No, it wasn't a change in the questions. It was that it went from a census to a sample. We know that people who are more privileged are more likely to answer those voluntary surveys than people who are less privileged, who can have low incomes and be racialized. As a result, the sample size was skewed. You had a lower sample of the racialized cohorts, and as a result, the data that were available and the data that we had to analyze were inconsistent with the other research and had some surprising results. We then had conversations with the officials at Stats Canada, and because the national household survey was an experiment, where they tried to adjust the data and correct the data so that.... I'm using non-statistical terms, but they did some statistical corrections to try to correct for the methodology, and those—
On June 8th, 2017. See this statement in context.