Principle 8: Appropriate Statistical Procedures

When European Statistics are based on administrative and other data, the definitions and concepts used for non-statistical purposes are a good approximation to those required for statistical purposes.

Administrative sources are used to produce many Official Statistics. The Nominative Social Declaration (DSN) for example, is used to compile statistics on employment or wages.

In order to move from administrative sources to the Official Statistics that it produces, INSEE implements processes to reconcile raw data with the usual statistical concepts. Thus, in the ÉSANE process (Elaboration of annual statistics of companies) for the creation of annual business statistics, specific processing operations aim to reconcile administrative data and those resulting from sample surveys. Similarly, the use of fiscal and social security sources, which feed into the systems producing statistics on income, poverty and inequality, requires the concept of the tax household (“foyer fiscal”) used in the fiscal sources to be reconciled with the concept of the household (“ménage”) used in Official Statistics surveys (including the SILC).

Processing chains for systems based on administrative sources include stages of checks, verifications, and, if necessary, adjustments. On completion of the processing operations, a selection of variables for statistical use are produced. In the case of systems using fiscal and social security sources, these include, for example, declared income, disposable income and variables for categorising households on the basis of their socio-demographic characteristics. The processing operations carried out and their results are covered in ad hoc publications made available on the INSEE website (e.g. FILOSOFI, a localised social security and tax data file).

Furthermore, where INSEE itself is responsible for an administrative source, the processing of that source is clearly distinguished from the production of Official Statistics. This is the case for the three inter-administrative registers for which it is responsible: the National Directory for the Identification of Natural Persons (RNIPP), the Single Electoral Register (REU) and the Business register identification system (SIRENE). Thus, for social statistics, the individual identifiers present in the registers of natural persons are not used directly, whereas for business statistics a specific register is produced (SIRUS).

In the case of statistical surveys, questionnaires are systematically tested prior to the data collection.

Survey questionnaires, whether recurring or not, which have been subject to amendments, are systematically tested prior to data collection by INSEE staff (designers, managers and specialist interviewers). The tests can take various forms (focus groups, standard tests, etc.).

The testing of questionnaires is one of the ten key recommendations of the Ouvrir dans un nouvel ongletOfficial Statistics Quality Label Committee for the constitution of dossiers for requesting opinions on compliance. Thus, prior to issuing an opinion on compliance, the Quality Label Committee reviews, in particular, the reports on the tests conducted. Any difficulties encountered are described therein and amendments to be made are proposed. The tests also serve to verify that the observed response times are acceptable in terms of cost and response burden.

In addition, in order to develop the questionnaires, in particular those of the European surveys, producer services rely not only on the guidelines provided by Eurostat, but also on external opinions and lessons learned from previous exercises.

Statistical processes are routinely monitored and revised as required.

Within INSEE, the Department of Statistical Methods provides guidelines, recommends appropriate methodologies and regularly reviews and revises them, where necessary. It draws up and shares many methodological documents, which include guidelines on sample coordination, balanced sampling, methods for estimating accuracy, correcting non-responses, small areas, etc. It can also provide support to producers as it did during the revision of the adjustment methods for handling non-responses to the SILC. The department also revises samples when necessary and calculates the accuracy of the results or provides the producing services with the necessary tools to do so.

In addition, the producing services draw up production assessments so as to be able to learn lessons from previous editions of surveys or production from administrative sources for the subsequent editions. The review of systems is built into the production process. Thus, for each edition, including for regular or re-edited surveys, the questionnaires are modified if necessary and tested, interviewers are trained before collection, after revision of the content and training materials, the collection documents are revised and the downstream processing operations (monitoring-checks, weightings and imputations) are adjusted if necessary. When major changes are found to be necessary, INSEE undertakes redesigns of its processes, such as that of the SILC system or the Labour force survey over the recent period.

Finally, as part of the examination prior to the issue of an opinion on conformity for any survey project, the Ouvrir dans un nouvel ongletOfficial Statistics Quality Label Committee studies the description of the sampling, collection method and the treatments and dissemination envisaged in depth. In particular, the method for updating the sample forms an integral part of the description of the sampling method submitted to the Quality Label Committee. In addition, it examines ongoing surveys every five years.

Metadata related to statistical processes are managed throughout the statistical processes and disseminated, as appropriate.

INSEE has established a statistical metadata repository (RMéS) that structures and centralises this information. Concepts, classifications, variables, code lists or any other information relating to statistical operations are described in accordance with international standards and can be returned to various users in a variety of formats that can be read by machines. For example, RMéS feeds into the “Definitions”, “Classifications” and “Sources and Methods” pages of the insee.fr website.

The use of the metadata contained in the repository can go beyond their role of description and supporting understanding. Thus, complementary tools have been developed making it possible, for example, to generate the collection media (questionnaire in PDF format, online collection medium) from the metadata and to view the medium thus obtained.

The various tools put in place are the subject of training or mentoring activities for producers so that they are able to use them effectively. An expansive internal communication campaign was also launched to promote these tools, culminating in the Ouvrir dans un nouvel ongletK’RMéS which took place in autumn 2020 and devoted a day to the presentation of the repository and its benefits.

Revisions follow standard, well-established and transparent procedures.

The General Revisions Policy, formalised by INSEE in 2019, presents the different types of revisions and the reasons behind them. An internal revision announcement protocol was drawn up at the same time. It specifies the framework for such announcements and whether communication regarding the revision is necessary or preferable. Four decision-making criteria are taken into account: scope, nature, origin and sensitivity of the subject.

This system inspired the Common Revision Policy of the Ministerial Statistical Offices (MSOs), which was developed in 2020 and is available on the websites of the MSOs.

Routine revisions are thus disseminated according to predefined calendars that are kept up to date. The Indicators Calendar provides the publication dates of the various versions of the main economic indicators.

Redesigns are announced to the general public, as far as possible, before the publication of the revised statistics. Thus, the dissemination in 2019 of the data related to the elaboration of annual statistics of companies (ÉSANE), for the financial year 2017 was accompanied by a Methodological Note explaining the impact of the new definition of “enterprise” in line with the economic concept in Council Regulation (EEC) No 696/93.

The description of statistical sources and indicators on insee.fr also contains a section on revisions.

Agreements are made with holders of administrative data which set out their shared commitment to the use of these data for statistical purposes.

Access to or re-use of data by INSEE is always subject to an agreement with the administrative bodies or private companies that hold data useful for Official Statistics. These agreements establish reciprocal rights and obligations and mention the characteristics of the source in question, the provision circuit, the calendar, the variables to deliver, the formats to use, the documentation to provide with the data, the conditions of use and dissemination, the roles of the different stakeholders, the rules to follow with regard to confidentiality, etc.

An example of the agreements concluded with administrative data holders is the agreement signed in 2018 by INSEE and the Directorate-General for Public Finance concerning value added tax (VAT) data from returns filed by companies. The agreement specifies, in particular, the variables transmitted, the frequency and the processing and adjustments carried out by the Directorate-General for Public Finance, such as the addition of calculated variables useful for statistical use and the enrichment of the file with activity codes from the official classification.

In addition to agreements to access private data sources for study, analysis or research purposes, partnerships with private companies can facilitate the response to Official Statistics surveys. Thus, the tripartite agreement signed in 2018 by INSEE, the Airbus Group and the Customs Directorate General (DGDDI) formalises the conditions for the retransmission to INSEE by the DGDDI of data relating to Airbus, as part of the survey on observation of prices in industry and services. This exchange of data reduces the survey burden for respondents from the Airbus Group.

Statistical authorities co-operate with holders of administrative and other data in assuring data quality.

The French Official Statistical Service (SSP) works with the holders of administrative and private data to continuously improve the quality of the data transmitted.

Anomalies detected during production are, to the extent possible, discussed with the data provider as soon as they are discovered. In general, the data provider provides explanations making it possible to validate the data or reports a verified defect in the data, which may, depending on the case, be corrected by a new data delivery, rectified by INSEE through reprocessing (e.g. imputation) or mentioned as a limitation when disseminating statistical results taken from the data. At the end of the campaign, data providers may also receive assessments (e.g. production assessments) that could help them to improve the data over time.

Meetings are usually organised between the administrative bodies holding the data and the official statisticians prior to each delivery of data or in the case of an expected substantial change in the content of a file. These meetings, which are often annual, provide an opportunity to perform an assessment of the past campaign, to report any difficulties encountered, to plan and organise future data exchanges and, sometimes, to consider changes to the agreements to allow access to new data.

Working together with data providers allows official statisticians to gain a better knowledge of sources over time. Through interactions between the services collecting the data and the services using the data for statistical purposes, this collaborative work contributes to improving the quality of the statistics produced and optimising the use of existing data by Official Statistics. Discussions between INSEE and the Directorate-General for Public Finance to compensate for the abolition of the housing tax are an example of this.