An economic model for the cost of electricity service interruption in South Africa

Utilities Policy xxx (2017) 1e10

Contents lists available at ScienceDirect

Utilities Policy journal homepage: www.elsevier.com/locate/jup

An economic model for the cost of electricity service interruption in South Africa U.J. Minnaar*, a, W. Visser b, J. Crafford b a b

Eskom Holdings, South Africa PrimeAfrica, South Africa

a r t i c l e i n f o

a b s t r a c t

Article history: Received 4 August 2016 Received in revised form 24 August 2017 Accepted 24 August 2017 Available online xxx

Cost of unserved energy provides an economic measure of the cost of electricity interruptions to electricity customers. These values inform investment and refurbishment decisions related to the power system, with the aim of optimizing network reliability. This paper reports on the development of an economic model to estimate the value of electricity reliability in South Africa by Eskom to meet regulatory requirements. The model allows for the scalability of COUE from national level to more detailed resolutions. The impact is that decisions for power system investment can be made to meet the needs of planning and regulatory applications. © 2017 Published by Elsevier Ltd.

1. Introduction In South Africa, the Cost of Unserved Energy (COUE) is used to provide an economic measure of the cost of electricity interruptions to electricity customers and the economy as a whole. These values are used to inform a number of investment and refurbishment decisions related to the electrical power system, with the aim of optimising network reliability. The economic cost of electricity service interruptions has been of considerable interest due to load-shedding in South Africa in the period 2008 to 2015. The benefit of reducing the frequency and duration of electricity interruptions is quantified in economic terms for planning and regulatory purposes so that the business case for network investment and refurbishment can be defined and optimal levels of reliability can be engineered for the needs of the South African economy. COUE is defined as the value (in South African Rands per kWh) placed on a unit of electricity not supplied due to an unplanned interruption of a short duration. These unplanned, short-duration outage events are expected in a well-planned system with an adequate reserve margin as a result of random failures of equipment. Typically, a power system planner would balance the total COUE against the cost to supply the energy not served in order to make optimal planning decisions. It is assumed that businesses and households experience theses outages infrequently, irregularly, and * Corresponding author. E-mail address: [email protected] (U.J. Minnaar).

of short duration; and therefore little or no mitigation measures are implemented by customers. However, over the long-term, the economic impact to customers of regular short-duration interruptions or the quantifiable “nuisance value” to customers should be factored into planning and investment decision-making to optimize systems and improve reliability. The contributions of this paper include the development of a model for COUE for the residential sector and the economy at large, focusing on South Africa. The methodology takes into account both the direct and consequential indirect economic effects of electricity interruptions. We develop a macro-economic approach that utilises publicly available data and enables repeatable, transparently calculated numerical values to represent the COUE. 2. Where and how is COUE used? Electricity interruption cost studies have been conducted in numerous jurisdictions worldwide. Billinton (2001) and more recently Electricity Reliability Council of Texas (ERCOT, 2013) reviewed studies from different regions worldwide. ERCOT concluded, however, that the findings across regions are limited by lack of comparability. Nooij et al. (2007) identified two decision-making applications for interruption-cost values: first, to make socially optimal investment decisions, and second, to decide which customers should be cut off in times of electricity supply shortages. COUE values have been applied in transmission planning and investment in Australia (Hicklin, 2010), Canada (Bhavaraju, 2004)

http://dx.doi.org/10.1016/j.jup.2017.08.010 0957-1787/© 2017 Published by Elsevier Ltd.

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010

2

U.J. Minnaar et al. / Utilities Policy xxx (2017) 1e10

and Germany (Praktiknjo et al., 2011) as well as in tariff design in Sweden (Carlsson and Martinsson, 2008), Thailand (Energy Planning and Policy Office Thailand, 2001), and India (TERI, 2001). COUE is also used for generation investment planning and policy decisions in Great Britain (London Economics, 2013), Spain (Linares and Rey, 2013), and Ireland (Leahy and Tol, 2010). In the South African context, COUE values are used exclusively to inform socially optimal investment decisions for utility systems. Both the South African Transmission Grid Code and the Distribution Network Code require a regulator-approved method of determining the COUE as an economic parameter for network investment criteria. Additionally, COUE is defined as an input parameter to South Africa's Integrated Resources Plan for future generation investment. In practice, these values are used for different applications with respect to generation, transmission, and distribution of electricity in South Africa. In generation planning, COUE is used to assess the risk of economic damage (at macro-economic level) as a result of generation capacity inadequacy. This planning is concerned both with constrained economic growth as well as total losses in the economy resulting from interruptions. Investment in the transmission system in South Africa is primarily based on a deterministic (n-1) criterion. The economic impact of losing load or not being able to supply load is considered before a decision is made regarding new investment. COUE is commonly used in comparative analyses to prioritize investment options. COUE is also used to justify the capital expenditure required to implement utility projects and identify the least economic cost (Marais et al., 2006). Refurbishment projects require COUE because the level of impact on some refurbishment projects will depend on the customer profile of the individual network. For electricity distribution, COUE is used for load forecasting, reliability based planning, and investment decisions. Load forecasting is premised on sub-zone classification and customer class building up from the sub-station level. This type of analysis requires economic impact measurement disaggregated by sub-station and by economic sector. Reliability-based planning uses COUE values to inform breakeven planning for capital investments. For distribution network planning, COUE is used to identify life-cycle least economic cost infrastructure investments among various alternatives. The cost of a project to the electricity utility is weighed against the cost impact if the project is not implemented and customers are not served. In addition to the application of COUE to investment and refurbishment decisions, a need also exists within South Africa's electricity distribution sector to ensure consistent and optimal decision-making across entities, including the vertically integrated national utility (Eskom) as well as municipal utilities. Approximately half of South Africa's electricity distribution is delegated to municipalities. The municipalities, however, “lack appropriate, politically-insulated commercial structures for the management of distribution and supply, and which, in many cases, have failed to maintain infrastructure and retain suitably skilled staff” (Newberry and Eberhard, 2008). The need for a standardised COUEvalues to improve selection and prioritisation of distribution infrastructure projects was recognised by Cameron and Van der Merwe (2014). 3. Methods to estimate the value of service interruptions International researchers have developed several approaches to determining the cost of interrupted electrical power service. Primary amongst these are the production-function and customer surveys (Nooij et al., 2007).

3.1. The production-function method The production-function method uses official, published macroeconomic data, such as gross domestic product (GDP) and gross value added (GVA), and household expenditure measures. This method estimates the impact of electricity interruptions due to lost production for businesses or lost utility for households. According to Billinton (2001), the production method has several advantages: it is feasible and simple to implement as a result of data availability; it uses official, publicly available data, and is thus, transparent, verifiable and repeatable; it is consistent with the System of National Accounting (SNA) methodology of the United Nations (United Nations, 1993); it enables and supports macro-economic modelling, and it allows for scalability of COUE measures from a national level to a more detailed resolution (e.g., the municipal level). The key disadvantage of the production approach is that it assumes that macro-economic indicators are a reasonable proxy for costs of unserved energy. This is because the approach is based on macro-economic estimations and no data are collected directly from customers, and it is insensitive to variations in costs associated with time-of-day, day-of-week, and time-of-year in which interruption occurs (Billinton, 2001). The impact of time dependency on interruption costs is well documented (ERCOT, 2013) and methods to account for this effect include using weighting factors (Billinton and Wangdee, 2005) as well as associating interruptions with time-of-day and seasonal intervals (Herman and Gaunt, 2010). These methods are used in conjunction with customer surveys to account for the impact of the time dependence of interruptions and its potential relevance to planning and investment decisions. While the relevance of interruption timing is relevant, the timeconsuming nature and high cost of surveys to collect relevant data are also considered when assessing interruption costs. As a result, the production approach is often used because it does not account for time dependence. Interruption impact studies using the production function method have been conducted in the Netherlands (Nooij et al., 2007), Ireland (Leahy and Tol, 2010), Germany (Praktiknjo et al., 2011), and Spain (Linares and Rey, 2013). 3.2. Proxy indicator methods It is possible to estimate COUE from proxy indicators using data collected from business. These indicators relate to purchases of backup generation and other relevant measures. The key assumption is that customers would invest in backup generation to avoid interruptions at the point at which the value of generation equals the cost of unreliability. Disadvantages associated with this method include high data collection costs and the inability to disaggregate the data to a more detailed spatial resolution. 3.3. Blackout case studies Past blackout events can be used to estimate the cost of electricity interruptions by conducting after-the-fact case studies (Billinton, 2001). The advantage of this method is that estimations are based on real events and not hypothetical situations. However these findings are limited to the characteristic of the events studied, and these experiences can usually not be generalised to other events (Targosz and Manson, 2007). 3.4. Customer surveys This method uses survey questionnaires to collect data from customers on the impact of unserved energy. It is based on the

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010

U.J. Minnaar et al. / Utilities Policy xxx (2017) 1e10

assumption that customers are in the best position to assess their particular costs. Survey questionnaires may either ask customers to estimate what happens in the event of hypothetical interruptions (stated preference), or alternatively use revealed preferences. According to Billinton (2001) various techniques exist for which the survey method data are used and the choice of technique also determines survey design:
Direct-cost measurement is typically used to survey commercial and industrial customers on the impact of interruptions on lost production, labour costs, damage to equipment, and so on.
Willingness-to-pay (WTP) is used to ask residential customers what they are willing to pay to avoid interruptions.
Willingness-to-accept (WTA) asks customers to report how much money they are willing to accept in the event of interruption scenarios.
Conjoint surveys ask customers to choose among varying electricity bills with various reliability conditions described.
Mitigation-cost surveys ask customers to report their actual expenditure on mitigating interruptions (e.g., alternative power supplies or other measures). The major advantage of customer surveys is that they enable more refined analysis and thus addresses the key weakness of the macro-economic method. In principle, these methods should provide relatively accurate COUE estimates. In addition, these methods can capture cost variations based on the time of day, day of week and time of year in which ab interruption occurs; collect direct feedback from customers; and enable specific customer populations to be targeted through stratified sampling techniques. At the same time, these methods have several disadvantages. First, responses are at risk to bias, especially where stated preference methods are used. Bias may have multiple sources. It is possible that customers may understate their true willingness to pay or overstate damage costs in an attempt to secure discounts on high tariffs (CEER, 2013). Second, the accuracy and repeatability of results are highly dependent on survey design and implementation. Customers who experience few interruptions have difficulty estimating the impact of interruptions, and therefore data received would be inherently unreliable. Finally, customer surveys are expensive and time-consuming to conduct (Dzobo and Gaunt, 2012), especially if scalability to detailed resolution is required. The time and cost requirements of surveys result in extended time periods between updates. For example, Norway has a long history of using surveys for determining interruption costs. These surveys have been updated once each decade with surveys conducted in 1990e1991, 2000-2001, and subsequently (CEER, 2013). Similarly, in Italy, the only survey conducted was in 2003 and in the Netherlands, surveys were conducted in 2004 and the results were updated to reflect economic changes without conducting a survey in 2009 [5]. 3.5. Interruption costs for residential electricity customers The method applied by Growitsch et al. (2013) to determine the COUE to households considers substitutability between electricitybased and non-electricity-based leisure activities, and therefore they assume that power outages reduce the amount of welfare that households gain from leisure activities by 50% (factor of substitutability). To determine the amount of time households dedicate to leisure activities, they use labour market data and available information regarding the shares of time devoted by households to different activities. Computing the annual amount of leisure time across all households and multiplying by the factor of substitutability yields the time spent for electricity-based leisure activity.

3

Economic value is then assigned to leisure as per the work of Becker (1965); that is, the value of an hour of leisure is equal to the income from an hour of work. The leisure time of unemployed people, children, and pensioners, sick or disabled persons are assumed to be half the hourly income. Once the economic value of leisure is obtained, the COUE of the residential sector can be calculated as the ratio of this value and electricity consumption for the specified duration. Similarly, for households in Ireland, Leahy and Tol (2010) define the value of lost load as the value of time spent on non-paid work (leisure time) divided by electricity used over that time. They assume that all activity stops when there is no electricity. The opportunity cost of leisure time is assumed to be equal to the average wage, here assumed to equal the average GDP per working hour, after tax. For those who do not work, the opportunity costs are approximated by half the average wage. Linares and Rey (2013) assume that electricity is essential for some leisure activities. Consequently, in the absence of electricity, leisure time is lost. The Spanish National Statistics Institute (INE) provides information on the distribution of activities in an average day (in their Time use survey 2009-2010). This information is used to estimate the amount of time spent on leisure. They assume that domestic activities, computer activities, watching TV, listening to the radio, and social activities require electricity, and therefore the time employed in these activities is lost when there is an electricity interruption. Again, the substitutability of these activities is not taken into account. To monetize leisure time, they assume that the value of 1 h of leisure time equals the income per hour, that is, the net hourly wage as per Becker (1965), who states that a marginal hour of leisure time equals the income per hour. They assume that opportunity cost of leisure for inactive and unemployed people is lower and therefore 1 h of leisure time is equal to half of the average wage. They also acknowledge that the COUE for households might be overestimated using this method, but they do not quantify other costs such as food spoilage and personal damages, which would increase costs. 3.6. Interruption costs in South Africa In the South African context, limited work has been published with respect to estimating interruption costs. Herman and Gaunt (2008) as well as Dzobo et al. (2012) conducted customer surveys for residential and commercial customers sectors. The studies used questionnaires to measure Customer Interruption Cost (CIC) using direct and indirect measurement approaches. The direct approach uses specific questions relating to the most recent power failure that occurred during the past year, and the indirect approach uses hypothetical scenarios to measure customer's willingness to pay (WTP) to avoid interruptions. The motivation for their study was to develop appropriate methods to determine the CIC for the various sectors in South Africa. According to Herman and Gaunt (2008) the results of their surveys showed that the WTP is significantly lower than the direct cost of an interruption. The difference might represent customers inherently being willing to share the risk of interruptions, but whatever the reason, it is clear that the results of direct and indirect methods are not the same and need to be applied accordingly. In addition, the time of occurrence significantly affects the CIC. The difference according to time of day in the commercial sector can be extreme, with interruptions outside normal operating hours incurring virtually no cost, and there is significant variation according to the day of the month. CIC in the residential sector did not vary to the same extent with the time of occurrence but is clearly evident. Finally, the time of occurrence of outages in systems with chronic interruptions tends to correspond with the system demand

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010

4

U.J. Minnaar et al. / Utilities Policy xxx (2017) 1e10

because load shedding and network overloading occur when the demand is high. This means that time-averaged values of CIC cannot be used to determine the costs of system interruptions, as implied by many approaches to optimizing system reliability. Cameron and Van der Merwe (2014) focus on power distributor energy infrastructure investments in network infrastructure. They highlight the fact that a network performance strategy that is justified by achieving network service level improvements (measured in terms of System Average Interruption Duration Index (SAIDI) while ignoring the South African Grid Code compliance requirements can be extremely expensive to the nation's economy. They also demonstrate that decision information that depends on COUE is very sensitive to the specific COUE rates applied. In this context, a decision regarding capital expenditure that is justified in terms of the Grid Code is extremely sensitive to COUE assumptions. Similarly, the estimated network performance, and the performance levels considered for target setting, is also very sensitive to the COUE. Cameron and Van der Merwe (2014) empirical approach demonstrates the need for a nationally accepted, transparently determined, and “standardised” COUE values that can be applied by all stakeholders to inform key infrastructure and related investment decisions. They state that failure to have this information will most likely lead to suboptimal capital and other resource allocation decisions, which would be especially costly to South Africa's economy in its currently capital constrained environment. 3.7. Method applied We selected the production-function method to estimate COUE to meet Eskom's regulatory requirements. The COUE is primarily an input into national planning and investment decisions with regard to electricity production, and the government, regulators, and utilities need to be in a position to assess the macro-economic electricity service interruptions and electricity infrastructure investment. Moreover, since investment decisions to serve the public interest are made on an on-going basis, the production-function method has the benefits of taking a macro-economic viewpoint, being repeatable (annually), independently verifiable, and unbiased. This contrasts with customer surveys which have associated high costs as well results which cannot be benchmarked or verified. The Residential COUE is calculated by taking the household expenditure that depends on electrical energy use, and dividing it by the domestic use of electricity. The production-function approach has been accepted by the National Energy Regulator of South Africa (NERSA) and its results are applied by Eskom, the national electricity utility. 4. Production-function methodology and results The production-function method used in South Africa estimates an Economic COUE for economic sectors that use electricity for production purposes and a Residential COUE for households that use electricity for various household applications. The Economic COUE measures the value (in South African Rand GVA per kWh) placed on the unit of electricity not supplied due to an unplanned outage. Gross Value Added (GVA) is used as an indicator of economic activity. Annual GVA by economic sectors is officially measured and reported annually by Statistics South Africa. The Economic COUE is expressed both as direct and total impacts on the economy. Thus, the direct cost of power outages to the economy is measured in terms of production opportunity forgone,

as GVA/kWh per economic sector. The direct Economic COUE is disaggregated to 62 International Standard Industrial Code (ISIC) sectors (UN, ref) and 257 District and Local Municipalities. The indirect cost of power outages to the economy is measured as the indirect impact on the economy as a consequence of the changes in sales and expenditure in the whole economy resulting from direct costs. These indirect costs to the economy (the costs associated with complex cross-linkages in the economy), is also measured in terms of GVA. The Residential COUE is measured as the portion of South African household annual expenditures on goods and services that are electricity dependent, expressed as a ratio of these expenditures to annual residential electricity consumption. Residential lifestyles are increasingly electricity dependent for activities such as communications, personal care, security, education, household income generation, and leisure. Power outages result in an opportunity cost of foregoing these activities as well as disruption, discomfort, and nuisance. 4.1. Sectoral GVA analysis of electricity use in the economy The 1993 System of National Accounts (SNA) was adopted by the United Nations Statistical Commission and requires countries to compile annual SU-tables. Accordingly, the annual estimates of GVA and its components, as well as output, intermediate consumption expenditures, final consumption expenditures, and GDP estimates all have their origin in the annual SU-tables. Stats SA, the national statistics service in South Africa, uses the SU-tables framework to derive nominal estimates of GVA and GDP on a detailed, 62-sector, industry and commodity level. The supply table shows the resource origins of goods and services, and the use table shows the uses of these goods and services and the cost structure of the various industrial sectors. As a result, SU-tables report both the GVA generated and the electricity used by 62 different industries over a 12-month calendar year accounting period. Electricity is a necessary production input for each of the industries to generate its GVA. The tables are reported in monetary terms and therefore the cost of electricity, as an input to production, is known, for each industrial sector, and how much gross value to the economy is added by each sector. A summary of the GVA and electricity purchases by economic sectors for South Africa in 2013 is tabulated in Appendix 1. 4.2. Sectoral estimation of electricity use in economic and residential sectors No consolidated database of actual kWh sales by ISIC industrial sector exists at either national or a disaggregated municipal level in South Africa. This is a consequence of an increasingly complex electricity generation, transmission, and distribution system involving Eskom as well as District Municipalities (DM) and Local Municipalities (LM). Although Eskom sales data are available at an aggregated, national level, DM and LM sales data are not currently available. The only official data sources available are the electricity purchases by industries and the residential sector, as reported by the SU-tables. Estimates are required for deriving a national aggregate kWh consumption estimate. Deriving sectoral kWh electricity use requires the estimation of weighted average electricity tariffs by ISIC industrial sector. The national aggregate sales to economic production industries and residential users are estimated from Eskom sales data, shown

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010

U.J. Minnaar et al. / Utilities Policy xxx (2017) 1e10 Table 1 Summary of Eskom national electricity sales data for the 2013/14 financial year. Eskom customer categories

2013/14 No. of customers

Sales (GWh)

Net Revenue earned (R/m)

Average tariff (c/kWh)

Redistributors Residential Commercial Industrial Mining Rural/Agricultural Traction International Total

795 4,874,004 50,399 2789 1062 83,877 509 11 5,013,446

91,386 10,390 9519 51,675 31,611 5193 2996 13,791 216,561

49,891 9044 6972 23,543 17,620 5180 2057 5892 120,199

54.59 87.05 73.24 45.56 55.74 99.75 68.66 42.72 55.50

2013/14 202,770 146,687 56,083

in Table 1, and we assume that 50% of redistributor sales are to economic users. The remainder of sales are to residential users as shown in Table 2. The disaggregated sales to economic production industries are estimated from the SU-tables by using the monetary value of electricity purchases for 62 industrial sectors and transforming this to a kWh through a weighted average pricing assumption. This data is tabulated in Appendix 2. 4.3. Direct economic COUE at the national and disaggregated levels Direct Economic COUE (expressed in GVA Rands per kWh) is calculated cumulatively and for various industrial sectors by dividing the GVA (Rands) by the estimated electricity use (kWh). These COUE values represent the direct impact that a loss of 1 kWh in each industry would have on the economy. Direct COUE ¼ GVA (R million) / Electricity consumed (kWh)

national weighted average is R21.63. This number can be interpreted to reflect the weighted average direct economic production lost that can be expected, in an average year, as a result of manifold, short-duration, unplanned power outages across the country. The COUE varies considerably among sectors depending on industry energy intensity. 4.4. Total economic COUE estimate

Table 2 Electricity use by economic and residential sectors. The estimate assumes that 50% of Redistributor sales are to economic users and the remainder to residential users.

Eskom Domestic Sales (GWh) Electricity Used by Economic sectors (GWh) Electricity Used by Residential sector (GWh)

5

(1)

Next, a weighted average of these direct impacts is used to calculate the national Economic COUE. This weighted average was calculated by weighting the COUE of all industries to the amount of electricity used in each industrial sector. Table 3 and Appendix 3 summarises the aggregated and disaggregated Direct COUE estimates for South Africa for 2013. The

The total effect of a 1 kWh interruption on the economy is estimated through an InputeOutput (I-O) model. This is a quantitative economic technique that represents the linkages and interdependencies between different sectors of a national economy. As an analytical tool, the supply-and-use (SU) tables are integrated into macroeconomic models in order to analyze the link and interaction between final demand and industrial output levels. In this way, Statistics South Africa's SU-tables provide the foundation for development of an I-O model and for analysis of the Total COUE effect. The COUE model adopts the methodology proposed by Bouwer (2002) for the construction of an I-O model from the SUtables, and for analysis of results. The tables further report on all the other inputs from all the other sectors that are used for a specific industry to achieve this GVA. These are called intermediate inputs and explain the interindustry relationships that exist in the economy. Consequently, the COUE model uses the SU-tables to construct a single I-O table to establish the linkages and interrelationships between industries, products and other economic variables. This IO model enables the estimation of both direct and total effects of 1 kWh of electricity on the economy. The I-O enables the derivation of the Leontief inverse matrix which reflects not only the direct effects on the production process, but also incorporates the indirect effects on the production process, resulting from a change in demand for a specific product (Leontief, 1966). The Leontief inverse thus measures all the linkage effects and interrelationships between industries and final consumers and thus also the total impact on the economy. The Leontief inverse matrix enables the assessment of a scenario where 1 kWh is forgone/or gained in the economy. This is implemented by changing the final demand for products that use electricity as inputs by the equivalent of 1 kWh, which is equivalent to a direct effect of R21.63/kWh. As a result, due to the interrelationships between industries in the economy, the change in final demand multiplies throughout the economic system, and changes not only the inputs and outputs of a specific industry concerned, but also the inputs and outputs of other industries in the economy. The resultant total effect for 2013 is a weighted average R77.30/kWh.

Table 3 Economic COUE for South Africa at aggregated national economy level. Industrial sector

GVA Current Prices (R millions)

% of National GVA

Total economic electricity use (GWh)

Direct Effect (R GVA/kWh)

Agriculture Mining Manufacturing Electricity and water supply Construction Trade Transport and communication Finance Community services General Government Total Economy

73,310 284,363 409,509 117,466 105,124 378,674 289,975 614,901 631,566 268,073 3,172,961

2.3% 9.0% 12.9% 3.7% 3.3% 11.9% 9.1% 19.4% 19.9% 8.4% 100.0%

5248 31,611 70,672 16,001 535 3739 3790 6592 4269 4229 146,687

13.97 9.00 5.79 7.34 196.47 101.26 76.52 93.28 147.94 63.39 21.63

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010

6

U.J. Minnaar et al. / Utilities Policy xxx (2017) 1e10

5. Conclusions and policy implications

Table 4 Direct and Total Economic COUE effects for 2013. Industrial sector

Direct Effect (R GVA/kWh)

Total Effect (R GVA/kWh)

Agriculture Mining Manufacturing Electricity and water supply Construction Trade Transport and communication Finance Community services General Government Total Economy

13.97 9.00 5.79 7.34 196.47 101.26 76.52 93.28 147.94 63.39 21.63

51.06 36.43 57.29 27.60 376.83 124.98 307.39 341.79 290.46 73.93 77.30

The direct and total economic COUE for the various economic sectors is shown in Table 4. The results for the national economic Direct COUE and Total Economic COUE are shown in Table 4.

4.5. Residential COUE at a national level Whereas the estimated Economic COUE provides a proxy measure for the dependence of economic productivity on electricity, the Residential COUE provides a proxy measure for the utility of using electricity in the household. Consequently, shortduration power outages result in disruption, discomfort, or nuisance. In well-planned power systems, for which the COUE measure is relevant, it is extremely unlikely that household behaviours and expenditures will change as a result of household anticipation of possible power outages. Economic COUE and Residential COUE are not additive because they measure different impacts (production loss vs. discomfort) and because they have different units (GVA vs. household expenditure). The discomfort caused by power outages can be approximated using annual household expenditures on goods and services that require electricity for its use. StatsSA conducts extensive household income and expenditure surveys every five years. The most recent version is for 2010/11 and contains approximately 870 household expenditure categories. The Residential COUE model was constructed based on linking these categories to electricity usage. It was found that 153 of those categories relate to electricity. Together, these expenditures comprise 14.3% of income earned from “compensation of employees” as reported in the SUtables. Power outages would make these expenditures temporarily wasteful and disruption, discomfort, or nuisance. Residential COUE is estimated by dividing electricity-dependent household expenditure (in Rand) by residential electricity use. We estimate expenditure per kWh at R4.12/kWh, as shown in Table 5.

Table 5 Residential COUE estimate.

We developed a method to calculate an Economic and Residential COUE. The Economic COUE is based on gross value added (GVA) produced in the economy and the electrical energy consumed to produce value added. The direct Economic COUE is also disaggregated by industrial sector as defined by the ISIC system. The direct COUE effect for 2013 is calculated as R21.63/ kWh while the total Economic COUE for 2013 is calculated as R77.30 GV A/kWh. The Residential COUE for 2013 is calculated as R4.12 household expenditure/kWh.The Residential COUE assumes that households receive utility from electrical energy, as measured by expenditure on items that are electricity dependent. This utility is lost during an outage event. Economic production is not affected, as in the case of the Economic COUE but causes disruption, discomfort, and nuisance. The Residential COUE is calculated by taking the household expenditure items that depend on electrical energy use and dividing it by the domestic use of electricity. The development of a COUE methodology based on macroeconomic principles and data provides a basis for investment decision-making in South Africa's power system which is consistent with the macro-economic considerations used in national planning decisions. It provides a common basis for assessing impacts of planning and investment decisions. The availability of these results removes the uncertainty that existed within the South African electricity supply industry concerning appropriate and commonly accepted valuations of electricity service interruptions to the economy. The COUE methodology and the results it yields meet the need, as identified by Cameron and Van der Merwe (2014), for a nationally accepted, transparently determined and standardised set of COUE rates to inform key infrastructure and related investment decisions. This provides the basis for optimal capital and resource allocation, which is important to the economy in South Africa's capital-constrained environment. This approach also meets the requirements of the Transmission Grid Code and Distribution Code of South Africa. It has been accepted by the National Energy Regulator of South Africa (NERSA), and work is in progress to adopt it as the common methodology to be applied by all electricity utilities, both national and municipal. The implication is that all investment decisions based on economic impact will have a common and comparable basis across utilities. We recognise that differences exist among the impacts of electricity service interruption at local, regional, and national levels and appropriate application of the COUE within these contexts is an area that is open for further research. While this work was developed in response to a need for a regulator-approved method to determine COUE values for regulatory and investment purposes, it can also be used to inform decision-making about load shedding. This is particularly pertinent in light of the generation shortage that South Africa has experienced in recent years. However, the application of COUE for planned interruption also requires further research as the production method does not account for the impact of planned interruption with customer forewarning.

COUE: Household Effect

Residential

Total household income (compensation of employees) Portion of household expenditure on electricity use Electricity-dependent household leisure and convenience expenditure (R millions) Total residential electricity use (GWh) Residential COUE (R HH Expenditure/kWh)

1,610,647 14.3% 230,823

Acknowledgements

56,083 4.12

The authors would like to thank Eskom and the National Energy Regulator of South Africa for their inputs and support in preparing the model and results.

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010

U.J. Minnaar et al. / Utilities Policy xxx (2017) 1e10

7

Appendix 1. Summary of GVA and Electricity purchases for 2013 as reported by the Statistics South Africa SU-tables

Industrial sector

GVA Current Prices (StatsSA)

Electricity Purchases Current Prices (StatsSA)

Units

million R

million R

Agriculture Forestry Fishing Mining of coal and lignite Mining of gold and uranium ore Mining of metal ores Other mining and quarrying Food Beverages and tobacco Spinning, weaving and finishing of textiles Knitted, crouched fabrics, wearing apparel, fur articles Tanning and dressing of leather Footwear Sawmilling, planing of wood, cork, straw Paper Publishing, printing, recorded media Coke oven, petroleum refineries Nuclear fuel, basic chemicals Other chemical products, man-made fibres Rubber Plastic Glass Non-metallic minerals Basic iron and steel, casting of metals Basic precious and non-ferrous metals Fabricated metal products Machinery and equipment Electrical machinery and apparatus Radio, television, communication equipment and apparatus Medical, precision, optical instruments, watches, and clocks Motor vehicles, trailers, parts Other transport equipment Furniture Manufacturing n.e.c, recycling Electricity, gas, steam and hot water supply Collection, purification, and distribution of water Construction Wholesale trade, commission trade Retail trade Sale, maintenance, repair of motor vehicles Hotels and restaurants Land transport, transport via pipe lines Water transport Air transport Auxiliary transport Post and telecommunication Financial intermediation Insurance and pension funding Activities to financial intermediation Real estate activities Renting of machinery and equipment Computer and related activities Research and experimental development Other business activities Government Education Health and social work Sewerage and refuse disposal Activities of membership organisations Recreational, cultural and sporting activities Other activities Non-observed, informal, non-profit, households, Total (economic activity only)

62,678 7319 3313 63,998 52,676 129,350 38,339 72,278 28,816 4959 5194 1628 1476 17,018 19,713 13,417 35,522 16,488 24,697 5360 11,628 3302 13,175 14,634 10,715 22,859 26,349 6648 3577 2586 22,611 4619 4187 16,053 92,398 25,068 105,124 155,599 122,125 72,517 28,433 175,878 1384 12,582 34,344 65,787 151,585 75,196 72,109 184,224 6923 6962 7141 110,761 532,122 35,129 63,452 864 1859 8030 959 257,225 3,172,961

3545 369 42 1943 11,102 11,506 1570 2936 322 576 204 30 53 491 1518 228 837 8868 4201 237 329 490 816 7954 2723 938 546 305 41 61 1180 163 163 302 8738 276 569 1244 1818 728 936 2664 5 32 278 454 502 66 27 7735 129 296 57 4913 1466 817 1328 2 40 329 66 1296 103,398

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010

8

U.J. Minnaar et al. / Utilities Policy xxx (2017) 1e10

Appendix 2. Estimation of weighted average tariffs for the economic production industries

Sector

Basic Tariff Assumptions (c/kWh)

Assumptions

Weighted tariffs (c/kWh)

Agriculture Forestry Fishing Mining of coal and lignite Mining of gold and uranium ore Mining of metal ores Other mining and quarrying Food Beverages and tobacco Spinning, weaving and finishing of textiles Knitted, crouched fabrics, wearing apparel, fur articles Tanning and dressing of leather Footwear Sawmilling, planing of wood, cork, straw Paper Publishing, printing, recorded media Coke oven, petroleum refineries Nuclear fuel, basic chemicals Other chemical products, man-made fibres Rubber Plastic Glass Non-metallic minerals Basic iron and steel, casting of metals Basic precious and non-ferrous metals Fabricated metal products Machinery and equipment Electrical machinery and apparatus Radio, television, communication equipment and apparatus Medical, precision, optical instruments, watches, and clocks Motor vehicles, trailers, parts Other transport equipment Furniture Manufacturing n.e.c, recycling Electricity, gas, steam and hot water supply Collection, purification, and distribution of water Construction Wholesale trade, commission trade Retail trade Sale, maintenance, repair of motor vehicles Hotels and restaurants Land transport, transport via pipe lines Water transport Air transport Auxiliary transport Post and telecommunication Financial intermediation Insurance and pension funding Activities to financial intermediation Real estate activities Renting of machinery and equipment Computer and related activities Research and experimental development Other business activities Government Education Health and social work Sewerage and refuse disposal Activities of membership organisations Recreational, cultural and sporting activities Other activities Non-observed, informal, non-profit, households,

100 100 73 56 56 56 56 103 103 103 103

100% supply by Eskom at Rural/Agriculture Tariff 100% supply by Eskom at Rural/Agriculture Tariff 100% supply by Eskom at Commercial tariff >70% supply by Eskom at Mining tariff >70% supply by Eskom at Mining tariff >70% supply by Eskom at Mining tariff >70% supply by Eskom at Mining tariff 100% supply by Municipalities at Eskom's Commercial 100% supply by Municipalities at Eskom's Commercial 100% supply by Municipalities at Eskom's Commercial 100% supply by Municipalities at Eskom's Commercial

75 75 76 83 83 83 83 106 106 106 106

103 103 64 64 46 46 46 46 46 46 46 46 46 46 46 103 64 64

100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%

46

100% supply by Eskom at Industrial tariff

47

46 46 46 46 55 55 103 122 122 122 122 81 81 122 122 122 122 122 122 122 122 122 122 122 122 122 122 122 122 122 122 122

100% supply by Eskom at Industrial tariff 100% supply by Eskom at Industrial tariff 100% supply by Eskom at Industrial tariff 100% supply by Eskom at Industrial tariff Redistributor price Redistributor price 100% supply by Municipalities at Eskom's Commercial tariff þ Markup 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 77% suply be Eskom at Traction tariffs þ 23% supply at Municipal tariffs 77% suply be Eskom at Traction tariffs þ 23% supply at Municipal tariffs 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up 100% supply by municipalities at Residential tariff þ mark-up

47 47 47 47 56 56 106 126 126 126 126 84 84 126 126 126 126 126 126 126 126 126 126 126 126 126 126 126 126 126 126 126

supply supply supply supply supply supply supply supply supply supply supply supply supply supply supply supply supply supply

by by by by by by by by by by by by by by by by by by

Municipalities at Eskom's Municipalities at Eskom's Municipalities at Eskom's Municipalities at Eskom's Eskom at Industrial tariff Eskom at Industrial tariff Eskom at Industrial tariff Eskom at Industrial tariff Eskom at Industrial tariff Eskom at Industrial tariff Eskom at Industrial tariff Eskom at Industrial tariff Eskom at Industrial tariff Eskom at Industrial tariff Eskom at Industrial tariff Municipalities at Eskom's Municipalities at Eskom's Municipalities at Eskom's

tariff tariff tariff tariff

Commercial tariff Commercial tariff Industrial tariff þ Industrial tariff þ

þ þ þ þ

Markup Markup Markup Markup

þ Markup þ Markup Markup Markup

Commercial tariff þ Markup Industrial tariff þ Markup Industrial tariff þ Markup

106 106 66 66 47 47 47 47 47 47 47 47 47 47 47 106 66 66

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010

U.J. Minnaar et al. / Utilities Policy xxx (2017) 1e10

9

Appendix 3. Disaggregated COUE by 62 industries for 2013

Sector

GVA (StatsSA)

Electricity Consumed

Units

million R

GWh

R GVA/kWh

Agriculture Forestry Fishing Mining of coal and lignite Mining of gold and uranium ore Mining of metal ores Other mining and quarrying Food Beverages and tobacco Spinning, weaving, and finishing of textiles Knitted, crouched fabrics, wearing apparel, fur articles Tanning and dressing of leather Footwear Sawmilling, planing of wood, cork, straw Paper Publishing, printing, recorded media Coke oven, petroleum refineries Nuclear fuel, basic chemicals Other chemical products, man-made fibres Rubber Plastic Glass Non-metallic minerals Basic iron and steel, casting of metals Basic precious and non-ferrous metals Fabricated metal products Machinery and equipment Electrical machinery and apparatus Radio, television, communication equipment and apparatus Medical, precision, optical instruments, watches, and clocks Motor vehicles, trailers, parts Other transport equipment Furniture Manufacturing n.e.c, recycling Electricity, gas, steam and hot water supply Collection, purification, and distribution of water Construction Wholesale trade, commission trade Retail trade Sale, maintenance, repair of motor vehicles Hotels and restaurants Land transport, transport via pipe lines Water transport Air transport Auxiliary transport Post and telecommunication Financial intermediation Insurance and pension funding Activities to financial intermediation Real estate activities Renting of machinery and equipment Computer and related activities Research and experimental development Other business activities Government Education Health and social work Sewerage and refuse disposal Activities of membership organisations Recreational, cultural and sporting activities Other activities Non-observed, informal, non-profit, households, Total

62,678 7319 3313 63,998 52,676 129,350 38,339 72,278 28,816 4959 5194 1628 1476 17,018 19,713 13,417 35,522 16,488 24,697 5360 11,628 3302 13,175 14,634 10,715 22,859 26,349 6648 3577 2586 22,611 4619 4187 16,053 92,398 25,068 105,124 155,599 122,125 72,517 28,433 175,878 1384 12,582 34,344 65,787 151,585 75,196 72,109 184,224 6923 6962 7141 110,761 532,122 35,129 63,452 864 1859 8030 959 257,225 3,172,961

4703 490 55 2351 13,435 13,924 1900 2761 303 542 192 28 50 742 2295 484 1780 18,863 8935 504 699 1041 1736 16,918 5792 1995 513 462 62 129 2509 346 347 643 15,511 490 535 984 1439 576 740 3180 6 26 220 359 397 52 21 6121 102 234 45 3888 1160 647 1051 2 32 260 53 1025 146,687

13.33 14.95 59.70 27.22 3.92 9.29 20.17 26.18 95.11 9.16 27.04 58.66 29.56 22.94 8.59 27.70 19.96 0.87 2.76 10.64 16.63 3.17 7.59 0.87 1.85 11.46 51.34 14.40 57.30 20.01 9.01 13.34 12.05 24.98 5.96 51.11 196.47 158.07 84.87 125.96 38.40 55.32 239.19 493.40 156.31 183.18 381.38 1433.97 3376.53 30.10 67.59 29.75 158.69 28.49 458.80 54.32 60.38 569.64 58.44 30.84 18.25 250.92 21.63

References Becker, G., 1965. A theory of the allocation of time. Econ. J. 75, 493e517. Bhavaraju, M., 2004. A New PJM Capacity Construct Reliability Pricing Model (RPM). September 2004. http://www.misostates.org/files/WorkGroups/RPM-Presentation_ to_MISO_9-13-04.pdf. Last accessed September 2014.

COUE

Billinton, R., 2001. Methods to Consider Customer Interruption Costs in Power System Analysis. Conseil International de Grandes Rseaux Electriques (CIGRE), Paris, France. Billinton, R., Wangdee, W., 2005. Approximate methods for event-based customer interruption cost evaluation. IEEE Trans Power Syst. 20 (2), 1103e1110. Bouwer, G., 2002. The role of supply and use tables in South Africa as a tool for

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010

10

U.J. Minnaar et al. / Utilities Policy xxx (2017) 1e10

economic analysis. Proceedings of the 14th International Conference on InputOutput techniques. Cameron, M.J., Van der Merwe, J., 2014. Energy Infrastructure Investment Decisions in South Africa e the Role and Need for a Standardised Set of Cost of Unserved Energy (COUE) Rates Illustrated. Paper developed for the 2nd South Africa Economic Regulators Conference (SAERC), 18e19 March 2014. Carlsson, F., Martinsson, P., 2008. Does it matter when a power outage occurs?eA Choice experiment study on the willingness to pay to avoid power outages. Energy Econ. 30, 1232e1245. Council of European Energy Regulators, 2013. Guidelines of Good Practice on Estimation of Costs Due to Electricity Interruptions and Voltage Disturbances. C10-EQS-41e03. http://www.energy-regulators.eu (accessed 30.06.2013). Dzobo, O., Gaunt, C.T., 2012. Investigating the use of probability distribution functions in reliability-worth analysis of electric power systems. Int. J. Electr. Power & Energy Syst. 37, 110e116. Dzobo, O., Gaunt, C.T., Herman, R., 2012. Reliability worth assessment of electricity consumers: a South African case study. J. Energy South. Afr. 23, 31e39. Electricity Reliability Council of Texas, 2013. Estimating the Value of Lost Load. http://www.ercot.com/content/gridinfo/resource/2014/mktanalysis/ERCOT_ ValueofLostLoad_LiteratureReviewandMacroeconomic.pdf. Last accessed June 2017. Energy Policy and Planning Office (Ministry of Energy). Customer Outage Cost Study, 2001. (http://www.eppo.go.th/power/ERI-study-E/ERI-ExeSummary-E. html), Last accessed Sept 2014. Growitsch, C., Malischek, R., Nick, S., Wetzel, H., 2013. The Costs of Power Interruptions in Germany e An Assessment in the Light of the Energiewende. Institute of Energy Economics at the University of Cologne (EWI), Cologne. Working Paper 13/07. Herman, R., Gaunt, C.T., 2008. Direct and Indirect Measurement of Residential and Commercial CIC: Preliminary Findings from South African Surveys. Probabilistic Methods Applied to Power Systems, 2008. Herman, R., Gaunt, C.T., 2010. Probabilistic Interpretation of Customer Interruption Cost Applied to South African Systems. Probabilistic Methods Applied to Power Systems, 2010.

Hicklin, R., 2010. Value of Customer Reliability Background. www.aemo.com.au/ planning/0400-0017.pdf. Last Accessed September 2014. Leahy, E., Tol, R., 2010. An Estimate of the Value of Lost Load for Ireland. ESRI working paper no 357, October 2010. Leontief, W., 1966. Input-output Economics. Oxford University Press, New York. Linares, P., Rey, L., 2013. The costs of electricity interruptions in Spain. Are we sending the right signals? Energy Policy 61, 751e760. London Economics, 2013. The Value of Lost Load (VoLL) for Electricity in Great Britain. Marais R., Zivanovic R., Estment R., 2006. Developing a model to simulate least economic cost of transmission expansion investment. Proceedings of the 9th International Conference on Probabilistic Methods Applied to Power Systems, Stockholm, Sweden. Newberry, D., Eberhard, A., 2008. South African Network Infrastructure Review: Electricity. http://www.gsb.uct.ac.za/files/SAElectricityPaper08.pdf. Last accessed 04 April 2016. Nooij, M., Koopmans, C., Bijvoet, C., 2007. The value of supply security: the costs of power interruptions: economic input for damage reduction and investment in networks. Energy Econ. 29, 277e295. Praktiknjo, A.J., Hahnel, A., Erdmann, G., 2011. Assessing energy supply security: outage costs in private households. Energy Policy 39, 7825e7833. Targosz R., Manson J., 2007. Pan European LPQI Power Quality Survey. Proceedings of the19th International Conference on Electricity Distribution. Paper 0263. TERI, 2001. Cost of Unserved Energy. Tata Energy Research Institute, New Delhi, 278pp. [TERI Project Report No. 98PG42]. United Nations, International Standard Industrial Classifications Rev. 4, http:// unstats.un.org/2C3D8FDA-B465-4488-967C-D522DCA701AE/FinalDownload/ DownloadId-620716733B76DBFF5234CF1419F891F5/2C3D8FDA-B465-4488967C-D522DCA701AE/unsd/publication/seriesM/seriesm_4rev4e.pdf (accessed 01.02.2016). United Nations, 1993. Inter-secretariat Working Group on National Accounts. 1993. System of National Accounts (Brussels: Eurostat, New York: UN, Paris: OECD, Washington, D. C.: IMF and World Bank).

Please cite this article in press as: Minnaar, U.J., et al., An economic model for the cost of electricity service interruption in South Africa, Utilities Policy (2017), http://dx.doi.org/10.1016/j.jup.2017.08.010