My comments on the paper are based on comparing and contrasting the paper’s findings and recommendations with my experiences as Chief Statistician of Canada. I had the opportunity, during my tenure, to develop amendments to Canada’s Statistics Act specifically intended to strengthen the autonomy of Statistics Canada.
While I agree that national statistical offices are key institutions in the democratic process, informing voter decisions on electoral issues, in reality I cannot think of a single instance when a national statistical office has been funded explicitly for this purpose. Statistical offices programs are usually funded under policy initiatives to provide needed information for that policy’s development, evaluation and administration. So policy departments are the key players in deciding what statistical programs are developed and executed. Informing citizens is, in effect, a residual use of information developed for other purposes, though I believe the residual to be more important than the primary use.
Statistical offices have little control over what statistical programs are developed and funded. What they must control, however, in order to ensure the credibility of the statistical program, is how the statistics are produced and disseminated. Hence, the requirement for a high degree of autonomy in their operations as the paper notes. Ideally, statistics developed for public use should be developed by the statistical office and the statistical office should be required to make all information produced public subject to confidentiality rules.
Statistical offices, unfortunately, have no special claim to exemption from budget reductions in times of budget restrictions or shifting government priorities. They should, however, defend themselves from disproportionate cuts and particularly those intended to intimidate or compromise key statistical programs. When cutting the spending of the statistical office, it is generally preferable to cut low priority programs rather than weaken all programs.
In Canada, the national statistical system is highly centralized in a single office that reports to Parliament through a Minister but without being embedded in a policy department. Small amounts of statistical activity continues in other departments and agencies but with no real independence and little transparency. If data needed to evaluate the work of the department is produced under the department itself there is an obvious conflict of interest.
As argued in the paper, the autonomy of a statistical office is best and most convincingly established by creating the office in statute law. Anything less cannot meaningfully protect the statistical office from determined politically motivated interventions. The statute of course must set out the scope of the subject matters to be addressed, the powers to compel production of records and response to its inquiries, the obligation to protect respondent confidentiality, and its right and obligation to publish its findings.
To establish the autonomy of the statistical office, all administrative powers relating to the office should be assigned to chief statistician of the statistical office or their delegates, including staffing, with the requirement to manage the office in accordance with the professional standards. The role of the chief statistician (head of the office) is critical to achieving and maintaining the office’s autonomy. Ideally, the chief statistician should be appointed, under the statute to a fixed, if renewable, term. They should be removable from office only for bad behavior through a public process. It is inevitable that the chief statistician will report to the government and legislature through an elected official. Canadian legislation provides that the minister must provide a public instruction in order to force the chief statistician to do something they are not otherwise willing to do, and that it belongs to the chief statistician to decide to require such an instruction be issued. These provisions dramatically increase the political cost of inappropriate intervention and, in Canada at least, are strongly dissuasive.
Another such device is the creation in the founding statute of an external advisory body on the health of the statistical program to the political appointee responsible for the statistical office.
For its autonomy and for its ability to protect the confidentiality of respondent information, it is vitally important that the statistical office have full control over its informatics systems and the security measures surrounding them. If government informatics systems are centralized and the statistical office is forced to use them, they constitute a means to control office activities and make it impossible to convince the public that their confidential information is genuinely secure, particularly from government police and security agencies.
As is hinted at in the paper, a useful adjunct to statute law in protecting autonomy, are robust, written and publicly available policies developed by the statistical office itself governing the subtleties of behaving in a objective and trustworthy way, such as the establishment and respect of preannounced release dates whenever possible, the release of un-preannounced data during electoral periods, provision of documentation on methods and data quality coincident with data releases, release of data deemed of insufficient quality for formal release, controls in place to protect confidentiality and practices regarding the brief prerelease of financial market sensitive data to fiscal and monetary authorities to allow them to prepare policy response (but not to propose changes to data presentation and analysis, or to delay the release). All of this helps to create a strong culture supporting the autonomy of the office. It allows staff to deal with officials of other departments and political appointees with great confidence.
Gaining and keeping public trust in official statistics in the new world of social media is increasingly difficult. Where we could once say that people are entitled to their own opinions, but not their own facts, this seems no longer to be a widely held view. Statistical offices and the governments that create them must do everything they can to earn public trust in official statistics.