Statistics Canada Abandons the Class War?

In contrast to last week’s Census release highlighting stagnant individual earnings, today’s Survey of Labour and Income Dynamics (SLID) release trumpets rising family income. The political right critiqued coverage of last week’s release for emphasizing individual rather than family income and for not capturing the tax-and-transfer system’s equalizing effects. In particular, the Prime Minister indicated that his government had cut taxes to improve after-tax incomes. Today’s release focuses on family income and the tax-and-transfer system.

It’s almost as though Statistics Canada has responded to pressure from the National Post (or perhaps the Prime Minister’s Office). In the agency’s defence, I note that today’s release follows the same format as previous SLID releases.

However, the language seems more optimistic. Reporting a 1.6% increase, last year’s release began “Median after-tax income rose slightly for most Canadian families.” Today’s release reports a 2.1% increase “As a result of strong economic growth.” Does 0.5% make the difference between slight and strong?

What changes produced the 2.1% after-tax gain from 2005 to 2006? SLID reports that market incomes rose 1.4%, tax payments rose 3.4%, and government transfers rose 12.5%. These figures hardly support the Prime Minister’s narrative of tax cuts for the median Canadian family.

Median Family Income from SLID





 Market Income




“Other” Income

 $ 3,000

 $ 3,200


 Income Taxes

 $ 8,700

 $ 9,000


 Gov. Transfers

 $ 4,000

 $ 4,500


 After-Tax Total




Indeed, I am at a loss to explain the increase in taxes and transfers. The new transfer mentioned in Statistics Canada’s release is the “Universal Child Care Benefit”. However, this $100 monthly payment was only in effect for the second half of 2006 and is only paid to families with young children, so it seems unlikely to have singlehandedly increased the median family’s income by $500 in that year.

Much as a mean can be driven by extreme values at either end of the distribution, a median can be shifted by relatively small movements in the middle. Perhaps these year-over-year comparisons need to be taken with a grain of salt or perhaps there is a good explanation that I am missing.

Looking at the larger, longer-term picture, it is clear that individual earnings remained constant due to stagnant pay rates and that family incomes rose due to increased employment. While increased employment is welcome, it does not excuse stagnant pay rates. The fact remains that Canadian workers have been deprived of the proceeds from a quarter-century of rising labour productivity.

UPDATE (May 6): This post is quoted today in CanWest’s report on SLID and in Terence Corcoran’s column.


  • From what I’ve heard, the skewing of the income distribution towards the higher earners has resulted in windfall revenues for governments. Because the higher income earners have higher marginal income tax rates, if they are the only ones earning more money, a relatively larger fraction of the income growth would end up in government coffers. Then growth in the number of seniors would increase the value of transfers.

  • Hey Erin,

    You do know that old Corky (Terence Corcoran) took a shot at you yesterday and made reference to this post.

    It seems as though one will never get to the bottom of wealth distribution. There are so many ways of measuring it, economic families, census families, individuals. before transfers, after transfers, on and on. It is definitely a complicated process. However we do need to communicate a culturally acceptable construct to evaluate income distribution. It is times like these that one should rely on its Statistical agency to provide unbiased, accurate information to fulfill this need.

    You are quite correct in some of your interpretations with regard to the less than transparent process that Statcan provides within this process. However, to infer as Mr. Corky did within that thing they call a newspaper, (is that what the nazi’s called their propaganda vehicles, newpapers?) is so far out to lunch. I love those right wingers and their attempts at discrediting a report.

    Individually there is a large variance to distributional aspects of income. How can we just combine it to families and that distribution somehow disappears. I not understanding that logic really. Economic families is an awfully difficult measure and so is census family. The purest measure is individual. Corky wants to somehow combine income with expenditure items. A soona s you move from an individuals income to a family income you automoatically jump the traverse from looking at income to looking at expenditures because families usually result in offspring and hence dependency comes into play. So yes family income may be up but so too does the expenses factor into that. So looking at family income makes one automatically reach for the family expenditure data. Which leads to a whole new measure to determine wealth or some such other construct that looks at both income and expenditure which brings us full circle back to the LICOs and that whole notion.

    To me there is no simplistic measure for this exercise. No matter what single measure is thrown out there it is a mere abstraction of a much more complicate problem. However the culture begs simplicity to these complexities.

    I will say this, do not use SLID for measuring this crossectional component. It is panel data and is best used for longitudinal investigation, i.e. what happens to income through transitional aspects of labour force activities. It may state that it can be used for cross sectional purposes, but I would stay away from it for that. Same advice goes to Corky.

    As an alternative I would use the small area data file produced by statcan. It is based on admin data and not a longitudinally designed survey methodology. THe tax data also has family data so have a look there. I don’t think they do much of an official release. Statcan used to have the Income Survey of what was called the Survey of Consumer Finance but it was dropped. I will look into it for you if you like, and see what other measures could have a look into this family versus individual income measure.

    Corky is running on a bit at the mouth, I wonder if he is foaming, has that guy had his shots?


  • Andrew Jackson

    Census data based on large samples are clearly better able to get at top tail driven inequality than SLID – whcih has a much smaller sample and under samples at both the low and very high end. I’d take year to year changes in SLID with a considerable does of salt.

  • Yes I would concur that the census long form data is by far the best source for income data. I recall working on the consumer finance data some years ago and had a real problem with oversampling of millionaires on the top end. Too many clustered within a specific geography and of course with a sample based on geography like the way it is stratified in the LFS sample one gets a lot of fuzziness on the top tail. (sometimes over representation and sometimes under)

    The bottom tail, if you wanted to argue with Corky even more, is not covered by any survey or census. That is none account for the the homeless or the transient. Try doing a CATI (telephone interview) on them or a how about a CAPI. Anybody have an estimate on the size of the homeless and transient population in Canada. Take that number and add it onto the already difficult lower tail to measure.

    I thought Linda McQuiag had an interesting article today in the Star on this topic.

Leave a Reply

Your email address will not be published. Required fields are marked *