The New Zealand Conference on Database Integration and Linked Employer-Employee Data.

• Author: Spittal, Matt

The New Zealand Conference on Database Integration was held at Te Papa, Wellington, in March 2002. The purpose of the conference was to provide a forum for a wide range of international and local speakers to discuss the added value associated with database integration. Key themes to emerge from the conference were: (1) that international experience has shown that integrated data can provide decision makers with policy-relevant data in a timely fashion without increasing respondent burden; and (2) that issues of privacy and confidentiality must also be addressed when integrating data.

Data integration refers to the linking of two or more data sets, usually to obtain longitudinal information. Broadly, this can take the form of linking administrative data, which is essentially non-research data collected in the course of programmatic activities for client-level tracking, service provision, or decision making; and linking survey data, which is collected for answering specific research questions. Integrated data offers a number of advantages for research and policy advice over cross-sectional data, such as a richer insight into phenomena such as the interactions between employers and employees; between unemployment, training and employment; and between health, education and economic outcomes. Data integration can provide decision makers with information relevant to policy advice without increasing respondent burden, and it offers the possibility of addressing questions that cannot be answered by cross-sectional data.

An important component of research into any source of data is to understand the limitations of the data. With respect to integrated data, several limitations are evident. Administrative records, which form the basis of many integrated data sets, are not developed for the purpose of research and may be unsuitable for various reasons, including:

* a lack of adequate control variables;

* the fact that all outcomes of interest are not measured;

* the data may only be available for the periods that the person is in a programme; and

* the undetermined reliability of the administrative data.

Sometimes administrative data is difficult to access because of confidentiality issues, such as obtaining informed consent. Many of these problems may also occur with the use of survey data (e.g., incomplete records).

The conference provided a forum for international and national speakers to discuss the usefulness and applicability of data integration...