Dissemination of sensitive group-specific results Which fundamental principles of official statistics are applicable?

UNECE Statistical Division Heinrich Brüngger UNECE Statistical Division

Outline of presentation What is the problem with sensitive group-specific results? Crude methods for dealing with (avoiding) the problem Balanced approach: using selected fundamental principles and ensuring that such results receive a higher level of professional scrutiny The use of the 4th principle as safety net

What is the problem? Sensitive group-specific results (SGSR) from official statistics may reinforce or help create negative perceptions against certain groups On the hand, official statistics should disseminate « relevant « results Should the impact of results (not only SGSR) planned for dissemination have an influence on the decision by the NSO whether to disseminate them at all?

Crude methods Avoid collecting characteristics that can be used for delimitating sensitive groups If collected (also as part of administrative sources), ban the use in official statistics For exhaustive sources: define a very high threshold in terms of minimum number of persons/households for results to be released (higher than what is required for protecting confidentiality of single units)

Crude methods (ctd.) For samples: define a very high quality threshold for accuracy All these methods make it impossible to release some relevant aggregates, including in the context of tailor-made statistical services Preventive methods have a high cost in terms of restriction of access If regulated by law, one of the crude methods is likely to prevail

Balanced approach 1st step: use the same confidentiality, accuracy and quality standards as for all results of official statistics; if an SGSR does not pass one of these criteria, they should not be released at all (but not because they are sensitive) 2nd step: investigate from the professional point of view whether certain influencing factors that may not matter at the national level or for larger groups start to matter for small groups (whether sensitive or not)

Balanced approach (ctd.) If this is the case, fine-tune the indicator for small groups to eliminate statistical artefacts or structural effects before release as result of official statistics, and add carefully drafted comments (respecting the principle of impartiality) The unadjusted SGSR must remain available on request, but should only be handed out with a clear indication that a more relevant result exists and that the NSO has the right to react publicly against any misuse or erroneous interpretation

Balanced approach (ctd.) Balanced approach is ex-post, i.e. checking each small aggregate according to specified criteria and thresholds, but the 1st step is good practice of quality management anyway More resource-intensive, but ensures that most relevant results that pass the 1st step can be released in a form that fully reflects professional standards Simple disaggregation of national indicators may not be good enough for SGSR

Use of 4th principle: safety net against misuse by third parties Possible impact of releasing new results should never be a criterion for whether or when to release (professional independence), but it may influence the release product (comments) NSOs should have a defined policy when to make use of the 4th principle, and a defined mechanism for applying them Such public comments are released under the sole authority of the chief statistician

Use of 4th principle (ctd.) Occasional use of the 4th principle implies that the use of statistics is regularly monitored and checked for misuse/erroneous interpretation Use is not only referring to media, but also to government reports e.g. Discussion about national experience and policies in making use of the 4th principle, and about monitoring use from this perspective, might be a topic for further meetings

Questions for discussion Concerning the issue of release of small aggregates, do you have special rules or a special policy for SGSR? If yes, what are they? What type of comments did you add to the last release of a SGSR? What was the reaction by media and users? Were some of these interpretations erroneous, and if yes did the NSO react publicly?

THANK YOU FOR YOUR ATTENTION