I think the census, as it stands, underestimates the number of multiple responses to questions that exist for the reasons I mentioned. The other two questions provide for multiple responses. The question concerning the mother tongue, which comes immediately afterward, does not provide for multiple responses. The message sent to respondents is that a choice must be made. The instruction then says that the language most often spoken must be indicated. This seems to discourage multiple responses. Also, the Statistics Canada data shows that there are two to three times more multiple responses. Remember that, until 2011, the short questionnaire contained a single question on the mother tongue. In this case, there were more multiple responses than when they're juxtaposed with the other two questions. This seems to be a discouraging factor.
On February 14th, 2017. See this statement in context.