Benchmarking Transnet Limited s Petroleum Pipelines

Benchmarking Transnet Limited s Petroleum Pipelines December 2011

Table of Contents Page CHAPTER 1... 1 1.1 Introduction... 1 1.2 Background... 2 1.3 Theoretical overview... 3 1.3.1 Literature Review... 3 1.3.2 What is Benchmarking?... 3 1.3.3 Applications of Benchmarking for Regulatory Purposes... 4 1.3.4 Performance Indicators and Benchmark Measures... 5 1.3.5 Statistical Benchmarking Methods... 6 1.3.6 Examples of Benchmarking Studies... 7 CHAPTER 2: Benchmarking the Efficiency of Transnet s Operations... 9 2.1 Ratio Analysis... 14 2.2 Data Envelopment Analysis... 25 Conclusion and Recommendations... 31 Bibliography... 32 Annexure A... 33

CHAPTER 1 1.1 Introduction The National Energy Regulator of South Africa (NERSA or the Energy Regulator ) indicated in its Reasons for Decisions (RfD) on Transnet s 2011/12 petroleum pipeline tariff that it would investigate the possibility of meaningful benchmarking of Transnet s petroleum pipelines. To fulfil this undertaking by the Energy Regulator, the concept of benchmarking Transnet s petroleum pipelines has been explored and as a result this report has been produced to solicit public comment and guidance on this benchmarking initiative. This report contains a theoretical overview of benchmarking including reviews of international literature on benchmarking, precedents for the use of benchmarking in regulation, and an overview of the three most-used benchmarking methodologies in the energy sector. To build organisational understanding of benchmarking and its role in the energy sector, a two-day training course conducted by NERA on Benchmarking was held in Pretoria on 13 and 14 October 2011. Delegates included staff from the three divisions within NERSA, as well as other stakeholders affected by or interested in the business of regulation. A first attempt at benchmarking the efficiency of Transnet s petroleum pipelines operations has been conducted by comparing Transnet with a group of proxy companies. NERSA has conducted a ratio analysis of performance measures based on an international review of related literature and consultations with leading consulting firm in energy regulation, NERA Economic Consulting. 1

1.2 Background Since the 1990s, many regulators of infrastructure industries around the world have adopted incentive-based models of regulating natural monopoly activities, with the aim of promoting improvements in efficiency in the absence of market mechanisms 1. A central issue to be confronted when promoting efficiency is how the efficiency requirements are to be set. One approach is through the benchmarking of utilities based on their relative efficiency. Benchmarking identifies the efficiency levels of the population of firms in the sector and measures the relative performance of the target firm against these based on various metrics. Countries such as the Netherlands, the United Kingdom, Norway, Canada and Japan have adopted benchmarking as part of their regulatory processes. Regulators can use cross-country benchmarking in order to evaluate the performance of utilities within the larger context of international practice. International comparisons enable regulators to measure the efficiency of utilities in comparison with international best practice 2. A review of literature on the use of benchmarking in economic regulation, an overview of the methods of benchmarking and examples of benchmarking are discussed in the theoretical overview section of this report. The second chapter of this report contains an overview of NERSA s approach to benchmarking, as well as the results of NERSA s first attempt at benchmarking. 1 Benchmarking and incentive regulation of quality of service: an application to the UK electricity distribution networks, Energy Policy 33 (2005) 2 International benchmarking and regulation: an application to European electricity distribution utilities, Energy Policy 31 (2003) 2

1.3 Theoretical overview 1.3.1 Literature Review This report focuses on utility benchmarking studies in the regulatory arena. There is a vast amount of literature on regulatory benchmarking, including published academic literature and work done by research firms that are specialists in the field. This report relies heavily on publications by the Pacific Economic Group (PEG) 3, Frontier Economics 4 and First Quartile Consulting and LLC (1QC) Elenchus Research Associates (ERA), Inc 5. These are leading consulting firms in utility regulation and have done some benchmarking of energy utilities for Regulatory bodies in North America and Europe. 1.3.2 What is Benchmarking? Broadly, benchmarking can be defined as a comparison of some measure of actual performance against a reference or benchmark performance 6. Frontier Economics defines a benchmark as a standard by which something may be measured or judged and benchmarking as the process through which a benchmark is identified 7. The Pacific Economic Group (PEG) describes benchmarking as a scientific approach to performance measurement that makes extensive use of data on utility operations. Indicators that reflect important dimensions of company performance are chosen. Company values are then compared to benchmarks that reflect the performance of other utilities 8. 3 Pacific Economics Group (PEG), LLC report Benchmarking The Cost of Ontario Power Distributors for further discussion, 20 March 2008 http://www.pacificeconomicsgroup.com 4 The Future Role of Benchmarking in Regulatory Reviews May 2010 5 CAMPUT Benchmarking for Regulatory Purposes Prepared by: First Quartile Consulting, LLC Elenchus Research Associates, Inc April 2010 6 Jamas, T, Pollitt, M, 2001. Benchmarking and regulation: International electricity experience. Utilities Policy 9, 107 130 7 The Future Role of Benchmarking in Regulatory Reviews May 2010 8 (www.peg.com; accessed on 01 Aug 2011) 3

PEG further describes benchmarking as a term that has been used more generally to indicate something that embodies a performance standard and can be used as a point of comparison in performance appraisals. PEG states that benchmarks are often developed using data on the operations of agents that are involved in the activity under study and statistical methods are useful in both the calculation of benchmarks and the comparison process 9. Benchmarking treats firms as production entities which transform inputs into outputs. The variables used may be physical or monetary units; monetary values of input costs are preferable in a regulatory context. 1.3.3 Applications of Benchmarking for Regulatory Purposes First Quartile Consulting, LLC (1QC) and ERA, Inc. (ERA) conducted a study for CAMPUT (Canada s Energy and Utility Regulators) in June 2009, which looked at different options available for using benchmarking as a regulatory tool for Canadian utilities 10. Their report argues that from a regulator s perspective, benchmarking can be used for the following reasons: a) Reducing Information Risk to mitigate the risk associated with the imperfect and incomplete information that regulators must rely on in making regulatory decisions; benchmarks can provide an independent check on the reasonableness of the information available to regulators. b) Monitoring to determine the utility s accountability (delivering performance to customers), individual efficiency assessment (delivering value to ratepayers), and utility industry s efficiency assessment (performing within the range of acceptable values for the industry). Benchmarking in this context is for data collection and to help establish a range of acceptable values or identify areas that require additional review. 9 (www.peg.com; accessed on 01 Aug 2011) 10 CAMPUT Benchmarking for Regulatory Purposes Prepared by: First Quartile Consulting, LLC Elenchus Research Associates, Inc April 2010 4

c) Audit to support the financial and operational review of utility performance including a systematic review and verification of results. In this case, benchmarking provides standard definitions of performance and expected results. d) Compliance to ensure that a utility is compliant with regulatory requirements. This may involve assessing whether the utility meets the requirements of accepted practices, legislation, rules and regulations in the form of specific standards or the terms of a contract. Benchmarking can be used to identify the validity of an approach and best practices from other companies in the industry. e) Rate-making to assess the validity of the information presented and used to set rates; address concerns about information risks and ensure that utilities are performing as efficiently and effectively as possible. Benchmarking provides valid comparison points across similarly performing utilities. 1.3.4 Performance Indicators and Benchmark Measures The CAMPUT benchmarking report published by First Quartile Consulting and ERA further outlines performance metrics useful for benchmarking utilities performance by regulators. The metrics are designed to provide an overview of the information required by regulators to understand the performance of the utility and in their view should ideally cover the following areas: Costs: to ensure that there is a prudent use of resources to ensure reasonable rates, but also that enough is invested in the organisation to ensure continued delivery of quality service. Asset management: to address the balance between investing in new infrastructure and establishing a robust maintenance programme to avoid interruptions in service or unexpected repair costs. 5

Customer care: allows regulators to understand how a utility is delivering on the promise it is making to customers and meeting the objective of utility accountability. Operations: focus on the efficient delivery of the product and timely installation of new connections; it refers to the reliability of the system and the way it is managed. 1.3.5 Statistical Benchmarking Methods The three most commonly used statistical benchmarking methodologies are: econometric modelling, indexing and data envelopment analysis (DEA). 11 Indexing (Unit Cost & Productivity Indexes) involves the comparison of a company s unit cost or productivity to historical values of such key performance indicators for a peer group. The challenge in using a unit cost approach is deciding which measure of output should be used. Accuracy also hinges on the degree to which the cost pressures faced by the peer group resemble those faced by the subject utility. Econometrics (Cost & Quality Models) Econometric cost models explain the relationship between utilities' costs and model parameters which are estimated using the historical cost drivers of a sample of utilities. When feasible, Econometric Cost Models have advantages over unit cost and productivity metrics in performance measurement in that econometric models can be used to predict the change in a company's cost given expected changes in local business conditions (for example, input price inflation and customer growth). 11 See Pacific Economics Group (PEG), LLC report Benchmarking the Cost of Ontario Power Distributors for further discussion, 20 March 2008 http://www.pacificeconomicsgroup.com 6

Data Envelopment Analysis (DEA) uses linear programming techniques to envelop data on sample firms that relate outputs to inputs. It is therefore essentially a technique for identifying what is known in economics as isocost and isoquant curves. Efficiency is measured as the distance from the best attainable curve. DEA is a frontier-oriented method of benchmarking. It measures the performance of firms against an efficient frontier or best practice. From a regulatory perspective, frontier methods can be used to identify performance gaps, particularly in the initial years of regulatory reform. 1.3.6 Examples of Benchmarking Studies Utilities and regulators (in Europe) have been using benchmarks in support of rates and regulatory proceedings. Sometimes these utilities choose to do so, and sometimes the regulators require it. The reasons and the approaches vary by jurisdiction and by utility. Below are some benchmarking studies that have been undertaken in different countries. Most benchmarking studies in the energy sector are on the electricity and gas industries. Benchmarking studies in the petroleum pipelines sector are not common. 1. The use of large-scale benchmark studies in rate proceedings British Columbia Hydro (BC Hydro) has been using a large-scale benchmarking approach in rate proceedings for the past ten years. It chooses specific type of information to demonstrate specific points about its capital investment levels, as well as to highlight its operating and maintenance approach. BC Hydro considers the benchmarks as a support tool, rather than a primary tool on which it bases its decisions. 12 12 See BC Hydro 2011 Revenue Requirements Exhibit B-1, the report has a comprehensive list of benchmarking studies conducted by BC Hydro during F2009 and F2010 http://www.bcuc.com/documents/proceedings/2010/doc_24719_b-1_bchydro-f11rr- Application.pdf 7

2. Benchmarking the Cost of Ontario Power Distributors 13 The Ontario Energy Board (OEB) consulted Pacific Economic Group (PEG) to help it develop an operational benchmarking method for rate making. The study by PEG considers the impact of service quality and capital use on operation, maintenance, and administration (OM&A) expenses and explores the potential for benchmarking capital costs. While the econometric method of benchmarking is adopted in the regulation of the OM&A expenses of Ontario s numerous power distributors, the study explains in detail other most commonly used benchmarking methods. 3. The Future Role of Benchmarking in Regulatory Reviews - prepared by Frontier Economics for the Office of the Gas and Electricity Markets (OFGEM), May 2010 14 OFGEM, which regulates the electricity and gas markets in Great Britain, requested Frontier Economics to conduct a study and write a report (the Frontier Report) on the future role of benchmarking in regulatory reviews for electricity distribution and transmission, as well as gas distribution and transmission. OFGEM considered adopting a high-level DEA benchmark of the recent historic costs of transmission operators amongst a small number of European peers. Given the limitations on data availability, Frontier Economics recognised that this approach was unlikely to provide definitive results. 13 Ontario Energy Board (OEB) report by: Pacific Economics Group (PEG) 20 March 2008 14 For more detailed theory see report prepared by Frontier Economics for OFGEM: The Future Role of Benchmarking in Regulatory Reviews May 2010. The report addresses areas like the context for benchmarking, criteria and approaches followed by companies to benchmarking. 8

CHAPTER 2: Benchmarking the Efficiency of Transnet s Operations Due to the complexity of tariff-on-tariff comparison as a result of the various factors that contribute to tariff administration, NERSA has considered the possibility of benchmarking the efficiency of costs in Transnet pipelines operations against that of international pipeline companies. The steps to a benchmarking process are as follows 15 : a) Select a benchmark (for example, distribution cost per customer). b) Compare the chosen metric for the utility (Transnet) to the average for the proxy group. c) Consider whether the cost pressures faced by the peer group resemble those faced by the subject utility (Transnet). d) Numerators should reflect the fixed and variable costs of running a pipeline: - the fixed cost component would be net plant ; and - the variable cost component would be the various categories of operating expenses. e) Denominators for consideration would be: - volume of throughput; - kilometres of pipeline; and - volume per kilometre. Using a consistent denominator helps in normalising the data. To benchmark the efficiency of costs requires an understanding of the cost drivers in the industry. This is crucial to ensure that appropriate benchmarking metrics are selected. Some of the cost drivers for a pipeline company are: Capacity Distance (km) 15 Graham Shutterworth, Wayne P. Olson. Introduction to Benchmarking 9

Terrain/Elevation Flow rate (m 3 /h) Number of entry and exit (off take) points Pipeline diameter The above are reflected in the value of assets, as well as the costs and revenues of the companies. Differences in costs can be attributed to the differences in economies of scale between the companies, as well as the geographic impact, which can result in significant differences in the cost of the assets over similar distances. Three ratios that were suggested by NERA for use in the benchmarking analysis are: a) This is a measure of the capital intensity of the company. Throughput is, however, a function of economic growth. In the instance where international comparisons are made, adjustments for the differences in the countries economic factors such as labour productivity, cost of labour, investments and other cyclical components that impact on economic growth, would be incorporated into the analysis. b) A key consideration in the analysis of costs is the reasonableness of costs, i.e. the costs prudently incurred? To make the results more comparable, an assessment of what percentage of allowable revenue (AR) is operating expenditure (OPEX) is required. c) This is also a measure of efficiency of operations. 10

Further assessment on whether Transnet s financial statements and performance reflect a sound business is conducted using its efficiency and productivity ratios. Financial ratios worth considering include: d) Profitability Ratios - Return on Asset - Return on Equity These show the efficiency with which assets and equity are used to generate Net Income. e) Liquidity Ratios - Current Ratio - Quick Ratio These show the company s short-term solvency and financial flexibility. f) Debt Utilisation Ratios - Debt: Equity Ratio This indicates what proportion of equity and debt the company is using to finance its assets. Companies in capital intensive industries usually have high debt: equity ratios, which reflect the financing of growth with debt. - Asset: Equity Ratio An increasing asset: equity ratio indicates that assets are increasing faster than equity. The increase therefore reflects an expansion of assets to generate earnings. g) Asset Utilisation Ratios - Asset Turnover Asset turnover measures a company s efficiency at using its assets in generating revenue; the higher the number the better. 11

NERSA has decided to conduct a ratio analysis of Transnet s performance relative to relevant international benchmarks as a first step in the benchmarking exercise. These ratios are also used in a preliminary DEA model for benchmarking. NERSA has decided to use the data of the companies used to determine a proxy beta for calculating the value of the market risk premium (MRP) in the formula for determining Transnet s cost of equity (Ke). The data was obtained from the United States of America s Federal Energy Regulatory Commission (FERC). The proxy companies considered by NERSA in the determination of the industry beta for 2010/2011 tariffs were: 1 EQT Corporation (US) 2 Enbridge Inc. (CN) 3 El Paso Corporation (US) 4 Magellan Midstream Partners, LP (US) 5 Plains All American Pipeline, LP (US) 6 New Jersey Resources Corporation (US) 7 Piedmont Natural Gas Company, Inc. (US) 8 AGL Resources, Inc. (US) 9 Cabot Oil and Gas Corporation (US) 10 Crosstex Energy Inc. (US) 11 Devon Energy Corporation (US) 12 Encana Corporation (US) 13 Enterprise Products Partners, LP (US) 14 EOG Resource, Inc. (US) 15 Laclede Group, Inc. (US) 16 National Fuel Gas Company (US) 17 Nicor Inc. (US) 18 Provident Energy Ltd. (CN) 19 South Jersey Industries, Inc. (US) 20 Southern Union Company (US) 21 TEPPCO Partners, LP (US) 22 UGI Corporation (US) 12

23 WGL Holdings, Inc. (US) For the purpose of this report, the list was limited to four companies based on the availability of data from the FERC databases. The proxy companies against which Transnet is benchmarked in this report are: (i) Sunoco, Inc Sunoco, Inc., through its 34% ownership interest in Sunoco Logistics, has approximately 7,900 miles of crude oil and refined products owned and operated pipelines; and approximately 40 product terminals. (ii) Enbridge Inc. Enbridge Inc. provides energy transportation, distribution and related services in North America and internationally. The company operates a crude oil and liquids pipeline system, and is involved in international energy projects, natural gas transmission and midstream business. The company also distributes natural gas and electricity, and provides retail energy products. (iii) Kinder Morgan Through its Products Pipelines business unit, Kinder Morgan transports over two million barrels per day of gasoline, jet fuel, diesel, natural gas liquids and other fuels through more than 8,000 miles of pipelines. The company also has approximately 50 liquids terminals in this business segment that store fuels, and offers blending services for ethanol and other products. (iv) Magellan Midstream Partners, LP Magellan Midstream Partners, LP is primarily involved in the storage, transportation, and distribution of refined petroleum products and ammonia. The company assets 13

include a pipeline system serving the mid-continent region of the United States (US), petroleum products marine terminal facilities, petroleum products terminals, and an ammonia pipeline system. The following assumptions have been made in terms of the conversion from United States metrics to South African metrics: Miles to Km 1.609344 Exchange Rate (R/US$) 7.04 (this is the 2007 average rate, used as a base to convert all the periods. A constant rand-dollar rate was utilised to eliminate the fluctuation of currencies and to simplify the analysis) Litres per Barrel 158.99 All monetary values used are real values with US$ values converted to Rand values, as Transnet operations are Rand based. 2.1 Ratio Analysis The ratio analysis compares Transnet and the proxy companies ratios over the years 2008, 2009 and 2010. A primary challenge faced in developing ratios is data availability. While NERSA has access to Transnet s information, obtaining the required comparable data for the proxy companies has proved to be a challenge. This has resulted in the analysis being limited to the following ratios: 1. Asset Turnover = where RAB refers to the Regulatory Asset Base 2. 14

3. 4. 5. 6. a) Asset Turnover Asset turnover measures a firm's efficiency at using its assets in generating sales or revenue - the higher the number the more efficient is the firm. The ratio measures the revenue that is generated for every Rand of asset owned by the company. For most companies, their investment in fixed assets represents the single largest component of their total assets. The same applies for the capital intensive. For a capital intensive company, capital productivity should result in low tariffs. Figure 1: Transnet s Asset Turnover 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 Transnet: Asset Turnover 2008 2009 2010 Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application NOTE: The Transnet Annual Financial Report for 2008 presents Transnet's performance for the year ended 31 March 2008. The same applies for the years 2009 and 2010. 15

In 2008, Transnet returned R0.38 in revenue on each Rand of asset owned. By 2010, this had dropped to R0.12. These asset turnover ratios are low, which is understandable as Transnet s asset base has been increasing due to the capitalisation of the NMPP. Figure 2: Asset Turnover - Transnet and Proxy Companies 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 Enbridge Inc. Kinder Morgan Magellan Midstream Partners, LP Sunoco Transnet Pipelines 0.00 2008 2009 2010 Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application http://www.ferc.gov/docs-filing/forms/form-6/data/asp Asset turnover for all four companies has dropped since 2008, with the exception of Enbridge, whose asset turnover increased in the period 2009 to 2010. Transnet, however, has a lower asset turnover in comparison to the proxy companies. 16

m 3 /R m 3 /R b) Throughput/RAB Figure 3: Transnet s Throughput/RAB 6.00 5.00 4.00 Throughput/RAB 3.00 2.00 1.00 0.00 2008 2009 2010 Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application Transnet s capital intensity increased by 62% in the period from 2008 to 2010. These results are consistent with the results in 2.1.1 above. Figure 4: Throughput/RAB Transnet and Proxy Companies Throughput/RAB 45 40 35 30 25 20 15 10 5 0 2008 2009 2010 Enbridge Inc. Kinder Morgan Magellan Midstream Partners, LP Sunoco Transnet Pipelines Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application http://www.ferc.gov/docs-filing/forms/form-6/data/asp 17

The trend is quite different from what is seen in Figure 2 above. Magellan s capital intensity is relatively constant throughout the three years. While Enbridge s asset turnover dropped in the period 2009 to 2010, a similar trend is not evident in this case. An investigation into the underlying reasons will be conducted in the follow-up discussion paper. c) Operating Expenditure/RAB Figure 5: Transnet s OPEX/RAB Opex/RAB 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 2008 2009 2010 Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application Both OPEX and RAB increased in the period 2008 to 2010. The downward sloping curve is therefore a result of an increase in RAB that is growing faster than the increase in OPEX. 18

R/km Figure 6: OPEX/RAB Transnet and Proxy companies 0.6 0.5 0.4 0.3 0.2 0.1 0 OPEX/RAB 2008 2009 2010 Enbridge Inc. Kinder Morgan Magellan Midstream Partners, LP Sunoco Transnet Pipelines Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application http://www.ferc.gov/docs-filing/forms/form-6/data/asp All companies have a downward sloping curve except for Enbridge, with Transnet having the lowest ratio. d) Net Plant/Kilometres of Pipeline Figure 7: Net Plant/Kilometres of Pipeline Transnet 4500000 4000000 3500000 3000000 2500000 2000000 1500000 1000000 500000 0 Transnet: Net Plant/Kilometres of Pipeline 2008 2009 2010 Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application 19

R/km This ratio reflects an increase in Transnet s plant over the three years under review. This increase is due to the massive infrastructure expansion programme Transnet has embarked on. Figure 8: Net Plant/Kilometre of Pipeline Transnet and Proxy Companies Net Plant/Kilometres of Pipeline 4500000 4000000 3500000 3000000 2500000 2000000 1500000 1000000 500000 0 2008 2009 2010 Enbridge Inc. Kinder Morgan Magellan Midstream Partners, LP Sunoco Transnet Pipelines Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application http://www.ferc.gov/docs-filing/forms/form-6/data/asp Transnet and Enbridge have a higher plant per kilometre than the other three companies. There was a drop in Sunoco s Net Plant in 2009, followed by an increase in 2010. Table 1: Net Plant and Kilometres of Pipeline values Kilometres of Pipeline Net Plant 2008 2009 2010 2008 2009 2010 Mean 5 402 5 829 5 598 6 080 451 078 8 511 757 822 11 488 043 581 Min 681 681 681 57 500 092 57 280 602 74 717 497 Max 13 593 13 593 13 670 14 696 558 352 22 291 529 919 30 552 051 587 Enbridge Inc 6 233 6 233 7 139 14 696 558 352 22 291 529 919 30 552 051 587 Kinder Morgan 681 681 681 57 500 092 57 280 602 74 717 497 Magellan 13 593 13 593 13 670 8 776 552 982 10 444 954 109 13 125 466 381 Sunoco Inc 4 080 6 217 4 080 3 397 543 962 3 785 954 478 4 086 682 440 Transnet 2 423 2 423 2 423 3 474 100 000 5 979 070 000 9 601 300 000 20

R/m 3 /km From Table 1 we see that in 2010, Enbridge s regulatory asset base (Net Plant) is three times the average and is the largest in comparison to the other companies, with a relatively long average pipeline length. Transnet has a large asset base but the length of its pipelines is less than half the average. While the asset base has been increasing over the three years, the length of its pipelines has not changed. The length of Sunoco s pipelines increased in 2009, but dropped to the 2008 value in 2010. This is a possible data capturing error and will be investigated in the follow-up report. e) Operating Expenditure/Volume per kilometre Figure 9: Operating Expenditure/Volume per kilometre Transnet Transnet: Operating Expenditure/Volume per kilometre 80 60 40 20 0 2008 2009 2010 Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application The increase reflected over the three years is due to an increase in operating expenses, which is likely to be due to the increasing costs of maintenance on the Durban to Johannesburg Pipeline (DJP). NERSA needs to further investigate what the desirable value for this ratio is. 21

R/m 3 /km Figure 10: Operating Expenses/Volume per kilometre Transnet and Proxy Companies Operating Expenditure/Volume per kilometre 900 800 700 600 500 400 300 200 100 0 2008 2009 2010 Enbridge Inc. Kinder Morgan Magellan Midstream Partners, LP Sunoco Transnet Pipelines Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application http://www.ferc.gov/docs-filing/forms/form-6/data/asp Transnet s performance is in line with two other companies. Magellan s performance over the three years and Enbridge s performance in 2010 are outliers in this analysis. Table 2: Operating Expenditure and Volume per Kilometre values Operating Expenditure (R) Volume per kilometre (m 3 /km) 2008 2009 2010 2008 2009 2010 Average 981 066 896 1 087 654 438 1 858 814 157 9 504 312 8 364 436 9 517 690 Min 28 653 487 19 047 985 18 087 401 3 484 747 3 074 595 2 474 968 Max 1 923 355 088 2 185 507 688 6 388 956 935 17 765 056 15 894 402 20 020 498 Enbridge 1 800 445 687 2 185 507 688 6 388 956 935 15 689 942 15 894 402 13 872 321 Kinder Morgan 28 653 487 19 047 985 18 087 401 3 610 293 3 074 595 3 894 620 Magellan 1 923 355 088 1 816 681 480 1 435 565 828 3 484 747 3 509 015 2 474 968 Sunoco 874 853 311 1 029 546 708 1 002 754 962 17 765 056 12 257 951 20 020 498 Transnet 278 026 906 387 488 329 448 705 657 6 971 523 7 086 215 7 326 042 Magellan started with a high operating expenditure in 2008 compared to the other companies, but has since decreased its operating expenditure steadily over the three years. Magellan s volume per kilometre is on average three times smaller than the average, and has decreased over the period 2009 to 2010. This results in an inflated 22

value for this ratio. Enbridge s operating expenditure tripled from 2009 to 2010, while its volume per kilometre decreased during the same period. An investigation into the underlying factors will be considered in a follow-up report. f) Operating Expenditure/Volume Figure 11: Operating Expenditure/Volume - Transnet 0.03 Transnet: Operating Expenditure/Volume 0.025 0.02 R/m 3 0.015 0.01 0.005 0 2008 2009 2010 Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application The ideal trajectory for this graph in terms of efficiency is a downward slope. The results suggest that Transnet could be becoming less efficient. 23

Figure 12: Operating Expenditure/Volume Transnet and Proxy Companies 1.2 Operating Expenditure/Volume R/m 3 1 0.8 0.6 0.4 0.2 Enbridge Inc. Kinder Morgan Magellan Midstream Partners, LP Sunoco Transnet Pipelines 0 2008 2009 2010 Sources: Transnet Annual Financial Report for 2008, 2009 and 2010 NERSA Reasons for Decision, Transnet Pipelines Tariff Application http://www.ferc.gov/docs-filing/forms/form-6/data/asp Transnet s performance, however, remains in line with three of the proxy companies. Magellan s performance is consistent with the results from the above ratios, which shows that the company s operating expenditure is high in comparison to volumes. The ratio analysis shows that while there are some possible inefficiencies in Transnet operations, its performance is comparable with that of the proxy companies. 24

2.2 Data Envelopment Analysis Data Envelopment Analysis (DEA) is a non-parametric method and uses piecewise linear programming to calculate (rather than estimate) the efficient or best-practice frontier of a sample. The decision-making units (DMUs) or firms that make up the frontier envelop the less efficient firms. The efficiency of the firms is calculated in terms of scores on a scale of 0 1, with the frontier firms receiving a score of 1. DEA models can be input- or output-oriented, and can be specified as constant returns to scale (CRS) or variable returns to scale (VRS). The CRS hypothesis suggests that companies are flexible to adjust their sizes to the optimal firm size. In contrast, the VRS approach is less restrictive since it compares the efficiency of companies only within similar sample sizes. This approach is adopted if the companies are not free to choose or adapt their size. The comparison between the two approaches also provides some information about the underlying technology: if the results of the CRS and the VRS approaches are similar, then the returns to scale do not play an important role in the process. Output-oriented models maximise output for a given quantity of input factors. Conversely, input-oriented models minimise input factors required for a given level of output. Given that most distribution utilities have an obligation to meet demand, they can only become more efficient by providing a predefined output level with fewer inputs. Over the past 20 years, DEA has attracted much attention from among the wide spectrum of energy and environmental modelling techniques. DEA has been accepted as a major frontier technique for benchmarking energy sectors in many countries, particularly in the electricity industry. An international survey on regulatory benchmarking for distribution companies found that Chile, Columbia and Brazil were all employing benchmarking, with DEA analysis 25

being the most popular approach 16. Similarly, the Finnish Energy Market uses a DEA model for distribution company efficiency benchmarking. In Norway there are a large number of utilities (approximately 180), and the regulator uses the DEA technique with multiple inputs and outputs and directly converts the benchmarking scores into price caps. A growing number of studies demonstrate the application of DEA in the benchmarking of electricity distribution, gas distribution and water utilities. An important step in DEA is the choice of appropriate input and output variables. The variables should, to the extent possible, reflect the main aspects of resource-use in the activity concerned. DEA can also control the effect of environmental variables that are beyond the control of the management of firms but affect their performance. Also, the basic DEA model illustrated above does not impose weights on model input and output variables, but it can be extended to incorporate value judgements in the form of relative weight restrictions imposed on model inputs or outputs. NERSA s preferred model is input-oriented and assumes constant returns to scale (CRS) so that the measured relative efficiency of firms is not affected by their size (costs vary with for example the units of energy delivered). The model uses a single cost input reflecting the OPEX of the distribution business of the utilities. As reported in the paper on benchmarking in electricity regulation by Jamasb and Pollitt 17, the most widely used output variables for modelling of electricity distribution utilities are: i. units of electricity delivered; ii. number of customers; and iii. length of network. The comparable variables in the pipelines industry are: i. volume/throughput ii. number of customers iii. kilometres of pipeline 16 Performance Benchmarks for Electricity Distribution Companies in South Asia, USAID SARI/Energy Program, November 2004 17 Jamasb, T. and Pollitt, M. (2001), Benchmarking and regulation: International electricity experience, Utilities Policy, Vol. 9/3, pp. 107-130. 26

In this benchmarking study, Transnet s petroleum pipelines performance is compared to that of the four proxy companies listed at the beginning of this chapter. In order to smooth the variables, a three-year average of each of the ratios is used in the model. Table 3: Summary Statistics for the dataset OPEX Volume Kilometres of Pipeline Target Company: Transnet Pipelines 371 406 964 17 270 966 667 2 423 Proxy Companies: Sunoco 969 051 660 76 786 546 107 4 792 Enbridge Inc. 3 458 303 437 98 652 434 961 6 535 Kinder Morgan 21 929 625 2 400 675 476 681 Magellan 1 725 200 799 1 899 887 715 13 618 Sample: Mean 1 309 178 497 39 402 102 185 5 610 Minimum 21 929 625 1 899 887 715 681 Maximum 3 458 303 437 98 652 434 961 13 618 The linear equations to be solved in the calculation of the efficiency scores are provided in Annexure A. Table 4: Model Specifications Variable Model 1 Model 2 Model 3 OPEX I I I Volume O O Kilometres of Pipeline O O Key: O Output Variable I Input Variable 27

The three DEA models were run on MaxDEA software developed by Cheng Gang and Qian Zhenhua (2011) 18. Model Results Efficiency scores for the five companies considered are reported in Table 7. Note that the efficiency of the companies is calculated in terms of scores on a scale of 0 1, with the frontier firms receiving a score of 1. Table 5: Efficiency scores using both constant returns to scale and variable returns to scale specifications Magellan Enbridge Inc. Kinder Morgan Midstream Partners, LP Sunoco Transnet Pipelines Model 1 CRS 0.260581 1 0.254287 0.723829 0.42478 Model 2 CRS 0.260581 1 0.01006 0.723829 0.42478 Model 3 CRS 0.060873 1 0.254287 0.159302 0.210158 CRS Average 0.194012 1 0.172878 0.535653 0.353239 All three models have Kinder Morgan as the benchmark company (efficiency score=1). In none of the models is Transnet the least efficient company. With the exception of Kinder Morgan, the efficiency scores for all companies are very low under Model 3, an indication of a possible inappropriateness of the model specification. On closer inspection, we notice that Kinder Morgan is a significantly smaller company in comparison to the other four companies, even in terms of the asset base as seen Table 6 below. 18 http://maxdea.cn 28

Table 6: Summary Statistics OPEX RAB Volume Kilometres of Pipeline Target Company: Transnet Pipelines 371 406 964 6 351 490 000 17 270 966 667 2 423 Proxy Companies: Sunoco 969 051 660 3 756 726 960 76 786 546 107 4 792 Enbridge Inc. 3 458 303 437 22 513 379 953 98 652 434 961 6 535 Kinder Morgan 21 929 625 63 166 064 2 400 675 476 681 Magellan 1 725 200 799 10 782 324 491 1 899 887 715 13 618 Sample: Mean 1 309 178 497 8 693 417 493 39 402 102 185 5 610 Minimum 21 929 625 63 166 064 1 899 887 715 681 Maximum 3 458 303 437 22 513 379 953 98 652 434 961 13 618 We therefore ran a fourth model excluding Kinder Morgan with OPEX as the input and volume as the output. The results are as follows: DMU Score Enbridge 0.360003 Magellan 0.013898 Sunoco 1 Transnet 0.586852 Under the model Sunoco is the benchmark company with Transnet s score above 50%. Figure 12 illustrates the main features of our model. The figure shows the five companies used in our analysis with OPEX and RAB as the inputs and volume for the output. The vertical and horizontal axes represent the OPEX and RAB per unit of output respectively. 29

RAB/Volume RAB/Volume Figure 11: Data Envelopment Analysis All Five Companies 6 4 E 2 0-2 D B A C 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 OPEX/Volume A - Transnet B - Sunoco C - Enbridge D - Kinder Morgan E - Magellan This graph, however, shows Magellan to be the outlying company. Drawing the graph without Magellan we see the following: Figure 12: Data Envelopment Analysis Excluding Magellan 0.4 0.3 0.2 A C 0.1 0 D B 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 OPEX/Volume Once again we see clearly that Kinder Morgan uses less RAB and OPEX per unit of output. Transnet requires more RAB per unit of output than all the companies. While the results of this analysis are not definitive, they reinforce the results of the ratio analysis: the efficiency with which Transnet is using its assets must be investigated. 30

Conclusion and Recommendations Results from the ratio analysis in Chapter 2 suggest that Transnet is within the range of the proxy companies in terms of efficiency. However, while its performance is in line with the proxy companies, the upward-sloping operating expenditure curves (suggesting that Transnet is becoming less efficient), as well as the downward sloping cash flow curve (another sign of inefficiency) are an indication of the need for further investigation into Transnet s performance. The next step in the benchmarking exercise could include an investigation into the underlying factors driving the trajectory of the various ratio curves, the desirable slope for the curves in terms of efficiency, as well as further analysis to establish desirable benchmark values for each of the ratios. NERSA decided to investigate the possibility of conducting meaningful benchmarking. It is possible that this preliminary analysis could be made more meaningful if it were to be customised to suit local conditions, but this too faces data and other challenges. This will be further considered following input from stakeholders. The use of benchmarking as an instrument for the setting of South African petroleum pipeline tariffs is not recommended as benchmarking does not easily provide definitive results. Rather, the results of a benchmarking exercise may be a useful step in a debate about the cost efficiency of a utility. Benchmarking, therefore, may be more useful in pointing to the questions that could be posed in relation to the cost performance of a target utility. NERSA s efforts thus far point to the limited usefulness of benchmarking for South African petroleum pipelines given the limited data availability. However, this benchmarking exercise has suggested that benchmarking may be a more useful tool in efforts to improve the efficiency of petroleum storage and loading licensees where more local data should be available. This will be considered in future. 31

Bibliography 1. Pacific Economics Group, LLC (20 March 2008): Benchmarking the Costs of Ontario Power Distributors 2. Frontier Economics (May 2010): The Future Role of Benchmarking in Regulatory Reviews 3. First Quartile Consulting, LLC Elenchus Research Associates, Inc (April 2010): CAMPUT Benchmarking for Regulatory Purposes 4. Cheng Gang (June 2011): MaxDEA manual version 5.2 5. Amit Kabnurkar (14 March 2001): Mathematical Modeling for Data Envelopment Analysis with Fuzzy Restrictions on Weights 6. BC Hydro 2011 Revenue Requirements Exhibit B-1. http://www.bcuc.com/documents/proceedings/2010/doc_24719_b-1_bchydro- F11RR-Application.pdf 7. Transnet Annual Financial Report 2008 8. Transnet Annual Financial Report 2009 9. Transnet Annual Financial Report 2010 10. NERSA Reasons for Decision, Transnet Pipeline Tariff Applications. www.nersa.org.za/petroleumpipelines/tariffs/pipelines/tariffdecisions/current 11. Form 6/6-Q - Annual/Quarterly Report of Oil Pipeline Companies. http://www.ferc.gov/docs-filing/forms/form-6/data/asp 12. Jelena Zorić, Nevenka Hrovatin, Gian Carlo Scarsi (April 2009): Gas Distribution Benchmarking of Utilities from Slovenia, the Netherlands and the UK - an Application of Data Envelopment Analysis 13. Tooraj Jamasb, Michael Pollitt, (2003): International benchmarking and regulation - an application to European electricity distribution utilities. Energy Policy 31 (2003) 1609 1622 14. USAID SARI/Energy Program (November 2004): Performance Benchmarks for Electricity Distribution Companies in South Asia. www.sari-energy.org 15. Working Paper CMI EP 19/DAE 0312, January 2003, Dept. of Applied Economics, University of Cambridge 16. Jamasb, T and Pollitt, M. (2001), Benchmarking and regulation: International electricity experience, Utilities Policy, Vol. 9/3, pp. 107-130. 32

Annexure A Assume there is information on K inputs and M outputs for each of N firms. For the i- th firm, these are represented by the column vectors x i and y i, respectively. The K N input matrix X and M N output matrix Y represent the data for all N firms. The linear programme of input-oriented CRS envelopment model is formulated as follows: min θ, λ θ st -y i + Yλ 0 θx i Xλ 0 (1) λ 0, where θ is a scalar and λ is a Nx1 vector of constants. The value of θ obtained will represent the technical efficiency score of the i-th firm. The linear programming problem must be solved N times, once for each firm. Essentially, the problem takes the i-th firm and then seeks to radially contract the input vector x i as much as possible, while still remaining within the feasible input set. The inner-boundary of this set is a piece-wise linear isoquant, determined by the observed data points. Since θ is a feasible solution to (1), the optimal value θ 1. If θ = 1, the current input levels can no more be proportionally reduced, indicating that a firm is on the frontier. Otherwise, if θ < 1, then the firm is dominated by the frontier. In the VRS DEA model, a convexity constraint is added to (1): (2) This additional constraint ensures that the firm is compared with other firms of a similar size. When not all the firms are operating at the optimal scale, then technical efficiency as calculated by the constant returns to scale model (TE CRS ) will include pure technical efficiency (TE VRS ) as well as scale efficiency (SE): TE CRS = TE VRS x SE (3) 33

By conducting both CRS and VRS DEA, one can obtain a scale efficiency measure for each firm. 19 19 Gas Distribution Benchmarking of Utilities from Slovenia, the Netherlands and the UK: an Application of Data Envelopment Analysis 34