Canadian Census will overestimate prevalence of religious belief

Canada is currently conducting its census. One question on the long form (sent to 25% of households) attempts to identify respondents’ religious beliefs. However, it is worded in a manner that will likely underestimate the number of people who currently hold no such belief:

What is this person’s religion?
Indicate a specific denomination or religion even if this person is not currently a practising member of that group.
For example, Roman Catholic, United Church, Anglican, Muslim, Baptist, Hindu, Pentecostal, Lutheran, Presbyterian, Sikh, Buddhist, Jewish, Greek Orthodox, etc.
For additional examples of denominations and religions, visit

    • Specify one denomination or religion only


    • No religion

The Centre for Inquiry is encouraging “atheist, agnostic, non-believing, freethinking, and spiritual-but-not-religious Canadians” to answer “No religion” to this question, which I think would provide more accurate data.


It’s common for long term survey and census questions to remain unchanged so that answers are comparable over time, even if the cultural definition may have moved. We see this in the Pew survey question about religious Creationism, even though most people answering “Yes” aren’t Young Earth creationists.


That is not the case here. The question was worded differently the last time it was asked, in 2011:

22 What is this person’s religion?

  • 1: Specify one denomination or religion only.
  • 2: No religion

1 Like

Hmm, I wonder if this could cut both ways:

  1. By asking for a specific denomination/religion it might push the “spiritual-but-not-religious” / “I don’t do organized religion” / None crowd into the “no religion” pile.
  2. By adding “even if this person is not currently a practicing member” it seems like it could be pushing “cultural” identities (non-observant Jews, northern European Lutherans, southern European Catholics, etc.) into the “religion” pile.

I have no idea what the relative numbers would be between those though.


In those situations, wouldn’t it make more sense to parallel the old question, and a new, more culturally relevant, question for a few years, until you develop a sufficient minimum of time-comparability for the new question (at which point you can abandon the old one), rather than sticking with an increasingly less relevant, but comparable, question? Surely a less comparable, but more relevant, question is of more use than a more comparable, but less relevant, one?

1 Like

Spoken like a true Statistician! :slight_smile:

The answer is yes, of course, and I think the major polling companies are probably doing this, but at a very slow pace (over decades). Change things too fast and no one will understand (or believe) the results, and most of the audience is not too statistically savvy.

1 Like

So they leave it buried in the appendices, where only statisticians will notice it, until they’re ready for the changeover. :wink:

1 Like

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.