The importance of sharing survey data

2Activities to facilitate access to survey data are of key importance for the social sciences. It is thanks to such efforts that surveys can be shared between the teams involved in their design and with the broader research community. Sharing survey data is essential for the social sciences as it provides opportunities for secondary analyses, and for the testing and replication of studies. This enables researchers to understand, evaluate and build upon existing research, thereby contributing to the progress of their disciplines (King, 1995, 2006; ICPSR, 2012; Silberman, 1999). Sharing survey data also discourages scientific fraud and is useful for teaching analytical methods. It benefits survey producers by raising awareness of their work through citation, provides justification, through data re-use, of the high costs of surveys, and allows for further testing of data collection methodologies (Silberman, 1999).

3Data sharing activities concern both data producers and survey data archives (ICPSR, 2012). Producers process the data for subsequent use and create coherent data files. They also prepare survey documentation, grant access to data files and archive them to assure their reusability. Survey data archives, i.e. archives that mainly deal with data at the individual (micro) level, serve the entire research community, and must provide accompanying documentation that is as exhaustive as possible. The role of data archives is to review data quality, create exhaustive metadata records, publish data and metadata files in online catalogues, manage users’ requests for data access, provide assistance to users and liaise with data producers to report on survey data use (ICPSR, 2012). [1]

4Data documentation is paramount to data sharing, because “without adequate documentation, scholars often have trouble replicating their own results months later” (King, 1995, p. 444). However, there is generally no specific budget for the preparation of metadata, even though produced data can be used by other research teams, avoiding the need to duplicate data collection operations. Moreover, researchers are reluctant to dedicate much time to this activity. The majority hastily assemble the documentation just before depositing the data in an archive. Often “researchers still wrestle with the idea that others may benefit from using their painstakingly gathered datasets and, perhaps even more important, they also fear that by making their datasets public, mistakes in the collection or processing of their data and their (to be) published results may be discovered” (De Moor and Van Zanden, 2008, p. 68). At the same time, in most western countries, the practice of depositing data has become a mandatory requirement of funding agencies, and a growing number of journals require authors to make available all data referenced in an article (De Moor and Van Zanden, 2008; Mochmann and Vardigan, 2011). Nonetheless, archivists who assemble the documentation and manage data access are often confronted with issues of metadata availability.

5This article acknowledges the importance of data sharing and describes the archival activities carried out by the Surveys Department (SES) [2] of the French Institute for Demographic Studies (Institut national d’études démographiques, INED) to provide access to INED surveys. These surveys consist of quantitative individual (micro) level data produced for non-commercial purposes by researchers with public funds, often with the collaboration of other public bodies. It examines INED’s activities in both the international and French contexts of access to quantitative social science survey data (hereafter referred to as survey data). It does not cover other types of social science research data that fall beyond its scope, such as qualitative survey data, electronic texts, linguistic corpora, historical and archeological data, administrative data, or data produced for commercial purposes.

6The article begins by tracing the origins of survey data archives. It then illustrates how international standards for documenting survey data were established and describes the development of survey data archives and the provisions regulating access to social science data in France. Last, it reviews the development and the current organization of the activities that provide access to survey data at INED. It ends with some reflections on future developments in archiving and sharing of survey data in the social sciences.

Survey data archives and regulations on data access up to the twentieth century

12Although some French researchers took part in early discussions about sharing survey data, no survey data archives were created in France before the 1980s (Silberman, 1999). There were two main reasons for this. First, there was little institutional and structural support in France for the large-scale university surveys which, as discussed above (Section I), prompted the development of survey data archives in other countries. Second, the surveys of the National Institute of Statistics and Economic Studies (Institut national de la statistique et des études économiques, INSEE) covered a large spectrum of topics, and French researchers could use the aggregated data published by the Institute. It was also possible to access its individual-level data files, [11] but this was difficult due to a legal framework which placed strong emphasis on personal data protection.

13Individual-level data were protected under two separate laws. First, the Act of 7 June 1951 prohibited the communication of individual data collected by official statistics (INSEE, and Ministerial Statistical Departments). [12] These data could be disclosed only in aggregated forms and in anonymized files. From 1984, some exceptions were granted, but exclusively for data on companies, with the creation of a committee for confidentiality of corporate statistics (Comité du secret statistique concernant les entreprises) [13] (Gaeremynck, 2009; Silberman, 2011). Second, the Data Protection Act of 6 January 1978 established that, whenever identification was possible, personal data could be collected and processed only within a limited time, after notification of the French Data Protection Authority (Commission nationale de l’informatique et des libertés, CNIL), [14] and for a specific purpose. This Act, which grew in importance with the development of computer technologies, [15] made it difficult to re-use personal data for purposes other than those that justified their initial collection (Riandey, 2000; Silberman 1999, 2011).

14Against this legal background, [16] data archives were first created to facilitate the sharing of anonymized individual data files (Silberman, 1999). A socio-political database (Banque de données sociopolitiques, BDSP) was established at the Institute of Political Studies in Grenoble and incorporated into a centre for computerization of socio-political data (Centre d’informatisation des données sociopolitiques, CIDSP). [17] This archive contained socio-political data produced by private and public statistical offices, as well as data produced by academic researchers. Another archive, run by a research unit of the National Centre for Scientific Research (Centre national de la recherche scientifique, CNRS), the Laboratory for Secondary Analysis and Applied Methods in Sociology (Laboratoire d’analyse secondaire et de méthodes appliquées à la sociologie, LASMAS), which has since become the National Archive of Data from Official Statistics, (Archive de données issues de la statistique publique, ADISP) of the Centre Maurice-Halbwachs (CMH), [18] was created in Paris. One of its missions was to provide access to survey data gathered by public statistical bodies. These archives were established on the initiative of individual specialists rather than as a common institutional effort [19] (Silberman, 1999).

15Access to stored data by researchers was regulated by a number of agreements (Rhein, 2002; Silberman, 1999), such as the one signed between INSEE and the CNRS in 1986 that gave LASMAS the right to disseminate a few anonymized INSEE surveys to all CNRS researchers. CNRS also signed similar conventions with other public data producers, such as the Ministry of Education. French research institutions, such as INED and the French National Institute for Agricultural Research (Institut national de la recherche agronomique, INRA), signed analogous agreements with the public data producers, in order to gain access to their data. Agreements for data produced by researchers and archived at CIDSP were less formal and established on a more individual basis, but licenses for dissemination to researchers were similar. The general principles regulating access to data produced with public funding were clarified in 1994, in the so-called “Balladur Circular”. While access to these data was free of charge, institutions wishing to acquire them had to pay the cost of the service.

16In the late 1990s, Claude Allègre, minister of education, research and technology, noted a growing need among social science researchers for access to survey data, and commissioned Roxane Silberman, the director of LASMAS, to conduct a study on French social sciences and their data (Allègre, 1999). Based on a survey among CNRS and university laboratories, Silberman (1999) identified three main groups of problems. The first group concerned data access. Data files in France were not regularly upgraded to the latest computer innovations. Furthermore, some surveys were not fully documented or not documented at all, which made their re-use impossible. Above all, there were no provisions in the agreements between CNRS and INSEE entitling university researchers to access data produced by public data producers, so many scholars had to rely on personal contacts with INSEE or other administrations. There was also uncertainty about university researchers’ intellectual property rights and their obligations to provide access to the data they produced. In the absence of any regulation, surveys co-produced by public institutions and universities were not made available by LASMAS. Furthermore, to protect citizens’ personal data, an increasing level of anonymization, especially for census micro data files, prevented French social researchers from making detailed analyses. Some of the BDSP studies were protected and could not be accessed. The second group of problems concerned data use. Compared with other countries, French quantitative sociology was less developed and less well-equipped in terms of computers and software for data processing. The third group related to data production. Large-scale surveys produced entirely by researchers in the academic world were still rare in France.

17In response to these problems, Silberman (1999) called for the creation of a “purpose-built archival structure” (p.47) and for a major reform of French policy on access to social science data, including legal provision for access to data for research purposes, and measures to make all survey data produced with public funds available for re-use. Experience in other western countries had shown that archives played a key role in stimulating the development of quantitative sociology and in involving researchers in survey production. A large archival structure was needed to harmonize survey documentation and to enable France to play an active role at CESSDA. The Silberman report laid the foundations for the creation of the Centre Quetelet (Chenu, 2011) which today manages access to most of the social science survey data produced in France and to some international surveys (see Box 3).

The Réseau Quetelet: creation, objectives and data access

18The Centre Quetelet was set up by decree on 12 February 2001 (Chenu, 2003). Together with the university data platforms (Plateformes universitaires de données, PUD), it implements the policy of the coordinating committee for humanities and social science data (Comité de concertation pour les données en sciences humaines et sociales, CCDSHS). This committee is responsible for defining national public policy on data in the social sciences and humanities, with three main purposes: 1) to facilitate access to data that are useful for research, 2) to develop the use of these data, and 3) to provide support for the production of large-scale surveys in the social sciences. The university data platforms provide local assistance to users and support for survey implementation. The Centre Quetelet is responsible for collecting, managing and archiving social science data and for providing user training in the technical and scientific innovations in this field.

19The Centre Quetelet was founded in December 2001 as a CNRS institution in partnership with three other institutions: École des hautes études en sciences sociales (EHESS), INED, and the University of Caen (Arduin, 2004; Chenu, 2003; Riandey, 2003). It had three founding members, BDSP (current CDSP, Footnote 17), which provides socio-political surveys; LASMAS (current ADISP), responsible, in particular, for data from public data producers; and INED through its Surveys Department, which provides the socio-demographic surveys produced by INED since its creation in 1945 (Box 4 and Section IV). In 2005, the Centre Quetelet was transformed into a network with the name of Réseau Quetelet (Chenu, 2011). It is a member of CESSDA and, since 2013, has been the French service provider within the new CESSDA set up as a European research infrastructure with a legal status (see Footnote 9). It is also a member of PROGEDO, the French social science research infrastructure which brings together the main actors involved in quantitative surveys.

20The Réseau Quetelet has three main missions: data documentation, data access and data dissemination (Arduin, 2004). Its partners are responsible for gathering all the information available about each archived survey and of restructuring it in accordance with international data documentation standards. The Réseau Quetelet prepares and sends the datasets to users in compliance with deontological rules. It also advises researchers on how to use datasets and informs them about data availability and the latest technological innovations in software and methods. It maintains ties between users and data producers at both national and international levels.

21In order to accomplish its missions, the Réseau Quetelet runs a website [20] that catalogues and documents all the data provided by its partners. In line with CESSDA recommendations, each of its partners has adopted DDI as a standard to document the data, implemented through the Nesstar software. Efforts are being made to provide survey metadata both in French and English. Some international surveys, such as the INED survey Migrations between Africa and Europe (MAFE), are documented in English only (Box 5).

22Access to data managed by the Réseau Quetelet is regulated through two types of conventions: one with data providers (its partners) and the other one with data users. The general principle is that access is provided free of charge and exclusively for non-commercial research purposes. Access procedures are available both in French and in English. They may vary depending on the status of users, their institutional affiliation, and the nature of the requested data files. While procedures are open to users of all nationalities, more detailed information (e.g. an extended description of research projects) may be required from users who do not belong to a French university or public research institution. Data requests can be submitted via the Réseau Quetelet website which was overhauled in March 2014. Several datasets from different data producers can now be requested in a single application. Users undertake to comply with a set of rules. [21] The mean waiting time for access to data may vary from a couple of days to several weeks.

23Soon after the creation of the Réseau Quetelet, the legislation on data protection was modified. Increasing requirements for data anonymization were making it more difficult to analyse French statistics. Spatial variables and sensitive variables (especially on nationality and country of birth) were being provided in increasingly aggregated forms (Riandey, 2000; Silberman, 2011). First concerning personal data, the 1978 Data Protection Act was reformed in 2004 (Act no. 2004-801 of 6 August), following the European Parliament and Council Directive 95/46/EC of 24 October 1995 – effective in member states since 1998 – on the protection of individuals with regard to the processing of personal data and on the free movement of such data. It introduced the possibility of using personal data collected for other purposes to carry out statistical or historical research (Silberman, 2011). [22] The re-use of personal data that might lead to direct or indirect identification and/or that was of a sensitive nature had to be submitted to the CNIL (Footnote 14). Second, with regard to official statistics, the 1951 Act (Footnote 12) was reformed by the Archives Act of 2008 to permit the re-use of very detailed personal data collected by official statistics for statistical, historical and research purposes (Gaeremynck, 2009). [23] It is also worth mentioning the Act of 22 July 2013 on higher education and research which opened access to fiscal data for scientific research purposes. As already established for data on companies (Footnote 13), and for data on public bodies (Order no. 2004-280 of 25 March), access to very detailed personal data was subject to the approval of the committee for statistical confidentiality (Comité du secret statistique – the reference to corporate statistics having been dropped from its title).

24These developments opened the path for organizing secure access to very detailed data from official statistics. The Secure Data Access Centre (Centre d’accès sécurisé aux données, CASD) was created in 2010 by the Group of National Economics and Statistics Schools (Groupe des écoles nationales d’économie et statistique, GENES), at that time an INSEE department, and it became a partner of the Réseau Quetelet (Le Gléau and Royer, 2011; Silberman, 2011). It also became the French data provider in the European Data Without Boundaries (DWB) project, which supports equal and easy access to official micro-data in Europe (Silberman, 2013). Contrary to the access provided by the other partners of the Réseau Quetelet, access to the CASD is not free of charge. Moreover, the procedure involves obtaining approval from the committee for statistical confidentiality, the data producers and, in the case of personal data, the CNIL. [24] It can take several weeks, depending on the number of authorizations required.

From archiving to providing access to surveys

25Since its creation in 1945, INED has carried out numerous socio-demographic surveys, sometimes in collaboration with other public bodies (Box 5). For many years, access to these surveys was managed in an informal manner. [25] From the 1970s, surveys and their documentation were archived and transmitted to the French national archives (Archives nationales) thanks to the work of Suzanne Helgoual’ch and Henri Bastide (Comité d’archivage de l’Ined, 2001a). However, these activities were not sufficient to fully accomplish INED’s mission of disseminating knowledge across the research community, to the public authorities and the public at large, as laid down in its statutes of 1945 and 1986 (amended in 2001) and in its strategic orientations (INED, 2002).

26Initial discussions as to whether and how INED should engage in activities for providing access to its surveys began in the early 1990s. Surveys were becoming more costly and requests to access the data for secondary analyses were increasing (Bozon, 1995). Jacques Magaud, director of INED at that time, commissioned Michel Bozon, then head of the Surveys Department, to draw up a report on the management of access to INED surveys. Bozon consulted various INED researchers and recorded a variety of different, sometimes divergent, proposals, as well as certain misgivings (Bozon, 1995). He concluded that certain principles of data access, previously tacitly understood, needed to be formalized and that INED surveys needed to be made accessible for re-use, accompanied by full documentation.

27Defining rules for accessing INED surveys was one of the objectives laid down in INED’s strategic orientations for the years 2002-2005. It was important to respect the intellectual property rights of survey producers, whether INED researchers or not, but also to factor in the costs of data access. Priority access was to be reserved for partners who had contributed to survey funding and design. INED, unlike other data archives, only disseminated and provided access to surveys conducted by INED alone or in partnership with other institutions. Except in some special cases, INED was not required to provide access to data files produced by others without its collaboration.

28Against this background, and spearheaded by François Héran, INED’S director at that time, the Institute was a co-founding partner in the creation of the Réseau Quetelet in December 2001, in which it occupied an intermediate position between a producer of official statistics and an academic research body; INED researchers not only produce surveys but also use them. Specifically, the Surveys Department (Box 4) was given the task of applying the institute’s policy on documenting and providing access to its surveys, in close liaison with the Réseau Quetelet (INED, 2002). This policy did not take practical effect until 2004. [26] Anonymized data files, along with their documentation, could be requested and accessed by INED and non-INED researchers, provided that they accepted the accompanying terms and conditions and agreed to comply with the rules governing use of the data [27] (Service des enquêtes et des sondages de l’Ined, 2004). The data were then sent by post on a CD-ROM.

29The activities aimed at providing access to surveys were progressively ramped up (INED, 2006) and, from 2006, under the impetus of Francois Héran and Cécile Lefèvre, head of the Surveys Department at that time, additional resources were deployed (Comité d’archivage de l’Ined, 2006). These activities included producing documentation and re-formatting data files, and extended beyond the tasks of data archiving per se. As a result of this policy, in 2008, some recent surveys became available before being sent to the National Archives.

30In 2009, the Surveys Department was also given the task of providing access to the Generations and Gender Surveys (GGS) that form part of the Generations and Gender Programme (GGP). [28] Furthermore, work was undertaken to determine the most appropriate ways of promoting the use of INED surveys, to better understand current and potential users, and hence to identify the surveys that should take priority in terms of user access (Comité d’archivage de l’Ined, 2009). This was made possible by adopting the DDI standard (Box 1) and posting the entire catalogue of INED surveys online using Nesstar software (Box 2).

Implementation of DDI and the Nesstar software

31After considering the option of using INSEE’s Statistical Data Dictionary (Dictionnaire de données statistiques, DDS) to archive and document surveys, it was decided to adopt the DDI standard in 2003. The DDI rules were screened and tested, along with the functionalities of the Nesstar software liable to facilitate implementation of the DDI-C standard (Comité d’archivage de l’Ined, 2003, 2004). It was not until the end of 2008, however, that work began to transfer the INED survey collection to Nesstar.

32With the Nesstar software, also adopted by the other members of CESSDA and the Réseau Quetelet, [29] an online catalogue has generated containing all survey records. The INED-Nesstar Catalogue (Catalogue Nesstar des enquêtes de l’Ined), [30] was officially launched in June 2012. It replaces the previous dissemination tool. Since it came online, more than 5,800 users (from 93 countries) have viewed more than 67,700 pages of the catalogue. In 2014, the number of users grew by 160%.

33The GGS surveys are managed through another catalogue, called GGP Online Codebook & Analysis, which was launched in 2010, and is available at the GGP homepage. [31] This catalogue has been visited by more than 5,100 users (from 74 countries), who viewed about 19.4 pages on average per session. The number of users increased by about 180% in 2014.

Surveys made available and data access procedures

34As of November 2014, the INED Nesstar catalogue contains 248 references. They cover a large spectrum of socio-demographic themes such as fertility, contraception, sexuality, marriage, migration, immigrants’ integration, discrimination, gender, generation, inequality, health, ageing, housing, employment, etc. The surveys in the catalogue are divided into two categories (each organized by decade and by year of production):

• Available surveys (55 references), i.e. whose data file(s) can be supplied to users;
• Unavailable surveys (193 references), i.e. whose data file(s) cannot be supplied to users (either because they do not exist, or because access has not yet been authorized).

35For the available surveys, the catalogue offers complete, downloadable data documentation, and enables users to perform and export some basic analyses. The catalogue can be consulted in both French and English. However, for most surveys the documentation is only available in French. Requests for access to data can be submitted via the Réseau Quetelet portal, and access is granted in accordance with its guiding principles. Thanks to the new online catalogue, the number of data access requests is increasing (+8% over the last year).

36The GGP Online Data Codebook & Analysis catalogue is entirely in English. As of November 2014, it offers data for two waves of the survey and for 17 countries (Australia, Austria, Belgium, Bulgaria, Czech Republic, Estonia, France, Georgia, Germany, Hungary, Italy, Lithuania, Netherlands, Norway, Poland, Romania, Russian Federation). The data files are regularly updated to the most recent versions of the surveys. Access to the GGP surveys is restricted to researchers and is managed via an online platform by the United Nations Economic Commission for Europe (UNECE). The mean waiting time for access is five days. The conditions are laid down in agreements signed by UNECE and GGP participant countries (see the GGP webpage for more information). This catalogue has also contributed to GGP data dissemination. The number of registered users of GGS micro-data files has grown by 33% over the last year (Generations and Gender Programme, 2014).

The work of data preparation and data documentation

37Before making a survey available for re-use, the sometimes “invisible” task of preparing and documenting the data must be performed. This involves gathering as much documentation as possible about both the survey itself and its data file(s). This documentation is often scattered: some documents may be stored in electronic files or in boxes of archives kept at INED, others may come from the research teams or take the form of articles and working papers based on the survey data. There are various possible file formats which do not represent the same workload. For the more recent surveys, documentation files are in “current” formats, i.e. Word or pdf documents from which information can be extracted and copied, and data files are in SAS, SPSS or Stata formats that do not need to be converted. In the case of old surveys, it is not rare to have only hand- or type-written paper documents, or files in text format, for example (or even no file at all [32]). Documents are sometimes incomplete, and variable names may be missing.

38Sometimes it is also necessary to select the correct information. There may be a lack of documents or, on the contrary, some information may be duplicated. At this stage, the collaboration of research teams involved in data collection is of crucial importance, both to optimize the gathering and selecting of information and to synthesize it. If they supply exhaustive documents and clean data files, the work of data preparation and documentation can be simplified and surveys can be made available much faster.

39Once all the data files and related documentation are ready, they are imported into Nesstar. The aim is to make all the information clear and understandable for all users. Metadata are documented according to the DDI fields chosen by the Surveys Department. These fields are classified in three groups:

• Document description: information on the Nesstar file (survey concerned, author, etc.);
• Study description: information on the survey (summary, researchers, producers, funding agencies, dates of collection, field, methodology, sampling procedure, etc.) and links to the questionnaire, other associated or connected surveys, and bibliography; [33]
• Description of data files: information on the datasets themselves (how they are constructed, missing data, notes on which variables are replaced with derived ones, etc.).

40In Nesstar, the variable labels and categories are also entered, so datasets that are already documented represent a significant time saving. In addition, every variable is examined in great detail: the texts of the question (text before the question, actual question wording, text after the question, instructions to interviewers), the universe (i.e. the persons to whom the question was asked) and who answered the question (interviewer or respondent). Other information may be added, such as the questionnaire from which the variable is taken (when there are several questionnaires) or, in case of derived variables, the variables and the software programs used in the calculation. To ensure maximum precision, each dataset is entirely reviewed, and variables are reorganized to follow the order of the questionnaire.

41This work of documentation, verification and harmonization is very time-consuming. A good codebook, and clean and anonymized files speed up the process considerably. A non-anonymized file requires additional work to detect potentially identifying variables which must be be recoded or deleted before dissemination. [34]

42The GGP surveys are documented using the same DDI fields as for INED surveys. The datasets are already anonymized and the variable labels and categories are already specified. Some of the survey metadata are common across all GGP surveys (e.g. how the study should be cited, data content keywords and summary). However, even for these surveys, the preparation of country-specific metadata (i.e. information that varies across harmonized surveys, such as data collection methods and processing, and specific variables) may prove to be time-consuming.