Why are some estimates missing from current or previously published Census of Agriculture tables?
Estimates may be missing from Census of Agriculture data tables for one of the following two reasons:
- Data suppression: One of the methods used to protect the information of individual members of the population is data suppression. This method was used for the 2016 Census of Agriculture. However, in 2021, data suppression was replaced with a new method called random tabular adjustment (RTA).
- Data quality: For the first time, the 2021 Census of Agriculture will publish a quality indicator for most of its estimates to account for the degree of uncertainty because of non-response, data processing and RTA in individual estimates. This indicator takes the form of a letter between A and F, where A-level estimates are considered to be the most reliable, and F-level estimates are considered to have so much uncertainty that they are too unreliable to be published. When an estimate cannot be published, the cell in the table will simply show the letter F.
What is a data suppression technique?
To protect sensitive statistical information, Statistics Canada typically uses suppression techniques. These techniques involve suppressing data points that can directly or indirectly reveal information about a respondent. This can often lead to the suppression of a large number of data points, significantly reducing the amount of available data. In Statistics Canada data tables, cells suppressed for confidentiality reasons are marked with an “x.”
What is random tabular adjustment?
Random tabular adjustment (RTA) is a new method being used by Statistics Canada to balance the need for more high-quality data outputs for users while protecting the confidential information of respondents in the release of economic data estimates. Using RTA, Statistics Canada can identify sensitive estimates that may reveal information about a respondent and randomly adjust their value instead of suppressing them.
How does random tabular adjustment differ from other data suppression techniques?
Random tabular adjustment (RTA) improves the utility of economic data tables released by Statistics Canada. While traditional suppression techniques use rules similar to RTA to determine whether a cell contains sensitive information, they will suppress, or not publish, a sensitive cell. With RTA, cell estimates can still be released as individual values are not disclosed. Using this method allows Statistics Canada to increase the amount of useful data it can publish while ensuring confidential information remains protected.
How does random tabular adjustment work?
Random tabular adjustment (RTA) identifies sensitive estimates and randomly adjusts their value, or adds noise, so an estimate remains confidential, allowing it to be released. In other words, instead of “suppressing” the data, the estimates are “perturbed.” The size of the adjustment is calculated to protect the confidentiality of the individual responses that contributed to the estimate.
After adjusting the value, Statistics Canada assigns a quality indicator (A, B, C, D, E or F) to the estimate to indicate the degree of confidence users can have in its accuracy. This indicator accounts for uncertainty throughout the data collection, processing and confidentiality steps; it is not just for uncertainty because of RTA.
Will random tabular adjustment impact all estimates in the same way?
Random tabular adjustment (RTA) will not impact all estimates the same way. RTA looks at every cell individually to make a determination as to whether or not it contains sensitive information. While most Census of Agriculture cells do not contain identifiable information and will not need to have RTA applied, a number of cells will have RTA applied as they contain information that could be directly attributable to one or more of the individual values making up the total estimate in the cell.
When RTA is applied to a cell, the program will make a determination on the amount of unique noise that will be added, and users will be unable to identify whether RTA has been applied to a cell or not. In this way, RTA provides a measure of protection for all cells in the table.
For the Census of Agriculture, RTA has not been applied to the following tables:
- Characteristics of farm operators: age, sex and number of operators on the farm, Census of Agriculture, 2021 (Table 32-10-0381-01);
- Characteristics of farm operators: farm work and other paid work, Census of Agriculture, 2021 (Table 32-10-0382-01).
In addition, RTA is not applied to count estimates, such as the number of farms with a certain characteristic.
What are the advantages of random tabular adjustment?
Random tabular adjustment allows for more data to be released, increasing the utility of data tables. Instead of “suppressing” the data, the estimates are “perturbed,” meaning that the sensitive information is randomly altered to protect the confidentiality of the individual responses contributing to the estimate. This is done by adjusting the estimate in question so a precise value cannot be assigned to an individual contribution.
Another added benefit of this method is that it does not affect cells that are not considered sensitive. Only sensitive cells and their aggregates are affected. In other data suppression techniques used by Statistics Canada, some cells have to be suppressed to protect the confidentiality of another cell. For example, if one part of a total is suppressed, another part must also be suppressed to ensure the confidential cell cannot be calculated from the total.
Where can I find more information on random tabular adjustment?
For more technical information on random tabular adjustment, please see Disclosure control and random tabular adjustment by Mark Stinner from the Statistical Society of Canada’s 2017 annual meeting.