One of my Faculty Members has a question about people who list multiple ethnic origins in the census. He has specifically asked that I forward his question to those who are knowledgeable about the Census.
"I am wondering whether the multiple origins option results in duplication, or whether some method is used to prevent this. For example, if an informant reports that her origins are French and Aboriginal, does she appear in the aggregate data twice, once as 'French, multiple origins' AND a second time as 'Aboriginal, multiple origins'? If she would only be aggregated into one of these categories, how is the determination made? I would like to have an answer to this that goes back to the introduction of multiple origins on the census (is that 1981?)."
I wonder if this client is referring to a variable in the PUMF file? If the PUMF contains a transformed ethnic origin variable from a multiple response (a variable for which the answer of a person can match one or more valid response(s)) to a standardized variable (a variable for which the answer of a person can match one and only one valid response), then there would be no issue of double counting because the category of French and Aboriginal origins refers to a single category -- and if the respondent had French and Aboriginal origins, that person would only show up in the data once in this category. For this type of variable, a person is matched to one and only one category of the variable and it is possible to add up the categories to obtain the total population. (Look at the variable "DethNic" in CAPSS or the e-dict).
However, if the client is simply referring to the ethnic origin variable as it is disseminated in standard tables on the Internet site, then there is indeed double/triple etc. counting occurring. You can not add up all of the ethnic origins to obtain the total population because people will be included in each and every category that they reported their ethnic origin to be, the sum would thus be greater than the total population count. For example, if the person reported French and Aboriginal origins, the person will be included in the total French origins as well as the total Aboriginal origins. That is why we have the Total, single and multiple indicators of the ethnic origin variable.
The ethnic origin variable is one of a few "multiple response" variables in the Census. (Other multiple response variables are Mother tongue, home language or population group). A multiple response variable is a variable for which the answer of a person can match more or more valid responses. Hence a person will match more than one category if that person has answered more than one response category to that question. The sum of all ethnic origin response categories will not add to the total population because respondents were permitted more than one response. We do not do any aggregations into one ethnic origin category if the respondent reported multiple origins. We could create single and multiple response categories to transform the ethnic origin variable from a multiple response variable to a standardized variable, but even in this transformation, we do not choose one or the other category to create the final categories, rather, we group the types of responses into similar categories. (see the
I suggest that the client consult the 2001 Census ethnic origin user guide for more information on the variable.