=== ires-2018-page-100.pdf === 87 Chapter VII Data collection and compilation 7.1. Energy data collection and compilation are difficult tasks, and country practices in this respect vary significantly. Countries should make efforts to learn from the experiences of others, share best practices and promote relevant standards and strategies that will improve the overall quality of energy data, including their completeness and international compar­ ability. To assist countries in these activities, this chapter discusses the role of legal frame- works and institutional arrangements in data collection, followed by a discussion of data collection strategies, data sources and data compilation methods. A. Legal framework 7.2. The existence of a strong legal framework is one of the most important prerequisites for establishing a sound national statistical system in general and a national system of energy statistics in particular. The legal framework is provided by statistical laws and other applica- ble national laws and regulations that, to different degrees, specify the rights and responsi- bilities of entities that collect data, provide data, produce statistics or use statistical outputs. For example, data obtained by conducting statistical surveys depend on statistical laws and energy-related legislation and regulations, while data on imports and exports of energy are subject to customs laws and regulations. 7.3. The establishment of a legal framework to make reporting of energy data mandatory through well-designed channels and instruments is of great importance to ensuring the com- pilation of high-quality energy statistics. Although many countries lack such a legal frame- work, it is important to recognize it as the preferred option. For such a framework, energy min- istries or energy agencies maintain administrative records relevant to energy statistics, while national statistical offices organize data collection from entities that produce energy products as a primary or secondary activity and from energy users. The legal framework should not only enable efficient data collection but deal adequately with confidentiality issues, providing necessary protection to data reporters (see chapter X for further discussion of confidentiality). 7.4. The legal framework should also describe responsibilities for collection, compilation and maintenance of the different data components among different government bodies, tak- ing into account the variety of public policy objectives, and the changes brought about by market liberalization that often increase the difficulty of obtaining data given the growing number of participants in energy industries and the commercial sensitivities around data disclosure in an ever more competitive market. 7.5. It is recommended that, whenever appropriate, national agencies responsible for the compilation and dissemination of energy statistics actively participate in the discussions on national statistical legislation or relevant administrative regulations in order to establish a solid foundation for high-quality and timely energy statistics, with a view to mandatory reporting, whenever appropriate, and adequate protection of confidentiality. Also, such par- ticipation would strengthen the agencies’ responsiveness to the data requirements and priori- ties of the user community.  === ires-2018-page-101.pdf === 88 International Recommendations for Energy Statistics (IRES) B. Institutional arrangements 7.6. The legal framework creates a necessary, but not sufficient, foundation for energy statistics. To ensure that these statistics are collected and compiled in the most effective way, establishing appropriate institutional arrangements among all relevant governmental agencies is of paramount importance. 7.7. Members of a national system of energy statistics. A national system of energy sta- tistics consists of various governmental agencies engaged in the collection, compilation and dissemination of energy statistics. The most important members of such a system are national statistical offices and specialized governmental agencies responsible for the implementation of energy policies (e.g., energy ministries). However, the complex and vast nature of energy supply and use and the liberalization of the energy markets in many countries have resulted in an increasing number of governmental agencies and other organizations collecting data and maintaining databases on energy, such as chambers of commerce, industry associations, regional offices, etc. This represents, on the one hand, a great potential for reducing the response burden and improving timeliness of the data, but, on the other, poses great chal- lenges to ensuring the harmonization of data as the underlying concepts, definitions, methods and quality assurance applied by different agencies might vary significantly. 7.8. Purposes of institutional arrangements. To function efficiently, a national system of energy statistics should be based on appropriate institutional arrangements among many relevant agencies. Such arrangements should allow for the collection, compilation, stand- ardization and integration of information scattered among different entities and for the dis- semination of the compiled statistics to users through a coherently networked information system or a central energy database. The institutional arrangements should also promote harmonization with international standards and recommendations to enable the collection of high-quality and internationally comparable official energy statistics. Last but not least, effi- cient institutional arrangements will not only minimize the data collection cost for agencies involved by avoiding duplication of work and enabling the sharing of good practices, but will also result in a reduced response burden on data reporters due to improved communication and coordination among data collectors. 7.9. Governance of the national system of energy statistics. A key element of the insti- tutional arrangements is the establishment of a clear, efficient and sustainable system of gov- ernance of the national system of energy statistics. Depending on a country’s legislation and other national considerations, various agencies might lead the system and be responsible for official energy statistics. These can be national statistical offices, energy ministry or agency, or another specialized governmental agency. It is imperative that the lead agency ensures the necessary coordination of work, thus resulting in energy statistics that comply with the qual- ity standards as described in chapter X. 7.10. Mechanism for functioning of the system. In order to guarantee the successful func- tioning of a national system of energy statistics, it is vital that all stakeholders are actively involved. It is recommended that countries develop an appropriate interagency coordina- tion mechanism that, while taking into account existing legal constraints, would systemati- cally monitor performance of the national system of energy statistics, motivate its members to actively participate in the system, develop recommendations focused on improving the system’s functioning, and have the authority to implement such recommendations. Such a mechanism should address, among others, the issue of statistical capacity, as the lack of funding and human resources is a persistent problem in many countries. In this context, the proper allocation of responsibilities to agencies, as well as the convening of joint training courses and workshops on energy matters to further develop the skills and knowledge of the staff, can be of great help.  === ires-2018-page-102.pdf === Data collection and compilation 89 7.11. Models for the organization of a national system of energy statistics vary from a cen- tralized system, in which one institution is in charge of the whole statistical process (from the collection and compilation to the dissemination of statistics), to a decentralized system, in which several institutions are involved and are responsible for different parts of the process or different components of energy statistics. 7.12. It is recognized that different institutional arrangements (depending on the structure of a country’s government, legal framework and other national considerations) can result in high-quality energy statistics, provided that the overall national system follows internation- ally accepted methodological guidelines, utilizes all available statistical sources and applies appropriate data collection, compilation and dissemination procedures. Effective institutional arrangements are usually characterized by: (a) The designation of only one agency responsible for the dissemination of official energy statistics or, if this is not possible, the identification of agencies responsible for the dissemination of specific data subsets and mechanisms that will ensure overall consistency of energy statistics; (b) A clear definition of the rights and responsibilities of all agencies involved in data collection and compilation; (c) The establishment of formalized working arrangements between them, including agreements on holding inter-agency working meetings and on access to relevant microdata collected by those agencies. The formal arrangements should be com- plemented by informal agreements among the involved agencies and institutions as required. 7.13. It is recommended that countries consider the establishment of the institutional arrangements necessary to ensure the collection and compilation of high-quality energy sta- tistics as a matter of high priority and periodically review their effectiveness. 7.14. Whatever the institutional arrangement, the national agency that has the overall responsibility for the compilation of energy statistics should periodically review the defini- tions, methods and the statistics themselves to ensure that they comply with relevant inter- national recommendations and recognized best practices, are of high quality, and are avail- able to users in a timely fashion. If such an agency is not designated, then an appropriate mechanism should be put in place to ensure that those functions are performed consistently and effectively. C. Data collection strategies 7.15. The collection of energy data can be a complex and costly process, which very much depends on a country’s needs and circumstances, including the legal framework and insti- tutional arrangements. Therefore, it is important that countries undertake it on the basis of well-thought-out strategic decisions regarding the scope and coverage of data collection, organization of the data collection process, selection of the appropriate data sources and use of reliable data collection methods. 1. Scope and coverage of data collection 7.16. The scope and coverage of the collection of energy statistics are defined according to: (a) Conceptual design, which includes the objective and thematic coverage; (b) Target population; (c) Geographical coverage;  === ires-2018-page-103.pdf === 90 International Recommendations for Energy Statistics (IRES) (d) Reference period of data collection; (e) Frequency with which data are to be collected; (f ) Point in time of collection. 7.17. Conceptual design. The overall objective of the data collection should be defined clearly. The thematic coverage must take into account the type of statistics to be collected, for example, the flows and stocks of energy products and the units of measurement. International standards should be applied in the conceptual design process. 7.18. Target population. Good knowledge of the main groups of data reporters is required for an efficient data collection, so that data collection methods can be customized as neces- sary. It is recommended, as applicable, that at least the following three reporter groups be distinguished: energy industries, other energy producers and energy consumers. 7.19. Energy industries (see chapter V for definition) are represented by various entities whose principal activity is directly related to energy production and which often concentrate on one particular fuel or one part of the overall energy supply chain. Detailed information is compiled by the energy industry entities themselves on a regular basis for management pur- poses, as well as for reporting to government regulatory bodies. Therefore, statistical data can often be obtained without too much delay from those entities directly or from administrative records maintained by regulatory bodies, when the proper data collection mechanisms exist. 7.20. The entities belonging to the energy industries can be differentiated, according to their ownership status, as private industries, public industries and public-private industries. The degree to which a central government is directly involved in the industries can have a significant effect on both the ease with which data may be collected, and the range of data that will be considered reasonable to collect. Given that such industries can provide data on most energy flows, they need to be treated with special attention and be fully enumerated in statistical surveys or covered using appropriate administrative sources (see section on data sources for details). When the number of energy industry entities is large and the energy statistics compiler has no direct contact with the original sources, it is common for industrial associations, regional offices or civil organizations to act as intermediate data collectors and reporters to simplify the data collection process. However, in such a case, efforts should be undertaken to ensure that data quality is not compromised. 7.21. Other energy producers include those economic units (including households) that produce energy for self-consumption; they may sometimes supply it to other consumers, but not as part of their principal activity (see chapter V for details). Since these activities are not the principal objective of these companies and since they may be partially or fully exempt from the provisions of energy legislation and regulations, it cannot be expected that they will have the same amount of detailed information readily. 7.22. Even though in most cases other energy producers account for only a small part of the national energy production, it is important that they are included in national energy statistics to properly account for their energy needs and measure their energy efficiency. In countries where other energy producers play a significant role in the national aggregate of energy supply and consumption, appropriate procedures have to be devised to obtain more adequate data from them. In some countries, the auto-production of electricity and heat (see chapter V for details) requires governmental authorization, which facilitates the monitoring of these companies and creates the means for obtaining the required data. 7.23. Energy consumers can be grouped according to the energy needs of the economic activity under which they are classified, such as industry, households, etc. (see chapter V for details). Data collection from energy consumers is complex since it has to take into account  === ires-2018-page-104.pdf === Data collection and compilation 91 their diversity, mobility and multipurpose forms. To facilitate this task, specific methodolo- gies and compilation strategies must be designed for the different subgroups of consumers, given their particularities. 7.24. It is usually the case that energy producers can provide data on how much energy in total is being delivered to energy consumers, and often may also be able to provide a break- down of total deliveries by the various consumer groups taking into account differences in applicable tariffs and/or taxes. However, in order to fill the remaining data gaps and obtain more detailed information (e.g., in the case of energy balance compilation), direct consumer surveys might be necessary. The coherence between data based on the information about energy deliveries to final consumers and the information reported by consumers must be ensured. In some other cases, for example solid biomass fuels, information will most likely be obtained through surveys and consumer-derived measurements, rather than from energy producers, thus avoiding potential producer-consumer data mismatch. 7.25. Collection of energy data from the informal sector. The informal sector has been defined by the Fifteenth International Conference of Labour Statisticians 59 according to the 59 Resolution concerning statistics types of production units of which it is composed. It consists of a subset of household unin- of employment in the informal corporated enterprises, with at least some production for sale or barter, operating within the sector, adopted by the Fifteenth International Conference of production boundary of the SNA. As household production units, these enterprises do not Labour Statisticians (January constitute separate legal entities independent of the households or household members that 1993). Available from www.ilo own them, and no complete sets of accounts (including balance sheets of assets and liabil­ .org/public/english/bureau/stat ities) are available that would permit the production activities of the enterprises to be clearly /download/res/infsec.pdf. distinguished from the other activities of their owners, nor any flows of income and capital between the enterprises and the owners to be identified. 7.26. An energy producing unit in the informal sector may be defined as a household enter- prise with at least some production of energy for sale or barter that meets one or more of the following criteria: limited size in terms of employment; non-registration of the enterprise; and non-registration of its employees. The informal sector thus defined excludes household enterprises producing energy exclusively for own final use. The area-based enterprise survey approach is commonly used to collect data from such enterprises, as a satisfactory list of such 60 enterprises is normally not available.60 For details on issues such as the identification of statistical units 7.27. Geographical coverage. The geographical coverage identifies the area for which the applicable in the case of the informal sector and the organi- statistics are collected. In general, for policy purposes it is fundamental to collect statistics zation of surveys of the informal at the national level. However, for analytical and policymaking purposes it is often the case sector, see the International that countries compile their energy statistics at a subnational level, which implies a more Recommendations for Industrial detailed geographical coverage. The collection of energy statistics at the subnational level is Statistics (UN 2009b) (chapter 2, often essential in planning future infrastructure, as it allows the different locations of produc- section F, and chapter 6). tion and consumption to be taken into account. Where consumption is concerned, regional disaggregation is necessary, since energy use could vary significantly according to climate, local behaviour, customs, economic activities, incomes, availability of energy products, etc. The collection of such detailed information often implies a higher data collection cost and requires additional methodological efforts to ensure that there are no omissions or double- counting in the results when collating regional data up to the national level. 7.28. Reference period of data collection. The reference period of the energy data collected refers to the time period that the data relate to. For example, oil production data may have a reference period of one month, energy use data collected from households may have a refer- ence period of one quarter and energy behaviour data (e.g., data on measures taken to reduce energy use) may have a reference period of one year.  === ires-2018-page-105.pdf === 92 International Recommendations for Energy Statistics (IRES) 7.29. Frequency of data collection. The frequency of collection of energy data adopted by a given country is the result of a balance between users’ needs, the priority given to timeliness of particular data items, the level of detail required, the availability of data and the available resources. Comprehensive annual data should be the initial objective when setting up an energy statistics programme. However, higher frequency (infra-annual) collections are critic­ ally important for the timely assessment of a fast-changing energy situation, and countries are encouraged to conduct them on a regular basis within the identified priority areas of energy statistics. The various frequencies of data collection are further outlined below. Annual data collections. These collect energy data relating to the basic and most appropriate information needs. In general, they cover production, supply and con- sumption at a detailed level of disaggregation for any energy products that make up a significant share of total energy supply. Infra annual data collections (quarterly, monthly, etc.). These are conducted when the need for frequent data is of high priority (e.g., monthly oil production and trade) but are usually more restricted in the level of detail (for example, total consumption rather than consumption by consumer groups) than those collections carried out annually, as higher frequency leads to increased costs and reporting burden. Infrequent data collections (less frequent than annual). These are generally con- ducted by countries, either for specialized topics (e.g. deployment of fuel cells), to fill in gaps in the data collected annually or infra-annually (e.g. more accurately ascertaining sub-sector breakdowns of minor products), to provide baseline information, or where data collection is particularly expensive (e.g., large consumer surveys or censuses). 7.30. Point in time of collection. The point in time when the collection is carried out should also be carefully considered, since this may have an impact on the response rate (e.g., avoid sending questionnaires during holiday periods, overlap with other surveys, as well as with administrative data collections such as tax filing). 2. Organization of data collection 7.31. Proper organization of the data collection process is fundamental for official energy statistics. The first important step in data collection is to identify the production, supply, transformation and consumption flows for each fuel in order to clarify the processes, pro- cedures and the statistical units involved. Then it is necessary to outline the potential data sources for each flow to determine whether it is feasible to obtain accurate information on a regular basis from them, making use of the information they already hold for their own management purposes. From these descriptions it can be determined what kind of energy data can be obtained from different sources, and the process can be planned accordingly. 7.32. In general, data collection relies on the legal framework and institutional arrange- ments of the country, as well as on the use of agreed collection methods, such as the use of statistical business registers, administrative data and census or sample surveys, to obtain comprehensive data. The most appropriate collection method should be selected, taking into consideration the nature and specific characteristics of the given energy activity, the avail- ability of the required data, and the budget constraints that might affect the implementation of the collection strategy. 7.33. An integrated approach to collection of energy statistics. The collection of energy data should be seen as an integral part of the data collection activities of the national statistical system in order to ensure the best possible data comparability and cost efficiency. In this con- text, close collaboration between energy statisticians and compilers of industrial statistics, as well as statisticians responsible for conducting household, labour force and financial surveys,  === ires-2018-page-106.pdf === Data collection and compilation 93 is of paramount importance and should be fully encouraged and systematically promoted. A collaborative relationship will create a better understanding of the information, provide an opportunity to incorporate energy items into non-energy specific questionnaires, taking into account the priorities and specific needs of the energy industries, and facilitate the conduct of a cost-benefit analysis. 7.34. The establishment or improvement of the regular programme of energy data collection should be part of a long-term strategic plan in the area of official statistics. Such a programme should be properly designed and executed in order to obtain the widest possible coverage and ensure the collection of accurate, detailed and timely energy statistics. 7.35. An integrated approach is especially important for the collection of data on energy consumption, as many different data sources can be used. Data can be obtained directly or indirectly from appropriate economic units (i.e., enterprises or establishments and house- holds) by means of censuses, surveys and/or administrative records. Given that the number of energy consumers is larger than energy suppliers, it may be necessary to exploit existing business surveys to identify those establishments that will be required to answer specific ques- tions on energy consumption. Consistency of data on energy consumption collected from various sources should be ensured. D. Data sources 7.36. The generation of energy statistics is based on data collected from two main sources: Statistical data sources that provide data collected exclusively for statistical purposes from censuses and/or sample surveys; and Administrative data sources that provide data created originally for purposes other than the production of statistical data. 1. Statistical data sources 7.37. The typical statistical data sources for compiling energy statistics are surveys of the units in the population under consideration. The surveys are done either by enumerating all the units in the population (census) or a subset of representative units scientifically selected from the population (sample survey). 7.38. In general, censuses represent a time-consuming, resource-intensive and costly exercise for the collection of energy statistics and imply a high overall response burden on the popula- tion. For these reasons, it is unlikely that censuses will be used very often. However, depend- ing on the population of interest, the available resources and the particular circumstances in a country, conducting a census may be a viable option for collecting energy statistics. A complete census of units in the energy industry may be appropriate when, for example, a particular country does not maintain an up-to-date business register, there are few energy producers (in such a case they should be included in a “take all” stratum of appropriate sur- veys), or there is significant user interest for detailed energy data. 7.39. Sample surveys are used to collect information from a portion of the total population, called a sample, in order to draw inferences on the whole population. They are almost always less costly than censuses. There are different types of surveys that can be used in energy statistics depending on the sampling units: (i) enterprise surveys, (ii) household surveys and (iii) mixed household-enterprise surveys. In general, it is recommended that countries make efforts to establish a programme of sample surveys that would satisfy the needs of energy statistics in an integrated way (i.e., as part of an overall national sample survey programme of enterprises and households) to avoid duplication of work and minimize the response burden.  === ires-2018-page-107.pdf === 94 International Recommendations for Energy Statistics (IRES) Survey design 7.40. Before carrying out a survey, it is fundamental to have a proper survey design. To achieve this goal, a number of steps are needed. First, the particular information needs should be identified and the specific goals of the project established, placing special emphasis on priorities, feasibility, budget and geographic breakdown. In order to do so, it is necessary to make use of the experience gained in similar projects in other areas of statistics, and take into account relevant international recommendations (e.g., those published in International Rec- ommendations for Industry Statistics 2008) and the provisions of relevant applicable national laws and regulations. This phase requires the expertise of professionals in the specific subject area being covered, such as specialists in energy matters, as well as specialists in sample design, interviewing techniques, analysis procedures, etc. Given the above, participation and cooperation among national statistical office, different ministries and academic institutions are crucial. 7.41. Ideally, energy surveys must be designed to ensure their regular conduct. For this reason, it is recommended that the periodicity of such surveys be established from the very beginning. Countries are encouraged to ensure that the survey design is optimized, keeping in mind the desirable use and inferences from the expected results, while information not essential for survey purposes should be avoided as much as possible. Considering the cost of conducting such surveys, the survey must be designed in such a way as to guarantee the greatest benefits from the analytical results and ensure their consistency over time. 7.42. Once the specific topic(s) of the survey are determined, the next stage is to select the data items using those presented in chapter VI as a reference list and ensuring that the selec- tion is done according to an appropriate classification and precise definition of each of the concepts used in the data items definitions. 7.43. Selecting the target population or sample is critical to successfully meeting the goals of the survey. Within this phase, the number of units to be interviewed must be decided in order to ensure representativeness, taking into account the time availability, budget constraints and necessary degree of precision. The sampling technique used will depend on the population or populations being sampled, as well as on the information available from other regular survey programmes and business registers that may provide a better picture and context of the project being considered. 7.44. The design of the questionnaires and supplementary documentation should follow. Deciding on the interviewer’s profile, the interviewing method best suited for the survey’s purpose (personal interviews, telephone surveys, mail surveys, computer-direct interviews, email surveys, Internet surveys and others), the temporal scope of the data items and the way each of them and related concepts will be presented and asked are essential for good questionnaire design. Determining the type of questions and their sequence comes next, paying special attention to using clear, direct and straightforward language. The use of proper measurement units in terms of which the answers should be provided is also significant and depends largely on who is being interviewed. For example, small units of measurement such as kilowatt-hour, cubic metre, etc., are perfectly proper for consumers or gasoline stations, but not for energy supply industries. 7.45. Another important part of the survey design is the preparation of concise and clear instructions to help clarify any questions that potential respondents might have. It is impor- tant to mention that the survey design should be adapted as required according to the specific context, geographical scope, informant, interviewer and planned procedures. The question- naires must be tested in a context similar to the one in which they will be applied prior to the finalisation of the required adjustments. For example, interviewers need to be carefully  === ires-2018-page-108.pdf === Data collection and compilation 95 trained in the techniques to be used for measuring different fuels. In some cases, especially for measuring biomass, the availability of measurement instruments (e.g., scales for fuelwood and charcoal) for the physical measurement of fuels actually consumed is extremely important and should be ensured where possible. Enterprise surveys 7.46. Enterprise surveys are surveys in which the sampling units comprise enterprises (or statistical units belonging to these enterprises such as establishments or kind-of-activity units) in their capacity as the reporting and observation units from/about which data are obtained. They assume the availability of a sampling frame of enterprises. Depending upon the source of the sampling frame, such surveys may also be classified as either list-based or area-based. In a list-based survey, the initial sample is selected from a pre-existing list of enterprises or households. In an area-based survey, the initial sampling units are a set of geographical areas. After one or more stages of selection, a sample of areas is identified within which enterprises or households are listed. From this list, the sample is selected and data collected. In general, it is preferred to use list-based surveys as it may be difficult to enumerate the enterprises within an area, and area-based sampling is inappropriate for (large or medium-sized) enterprises that operate in several areas because of the difficulty of collecting data from just those parts of the enterprises that lie within the areas actually selected. A stratified sampling technique should be used whenever appropriate and feasible to improve accuracy of data. 7.47. Use of business register. In principle, the sampling frame should contain all the units that are in the survey target population, without duplication or omissions. The business reg- ister maintained by countries for statistical purposes provides such a population. In general, a statistical business register is a comprehensive list of all enterprises and other units, together with their characteristics, that are active in a national economy. It is a tool for conducting statistical surveys, as well as a source for statistics in its own right. The establishment and maintenance of a statistical business register in most cases are based on legal provisions, as its scope and coverage are determined by country-specific factors. It is recommended, as the best option, that the frame for every list-based enterprise survey for energy industries be derived from a single general-purpose statistical business register maintained by the statistical office, rather than from stand-alone registers for each individual survey. 7.48. For countries not maintaining an up-to-date business register, it is recommended that the list of enterprises drawn from the latest economic census and amended as necessary, based on relevant information from other sources, be used as a sampling frame. Ad hoc energy statistics surveys 7.49. Specially designed energy statistics surveys are extremely useful to compensate for the lack of information and gaps associated with the mechanisms and instruments mentioned above. Examples of ad hoc energy statistics surveys are energy consumption surveys designed specifically to measure the quantities of energy products consumed. The sampling unit is likely to be the household and possibly sites of small-scale rural industries below the normal threshold for sample enquiries. Data generally cover the weights (or volumes, if realistic conversions to weight can be made later) of different fuels consumed for different purposes. If there is a seasonal pattern of fuel usage, interviews will have to be spread over the entire year in order to be representative of all seasons. Results will need to be analysed by size of household in order to obtain a range of per capita consumption figures. 7.50. The design and implementation of such surveys may be demanding in terms of finan- cial and human resources and often require multidisciplinary expertise in order to identify  === ires-2018-page-109.pdf === 96 International Recommendations for Energy Statistics (IRES) the appropriate sample design, interviewing techniques and analysis procedures. In general, ad-hoc energy statistics surveys are very useful instruments for assessing energy consumption activities, monitoring the impacts of energy programmes, tracking the potential for energy efficiency improvements and targeting the feasibility of future programmes. Household surveys and mixed household-enterprise surveys 7.51. Household surveys are surveys in which the sampling units are the households. In mixed household-enterprise surveys, a sample of households is selected and each household is asked whether any of its members own and operate an unincorporated enterprise (also called informal sector enterprise in developing countries). The list of enterprises thus compiled is used as the basis for selecting the enterprises from which desired data are finally collected. Mixed household-enterprise surveys are useful for covering unincorporated (or household) 61 See IRIS 2008, paras. 6.19–6.24 enterprises which are numerous and cannot be easily registered.61 for a description of advantages and disadvantages of mixed 7.52. Even though household surveys are not designed specifically for energy data com- household-enterprise surveys. pilation, they can give a broad overview of residential energy consumption by end-use and, potentially, of energy production by households. Given the complexity of energy consump- tion characteristics in households, estimates and other measurements of energy consump- tion should be derived from such surveys, using the metadata provided by them. For energy purposes, useful information is related to the number and average size of households, appli- ance penetration and ownership, appliance attributes and usage parameters, fuels used for cooking and for heat and air conditioning, electricity sources (national grid, solar electricity, auto-production, etc.), and types of bulbs used for illumination, etc. It is to be noted that the characteristics of household appliances stock, such as age and efficiency, can also be deter- mined through the use of administrative registers or surveys on appliance sales. 7.53. The frequency of these household surveys is another key element in efforts to obtain information on a regular basis, given that the behaviour in this sector often shows high variation due to changes in prices, technologies and fuel availability. The appearance on the market of new household appliances creates new energy consumption habits that should be taken into account. 7.54. These surveys should be representative not only at national level but also in rural and urban areas and by regions in order to achieve a proper analysis of the data. 2. Administrative data sources 7.55. Publicly controlled administrative data sources. Data may be collected by diverse governmental agencies in response to legislation and/or regulations to: (i) monitor activities related to production and consumption of energy; (ii) enable regulatory activities and audit actions; and (iii) assess outcomes of government policies, programmes and initiatives. 7.56. Each regulation/legislation (or related group of regulations/legislations) usually results in a register of the entities (enterprises, households, etc.) bound by that regulation/legislation and in data resulting from application of the regulation/legislation. The register and related data are referred to collectively as administrative data. The data originating from administra- tive sources can be effectively used in compiling energy statistics. 7.57. There are a number of advantages in the use of administrative data, the most impor- tant of which include the following: reduction of the overall cost of data collection; reduction of the response burden; smaller errors than those arising from a sample survey (due to the complete coverage of the population to which the regulation/legislation applies); sustainability due to minimal additional cost and long-term accessibility; regular updates; possible absence  === ires-2018-page-110.pdf === Data collection and compilation 97 of a survey design, sample measure and data editing; possibility of cooperation between various agencies, which could lead to feedback on the compiling process and acknowledge- ment of diverse areas of interest; potential data quality improvement; potential recognition of administrative data uses; opportunity to link data from diverse sources; development of statistical systems within agencies; and possible use as a framework for statistical surveys. 7.58. However, since administrative data are not primarily collected for statistical purposes, it is important, when using such data, that special attention be paid to their limitations and efforts made to ensure that their description is given in the relevant metadata. Possible limita- tions in the use of administrative data include: inconsistencies in the concepts and definitions of data items; deviation from the preferred definition of the statistical units; the legislation/ regulation may differ from the desired survey population; poor quality data due to lack of quality assurance of the administrative data; possible breaks in the time series because of changes in regulations/legislations; and legal constraints with respect to access and confiden- tiality (see chapter X for further discussion of confidentiality). 7.59. It is important that compilers of energy statistics identify and review the available administrative data sources in their country and use the most appropriate ones for collecting and compiling energy statistics. This significantly reduces the response burden and survey costs. The relative advantages and disadvantages mentioned above are not absolute. Whether they apply and the extent to which they do depend on the specific country situation. Exam- ples of administrative data sources important for energy statistics include customs records (for imports/exports of energy products), value added tax (VAT), specific taxes (or excise duty) payable on specific fuels (gasoline and diesel for road use) or energy types (e.g., carbon tax), and regulated electricity and gas market meter operator systems. 7.60. Privately controlled administrative data sources. Data may be collected by privately controlled organizations, such as trade associations. This is typically done to assist the indus- try in understanding important aspects of its own operations. These data are often also important to governments and decision- and policymakers. The statistical agency responsible for energy statistics should work cooperatively with these private organizations to gain access to such data, in order to maximise their statistical value. This would keep the reporting burden on the industry to a minimum by not requiring the businesses to report to both the private organization and the statistical agency. However, if agreement cannot be reached, the statistical agency may need to require that the data be submitted to them directly. Every effort should be made to establish proper cooperation between private organizations and the statistical agency. Statistical agencies must ensure the quality and objectivity of data being provided by these organizations, as data collection is not their primary activity, and they may be functioning as industry advocates. E. Data compilation methods 7.61. Data compilation, in general, refers to the operations performed on collected data to derive new information according to a given set of rules (statistical procedures), with a view to producing various statistical outputs. In particular, data compilation methods cover: (a) data validation and editing; (b) imputation of missing data; and (c) estimation of popula- tion characteristics. These methods are used to deal with various problems with collected data, such as incomplete coverage, non-response, out-of-range responses, multiple responses, inconsistencies or contradictions and invalid responses to questions. These problems may be caused by deficiencies in the questionnaire design, lack of proper interviewer training, errors by the respondent providing the data and/or errors related to the processing of the data. It is advisable to periodically generate reports specifying the frequency with which each of  === ires-2018-page-111.pdf === 98 International Recommendations for Energy Statistics (IRES) the problems occurs, thus identifying the main sources of error and making the necessary adjustments in future data collection processes. A brief overview of the recommended data 62 More information on the differ- compilation methods is provided below.62 ent techniques used in data com- pilation is found, for example, in 7.62. Data validation and editing is an essential process for assuring the quality of the col- IRIS 2008. lected data and refers to the systematic examination of data collected from respondents for identifying and eventually modifying inadmissible, inconsistent and highly questionable or improbable values according to predetermined rules. It is important to define validation cri- teria that clearly and systematically confirm whether or not the data satisfy the requirements of completeness, integrity, arithmetic consistency and congruence, as well as guarantee their overall quality. Validation criteria are established by the statistical authority according to the nature of the data and the analysis of the variables of interest, taking into account magnitude, structure, trends, relationships, causalities, interdependencies and possible response ranks. 7.63. In recognizing the importance of the data validation and editing, it should be empha- sized that any arbitrary alteration of the data should not be allowed, and any changes in the collected data should be based on the relationship between variables and response values. To prevent out-of-range responses and inconsistencies, appropriate response ranges for each question and the congruence that must exist between responses from related questions must be established. For example, checking that the sum of available supplies equals the sum of recorded uses is an important validation criterion. This is also valid for routine questionnaires directed to the energy industries. 7.64. As validation and editing can be very expensive components of the survey process, attention should be focused on the most important areas and issues. For example, many survey responses may have minimal impact on the final results, and effort to correct errors in such responses may be ineffective. To maximize the effectiveness of the validation and editing process, the responses that will have the greatest impact on the final results should be identified prior to the start of the actual process, so that resources can be properly allocated. 7.65. Data imputation. Imputation refers to replacing one or more erroneous responses or non-responses with plausible and internally consistent values in order to produce a complete data set. It is used for estimating missing data values when, for example, the respondent has not answered all relevant questions but only part of them, or when the answers are not logic­ ally correct. There are a variety of imputation methods ranging from simple and intuitive to rather complicated statistical procedures. 7.66. The choice of methods for imputation depends on the objective of the analysis and 63 For more information on imputa- the type of missing data. No method is superior to others in all circumstances.63 In most tion options in the case of item imputation systems a mix of imputation methods is used. The following are the desirable non-response or unit non- properties of all imputation processes: response, see IRIS chapter VI.B.2. (a) The imputed records should closely resemble the missing or failed edit record, retaining as much respondent data as possible. Thus, the number of imputed data items should be kept to a minimum; (b) The imputed records should satisfy all edit checks; (c) The imputed values should be flagged and the used methods and sources of imputation described in the metadata. 7.67. It is recommended that compilers of energy statistics use imputation as necessary, with the appropriate methods consistently applied. It is further recommended that these methods comply with the general requirements as set out in international recommenda- tions for other domains of economic statistics, such as the International Recommendations for Industrial Statistics (UN 2009b).  === ires-2018-page-112.pdf === Data collection and compilation 99 7.68. Grossing up procedures. After the data have been validated and edited, and impu- tations corrected for non-response and erroneous responses, special procedures should be applied to the sample values to estimate the required characteristics of the total population (these are referred to as grossing up procedures). These procedures consist of raising the sample value with a factor based on the sampling fraction in order to obtain the levels of data for the sample frame population. In some cases, depending on the relationship with other vari- ables for which data may exist, more sophisticated statistical techniques can be used for this purpose. As the application of estimation procedures is a complex undertaking, it is recom- mended that specialist expertise always be sought for this task. 7.69. The treatment of outliers is an important estimation consideration, particularly in energy statistics. Outliers are reported data that are correct but are unusual in the sense that they do not represent the sampled population and hence may distort the estimates. If the sampling weight is large and the unadjusted outlier value is included in the sample, the final estimate will be inappropriately large and unrepresentative, as it is driven by one extreme value. The simplest way to deal with such an outlier is to reduce its weight in the sample so that it represents itself only. Alternatively, statistical techniques can be used to calculate a more appropriate weight for the outlier unit. The details on the treatment of outliers should be provided in the metadata.  === ires-2018-page-113.pdf ===