Page History

13.1.            As described in part II, several sources can be used in the compilation of data within the modes of services supply statistical framework. In practice, information from various sources will often have to be combined to obtain the required level of detail and quality, while also reducing excessive burden for respondents. The present chapter reviews the potential challenges compilers face when integrating data sources that cover the same activities or the same entities and provides guidelines on how to resolve those issues. More specifically, the chapter presents a summary of good practices and defines the relevant terminology (section A), and describes the integrated business survey program (section B) and integration, consolidation and merging of data from different sources (section C). A country experience is presented in section D. A.Summary of good practices

1. Integrating data from different sources is advised as the principle way to ensure the production of more detailed and comprehensive statistics as well as to reduce the burden on survey respondents. Compilers are encouraged to determine which data sources are most appropriate on a case-by-case basis, carefully considering the strengths and weaknesses of each data source. More specifically, compilers should be aware of the statistical units used in each source, the entities covered, the services categories identified, the variables compiled (e.g., the value of services exports/imports by EBOPS categories for the compilation of trade in services between residents and non-residents or output or turnover/sales and employment for the compilation of FATS, etc.), the availability of geographic breakdown, the time period of reference, the presence of thresholds and the survey frequency. Ideally, compilers should identify the sources in which they have more confidence, to be used as a benchmark for the other data sources.

1. To optimally exploit the advantages of data integration, it is recommended that the potential of using linked data be considered from an early stage and that decisions on survey design and sampling frameworks, for example, be made while keeping in mind that the data will be linked later on. Such forward thinking requires positive coordination across surveys, that is, compilers should ensure that the entities covered by different surveys and registers have sufficient overlap to avoid large shares of non-matched records, which make data integration more difficult and may create statistical biases. It is recognized that such coordination may be seen as a long-term strategy by countries whose capacity to compile statistics on the international supply of services needs significant strengthening.

1. The present Guide encourages national statistical agencies to adopt an integrated business survey programme, as outlined in the Guidelines on Integrated Economic Statistics. See Guidelines on Integrated Economic Statistics. Specifically, under those guidelines, it is recommended that the collection of statistical information move from a "stovepipe" approach, in which each statistical programme collects information for its own purposes, towards an integrated approach, which aims to integrate survey design and implementation across all statistical programme areas. That recommendation is particularly relevant in the context of statistics on the international supply of services, given that information needs cut across various statistical domains. Such integration of data collection procedures will reduce the burden on respondents. It will also create a common standardized statistical framework for presenting more coherent statistics for the entire economy, covering business statistics, short term statistics, national accounts and international statistics, for evidence-based policy making.

1. It is good practice to use the statistical business register (SBR) to identify the common statistical unit and to provide the sampling frameworks. In addition, it is advised that ITRS and the trade register, which includes merchandise trade statistics, be linked to the SBR, with the enterprise as common statistical unit. This is particularly relevant for services categories with a strong link between trade in goods and trade in services, such as manufacturing services.

1. Compilers should take note that for the purposes of the present chapter, the term "data integration" refers to bringing together information from two or more data sources, with the object of better understanding and presenting the nature of a transaction than would be possible on the basis of any one source alone. "Data consolidation" is understood as the summation of data from multiple sources, in which the sources generally provide information on non-overlapping parts of the total. "Merging data" refers to the process of combining data, especially data covering the same activities or entities, from different sources via common denominators, such as a unique identifier in the SBR, when individual sources provide incomplete information. Merging data is the final step in the process and creates a unified view of an entity or activity. Before merging data, compilers must be aware of the definitions and methodologies used across the various data sources to ensure internal consistency of the merged data and to prevent the duplication or overlap of records.

...

1. The idea of developing and achieving an integrated system of business and international statistics has a long history among national and international governing bodies of economic statistics. The motivating source of that work has been the policy and statistical benefits that can be obtained from an integrated statistical system for national and international coordinated economic policy planning in an interconnected global economy.

1. The benefits to data users, producers and providers, include, but are not limited to, the following:

...

1. The starting point for integrated economic statistics is the use of the internationally accepted standard for macroeconomic accounts, the System of National Accounts (SNA) (preferably its latest edition, the 2008 SNA) as the conceptual organizing framework. As a result of years of work in harmonizing macroeconomic statistical standards, the use of the 2008 SNA as the conceptual framework provides consistency and coherence with other internationally accepted standards and international recommendations, such as the Balance of Payments and International Investment Position Manual, 6^th ed. (BPM6) and other specialized manuals, such as MSITS 2010. See Guidelines on Integrated Economic Statistics, p. 17.

1. Standardized surveys across statistical programmes The most effective means of ensuring integration and consistency is through a holistic redesign of data collections, both surveys and administrative data, that minimizes inconsistencies through the use of common standards and methods, integrated survey design and a central business register. It is good practice for the collection of statistical information to move from a stovepipe approach towards an integrated approach. In a stovepipe model, in which each programme collects information on its own and for its own purposes, an entity surveyed may be asked the same questions with different definitions. A stovepipe approach, therefore, not only increases the survey burden, but is also likely to produce statistics that are difficult to align. In contrast, integrated questionnaire design and integrated surveys help to resolve such data-collection issues. Ibid., pp. 49-50. The integrated approach is particularly relevant for statistics on the international supply of services, which are compiled with information from various statistical domains.

1. Under an integrated approach, all economic data collections gradually change their objectives and statistical designs and enhance the coherence of statistical outputs. When designing a collection for various sets of economic statistics, compilers need to think beyond their current work to how those sets will integrate with other statistical outputs. Likewise, questionnaires must apply concepts and definitions that are consistent with those used in other surveys and administrative collections. Ibid.

1. A national statistics act may provide a national statistical office with the authority to approve the design of all statistical surveys operated by government agencies. Ibid., pp. 28-29. One practical approach to achieving integrated surveys is a reorganization of the statistical agency towards a more functional structure. A separate division with responsibility for developing and maintaining classifications and standards can lead a programme to apply consistent definitions and classifications to all statistical outputs. Creating a division that has responsibility for developing the SBR to be used for the production of all economic statistics is an effective way to ensure that units are selected and classified consistently across all statistical outputs. Establishing an independent statistical methods division can help to ensure the coherent design of all surveys. Ibid., p. 39.

1. It is important to note here that the subject matter expert or compiler of statistics on the international supply of services may not always have full control over the design of such integrated surveys, if the design is managed by an independent division. It is important for compilers to recognize that success in achieving integration across many programmes requires extensive collaboration and communication across the organization, support from senior leaders in the organization and adaptability and flexibility. A country example of the "Integrated Business Statistics Program at Statistics Canada", is available from http://unstats.un.org/unsd/nationalaccount/workshops/2014/St_Lucia/IBSP.ppt. Samples of integrated surveys from Statistics Canada are available on the website of the Statistics Division.

...

1. The compilation of data within the statistical framework for describing the international supply of services requires the use of a wide range of data sources, as some sources are complementary and their combined use can result in the production of more detailed and comprehensive statistics.

1. The integration of different sources and the reuse of surveys for several purposes could also reduce the burden for compilers, in particular in the context of compiling modes of supply data. Furthermore, linking data on a microlevel could allow for broader types of analysis of the resulting data. The potential of the approach of using information from other statistical frameworks therefore deserves further exploration by the compiler. For example, in the context of the compilation of inward FATS, it is possible to integrate information from foreign direct investment (FDI) surveys and domestic enterprise surveys. Other examples include using information from sources on migration, tourism, household expenditures, population or taxes (see chapter 15). That information could be used in the compilation of resident/non-resident trade in services data (see chapter 14) or to compile quantitative indicators on modes 2 or 4 (see chapter 16).

1. The consolidation and merging of data sources provides multiple advantages for compilers. First, it improves coverage and the diversity of information. Second, it usually reduces the resources required to collect the statistics. Third, it results in statistics of higher quality. However, compilers must be aware of the risk that some of the data sets resulting from the integration of various data sources may be internally inconsistent.

1. Consolidation is a straightforward summation of data from multiple sources in which the sources provide information on non-overlapping parts of the total. Meanwhile, the merging of data sources is usually not a simple task, as compilers must find common denominators between data from multiple sources that cover the same entity or activity. The business register is of vital importance in such linking of microdata.

1. Compilers should take into account all relevant dimensions of data sources before merging; for example, the content of each source could be more or less compatible with a certain services category definition. For such reasons, compilers should have in-depth knowledge of the methodologies and definitions used by the other data sources to ensure consistency and the comparability of the resulting merged data. Various dimensions of the different sources must be analysed by the compilers during merging, most notably: entities covered, services categories or activities identified, the availability of geographical breakdown and the period of reference.

1. Entities covered Sources could collect statistics at different structural levels of the entity. Some sources may target the legal entity, while others could survey entities responsible for the production. Companies could merge, be acquired, disappear or simply modify their organizational structure so that entities might end up being different from source to source if they have been surveyed at different points in their structural composition. There could also be differences between how the business register defines the structural organization of an entity and how that entity defines its own structural organization. That might result in the statistical compiler unwittingly incorrectly comparing entities from different sources.

1. It is good practice for compilers to analyse the coverage, possible overlap and potentially differing definitions of variables across the different sources available, and determine, on a case-by-case basis, which data sources are most appropriate to use.

1. Services categories Compilers should compare the services categories available from each source to be merged. Details and aggregations provided could be different from source to source. In the absence of clear definitions and guidance, respondents may interpret survey questions incorrectly and include invalid transactions or exclude pertinent ones.

1. Geographical breakdown Data sources could have different levels of geographical breakdown, or even none at all. For example, the country of transaction identified in the ITRS could be different from the country that actually purchased or sold the service.

1. Period of reference Values reported on a monthly basis from one source could differ from the annual value reported by another source (e.g., where there is an annual reconciliation or balancing process). Moreover, difficulties may arise when integrating data from administrative sources on a fiscal year basis with information collected on a calendar year basis. Similarly, there may be challenges in bringing together data reported on a monthly or quarterly basis with data reported on a through-the-year basis.

1. MSITS 2010 recommends the use of the accrual basis for determining the time of recording of transactions. The accrual basis provides the most comprehensive information because all flows are recorded, including non-monetary transactions, imputed transactions and other flows. The change of economic ownership is central in determining the time of recording on an accrual basis. See BPM6 Compilation Guide, para. 1.21.

1. Other factors, such as the threshold for certain sources or the frequency of the survey, may also affect the comparison between sources. If a source is available only once every two years, compilers should develop a strategy to preserve the comparability for the year that the source is not available. All such possible differences between sources may complicate the comparability and integration of data sources, at both the global level and the microlevel.

1. Possible approaches and solutions There is no quick fix for problems with merging data. Compilers must have in-depth knowledge of the data with which they work and need to recognize the strengths and weaknesses of each source. Data sources are rarely equivalent; one source may provide reliable information for some transactions but be weak for others. Ideally, if compilers can identify a source in which they have more confidence, that source could serve as a benchmark against which to compare the other sources.

1. Linking various data sources It is vital that compilers avoid approaching businesses multiple times with different surveys covering the same or similar information. Above all, statistical surveys should not request information that the business has already supplied in another data gathering mechanism. The central SBR should be linked to the trade register to enable the analysis of the effects of the international supply of services on production, employment and enterprise performance. See also chapter 16. The enterprise is the suggested statistical unit to be linked between trade and business statistics; thus, data collected and registered at the level of the declaring unit of trade operators can be aggregated to the level of the whole enterprise via characteristics available in the SBR. The linking of trade and business statistics enables the generation of relevant information on the structure of the international supply of services without collecting additional data from businesses.

...

1. Three different sources are used in Canada to generate trade in services statistics, except for travel and transport). First, the annual survey on trade in services (international transactions in commercial services (BP21S)) is sent to Canadian enterprises of varying size and economic sector. To supplement the International Transactions in Commercial Services (BP-21S), more than 200 entities from the insurance sector are also surveyed, with two surveys of Canadian incorporated insurance companies and Canadian branches of foreign insurance companies. Both surveys collect financial transactions as well as service transactions. The second data source is the set of approximately 15 annual industry surveys covering a large sample of establishments in service industries. The third source is administrative data on non-arm's-length annual transactions between residents and non-residents.

1. Statistics Canada integrates those three sources and is thereby able to improve coverage, make better use of limited resources and reduce the burden on respondents. The key to integrating the sources is that each entity that they cover is linked to the Canadian business register (CBR). The CBR contains information on ownership, industrial activity, organizational structure and the business identification number This identification number is assigned at the time of registration with the Canada Revenue Agency. For more information about the Statistics Canada Business Register, see http://www23.statcan.gc.ca/imdb-bmdi/document/1105_D2_T1_V3-eng.pdf. of businesses engaged in the production of goods and services in Canada.

1. The quality of the CBR directly influences the compiler's ability to integrate the different sources. A team at Statistics Canada works constantly to ensure that the CBR reflects the structure of the Canadian economy as well as possible.

1. The three sources cited above are not considered equal in terms of quality and the details provided. Data in the industry surveys are indirectly measured in a process by which respondents indicate the percentage of their total revenues related to their total trade in services. No breakdown is available for transactions conducted between related partners and transactions between unrelated entities. Furthermore, the geographical breakdown is available only for the total trade in goods and services. The administrative records cited present only trade with affiliated parties and aggregated groupings of services.

1. Both the industry surveys and the administrative data have limited details in term of EBOPS categories. However, the annual BP21S is very detailed, with 32 categories of services collected, broken down by partner country and affiliation.

...