Wednesday, February 08, 2012

Mandatory Response

Lorne Gunter's article on today's release of short form Census data is so bad that it necessitates a response. Point-by-point:

So the 2011 census results are being released today, or at least some results are. Is anyone else as surprised as I am that there is any data to announce? I mean in the summer of 2010, you would have thought the statistical world was coming to an end and taking much of the intellectual foundation of modern civilization with it just because the Harper government had decided to make the long-form census voluntary rather than mandatory.


To start off, the population data being released today is based on the short form Census, which remains mandatory (as does the farming Census). Why? Because there's nothing wrong with a little state-enforced coercion.


Experts were recruited from the United States to tell the Toronto Star that the Tories move would “lower the quality and raise the cost of information” gathered by StatsCan.


The funny thing about experts is they're often right. While I'm sure there were some American experts telling this to the Toronto Star (after all, the US quickly scrapped the idea of going the voluntary route after a disastrous trial), there were also experts of the Canadian variety, including the two most recent heads of Stats Canada, Alex Himelfarb, Don Drummond, 100-some organizations, and basically everyone besides Tony Clement. Oh, and the Toronto Star was joined by the pinkos at the National Post and Calgary Herald, among others, in decrying the demise of the Census.


It was Canadians’ “civic duty” to comply with government demands for information about ethnicity, education level, sources of income, types of housing, number and ages of children and their activities, sexual orientation, family relationships, divisions of household labour, recreation and so on.


Some of the above information is still mandatory - the short form, for example, asks for the names and ages of your children so that nefarious governments can build schools near them.

Sources of income is also asked on another mandatory form we'll all be filling out in the coming months.


On the other hand, if you were sceptical about government’s ability to solve big problems, no matter how accurate the inputs it uses to analyse the sources and solutions, you tended to think a voluntary census would be just as useful as a mandatory one, and far less destructive of individual rights in a democracy.


OK, let's say you hate big government and believe we'd be better off in a state of anarchy. If I'd lived under an Alberta PC government my entire life like Mr. Gunter, I'd be skeptical about government ever being a source of good too.

But the thing is, the long form Census is also used by hospitals to offer services and fight pandemics. Masters students, like Stephen Harper, use it to write thesis papers. Think tanks, like the Fraser Institute, use it to prove their kooky right wing theories. And businesses use it all the time – just think of restaurants and grocery stores that sell ethnic foods or cater to specific client demographics


But, really, are the figures produced by having StatsCan select 18.8% of homes based on pure statistical theory going to be so much more useful in setting public policy than the figures from 23.1% of self-selecting, voluntary homes?


Yes.

I don't want to turn this post into a statistics webinar, and I don't need to, because Gunter answers the question himself earlier in the paragraph when he mentions how aboriginals and immigrants are less likely to complete the Census. Also, there are studies on this topic (warning: these studies are by experts).


Besides, I have my doubts about how untainted data from the mandatory census was anyway. When I wrote about this issue two years ago, I received a handful of messages from former census planners telling me that it was routine practice at StatsCan to send long forms to the same households census after census. If a household had shown itself willing to fill out a long form before, it was likely to receive another the next time.


This is just factually inaccurate. I worked as a Census Rep in 2001, and every fifth household got the long form Census. So if house 2 got the long-form, houses 4, 6, 8, and 10 wouldn't, and house 12 would. Lather, rinse, repeat. Even if StatsCan wanted to employ faulty methodology, there's no way they could.


And my favourite: The Tories’ move was “enormously destructive” of morale at Statistics Canada. Huh!? How could that possibly matter?


Well, having competent employees resign on principle, and having others demoralized isn't good for any organization. I mean, just imagine how demoralized reporters like Mr. Gunter would be if newspapers started publishing factually inaccurate information. I mean, the entire industry would...well...never mind...

Labels: ,

14 Comments:

  • Ah, how nice it must be to get to rewrite history whenever you choose.

    The majority of the reporting - and the griping from opposition benches - claimed time and again that the census was being scrapped, and not the fact that the mandatory long-form survey was being made voluntary.

    As for the resignations from Stats Can, the headline speaker from the latest Liberal convention demonstrates that his, at least, was a partisan move, not a mathematical one.

    By Anonymous Anonymous, at 2:07 PM  

  • Great response, but you brushed off potentially his only valid point: the media did frame the story as "the census is being scrapped" - short-form, long-form, voluntary, mandatory, everything.

    You yourself used that language repeatedly.

    And while I agree completely with every single other word you've said on this subject, I still don't see why it means people should be criminally charged for thinking differently and refusing to participate.

    But hey - at least we'd find people to put in all those new jails they're building.

    By Blogger Robert Vollman, at 6:52 PM  

  • "The mandatory long-form survey was being made voluntary" doesn't fit well into a 15 second soundbite to grab the public's attention.

    And the griping is basically true, if a bit condensed. The census was scrapped, or rather, the information that you get from the long-form is as good as scrapped. You need to use a random sample to have any confidence in your data: not a self-selecting one.

    Now Stats Can's chief economic analyst has resigned. The Conservative partisans won't hear of any dissent regarding their wonderful scheme and its likelihood of failure. If making shit up out of whole cloth is what it'll take to avoid the embarrassment of failure, then that's how far they'll go.

    By Anonymous Anonymous, at 7:09 PM  

  • I live in Mississauga and got the long form twice in the mail. Third time, a man came to the verandah where I was reading and handed me the form again, saying he'd pick it up the next day when I didn't have the dollar amounts in full. Husband had to ask accountant for precise figures as anything so governmental needs to be as stated. When I read this was "voluntary" red lights started flashing. Who was this man? Who paid him? Were we cherry picked? I'm retired, my husband working at 66. Were we going to be skewed like the Tory polls? Guess who's not going to get to see good data that could be used for voter fodder? Since we don't have land lines anymore, how are Tories going to misuse the data they did get? Phoned the Star.

    By Anonymous sailsmart, at 5:26 AM  

  • I'm not sure if a mandatory Census is needed for optimal future health R+D or not. There is a great deal to be gained by assessing basic population health. Probably this can be attained with using cutting edge telecom surveys. Maybe an occasional quality-control check up. Seniors and poorer general population would take cash to record diets and wear novel blood pressure sensors and such forth. And probably a medical receptionist or LPN could record the data rather than an MD, if the Unions permit. There was a past 1950s survey in the USA that could be replaced much more cheaply now. Determining pandemic social distancing capacity could be done with surveys about # of rooms, duct tape, electricity sources and backups, pantry groceries, PPE....

    In any event this would necessitate some type of government bureaucracy. At least in this country. God din't protect America from the Spanish Flu. Maybe he doesn't like theists or Christians?!

    By Anonymous The Keystone Garter, at 2:55 PM  

  • "You need to use a random sample to have any confidence in your data: not a self-selecting one."

    There is absolutely no reason to believe that the previous data was any more accurate or true than the data collected during the current cycle.

    Heck, we already know that people lied when it was mandatory. How do you claim that gives you confidence in the data?

    By Anonymous Anonymous, at 10:03 PM  

  • "if house 2 got the long-form, houses 4, 6, 8, and 10 wouldn't, and house 12 would"

    That is most emphatically NOT a random sample.

    By Anonymous Anonymous, at 10:41 AM  

  • ...and I know for a fact the AB job market is not especially competitive nor efficient. It isn't as bad as Halifax. There you need to be friends with someone for millenia, or redheaded, or communist, or I don't really care to know what. I just know it should be possible to measure the inefficiency of the shipbuilt contract over time, due to communism. Should build in another country or upgrade another part of this one...
    AB gives employers the latitude to offer weird personality tests. Orwellian but harmless. Again, prefer not to hire losers. I see the downside of having so many young people in positions of power. Major USA chains are forced with inefficient hirings. Like when Cheech and the pothead from the 70's Show didn't like Fez as a worker.
    Again, from what I can tell it is all swimming pools and nice home interiors, and some careless truck drivers that like to hit pedestrians by accident. Oil makes you dumb: world.
    But becuase of the trickle-through and underlying prairie work ethic, not so bad.

    By Anonymous The Keystone Garter, at 6:33 PM  

  • ...for instance there isn't a few hundred thousands dollars in AB's budget for internet job search facilities. As telecom improves and cellphones and internet become more and more mandatory, the poor are marginalized. Clearly the poor in Canada are smarter than the rich and more capable of engendering a better world for all.

    By Anonymous The Keystone Garter, at 6:39 PM  

  • "Sources of income is also asked on another mandatory form we'll all be filling out in the coming months."

    Since you seem to equate your tax filing with the census, I presume you would not object to your personal tax filing being made public a few years down the road, with all of your personally identifying information attached?

    It's also worth noting that officials report 2.7 Million households filled out the long-form survey in 2011, which is up significantly over the 2.3 Million households who filled out the long-form census in 2006.

    By Anonymous Anonymous, at 1:23 PM  

  • All of you who are saying the long-form census still exists simply don't get it.

    There is no long-form census anymore. There is a long-form survey, that's all. The situation may have been imperfect in the past, but if there is simply no effort at all made to statistically control the sample, you cannot make the same claims that were made before. It's just logically impossible.

    That said I'm pretty surprised to hear what CG said about giving out the long-form census ("if house 2 got the long-form, houses 4, 6, 8, and 10 wouldn't, and house 12 would"), that is most certainly not random and is likely to create all kinds of statistical artifacts.

    By Blogger saphorr, at 3:45 AM  

  • Dan, I hope that annonymous has set you straight on statistical data and whether or not self-selecting samples are valid or not. He is talking Tory sense Arnold, tory sense. In other news, Vic Toews is a pervert and should not be in cabinet.

    By Anonymous Anonymous, at 9:04 PM  

  • I worked as a Census Commissioner in the '91 Census; I might be able to clear up a bit of the confusion regarding the "random-ness" of the long-form selection process.

    Each Census area is comprised of respondant households. Every 5th household got the long form, but the COUNT to five would start at a different household each time. So it's not designed to be the SAME 5th household every time. IF there are no changes in the number of households (admittedly rare) then each household would get the long-firm once every 25 years.

    It's far from ideally "random", but the design IS intended to be fair and to ensure that as complete a picture as possible is obtained.

    One obvious reason that people might get long form censi in two or more consequtive censuses is that neighbourhoods rarely are static and unchanging in a five year cycle. So if you were household 2 in one census, but your neighbour who was previously #1 built a basement suite and rented it out, you would now be household 3...and so on. So it IS possible to be asked to fill out the long-form census in two or more subsequent surveys.

    By Blogger Party of One, at 3:14 PM  

  • Montgomery triangle is awesome!!

    By Anonymous Anonymous, at 1:54 PM  

Post a Comment

Links to this post:

Create a Link

<< Home