the achievable scope for efficient cost frontier shift. We think this provides a better lower bound than other regulators’ decisions as it incorporates the historical characteristics of the Dutch economy. The similarities between the TFP and output price indices indicate that there is little difference in the input price change in the sectors relative to CPI.

The upper bound for efficient cost frontier shift, based on the other studies, may be around 2.6% for electricity distribution and 2.5% for gas distribution.² We have taken an average across the studies (set out in Section 6.6), rather than the highest value to mitigate any impact from measurement error. We consider that the other studies provide an appropriate higher bound, rather than estimates for the Dutch economy, because they are based on companies undertaking the same types of activities (rather than indirect comparators undertaking similar types of activities).

Therefore, an achievable ongoing efficient cost target may be around 0.5 to 2.6% per annum for electricity distribution and 0.5% to 2.5% for gas distribution. We note that these ranges are in line with the targets set in other jurisdictions. The upper bound of the ranges may contain some elements of catch-up efficiency and NMa should take this into consideration when setting the frontier shift target. The ranges do not take into account changes in capex required to support network growth (which is not initially funded through output growth) or increased depreciation requirements on existing assets and as such are not comparable with the RUTC estimates using Dutch DNO data.

We recommend that NMa investigate the feasibility of setting an explicit cost efficiency target with a separate adjustment for the fixed depreciation levels arising from the decision at the time of liberalisation. However, if NMa chose to continue using a RUTC type approach for setting the efficient cost frontier we would recommend that it investigates averaging annual changes over a longer period to reduce issues relating to short term volatility.

If NMa choose to set separate targets for opex, depreciation and cost of capital the following points need to be considered. For setting an efficiency target for opex it would be appropriate to look at partial productivity/cost measures. Table E.3 below provides ranges for the partial productivity measures. Assuming constant capital flows and selected industry comparators then a range of between 0.5% and 3.6% (the average for other regulated industries 11-15 years after privatisation) would be plausible. Setting the targets for depreciation and cost of capital would need to take into account the future capex spend of the companies’ (which could be based on historical trend).

2 The 2.6% for electricity distribution is based on the average across the ‘other studies’ i.e. (2.1% + 2.1% + 2.6% + 3.6%)/4. The 2.5% for gas distribution is based on the average across the ‘other studies’ i.e. (2.9% + 2.7% + 1.9%)/3.

(9)

6 Table E.3: Productivity growth and cost trend estimates as calculated (opex and/ or intermediate inputs)

Measure Range (% per annum) Frontier-shift only Cost measure or productivity measure

LEMS cost measure 0.5% Yes Cost measure

Other studies 1.2% (opex) Yes Productivity measure

RUOE -1.4 to 6.7% (3.6% average) No Cost measure

The assumptions required to apply a partial productivity/ cost measure are very strong and given the requirement on NMa of still setting targets, where possible, for depreciation and cost of capital we consider that NMa should focus on applying a total cost target.

Gas transmission

NMa’s current Method decision in setting GTS’s ongoing efficiency adjustment factor was to use a combination of Dutch output price indices and other studies to set an efficient cost target of 1% for GTS. NMa did not undertake an adjustment for catch-up efficiency due to a lack of comparator data.

In line with NMa’s current decision, we consider that the average change in output price indices for selected sectors in the Dutch economy provide an appropriate starting point for estimates of GTS’s scope for future efficient cost trends. We consider that basing the average output price change on selected industries which carry out similar types of activities as GTS may provide a more representative estimate for achievable cost trends than one based on the whole economy.

Our estimates for the average annual change in output price indices for the selected Dutch sectors carrying out similar types of activities to network companies over the period 1989 to 2007 are 0.5%.

The possible range for GTS’s scope for efficient cost reductions could therefore be between 0.5% and 2.1%.³ Setting the output price indices, of 0.5%, as the lower bound is in line with the empirical evidence from other jurisdictions showing much higher year-on-year efficiency gains.

Both bounds include input price inflation. While NMa currently make no adjustment for GTS’s notional income for catch-up efficiency, as the upper bound does not reflect catch-up efficiency, the ranges can apply even if NMa decide to make an adjustment for catch-up efficiency in future price controls.

NMa currently have a mechanism to adjust for changes in GTS’s capex requirements as such, if this mechanism was retained, a further adjustment factor would not be required if an output or productivity estimate were used to set the efficient cost frontier.

3 The 2.1% is based on the average from the ‘other studies’ i.e. (1.2% + 2.9%)/2.

(10)

7 Table E.4: Productivity growth and cost trend estimates (total expenditure/ costs)

Measure Range (% per

annum)

Cost measure or productivity measure

Other studies 1.2-2.9% Yes Mix

Electricity transmission

NMa set TenneT’s efficient price path using a combination of catch-up and frontier efficiency targets. These estimates were based on, for the most part, a set of reports prepared by Sumiscid (a consultancy). While TenneT has a similar depreciation arrangement as the DNOs the setting of a revenue cap allows NMa the flexibility to allow for this depreciation level while setting the efficiency cost target on the other costs. In addition, as a revenue cap is set for TenneT and a mechanism is in place to adjust for any ‘special’ capex, we do not consider that an adjustment factor for capex changes due to non-efficiency factors is required.

As with the DNOs and GTS, we believe that output price indices based on selected Dutch sectors undertaking similar types of activities as TenneT is an appropriate measure for the lower end of the scope for TenneT’s efficient cost frontier. Setting this as the lower bound is in line with the empirical evidence from other jurisdictions showing much higher year-on-year efficiency gains. The output price indices provide an estimate, of 0.5% cost reduction per annum, on what has been achieved in the Dutch economy in industries which have been operating in competitive markets but carrying out some similar activities. At the upper end of the range we consider that the other studies offer a good indication of the scope for frontier shift at 2.3% per annum on average.⁴ Note, none of the ‘other studies’ for electricity transmission include elements of catch-up efficiency.

Table E.5: Productivity growth and cost trend estimates (total expenditure/ costs)

Measure Range (% per

annum)

Cost measure or productivity measure

Other studies 2.1-2.5% Yes Mix

4 The 2.3% is based on the average across the ‘other studies’ i.e. (2.2% + 2.5% + 2.1% + 2.4%)/4.

(11)

8 Summary

In Table E.6 below we present possible ranges for the scope for setting efficient cost frontiers for each of the sectors that NMa regulates. These per annum efficient cost frontier shifts exclude any adjustments which may be required for changes in capex not funded through ‘within period’ volume increases and for changes in the DNOs’ depreciation (and TenneT’s if it is not addressed under the revenue cap). The output price indices comparison to TFP for the Dutch economy indicates that an adjustment for input price inflation differentials may not be required for the four sectors. In developing these ranges we began by considering the available evidence rather than first attempting to identifying estimates which sit within a tight range.

Table E.6: Efficient cost frontier sector ranges

Sector Range (% per annum)

Gas distribution 0.5 to 2.5%

Electricity distribution 0.5 to 2.6%

Gas transmission 0.5 to 2.1%

Electricity transmission 0.5 to 2.3%

In choosing a point within the range there are numerous factors that NMa should consider, for example, is there a separate adjustment for catch-up, are there any reasons why a Dutch network operator could not achieve, or even outperform, the efficiency gains seen in other jurisdictions, etc. Along these lines, we would expect the scope for outperformance to be much greater towards the bottom of the ranges in Table E.6 as evidence from other jurisdictions consistently show higher achieved ongoing efficiency. While at the upper end the scope for outperformance will be less, however these ranges are based on empirical studies and should set a challenging but achievable target. Towards the middle of the range we would expect the scope for under/out performance to be relatively symmetrical as these are in line with those set by regulators in other jurisdictions.

(12)

9

GLOSSARY

Abbreviation Definition

Capex Capital expenditure

CPI Consumer price index

DEA Data envelopment analysis

DNOs Distribution network operators

LEMS Labour, energy, materials and services

MEAV Modern equivalent asset value

Opex Operating expenditure

RAV Regulatory asset value

Repex Replacement capital expenditure

RUOE Real unit operating expenditure

RPEs Real Price effects

RPI Retail price index

RUTC Real unit total costs

TFP Total factor productivity

Totex Total expenditure

TSOs Transmission system operators

WACC Weighted average cost of capital

(13)

10

1. INTRODUCTION

1.1. Aim of the study

The Dutch Office of Energy Regulation, a part of the Dutch Competition Authority (‘NMa’) has engaged Cambridge Economic Policy Associates (‘CEPA’) to provide advice on setting an efficient cost frontier shift (see Section 2 below for more on definitions) for the four energy sectors that it regulates – electricity distribution and gas distribution (DNOs), electricity transmission (TenneT), and gas transmission (GTS). Cost frontier shift is also sometimes referred to in the regulatory literature as ongoing efficiency; it encompasses the productivity gains resulting from technical change over time and also taking account of changes in input prices over time. It does not include changes in productivity resulting from changes in the relative efficiency of firms within the regulated industry (sometimes referred to as catch-up efficiency). This is outside the scope of this project (except where it is relevant in computing frontier shift). NMa operates CPI-X regimes for each of the sectors and the frontier shift targets will be an input for the calculation of the X-factor in future price controls. NMa refers to the two concepts set out above, namely frontier shift and catch-up efficiency as dynamic and static efficiency respectively. For the remainder of the report we use the former terms, i.e. frontier shift and catch-up.

CEPA’s advice on possible efficient cost frontier shifts is to be applicable for the regulatory period commencing in 2014. NMa is still developing many aspects of the next regime and a method for setting ongoing efficiency targets and/ or changes in the level of inputs will help inform their development of other areas of the regime.

NMa’s current approach to price cap regulation means that its X-factor adjustment covers more aspects than just the forecast change in productivity, and in turn unit costs for each sector/ firm over time. It is used, in some instances, to adjust the notional income as required over the regulatory period which would include changes in depreciation and return on capital requirements (associated with the capital maintenance of the existing network and also new investment).

1.2. Previous studies

NMa has either commissioned or used a number of existing studies in determining appropriate productivity growth targets for each of its sectors. In this section we provide a brief summary of the most recent studies which relate to setting an efficient cost frontier target:

· Reckon (2011) – Productivity growth of GTS;

· Sumicsid (2009) – International Benchmarking of Electricity Transmission System Operators; and

· Oxera (2008) – Should DTe adjust expected productivity growth for catch-up effects when setting the X-factor.

(14)

11 1.2.1. Reckon (2011)

Reckon was commission to provide advice on GTS scope for productivity growth and how this could be incorporated in the X-factor. At the time of Reckon writing its report no decision on the method for setting the X-factor had been taken. As such, Reckon provided approaches based on two different possible methodologies:

· using a calculation that took account of total costs – operating expenditure, depreciation, and return on the regulatory asset base – for GTS at the start of the price control and then rolling this forward for each year of the regulatory period based on an adjustment factor (which Reckon referred to as ‘Y’). We discuss below how Y may be set; or

· setting the X-factor for a regulatory period by taking an estimate of GTS’s operating expenditure requirements at the start of the price control and rolling these forward by an annual adjustment factor (which Reckon referred to as ‘Z’). We discuss below how Z may be set. This approach would require separate adjustments for depreciation and return on capital.

Reckon noted that the annual factors were not intended to take account of changes over time in the volume of GTS’s outputs, for which other mechanisms would be required.

Based on these feasible potential methods for setting the X-factor Reckon identified four approaches (two for each) which were the most comparable to each method for estimating growth rates. These are summarised in Table 1.1 below. Reckon highlighted that, if the information that it presented was used as part of the decision for setting Z, then allowances for amortisation or capital expenditure should be on a consistent basis with that assumed for operating expenditure. The measures presented in the table above which are compatible with Z rely on the hypothesis that such allowances would cover the asset replacement needed to maintain the same amount and quality of services from capital assets (but no more) over the regulatory period i.e. if there was a constant flow of capital services.

(15)

12

Table 1.1: Reckon estimates for selected measures

Source: Reckon (2011) 1.2.2. Sumicsid (2009)

Sumicsid conducted a European benchmarking of electricity transmission system operators for the Council of European Energy Regulators (CEER). The benchmarking was carried out with support from 19 national regulatory authorities and 22 transmission system operators (TSOs).

Sumicsid’s efficiency analysis of the TSOs was based on a combination of “system science, engineering and econometrics”.⁵ The objective of the project was to produce estimates of frontier shift and changes in relative efficiency (or catch-up) that are robust and comprehensive and could be used, for example, to set X-factors. Sumicsid noted that the X-factor is, according to best practice, divided into a general productivity improvement factor (frontier shift) and an individual efficiency catch-up factor.

The efficiency estimation techniques used by Sumicsid depended on “the character of the underlying functions in terms of homogeneity, cost causality and production space”.⁶ The most extensive assessment that they used was based on a data envelopment analysis (DEA) type non- parametric frontier model with the assumption of constant or increasing returns to scale for a scope encompassing total expenditure for construction, maintenance, planning and administration (CMPA). Sumicsid’s scale assumption, non-decreasing returns to scale, protects grids below the optimal size from being compared with the most productive scale, while larger

5 Sumicsid (2009), page 4.

6 Ibid.

(16)

13 grids are compared to the entire set. The CMPA model consisted of three output variables: a normalised grid metric, connection density and the capacity of connected power for renewable energy including hydro.

Sumicsid’s results for frontier shift, produced from a panel covering 2003 to 2006, indicated a yearly productivity growth for best-practice electricity transmission operators in the range of 2.2- 2.5% in total expenditure for CMPA. Sumicsid compare these changes to studies of:

· European ECOM+ results for unit costs for 2000-2003 of 1.3% per annum reductions;

· Norwegian regional transmission operators net real frontier shift for 2001-2004 of 2.1%

per annum indicating a the scope for a reduction in costs of 2.1%; and

· American results for interstate transmission operators using FERC data for 1994-2005 of 2.4% net cost-weighted frontier shift per annum indicating a scope for reduction in unit costs of 2.4%.

We note that Sumicsid also produced a report in 2006 on benchmarking electricity transmission operators and a 2010 report for NMa on TenneT’s scope for catch-up efficiency.

1.2.3. Oxera (2008)

Oxera were commission by DTe (NMA) to investigate whether expected productivity growth should be adjusted for catch-up effects when setting the X-factor. Oxera’s conclusions are set out in Table 1.2 below.

Table 1.2: Oxera’s conclusion on adjustments for catch-up

Source: Oxera (2008)

Overall Oxera concluded that an adjustment was only required in the case of scenario 2b (more than 50% of companies were experiencing catch-up), or if there was full convergence towards the end of the previous period.

Oxera’ empirical analysis of both electricity and gas network operators led it to conclude that there had been very limited convergence in both sectors over the period examined. Therefore, it concluded that there was no need to adjust the X-factor for electricity and gas distribution operators.

(17)

14 1.3. Overview of approach

In order to provide NMa with advice on setting an efficient cost frontier target for companies within the sectors it regulates we adopted a three stage approach. This staged approach is set out in Figure 1.1 below.

Figure 1.1: Approach

1.4. Structure of report

The paper is structured as follows:

· Section 2 sets out some common concepts and definitions that are used throughout the report.

· Section 3 sets out common approaches used to calculate efficient cost frontier shifts and other international regulators’ decisions.

· Section 4 provides an overview of NMa’s approach to each of the sectors it regulates.

1. Identify approaches for setting ongoing efficiency

targets

2. Other complementary adjustments required

3. Applicability to Netherlands

• Regulator precedence

• Academic precedence

• K-factor to take account of capex growth

• Real price effects

• NMa’s requirements for the measure

• Data availability

4. Empirical analysis • Empirical analysis of applicable ongoing efficiency measures

4. Recommendation of applicable measure(s)

• Recommendation for setting an ongoing efficiency target

• Recommendation on other adjustment factors required to take account of changes in the level of output

Theoretical frameworkMethodology & empirical analysis

Stage 1Stage 2Stage 3 Advice

(18)

15

· Section 5 sets out the framework for choosing methods for measuring productivity growth applicable to the sectors regulated by NMa.

· Section 6 sets out the technical method for estimating each applicable approach and associated estimates.

· Section 7 provides our conclusions on the appropriate approach(es) that NMa might adopt to estimate productivity growth targets for the companies it regulates.

Annex A provides a theoretical definition of ‘efficiency’.

Annex B provides a discussion on liberalisation versus privatisation.

Annex C provides further details on the regulatory decisions discussed in Section 3

Annexes D, E and F provide detailed sensitivity results and analysis for TFP, Output price indices and LEMS (respectively).

Annex G contains the methodology behind - and the results of - our RUOE analysis for relevant comparator sectors, the summary of which is contained in section 6.

Annex H provides a summary of the academic and consultant studies set out in Section 6.

Annex I sets out the mathematical formulation of the methodologies used in this report.

(19)

16

2. CONCEPTS AND DEFINITIONS

2.1. Introduction

Throughout this report we refer to a number of concepts and definitions which assist in determining appropriate approaches to setting efficiency targets. In the sections below we define these terms and concepts in the context with which we have used them in this report.

2.2. Productivity

Productivity is defined as the ratio of output(s) to input(s); and may be expressed in terms of partial measures (e.g. labour productivity) or total factor productivity (the ratio of total outputs to total inputs based on a set of weights).^{7, 8} Productivity differences between firms, at a point in time, will be affected not only by differences in efficiency, but also by variations in the scale of production (sometimes referred to as scale efficiency), and differences in other environmental factors impacting on the production process.

When looking at productivity comparisons over time, an additional source of productivity change – technical change, or technological progress – also becomes possible. Technical change results from improvements in technology and may be represented as an outward shift in the efficiency frontier (or frontier shift). Frontier shift represents the movement in productivity over time that is achieved by the firms that are at the frontier of performance. Frontier shift is commonly measured as the change in the leading firm’s total factor productivity (TFP) over time.

Over time, total factor productivity (TFP) growth for any company can therefore be decomposed into four elements:

DTFP = DEfficiency (or catch-up) + DScale Effects + DEnvironment + DTechnology (1) Although it is not always possible to distinguish changes in environmental factors (DEnvironment) from technological progress (DTechnology).

Thus, whilst frontier shift is usually taken to mean technical change, it could also include changes in productivity associated with changes in environmental factors (such as quality obligations).

Technical change generally leads to a positive shift in the (production) frontier, however a negative shift is possible through. For example, improvements in quality that require more inputs but where quality is not included in the output measure.

The different elements of TFP growth shown in Equation (1) are important in a regulatory context. Economic regulators need to take a view on the future direction of each of these elements, in order to determine the scope for regulated companies to deliver improvements in TFP and in turn (after also taking account of changes in input prices) unit costs going forward.

7 See section 3.3 below.

8 The terms efficiency and productivity are often used interchangeably, for convenience, but they are not precisely the same. Annex A has a theoretical discussion on the definition of efficiency.

(20)

17 2.3. Illustration of frontier shift and catch-up

There can of course be firms within a sector, particularly in a recently liberalised or privatised sector, which are not at the frontier. An organisation which is considered to be inefficient in the present is deemed to fall short of the level of efficiency that is feasible (or achievable) with current technology and working practices (also known as the frontier of performance). In order to become more efficient based on current technology, the organisation would need to update its systems / working practices in order to catch-up to this frontier of performance.

Figure 2.1 illustrates the situation by comparing data over two years for a number of companies.

Two frontiers are therefore drawn on the diagram representing the two time periods (t=1 and t=2). In this example it will be seen that the cost frontier shifts downwards, indicating technological progress that reduces costs between the two periods. The frontier company, Z, moves its performance in line with the frontier. The inefficient company, Y, also improves its performance, but this time by more than the shift in the frontier, thus allowing it to move closer to the frontier. Company Y is therefore achieving ‘catch-up’, and sees an improvement in its efficiency score; although it still does not quite reach the frontier.

Figure 2.1: Cost frontier shift for given output

In the economy as a whole, or in sectors where there is assumed to be a reasonable amount of competition, if the sample of firms is both: (i) large; and (ii) random, it seems reasonable to expect that productivity improvements over time should be largely driven by frontier shift. That is, in competitive industries, inefficient firms will not survive and thus it might be expected that there would be no appreciable inefficiency. NERA (2006) make a similar point, noting that in competitive industries an equal number of firms ought to be moving closer to the frontier as

Ln (gas throughput) Frontier _{(t = 1)}

Ln (costs)

(t

Frontier _{(t = 2)} Z _{(t = 1)}

Z _{(t = 2)} Y _{(t = 1)}

Y _{(t = 2)}

(21)

18 those that are moving away from it, on average.⁹ In contrast, if the sample contains a significant proportion of companies that are commonly recognised to be experiencing catch-up, through the effect of privatisation, competition or comparative competition (via economic regulation), then it is appropriate to make an adjustment to the TFP improvement figure to recognise that not all the efficiency improvement is likely to relate to frontier shift.

Depending on the type of measure used, costs may change over time for reasons not solely associated with ongoing efficiency. For example, if accounting depreciation is used as part of a methodology to estimate TFP change based on changes in unit total costs then depreciation could vary period on period without a direct link to the actual flow of capital services computed on an economic basis. Accounting depreciation reflects the expense of using capital assets, while the change in the volume of productive capital stock can be assumed to reflect the flow in a firm’s capital services. These two measures may potentially be the same if the age-price profile and age-efficiency profiles of the assets are the same, however in practise these are likely to be different.¹⁰

2.4. Capital maintenance

In order to continue to provide services to their customers, regulated companies will need to maintain their asset base.¹¹ This process is commonly referred to as capital maintenance.

Infrastructure companies’ capital expenditure can be quite ‘lumpy’ as large projects are completed and the length of an asset’s life helps determine when needs to be replaced.

Therefore, this can lead to changes in a company’s capital maintenance requirements from one regulatory period to the next in order to cover the replacement cost of these assets (i.e.

depreciation costs).

Some measures of cost change will indirectly include changes in replacement expenditure by including depreciation in the measure (or directly, if a total cash expenditure figure is being computed). However, this will not provide information of the future capital maintenance needs of a company.

We note that depreciation is not in itself subject to efficiency improvements; however expenditure on new capital can lead to efficiency improvements which would in turn lead to lower growth in future depreciation (than if efficiency gains were not achieved).

2.5. Network growth

As well as changes in capital maintenance, a network company may also face altered costs due to changes in its network size/ scope. These changes could include expansion due to new customers being connected or increases in existing capacity. These changes will incur capex to implement and may lead to increases in opex related to the ongoing management, repairs, etc, of the new infrastructure. The growth of a network should be funded through increased volumes of

9 NERA 2006, High Level Efficiency Estimates for the Second Review Period, Report for the Office of the PPP Arbiter, p. 15.

10 OECD 2009, Measuring Capital – OECD Manual 2009 (Second Edition), pp 60-61.

11 There is a question around stranded assets, but that is outside the requirements of this definition.

(22)

19 services provided. However, as volumes increase over time, in order to ensure financeability the depreciation required may need to be profiled i.e., allowing higher nominal income earlier through higher per unit charges, but with these reducing over time as more units are distributed (over time volume growth should cover the expenditure).

2.6. Liberalisation versus privatisation

There is extensive evidence in the academic and regulatory literature that liberalisation in its various forms improves productivity. Competition is usually considered to be the most potent force driving improvements (whether publically or privately owned; see, for example, Caves and Christensen, 1980), but the general finding is that privatisation, with the associated profit maximising objective, adds a further incentive for productivity gains (see, for example, Rees (1984), Vickers and Yarrow (1988) and Alexander and Mayer (1997)). Other examples of the impact of economic regulation on productivity and efficiency include research carried out by Saal and Parker (2000). They found that tighter regulation in the water industry in England and Wales led to sharper cost reductions, an effect that was more important than the impact of privatisation itself.

The precedents from economic regulation of utilities in the UK and elsewhere in the world is that increasing the X factor, based on good benchmarking information, creates stronger incentives for cost reduction than simply pegging prices at CPI. Evidence from the UK also indicates that for the first control period, where regulation was not strong, gains were relatively small, and in subsequent price controls gains picked up, and then flattened out. In addition, in the UK, incentive regulation is applied to Network Rail and Welsh Water, which do not have shareholders, however evidence suggests efficiency improvements for these companies has not been as fast as in privatised companies.

Additional discussion and references on this issue are provided in Annex B.

2.7. Gross frontier shift versus net frontier shift

Some consideration needs to be given to determining whether a gross or net frontier shift should be applied in the X-factor. Gross frontier shift means the productivity gains that can be achieved by a firm or industry resulting from technical change, whilst net frontier shift refers to the difference between the technical change achieved in the firm or industry and that achieved by the economy as a whole.

Input prices need to be taken into account as well as productivity gains associated with frontier shift in order to get at cost-based trend measures. The equations for expressing cost trends (which can also be referred to as output price movements) for both a particular business and the economy are well established and can be represented in the following equations for the firm and the economy, respectively:

Firm: P_{F B} = P_{I B} – TFP _B (2)

Where variables in italics denotes growth rates, P is price movements, TFP refers to total factor productivity movements, B refers to a business, F refers to final outputs and I refers to intermediate inputs/outputs.

(23)

20 The economy-wide measure of inflation, represented by CPI, captures both changes in input prices and total factor productivity across the economy and this can be subtracted from both sides of equation (2):

P _{F B} – CPI = P _{I B} – TFP _B – CPI (3)

Rearranging, we have:

P _{F B} – CPI = (P _{I B} – CPI) – TFP _B (4)

The first term on the right hand side of (4) is termed differential inflation (or ‘real price effects).

The remaining term is a measure of the gross efficiency improvements that the regulated company can be expected to achieve. Therefore, the productivity adjustment term in the X- factor should reflect the difference between the sector’s productivity and that of the economy as a whole. Equation (4) can also be represented as (where E refers to the whole economy):

P _{F B} – CPI = (P _{I B} – P_IE) – (TFP_B - TFP_E) (5) The first term on the right hand side shows the difference in the input prices between the economy and the sector, while the second term represents net TFP.

2.8. Real price effects

Price cap regulation involves indexing prices to CPI, with any productivity savings being taken into account through an X-factor. However, the prices of inputs for the regulated companies may not rise in line with the input prices of the economy as a whole i.e. there could be differential inflation. This differential could lead to the regulated company’s costs changing at a greater rate than allowed for in its price. In turn this might lead to the companies being unable to cover their costs or, in a case where prices are falling relative to input price inflation in the economy as a whole, achieving more profit than intended by the regulator.

As set out in Section 2.7 above, if when using a gross TFP measure then input price inflation in the industry relative to CPI needs to be considered, but if using a net TFP measure then input price inflation in the industry relative to input price inflation in the economy needs to be considered.

Some measures of efficiency change can account for price inflation differentials, however if deemed necessary by the regulator then an explicit adjustment factor may need to be included in the price cap.

(24)

21

3. APPROACHES TO SETTING ONGOING EFFICIENCY TARGETS OR EFFICIENT COST TARGETS

3.1. Common methodologies 3.1.1. Introduction

In this section we summarise the different approaches used by regulators and academics to measure productivity growth. We discuss some of the strengths and weaknesses of these approaches with regards to NMa in more detail in Section 5. For the most part, these approaches have been covered in previous reports commissioned by NMa, including those by Reckon (2011)¹² and Europe Economics (2006).¹³

3.1.2. Törnqvist total factor productivity (TFP)

TFP takes into account all the factors of production (e.g. capital, and labour) used to produce goods and services. TFP growth therefore captures the component of the change in output that is not explained by changes in inputs. TFP indices provide a way of comparing the efficiency with which companies/ industries deploy their inputs in a multi-input, multi-output environment. Although as noted above they also reflect other differences, such as variation in the scale of production. They can be used both to compare firms/ industries at a specific point in time and over time.

A common type of TFP measure is based on a Törnqvist index which is a geometric index using the component’s share of total value to weight its movement e.g. the movement in the volume of labour is weighted using labour’s value share of GDP. Hereafter, when we refer to ‘TFP’ we are referring to a Törnqvist TFP index (as opposed to, for instance, a Malmquist TFP index which is discussed below).

The approach quantifies TFP change based on the residual or ‘unexplained’ component of output growth once the growth in inputs has been accounted for. In other words if output increased by 5%, but inputs only increased by 4% then TFP represents the growth in outputs over and above the growth in inputs. This residual does not identify whether the improvement in TFP is due to the firm(s) catching up to the frontier or the frontier itself moving (or indeed due to changes relating to scale effects). Instead, it is necessary to attribute the TFP improvement to either catch-up or frontier shift based on an a priori knowledge of the sample from which the data are drawn.

As noted earlier, in the economy as a whole, or where there is assumed to be a reasonable amount of competition, if the sample of firms is both: (i) large; and (ii) random, it seems reasonable to expect that the efficiency improvement should be largely driven by frontier shift.

In these circumstances, an equal amount of firms ought to be moving closer to the frontier as those that are moving away from it, on average. By contrast, if the sample contains a significant

12 Reckon, Productivity growth of GTS, March 2011.

13 Europe Economics, Research into Productivity Growth in Electricity Transmission and Other Sectors – A Report for DTe, March 2006.