STATISTICAL COMMISSION and COMMISSION OF THE ECONOMIC COMMISSION FOR EUROPE EUROPEAN COMMUNITIES CONFERENCE OF EUROPEAN STATISTICIANS EUROSTAT
Joint ECE/Eurostat Work Session on
reaction memberStatistical Data Confidentiality Working Paper No. 15 (Thessaloniki, Greece, 8-10 March 1999) English only
Topic (iii) - Administration and policy of statistical data confidentiality
INFORMED CONSENT: BUZZWORD OR PANACEA
Submitted by Statistics Netherlands1
Invited paper
I.INTRODUCTION
1.The production of official statistics has become considerably more complicated over the last decades. Many of these processes in the past started with a questionnaire and ended with a publicati
on. There were many of such stovepipes, parallel and simultaneous. If there was a need for new official statistics to be produced, the usual reaction was to design a new questionnaire, to collect new data, and to release a new publication.
2.Nowadays, a double transformation of these processes is taking place within Statistics Netherlands (Van Bochove 1996). Apart from and in addition to its (primary) data collection efforts, Statistics Netherlands makes increasing (secondary) use of whatever administrative data is already available. All data pertaining to one type of statistical unit (such as persons or companies) are loaded into one input database.2 At present, there are two of such databases under construction, one for persons (SSB or Social Statistical File) and one for companies (Microlab).3 For each member of the population in this database a record consists of variables from a variety of sources. These records in turn serve as the basis for further statistical processing and publication.
3.On the output side, Statistics Netherlands is in the process of creating one output database (called StatLine, accessible on the Internet at www.cbs.nl) containing in principle all publishable aggregates. Paper publications do not contain the traditional tabulations anymore. They focus on specific themes and topics from a variety of angles, thus accentuating the interrelations between economic and social dimensions in a concrete setting. In short, there is not a direct link anymore from
one statistical questionnaire to one particular tabular publication.
1Prepared by Joris Nobel. T he views in this paper are those of the author and do not necessarily reflect the policies of Statistics Netherlands.
2Matching of these data is done on either a unique identification number, or on a combination of detailed background variables such as postal code, housenumber, date of birth, and sex in the case of social data.
3Of course, these databases have a hierarchical structure (persons, households, families) and are related to each other (people work within companies). Other, similar databases for other types of statistical units (locations) will be created in due time.
4.Apart from the StatLine database and the thematic publications, the release of microdata for research purposes has developed into a powerful statistical outlet during the last decade. A microdata
file is one of several products along the production line of collecting, processing and releasing data. Microdata files are released for the major social surveys at Statistics Netherlands. As such they are
still very much input oriented conceived, along the traditional production lines. This new deviates from
a firm tradition of releasing aggregates only. It is both useful and legal, but nevertheless its legitimacy
in the eyes of the respondents may not be self-evident.
5.The playing field for Statistics Netherlands has changed hand in hand with the methodological approach:
a) Units (data subjects) contribute to statistical data collection in two roles: as respondents but also
(and increasingly) as the objects of administrations. The next section will give some examples of the combination of these two roles.
b) Since administrative data are primarily collected by administrative agencies, bilateral arrangements
have to be worked out.
c) Researchers and their organisations put pressure on government, parliament, and public opinion to
make access to microdata for research as easy as possible.
d) Legislation is available and further developed to protect the privacy of personal data. Apart from the
legislator, the national privacy authority (Registratiekamer) needs to be mentioned here.
6.In the remainder of this paper, I analyse the consequences of (1) the matching of survey and administrative data and (2) the release of microdata for research purposes for our relation with data subjects in the playing field mentioned above.4 In particular, I concentrate on the applicability and
utility of the concept of “informed consent”. In the second section I give a variety of recent examples of
the statistical use of the matching of administrative and survey data at Statistics Netherlands. I draw
them from the area of social statistics.5 It is the very variety of these practices that calls for a more
formal policy. I then describe, in the third section, frames of reference from (1) professional codes and ethical guidelines of professional associations, (2) international legislation, and (3) national legislation. From a strictly legal perspective, one could argue that there are no legal objections to either the
matching of survey and administrative data, or the release of microdata for research purposes. One
formal impediment turns out to be the conditions under which some administrative data is released by
its authorities to Statistics Netherlands. But from a broader ethical perspective, the legitimacy of
record matching and of the release of microdata is far from self-evident. Do respondents really understand what statisticians will do with their data and do they agree with such use? Some Dutch experiences are summarised in the fourth section. The fifth and final section summarises my findings
as of this moment.
7.It may be useful to observe here that survey response percentages in the Netherlands are well below averages for comparable surveys in other countries. The 1971 Population Census was the last survey to entail a response obligation for social statistics. In fact, one of the reasons why the Population Census Law was revoked since then, was popular resistance to the “Big Brother is watching you” perception. On the other hand, the Netherlands as a country is very much dependent on social research, modelling, and consultancy. The relative number of social scientists is high. And the importance of objective facts and research within the pillarised Dutch society is well documented. The Netherlands is probably unique in having three so-called planning agencies which assist, advise, and evaluate government policies in the economic, social-cultural and environmental areas (CPB, SCP, and RIVM). The population administration is very well developed. Nevertheless it has two particular characteristics: it is decentralised towards the six 4Jabine (1986) deals with the very same questions of planned linkages and release of microdata, within
the USA context.
5Privacy legislation refers to personal data. In statistics there often is separate legislation for economic microdata. But it is obvious that in many cases personal and company data coincide, as for companies without employees.
hundred odd municipalities in the Netherlands, and the personal identification number used in it is not the same as the one used by the fiscal and the social security authorities.
II.MATCHING DATA
8.The advantages of matching survey and administrative data are manifold (Kooiman and Nobel, 1997). Survey data can be ideally targeted to the research question at hand. But they are expensive in terms of the researchers’ budgets and processing time, and respondents’ time. Normally they pertain
only to a small fraction of the population. They may be hampered by non-response and by inexact answers due to the constraints of, e.g., human memory and social desirability. Administrative data, on
the other hand, is readily available for the whole population, in principle at marginal cost. The quality
of these administrations seems to vary, however, depending on the interests of those, whose data are registered, and those who use them within the administration. The registrations of the unemployment offices have a reputation of administrative pollution, whereas the Netherlands’ population
administration is excellent according to international standards.
9.With the combination of administrative and survey data one could in principle have the best of
both worlds. Registrations may serve the role of sampling frames, which makes statistical inference, including corrections for non-response, considerably easier. From rich registrations one could differentiate and target fieldwork strategies and operations in order to optimise response. Furthermore, data may be easily added to a sample survey from administrations because respondents do not have
them readily and exactly at hand.
10.If on the other hand one wants to use surveys in addition to statistical research based primarily
on administrative data, three roles can be distinguished: quality control; translation of administrative measurements into research concepts; and addition of missing data, be they records or variables. The
precise effect of these roles is to a large extent dependent on the correlation of the administrative data
with the target variables of the survey. If this correlation is almost zero, the administrative data are of
no relevance for the survey. On the other hand, if this correlation is almost unit, one can rely completely
on the administrative data and does not need a separate sample survey.
11.Now let us give some examples of matching of administrative data by Statistics Netherlands.
A sample of applications will be presented at the meeting, Joris Nobel.
It is easy and seductive to think of further applications. The use of fiscal data for routine matching of income variables with survey data is one of them. At least three practical and methodological reasons
are quite compelling: survey questions on income take a lot of interviewing time; moreover they induce
high item non-response; and finally there is a consistent downward bias in results.
III.FRAMES OF REFERENCE
12.Guidance may be derived from three kinds of sources that I will deal with successively.6 For statements about professional ethics we can turn to, e.g., the International Statistical Institute (1985),
the American Statistical Association (1989), and the American Sociological Association (1997).
Official statements of international relevance have been issued by the United Nations Economic Commission for Europe (1992), the Council of Europe (1997), and the European Union (1990, 1995,
and 1997). I will also briefly describe Dutch legislation (1988, 1996).
6I have no claim to completeness. Also my selection of texts may be somewhat arbitrary. Some Web addresses are given in the bibliography.
13.One should be aware of several qualifications. Codes and laws may have different levels of ambition. Jowell, the chair of the committee that drew up the ISI 1985 Declaration, made a distinction between an aspirational, a regulatory, and an educational code (Jowell 1981). He rejected the expression of unattainable ideas, and rules to govern behaviour based on the premise that there are universal (and enforceable) models of appropriate practice. Instead he preferred the latter kind of code that “would seek to describe and explain professional norms, expose inherent conflicts and give
guidance on possible approaches to their resolution.” Also, statisticians may be active in many other domains than official statistics, e.g., medical research, and many other activities than research, e.g., teaching and consultancy. It goes without saying that professional codes may have to pay attention to these other domains and activities as well. These qualifications are especially relevant as the origins of the theory of informed consent may be traced back to the Nuremberg Trials of the Nazi doctors’experiments.
III.1.1International Statistical Institute
14.The International Statistical Institute Declaration on Professional Ethics was adopted at its Centenary Celebration in 1985 after a seven year preparation by a Committee composed of, amongst others, Roger Jowell (chair), W. Edwards Deming, and Edmund Rapaport. The Declaration contains four sections, devoted to obligations to society, obligations to funders and employers, obligations to colleagues, and obligations to subjects, respectively. Of these, the latter section is almost as extensive as the former three taken together. The section about obligations to subjects is subdivided into six subsections. Of these, two pertain to obtaining informed consent and modifications to informed consent. The combination of these two subsections is longer than any single section of the other three sections. So without knowing any further, one might already conclude that the issue of informed conse
nt is both important and difficult for statisticians. It therefore seems wise to quote these two clauses in extenso:
4.2 Obtaining informed consent
Statistical inquiries involving the active participation of human subjects should be based as far as practicable on their freely given informed consent. Even if participation is required by law, it should still be as informed as possible. In voluntary enquiries, subjects should not be under the impression that they are required to participate; they should be aware of their entitlement to refuse at any stage for whatever reason and to withdraw data just supplied. Information that should be likely to affect a subject’s willingness to participate should not be deliberately withheld.
4.3 Modifications to informed consent
On occasions, technical or practical considerations inhibit the achievement of prior informed consent. In these cases, the subjects’ interests should be safeguarded in other ways. For example:
1) Respecting rights in observation studies. …
2) Dealing with proxies. …
3) Secondary use of records. …
4) Misleading potential subjects. …
15.Several arguments deserve attention. In Clause 4.2, the principle of obtaining informed consent is restricted to active participation of human subjects. In addition to Clause 4.2 the ISI Declaration contains a list of twelve “items of information that are likely to be material to a subject’s willingness to participate. … In selecting from this list, the statistician should consider not only those items that he or she regards as material, but those which the potential subject is likely to regard as such.” The ISI Declaration, however, also states that “One way of avoiding inconvenience to potential subjects is to make more use of available data instead of embarking on a new inquiry. For instance, by making
greater use of administrative records, or by linking records, information about society may be produced that would otherwise have to be collected afresh.” The text following Clause 4.3 explains that “A serious problem arises for statisticians when methodological requirements conflict with the requirement of informed consent.”
III.1.2American Statistical Association
16.The ASA Ethical Guidelines for Statistical Practice, published by the American Statistical Association in 1989, cover less than two pages. They do not mention the concept of informed consent explicitly. Instead, the ASA Guidelines do call upon statisticians to “inform each potential respondent about the general nature and sponsorship of the inquiry and the intended use of the data.”
17.The Guidelines are being revised at present and the draft guidelines are substantially longer. One (out of eight) general topic area(s) is devoted to responsibilities to research subjects, including survey respondents and persons and organisations involved with administrative records, as well as subjects of physically or psychologically invasive research. But the new clauses do not explicitly make mention of informed consent either. Extensive thought has been given to the right choice and phrasing of the new guidelines, as is evident from the various discussions at the Joint Statistical Meetings and in the various ASA publications, including its website. In fact, the notion of informed consent has been deleted from an earlier draft. It was argued that statistical practitioners can not personally accept the responsibility to “obtain” informed consent of human research subjects. The argument was advanced in the context of medical statistics where often statisticians are responsible for only a part of the larger research design. It was accepted, however, for the field of statistics in general.
III.1.3American Sociological Association
18.The ASA Code of Ethics, approved by the American Sociological Association membership in 1997, is quite elaborate and explicit. One out of twenty items concerns confidentiality, another item is explicitly devoted to informed consent. Both items take several hundreds of words. The two together make up a quarter of the full code. Informed consent is defined as “a basic ethical tenet of scientific research on human populations. Sociologists do not involve a human being as a subject in research without the informed consent of the subject or the subject’s legally authorised representative, except as otherwise specified in this Code.” Sociologists may seek waivers when the risks for research subjects are minimal or the research could not practicably be carried out were informed consent to be required. But such waivers require approval from institutional review boards or, in their absence, from another authoritative body with expertise on the ethics of research. Under such circumstances the confidentiality of any personally identifiable information must be maintained at any rate. In cases of doubt, consultation of institutional review boards is thought to be required, and confidentiality of any personally identifiable information is to be maintained a fortiori. In obtaining informed consent sociologists are to use language that is understandable to and respectful of research participants or their legal representatives. Also, they are to keep records regarding said consent.
III.2.1United Nations
19.In 1992 the United Nations Economic Commission for Europe accepted the Fundamental Principles of Official Statistics. The basis was laid by the Conference of European Statisticians. It attempted codification on behalf of the new democracies and their national statistical institutes in Central and Eastern Europe. Later, the United Nations Statistical Commission endorsed the principles. In its considerations the Resolution holding the Fundamental Principles recalls the ISI Declaration on Professional Ethics. Nevertheless, it does not mention the concept of informed consent explicitly. The two principles that come closest, are the following:
• Data for statistical purposes may be drawn from all types of sources, be they statistical surveys or administrative records. Statistical agencies are to choose the source with regard to quality,
timeliness, costs and the burden on respondents.
• Individual data collected by statistical agencies for statistical compilation, whether they refer to natural or legal persons, are to be strictly confidential and used exclusively for statistical purposes. III.2.2Council of Europe
20.The Council of Europe has accepted a Recommendation on the protection of personal data collected and processed for statistical purposes, in 1997. In its considerations the Council shows itself
explicitly aware of the necessity to balance the public need to know versus the individual right to privacy, but also of the advances in statistical methodology and information and communication technology over the past decade.
21.The Recommendation itself gives a number of requirements that should be met when collecting personal data for statistical purposes. When using administrative data for statistical purposes, proper notification should be given, unless this would be impossible or require unreasonable effort or provisions have been made in national law. If consent by the data subject is required, it should be free, express, and informed. The fact that the survey serves a legitimate goal may not outweigh the obligation to ask for consent, unless a serious public interest justifies such an exception.
22.Specific provisions have been taken to give legitimacy to proxy interviews and non-response analysis.7 Personal data may be collected from a member of the household rather than the data subject himself, depending on the nature of the survey. Data on data subjects that are necessary for the preparation and execution of the survey and data about non-response may be used, but only to assure the representative character of the survey.
23.Further provisions relate to the transmission of personal data for statistical purposes, i.e., the releas
e of microdata. A final feature is the call upon professions and institutions that produce statistics to establish and publish a code of ethics. One or more independent authorities should be given the task to ensure national upholding of the principles embodied in the Recommendation of the Council of Europe.
III.2.3European Union
24.Three pieces of European Union legislation bear on the production of official statistics, statistical confidentiality, and the protection of personal data. The oldest of these, Regulation 1588/90, made possible the transmission of confidential statistical data from the national statistical institutes to Eurostat, the EU statistical office. Under certain conditions it put aside national statistical confidentiality legislation that used to prohibit the release of national, confidential data to Eurostat. The text of the regulation does not mention data-subjects: Eurostat does not have a direct relation with them. The debate about this regulation has triggered many measures, including funding research into confidentiality methodology and software, however.
25.The EU Data Protection Directive (95/46/EC) went into force in 1995. It attempts to harmonise national legislation within the European Union so as to make it easier to exchange personal data witho
ut violating privacy constraints. One feature is the shift of attention from registrations to data.8 In the Directive personal data are defined as any information pertaining to an identified or
7These two activities should otherwise be considered to run counter to a strict application of the informed consent principle.
8As a consequence the Netherlands’ privacy authority, the Registratiekamer, will have to change its name.
版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。
发表评论