How the Census is Used in Canadian Elections
[Welcome, National Newswatch readers !]
Late one night last month on Twitter, a political party activist and journalist David Akin were debating how many muslims there were in Fort McMurray, Alberta, whether that number justified the local demands for a new mosque, and how recent the data was. I was called in by one of them when they wanted to know if the census religion data was from 2006 or 2001. In fact, the data on religion has not been collected since 2001 I was able to report, although data on many other new topics was reported in 2006 as part of the federal electoral district profiles.
Any experienced political organizer worth their salt can rhyme off the relevant census metrics in their riding from memory: the percent of tenants vs. home-owners, the proportion of immigrants, what languages are spoken and in what number, how many households there are, the age profile of the riding, the age of the housing stock, the median after-tax household income, the proportion of common-law versus traditional marriages, the ethnic origins of its residents, the number of first-generation Canadians, and of course the unemployment rate.
Elections Canada would use population numbers as part of their riding profiles, the census dissemination areas to help determine polling division boundaries, measures of transience and mobility to decide what areas need a special enumeration, and language data to determine how to deliver public education campaigns about voter registration to first-time voters, or the education data to estimate how many special ballots might be required under the special voting rules for post-secondary students.
Candidates use the data to decide how many brochures to print, and in what languages they need to translate them, what polls it’s safe to send women canvassers into at night or what polls canvassers will need to visit by car or would have more success reaching by telephone, what grade-level to write pamphlet copy in, and what issues to emphasize locally from out of the party’s national platform.
And the national parties not only make use of census data directly, but benefit from its use as the base of so much research into public opinion (for example, to properly weight sub-samples of larger samples) and marketing research.
As Conservative-affiliated pollster Greg Lyle of Innovative Research explained to Michael Valpy of the Globe and Mail during the 2008 election campaign, all political parties make use of geo-demographic and psycho-demographic data and surveys. This means that Census data, broken down to the smallest geographic units, is correlated with survey data on values and attitudes, a discipline known as psychographics. Psycho-demographics works from the assumption that people with similar values and beliefs (psychographics from surveys) also tend to live near one another (demographics from the Census), and it is now a staple of marketing strategies, including political ones.
The Conservative Party, then under the direction and analysis of marketing guru Patrick Muttart, put this data and approach together with their own vast resources to create a database assembling “geo-demographic and psycho-demographic surveys, huge-sample polling, and personal contacts made with voters through direct mail, e-mail, telephone calls and FRAN [Friends, Relatives, Acquaintances, and Neighbours] contacts”. According to their financial statements from 2003-2004, they invested about a quarter of a million dollars to building the information management system called CIMS. Susan Delacourt reminded us of its capabilities just last week.
In doing so, they elevated the practice to the most advanced level yet seen in Canadian politics. They knew which ethnic groups were more likely to support the Conservative Party (Chinese, South Asians, and Jews, amongst others), and not only what ridings to find them in, but what neighbourhoods. They “clustered” groups of related psychographic traits into personalities who were ranked according to likely Conservative support: Zoey, the urban latte-sipping organic-eating twenty-something, was terra unproductiva, while “economically beleaguered ‘battlers’ and a broad spectrum of ‘aspirational voters’ wanting more material gains for themselves and their children and feeling ripped off by the state, elites and big business” were the prime targets, such as “Steve, who owns his own business, and Heather, in their 40s with three children living in the suburbs”. Families with three or more children were more likely to vote Conservative, as were the children who went to community college rather than university. Catholics, historically most likely to support Liberals, are now increasingly likely to support the Conservatives, particularly those who attend church regularly, as are many new Canadians.
You might say that the Conservatives have, to date, benefited the most, electorally-speaking, from the very Census data their government now wants to stop collecting. It was employed to help them win the most seats in the 2006 general election, and to increase their seat count again in 2008.
In addition, the 2011 Census will also be used to help determine the boundaries for the next Redistribution of federal electoral districts (the next “Representation Order”). But now, it will only be able to take into account the six basic questions asked on the Short Form Census. [OK, technically there are 8, but two of them are the individual's name which is segregated once the data are entered and validated, and whether the individual agrees that their name can be re-associated with their census data in a hundred years for genealogical research purposes.]
The six questions on the mandatory Short Form questionnaire (opens PDF) which is given to 4 out of 5 Canadian households every five years are: age, sex, mother tongue, marital status, common-law status, and relationship to others in the household. Every other question is found on the mandatory Long Form questionnaire (opens PDF) which is given to 1 out of 5 Canadian households every 5 years, including other language questions (language spoken at work, or in the home), how recently you moved into your current address, whether you’re a tenant or homeowner, the age of your house, occupation, education, place of birth of your parents, hours spent providing care to children and/or the elderly, labour market information and household income, amongst others.
This data is tabulated to provide profiles of various cities and towns which are assiduously consumed by their municipal planning departments and local chambers of commerce, and also profiles of federal electoral districts, which you can find with handy pre-calculations and rankings on this website, as part of the riding profile pages of every 2003 representation order riding (for example, here are the Census Metrics for Parry Sound – Muskoka, ON), or on the “Search the Census” page which lets you rank ridings according to the Census metric of your choice.
I could not imagine building a database of election statistics without adding the census data, and when I did Macleans columnist Paul Wells was kind enough to call it “pure electoral geek heaven“. Census data is not only a staple of developing effective political strategies as he demonstrated in his book Right Side Up (see also Harper’s Team by Professor Tom Flanagan), but also for analyzing their success afterwards (for example, in the Canadian Election Studies). Indeed one of the first political books I ever read, and which was probably more responsible than any other for my budding interest in the field, was The Discipline of Power by Jeffrey Simpson, which contained a fascinating census political profile of Ottawa Centre for the 1978 by-election, written by then-Conservative pollster Allan Gregg.
I don’t take too many positions on issues in this blog, but I am going to state my strong belief that more, valid data is better for my field of study, and for our country as a whole. We obtain many, many rights, services and privileges from our citizenship in Canada, in return for which we have four duties: cast our ballots, obey our laws, pay our taxes, and complete our census surveys. While I recognize that there are some who will disagree with me on this point, I kind of want to ask them what they’re doing at a website all about data in that case.
To be crystal clear about exactly what will be lost if the Mandatory Long Form is to be terminated and replaced by a Voluntary National Household Survey, I have now added to Paul’s heavenly servings of census data on the riding profile pages, an indicator of which form the census measure came from (LF vs. SF). And on the newer “Search the Census” tool, I’m also showing which types and groupings of census measures come from which form, and which measures that includes.
For example, if you go to the Census search page here, and select “Language” as the Census Type from the first drop-down box, you’ll instantly notice that 3 of the Language Census Metric Groups come from the Long Form questionnaire, while just 1 comes from the Short Form. Similarly the first 5 Census Metrics under Language come from the Short Form, while the other 21 come from the Long Form.
The forms have been added so you can clearly see what valid data will be lost in 2011, and by implication what comparability to earlier years will be lost as well. Of the 285 available Census Metrics for the publicly available StatsCan Federal Electoral District profiles, across the 2001-06 censuses (including the relevant denominator measures), fully 249 are derived from the Long Form questions, while just 36 are based on the Short Form questions. Making the long-form questions voluntary means that they will no longer be statistically valid, since the self-selection inherent in voluntary surveys means they can no longer be said to constitute a representative sample. It’s like running a freep-able Internet poll … only spending millions and millions of dollars to do it.
If you want to express an opinion on that decision, here are several links to help you do that:
- The Government’s Digital Economy Consultation topic on saving the mandatory Long Form census questionnaire (you can vote up or down). At +305 it was running in second place for popularity as of Sunday night. But hurry: the consultation ends on Tuesday, July 13!
- An online petition to save the mandatory Long Form census questionnaire.
- You can find additional resources and information at the datalibre.ca website.
Tags: Census Data
“I’m hearing tonight that the announcement of a politically experienced candidate for Sydney – Victoria, NS is expected within the month.”
Let the record show the former Tory Premier Rodney Macdonald’s provincial seat overlaps with Sydney-Victoria.
Hi William,
I’m guessing that you intended that comment for this earlier blogpost on the Conservative nominations.
Thanks for raising and discussing this issue. I have voted and made a submission to the federal site. I just don’t understand “these people” and what they are thinking. I certainly don’t think they are being clear on why the change is necessary or what evidence there is to support the change. Thanks again.
I’d add jury duty to your list of responsibilities.
I think without question the data is useful. (Although there are legitimate questions about people simply making stuff up, guessing, or organized campaigns to get more services like the one where bilingual francophones claimed to have less english skills than they actually did).
The problem is that the usefulness of the data misses the point. Your answering a question of morality with a statement on utility.
Just so everyone understands where the objection comes from a certain segment of society views the census as overly intrusive. They don’t want to have to share private, personal information with the government.
For some reason this group’s opinions have carried the day around cabinet. One wonders whether the Tories did private polling on this issue ? Is this Tony Clement being his libertarian self ?
Do average Canadians care about this issue or is it just the elites ? Does it gain the Tories money, volunteers, and support from the base while losing them voters they never had anyways, the highly educated set ?
Lots of questions to consider as this issue unfolds.
Shadow, I accept your amendment to add “jury duty” to the list of citizenship responsibilities as a friendly one, without hesitation. We should probably add “giving evidence” when called to testify or assist in a policy inquiry, as well.
Minister Clement, when he was still commenting on the issue, did say that they hadn’t done any polling on the issue. I must say that if there had been such a groundswell of support for that position, we surely would have heard something more of it by now. Instead, conservative columnists like Colby Cosh and Andrew Coyne have come out on the side of collecting the data validly.
I also found Minister Strahl’s defence of the decision rather tepid last week, and Conservative blogger Stephen Taylor is saying tonight that he was “Disappointed w comms on census issue: The only point 2B made by gov: should ppl be fined/jailed for neglecting to fill out long-form? NO!!!!”.
Taylor’s probably hit on the compromise that will help the government ease its way out of this gracefully, if that’s what they want to do. They could reinstate the mandatory long-form census, but announce legislation to reduce some of the sting in the penalties (or perhaps it can be done by regulation, I don’t know).
There is a group representing everyone from big bank economists to the United Way to the retired Clerk of the Privy Council to the Canadian Centre of Policy Alternatives giving a news conference on the issue tomorrow. I think we’ve seen in the past that when a decision of the government’s results in assembling a coalition this comprehensive on the other side, it either backs off or calls an election.
Either possibility is equally possible at this stage, although I do note that the Liberal senators didn’t show up in sufficient numbers tonight to defeat the closure motion on C-9 in the Senate. We’ll see where we sit in the morning, I guess.
Anyways, back to the Census: if they’re going to change their minds, they should do so quickly before the drip-drip-drip goes on much longer.
On the philosophical question, you make the interesting point that I’m “answering a question of morality with a statement on utility.” The moral position you outline, that people don’t believe they ought to have to share personal information with the government, is not one I believe a government can endorse, when it already requires individuals to file their personal financial information with their taxes, or have their personal information stored in large databases used to support the country’s national security and law enforcement policies.
I view the rights and duties of citizenship as moral issues as well, a point I thought I made sufficiently well towards the end of the post, although my point in writing it, to be sure, was to demonstrate how political parties are using the data now.
I’ve seen it argued that individuals give their information willingly to political parties during election campaigns, but don’t feel they should be required to give it during the census. Except that the article outlined in the blogpost above goes into more detail on the “friends, relatives, acquaintances, and neighbours” approach to data collection the Conservatives currently use.
This will be a couple of occasions now where the Ottawa consensus (let’s call it the “view from Hy’s”) about what average Canadians will care about, has seriously underestimated what turned out to be the case. I realize people are going to use that as a way to assess the saleability of government decisions, but I really wish it would be the actual *reason* for them a lot less often.
As always, thanks for your thoughtful comments.
Alice it appears that this issue has just hit Britain too, they’re going to scrap their census altogether!
http://www.ctv.ca/CTVNews/Politics/20100711/census-consensus-100711/
Hopefully a reasonable compromise can be worked out.
To add in my two cents i’d simply scrap the long form, take the most important questions from it and add them to the mandatory short form. Maybe 15-20 questions in length instead of 50+.
You could even mail out 3 or 4 versions of the census with different questions. You’d get the entire set of questions answered with a large, random sample and people wouldn’t need to give their entire life story!
As is the long form feels like something of an audit – only a certain number of households receive it. Burdens of citizenship should be shared more evenly.
The length and certain types of questions are time consuming and lead to a lot of guessing and questionable data.
A medium length form that everybody answers and doesn’t take an hour to fill out seems like the most streamlined, cost effective, and fairest way to go in my view.
Have you ever had to fill out the long-form Shadow? It’s a one-in-five chance, every five years. One hour out of 25 years does not seem like such a huge burden to me.
You do propose some creative options to be sure, but they’d be sacrificing the comparability of the dataset over time, which seems like a terrible waste of resources already spent to me.
Jennifer Ditchburn from CP is reporting that the cost of the new voluntary census is already pegged at $30 million more than the old one, as there will be $5M more in distribution costs, and $25M more in public education encouraging people to fill it out. How cost-effective five different versions of a single form would be in comparison to the original regime is an empirical question of course, and I have no idea.
Alice I think there is near universal agreement that the new voluntary method is a waste of money and time.
With regards to the old method, I don’t think the question is whether you or I think the process is a burden. Most reasonable people probably don’t.
But without question there will be people who aren’t as literate, have language difficulties, attention span disorders, bad attitudes, etc who probably compromise the data pool. From some anecdotal accounts by census workers it appears some of the data is pretty dodgy, a guesstimate at best, and at worse just giving a random answer.
I’m not a statistician so I don’t know to what extent these people have an impact on the data. Maybe they just cancel each other out. Its an issue I hope is explored if this policy is re-worked.
Back the point about comparability of data using multiple forms as I suggested, i’m not sure there would be much difference at all. Each version of the medium length census would still reach about 1/5 of Canadians. You’d have the same number of random respondents to each question as there have been in the past.
Five duties: don’t forget jury service.
How about an online version (verified identity required) based on multiple-choice, with a section for feedback and improvements on the survey itself? So simple.
Politicians are like diapers. They both need regular changing. …for the same reasons.
Thanks for your comment, Swanee, but I have to tell you as an IT person that what you propose is anything but simple, although I think an opportunity to give feedback on the questionnaire is a good idea. Of course anyone could write to the Chief Statistician to give feedback already, but only 22 individuals did in 2006.
I realize your second comment is probably a bit more in jest, but we try to keep things respectful here. Anyone who offers themselves for public office has my respect, as does anyone who undertakes to serve the public in elected office. I do know that politicians with experience are often better able to say No than novices are. And remember that lobbyists don’t have terms limits; newbie politicians will be more easily influenced by them, in my experience, contrary to the common perception.