Te Taunaki Public Service Census is a survey of employees of departments and departmental agencies run by Te Kawa Mataaho | Public Service Commission (the Commission). The survey contains a wide range of demographic questions (sexual identity, religion, disability, caring responsibilities) that are not captured in normal data held by an agency about their employee. It also asks about the experiences of employees.
Information that is collected through the survey will be used for statistical or research purposes and will not be published in a form that could reasonably be expected to identify the individual concerned.
Processes and Practices
The Commission has procured a New Zealand-based research provider who will deliver the survey online. The research provider will collect responses online from public servants within participating agencies. The provider has Data and Sample Security Policies and Practices in place to ensure they meet their requirements identified in the New Zealand Information Security Manual.
Using administrative information provided by agencies
To conduct the survey, agencies supply the Commission with information about their staff before the survey starts. This information is temporarily combined with a unique identifier by the Commission. This ensures that no one inside the agency or inside the Commission can identify who has completed the survey and who has not.
Some variables are more accurate when sourced directly from agency systems and removing them as questions keeps the survey shorter reducing burden on participants. Therefore, the following information will be linked to survey responses:
- Business unit (for units with a minimum of 100 people)
- Management level (tier)
- Region
- Contract term (permanent/fixed term)
- Full time equivalent (proportion of full-time hours worked)
- Occupation (ANZSCO code)
- Start date (employment start date in organisation)- this is used to calculate tenure in organisation (e.g. 3 years to less than 5 years) which replaces start date in the final dataset
- Salary
Two further variables (email address and preferred/first name) are provided for the purpose of sending out the invitations to the survey but are removed before the dataset is finalised and provided to the Commission by the research supplier.
Only the unique identifier, preferred/first name and email address are sent to the research provider prior to the survey being conducted. At this same time preferred/first name and email address is removed from the dataset the Commission holds until the survey closes.
When the survey closes, the research provider will supply the list of unique identifiers for the survey respondents that participated to the Commission. The Commission will provide the variables associated with each unique identifier for those who participated in the survey to the research provider.
The Commission will then delete the dataset containing unique identifier and the information supplied by agencies, as well as the list of participants from the research provider. This ensures that the Commission cannot identify who participated in the final dataset, and no information is shared with the provider about people who chose not to participate.
The final dataset the research provider gives the Commission will have gone through a checking process to ensure the dataset is de-identified, accurate and complete. All reporting is generated from this final dataset.
This process was consulted with privacy officers in all participating agencies, the office of the Government Chief Privacy Officer, and with external experts through the Data Ethics Advisory Group managed by StatsNZ.
Reporting
System and agency level reporting on the Commission’s website will use a quality assured dataset where personally identifying information has been removed. Agency and demographic breakdowns will protect privacy through small count suppression, requiring a minimum number of responses, see Appendix for more details. The Commission will use its internal data management and release protocols and guidelines to ensure reporting is accurate and that no confidential information is reported.
In addition, a web-based reporting tool (“portal”) will be made available by the research provider to each agency. In this portal, agencies will be able to view summary level information about their results and compare results for their agency with results at a system level. Privacy will be protected through small count suppression requiring a minimum number of responses.
Large agencies will also be able to see results in the portal, for any organisational/business units they have with at least 100 people. Results at an organisational/business level will exclude demographic questions, which will only be reported at the agency level. Organisational/business units must have a minimum of 100 people to be included in the dataset, and therefore in reporting. See Appendix for more detail on minimum numbers required for reporting.
Data access
Any access to the dataset must be approved by either the Manager, Data Team or the Deputy Chief Executive, Strategy, Policy, and Integrity. Approved access is granted after reading the Access, Security and Reporting Protocols and completing a Confidentiality Agreement Form.
For Commission employees
To ensure privacy of individuals is protected, the Commission will restrict access to the unit record dataset stored in the Commission’s secure data warehouse. Access will only be used to manage the data or produce aggregated results for statistical or research purposes.
Access to de-identified individual level Te Taunaki data for bona fide researchers and analysts from other agencies
Government analysts and other bona fide researchers may request onsite access to de-identified unit record data for research purposes. Access will only be made available to relevant data required to answer the research question(s). If research questions require data that would expose individuals to risk through having their identity known, the research project will not be approved.
External researchers, including other government agency researchers, who are requesting access to de-identified unit record data must submit a research proposal containing the following:
- Information on the researcher, their affiliated institution(s), and research experience in the topic of study
- Summary outlining the proposed project and its benefits, and why the research is best completed using Te Taunaki data
- Proposed design/analysis, including details on what variables will be used, and how confidentiality will be maintained for the survey participants in any reporting
- How the analysis and reporting will support Māori data sovereignty principles
- Reporting/publication plans
Access will only be made available to limited subsets of the data, as required to answer the research question. This access is only available on site at the Commission’s offices in Wellington. The Commission will continue to be the owner of the data set, and no further copies will be made or shared.
Access to summary level information
Departments and departmental agencies receive summary information via the portal that has privacy rules applied by the Commission or Research New Zealand for all quantitative questions in the survey. Agencies are expected to limit access to their Census portal to those who require it for their role. Agency staff are not authorized to communicate findings to their agency or anyone outside it prior to public release on the Commission’s website.
Free text comments
De-identified free-text comments will be reported verbatim to a nominated contact within each agency, who has signed a confidentiality agreement. De-identified here means that the comments will not be linked to the other response fields when provided to the agency.
Agencies will be able to request comments for subgroups where the number of comments is large enough to meet the minimum reporting requirement i.e. where there are 30 or more comments from members of that group. These are the same rules for handling free-text fields that the Australian Public Service Census uses.
Communicating results
The Commission will publish results for the survey at the agency and system level. Agencies are to communicate the results of the survey to their staff in summary form and engage with staff about how they will use the results.
Appendix
Public Service Commission Data Release Rules
|
Counts (e.g. Headcount, FTEs) |
Magnitudes (e.g. Salary) |
Likerts* |
|---|---|---|
|
Suppress cells with counts below ‘4’ (this takes the possibility of collusion into account). Make them appear like cells with zero counts (i.e. either with a zero or by leaving blank). Totals and percentages should be calculated using supressed and rounded numbers. Additional rule if needed Apply random rounding to base 3, if suppressed values can be determined by looking across tables or a highly sensitive variable is involved (e.g. income). Note any use of random rounding. |
Suppress cells with counts below ‘4’ (this takes the possibility of collusion into account). Make them appear like cells with zero counts (i.e. either with a zero or by leaving blank).
Additional rule if needed Suppress cells with counts below ‘10’ in the following circumstances: The use of multiple variables (e.g. occupation by department) or detailed variables (e, g. occupation at very detailed levels) means the risk of identifying an individual’s data is high and the value is highly sensitive (e.g. income). |
Results for questions with counts of less than 10 are suppressed with a ‘s’. This rule is applied at the question (denominator) level, not the response (numerator) level. This is the approach taken for Likert questions in the Australian Public Service Census. Additional rule if needed Additionally apply the below ‘4’ rule to individual responses or random rounding to base 3 if the suppressed values can be determined by looking across tables |
* A likert question is one where someone is asked to rate their view on topic on scale (e.g. 5-point scale from strongly agree to strongly disagree). Typically, it is not possible to identify people from their response to a likert question (e.g. whether they were satisfied with their work-life balance). If it is possible (e.g. how much te reo they use at work, then suppress cells with counts below ‘4’ as well.