96. Cross-country comparisons are important for understanding the effectiveness of policies and programmes adopted by individual countries. Governments, in particular, find this useful for policy evaluation work. The benefits of providing access to anonymised microdata to researchers working on cross-country comparisons and international agencies are clear but there are also risks (e.g. passing on microdata without permission). With the exception of Eurostat, staff from international agencies are not subject to any national or international legislation other than the applicable staff rules of the organization. So some care has to be taken. The main difficulty is that there is more limited scope for retribution against breaches committed by staff of international agencies or researchers living in other countries. But, on the other hand, the probability of identification is much lower (as long as the researcher does not pass the microdata back to a third person in the host country). A further difficulty is that many countries do not have the legal authority to provide data to international agencies or researchers outside their country.
97. These guidelines have suggested moving towards a risk management rather than a risk avoidance approach in the provision of microdata. Risks are lower or perceived to be lower if the recipient agency has credibility and a trustworthy reputation. Risks are also more easily justified if the purpose for which the microdata are provided is in the interest of the country providing the data. This may be because the data are part of an international study being undertaken by an international agency or a respected international research collaboration (e.g. the Luxembourg Income Study).
98. Globalisation is increasing the importance of such international studies. It should be legitimate for NSOs to be able to support these types of studies through the provision of microdata. But it should be an enabling provision (i.e. the NSO should be able to decide whether to provide the data or not) that also specifies appropriate safeguards and conditions of release. NSOs should be more prepared to provide anonymised microdata where the risks are lower and the benefits are higher.
99. What are the options for researchers' access to data sets from other countries? How can international agencies obtain access to microdata for statistical and research purposes? The options include:
- (i) the data are collected directly by the international agency (or researcher) or through intermediaries (e.g. a specialist survey organization) in such a way that the forwarding of microdata is made known at the time of data collection:
- (ii) Public Use Files where they exist;
- (iii) Licensed Anonymised Microdata Files, where countries are able to do this;
- (iv) Remote Access Facilities with appropriate safeguards;
- (v) collaboration with a researcher based in the NSO or the NSO's country, who has access to the microdata.
100. From the point of view of microdata access, surveys of type (i) are preferable for international researchers. The PISA study is a good example. However, generally speaking, these data will not be collected under the statistics legislation operating in the various countries. The quality of the microdata, and particularly response rates, could suffer for some studies. This will depend on the type of study and the reputation of the organization undertaking the study. It also depends on the commitment of the data collectors in individual countries. This is a factor that international researchers need to consider before deciding to take this approach. There may be trade-offs between access and quality.
101. In some cases, it may be possible to satisfy the requirement of the study by providing very detailed data for analysis, but not microdata. This approach is used in the International Comparison Programme and the OECD/Eurostat Purchasing Power Parity study.
102. With respect to approach (ii), Public Use Files are only available for some countries. Licensed anonymised microdata files (i.e. approach (iii)) may be an option if not restricted by the legislation of the NSO. If NSOs are able to provide microdata in this way, factors that might be taken into account are:
- trust in the researcher and his or her institution;
- whether the study is of importance to the country; and
- whether such release would be consistent with promises made to respondents at the time of data collection.
103. It is likely that many countries will feel more comfortable releasing to particular international agencies or specific research projects rather than to the international research community more generally. Also, there may be some conditions that apply to particular researchers. For example, some countries may only feel comfortable releasing to researchers via the NSO of the researcher's home country. But in all cases of microdata release, it would be good practice to release only subject to specific conditions using either an undertaking or memorandum of understanding. Some of the conditions might include:
- limiting access to particular divisions of the international agencies and prohibiting the passing on of data to others;
- limiting the purposes for which the microdata could be used without permission;
- agreeing that the microdata must be returned on request (e.g. it may contain errors);
- allowing the opportunity to comment on published material which uses the microdata; and
- making clear the consequences of not following the conditions of release.
104. The most effective way of dealing with breaches of conditions of release will be to suspend further releases. The matter could also be taken up with more senior staff in the institution. For international agencies, this could be through diplomatic officials for more serious offences. But the key point is that breaches cannot be allowed to take place without some reaction. Otherwise, they will simply be repeated.
105. For many countries, the use of Remote Access Facilities (i.e. approach (iv)) may be the preferred route to provide access to international researchers. Under such arrangements, there are more controls, and the position of NSOs on international access to microdata is more easily defended if challenged. However, the usability of these arrangements for international access still needs to be improved. Experimentation is important.
106. There is another possibility. International researchers, including the international agencies, could work through networks of national researchers to achieve their aims (i.e. approach (v)) such as specified analysis against the microdata. Indeed, these national researchers could be located in the NSO for international studies of particular importance.
107. NSOs will need to decide whether they can provide access to international researchers or not, taking into consideration the range of issues discussed in this section. They should keep in mind that a risk management approach is being encouraged. For some research applications, the benefits may justify the risks involved, as long as the arrangement is legal. The risks may be lower for some institutions than for others. NSOs will also need to decide the most appropriate form of access in collaboration with the international researcher. To help ensure consistent decision making, countries should develop guidelines for access to international researchers and international agencies that are consistent with their own legislation. Or they may decide to change their legislation to enable access for justified cases.
108. Case Study 19 describes the arrangements used in the OECD PISA study and Case Study 20 describes a policy on international release of microdata for Australia.
109. The linking of data sets, whether by exact or statistical matching, can add considerable value to them. It can facilitate a much greater range of analyses. Health research, in particular, is an area where linked data sets can be of particular value. It is an appropriate function for NSOs to be involved in the linking of data sets for statistical purposes.
110. Increasingly, researchers are looking to utilise linked data sets that include links with the data sets of the national statistical office or other statistical agencies (including the population census in some countries). The statistical agency has to be the custodian for these linked data sets. There may also be situations where it is the preferred custodian of linked data sets even when the data sets come from outside the statistical agencies, because of the safeguards and public trust that already exist.
111. While there are clear benefits in data linking, there are also risks, particularly if the custodian of the linked file does not have confidentiality protection that often exists with the NSO. Identification risks are increased with linked data sets. Perceptions are also important. Studies in many countries show much public concern about linking databases. It is particularly important that the four principles outlined in Chapter 3 are followed for linked data sets.
112. In those countries where they exist, Privacy Commissions or equivalent bodies should support the arrangements for data linking.
113. Case Studies 21 and 22 describe how Canada and Sweden respectively manage data-linking arrangements in their countries.