The Conservative Government vs. The Canadian Census
by Peter Bursztyn July 27th 2010
Canada has been collective census data for at least a century. Moreover, Statistics Canada’s data (population, economic, health, environmental, etc) has won a reputation for excellence and timeliness. No other country publishes its data faster. No other country beats Statistics Canada for accuracy.
1. Statistics relies on sampling rather than measuring an entire population to collect data. It is considerably cheaper and far more practical.
2. To be valid, the sample MUST BE RANDOM. That means you must not choose the sample in any way, nor should people being sampled be allowed to excuse themselves.
As a simple-minded example, consider determining the heights of adult Canadians. Say you decide that a sample of 10,000 would be adequate. You set out to actually measure 10,000 people, but all the basketball players refuse to be measured. Your (flawed) result will suggest that Canadians are less tall than they really are.
In real life, Conservatives appear less likely to complete the long-form census than other Canadians. (Several Conservative MPs have actually said that they would refuse to complete a voluntary long-form census.) As a result, they will be undercounted. What other groups will be under-represented, and by how much?
Immigrants from countries with intrusive and nasty governments also tend to be suspicious of bureaucracy and are likely to avoid completing a questionnaire. Finally, people who are illiterate or literate in a language other than English and French will probably not do it either.
3. A “shoe-leather” census (as done in the USA), or a mail-in census (Canada) are old fashioned, and many other countries do it another way.
No statistician will claim that we must carry on counting Canadians exactly as we have done for many decades. There are more modern ways to do the job. Scandinavia comes to mind.
4. However, we have a very long, accurate (the judgment of other countries – not mine), trailing data set. That is hugely valuable to monitor changes, and hence to help predict (and plan for) the future.
Changing the sampling technique could render this data set almost useless.
5. If you wish to change your sampling procedure, there are statistically valid ways to do the job while preserving the value of your historic data set. This is actually very simple – in theory (in practice, it requires work).
You simply collect your data using both the old sampling method and the new one together for 2 data gathering cycles. You examine the differences in the results, and then apply a mathematical correction to align the two. That will allow you to continue using the old data with considerably confidence.
6. However, if you wished to change the sampling procedure you would never do what the “Con Men” (Federal Conservatives) plan; go to a self-selected sample and increase the sample size by 50%. The groups who don’t answer voluntary questionnaires would continue to be undercounted – and to an unknown degree.
It would also cost more! Why on earth would you pay more for less? Frankly, it would be more sensible to scrap the census altogether and rely on guesswork . . .
7. One of the modern methods of synthesizing a census is to “mine data” from other sources. Examples which come to mind include (a) income tax rolls, (b) drivers licenses, (c) home ownership rolls from property taxation (d) passports, (e) citizenship documents (including applications), (f) health cards, (g) telephone directories (less useful now than 5 years ago) . . . etc
The Scandinavian countries seem to have a negative income tax. If you earn little or nothing, the government pays you. That simple fact ensures that the great majority of people – even the destitute – have left their mark on the income tax data. We – of course – do not have this. The very poor would tend to be left out by data mining – no income tax records, no home ownership, no drivers license, etc . . .
8. Finally, “electronic data mining” is cheap, but far more “intrusive” than the current Canadian census . . . Also the people do not know that it is happening . . .
The decision to change the national census sampling technique is an excellent example of allowing ideology to trump science – just as the Catholic Church did when it threatened to excommunicate Galileo unless he recanted his suggestion that the earth orbited the sun. At that time, the church believed the reverse was the case! Two decades ago, the Catholic Church formally apologized to Galileo . . .
Statistics Canada census data does not intrude on privacy. The individual census form is identified only by postal code. Since hundreds of people share a typical postal code, there is no way to trace the data back to a specific individual.