How we keep data private

MSD is always improving the way we work with client information to make sure it’s kept confidential, while releasing the maximum amount of data to the public.

To ensure an individual’s information is safe the Ministry firstly anonymises (removes a person’s identifying information), then groups individual client information. This makes it difficult to identify a real person.

MSD then applies a range of confidentiality steps to protect the information.

The Ministry applies a cascading set of rules that tests how sensitive a data table is when considering its release. This can result in a data table being withheld or applying confidentialisation.

Two of the possible confidentialisation approaches are suppression and random rounding.

Suppression is only used for tables that are highly sensitive. Primary suppression removes all cells with numbers that are less than 6. Further suppression may be needed to avoid people successfully calculating the suppressed amount from other data.

Random rounding is applied to all publicly released tables and rounds all values to a multiple of 3, so a 5 could become a 3 or a 6. Random rounding is automated, so removes the risk of manual error in suppression and allows significantly more data to be published, increasing the value to the public.

This may affect the data people receive and, when seeking to interpret it, people should note:

• Rounding to a near multiple of 3 is applied to all cells in a table, including totals, but no cell or total will ever be more than 2 away from the original number.  Random rounding does not round down to zero. A value of one or two will always be rounded to three.
• Columns may not add to the totals but the totals will never be more than 2 away from the original number as the totals are calculated before random rounding is applied.
• Applying random rounding to the same number across different data requests may mean numbers that are slightly different. The difference will never be more than 2 from the original number. For large numbers the impact will be minor but the impact will be more obvious for smaller numbers.

A more technical explanation of this approach can be found on the  StatsNZ - Applying Confidentiality Rules webpage.