A New Era for Measuring Poverty in Canada
Last Thursday’s Statistics Canada release of individual and household income data for 2008 marks a new era in the study of poverty in Canada.
Instead of reporting only on the Low Income Cut Offs (LICO), as they used to, Statistics Canada reported on three of the most common measures of low income in the same publication (LICO, the low income measure and the market basket measure). Gone are the days of looking for different studies produced by different institutions to compare trends of low income in Canada.
Even more importantly for those of us looking for reliable and timely data on low incomes, Statistics Canada has now taken over producing the Market Basket Measure (MBM) from HRSDC.
The Market Basket Measure — an absolute measure of low income which captures the actual costs of living in 49 communities across Canada — was developed by HRSDC in early 2000s. This is the only Canadian low income measure that takes into account regional differences in cost of living, but it hasn’t been used a lot so far partly because data was only available back to 2000 and partly because there was no regular schedule for new MBM data releases so researchers never knew when the next one will be available. As a result, the latest available numbers often lagged 3-4 years behind.
The fact that Statistics Canada has now committed to make the data available annually makes this a much more reliable measure so get ready to hear a lot more about it in the near future.
Statistics Canada has long complained that LICO should not be used as a poverty line, but in the absence of a viable alternative the LICO became the default poverty measure in Canada.
It seems that the fine folks at Statistics Canada have finally realized that there is substantial interest in studying trends of low income and poverty in the country and are taking important steps to make more comprehensive data available, both on the breadth and on the depth of low income.
In fact, a more comprehensive report on the topic seems to be in the works: Incomes in Canada 2008 announced that “an Integrated Report on Low Income, which uses a multi-line, multi-index approach is forthcoming.”
Kudos to Statistics Canada for improving the quality and availability of data on low income and poverty trends in Canada. Better data will help policy makers make better decisions about tackling serious social problems like poverty. It will also make it easier for researchers to evaluate the impacts of government policies on the prevalence and depth of poverty in the country, paving the way for greater accountability of our governments.
This article originally appeared on http://www.policynote.ca, the CCPA’s blog on BC public policy issues.
It will be interesting to see how the MBM changes over time so we can see how inflation is different for people will low incomes verses average inflaction.
Wow this is great news! The LICO measure always bugged me because it was relative. It’s nice to hear StatsCan is starting to employ an absolute measure as well.
I think the MBM is very useful as a consumption based measure of poverty but I’d quibble with the oft cited view that it is an “absolute” measure. In fact the MBM measures consumption relative to a modest norm based on the consumption of others – which the designers were quite explicit about. A truly absolute line would only deliver a bare physical survival budget – and even Sarlos goes a bit further than that, dare I say.
Andrew, I agree that “consumption-based” measure is a more descriptive characterization of the MBM as it approaches the measurement of low income from the standpoint of “what can these people afford to buy with their income” rather than “how many dollars do they have”.
However, in the continuum of absolute to relative measures of poverty, it seems to be that the MBM certainly falls way closer to the absolute end because it measures the costs of a set basket of goods and services.
Obviously, what we put in the basket depends on the type of society we live in and what we consider appropriate levels of consumption, but I don’t think this is sufficient to make MBM a relative measure. We are measuring the absolute costs of having a certain minimum standard of living in our particular society and this is precisely what’s relevant when we consider how to target income supports in our society.
That said, perhaps the absolute vs relative distinction is not as important as the income vs consumption-based notion of measuring poverty or strained circumstances.
just a small point, a good proportion of people on the low end of the poverty scale, tend to be a bit transient and trying to measure these folk tends to be quite difficult, and I am not talking homeless. I am talking low wage workers, unemployed, underemployed, single parents and many others that find it difficult to maintain a permanent residence. (I mean more than yearly so that these is an actual point of contact) I believe there is quite a bit of bias within any survey that claims to have accurate measures on poverty when, most likely many of the targeted population is not in the data information universe. Sorry but there is a real problem here that needs to be mentioned.
Paul, I agree that surveys like the SLID do not capture the very top and the very bottom of the income spectrum in their entirety. Response rates on the SLID are around 80% if I recall correctly and there’s reason to believe that there is some selection bias there that no weighting system can completely account for.
David Green, one of my economics profs from UBC, wrote a paper with some Statistics Canada researchers a few years ago that showed that low income and inequality trends look much worse when you use the Census or tax filer data than when you look at the SLID. Even the Census doesn’t get us complete coverage – First Nations reserves being a case in point – but it gets a lot closer to the true distribution of income.
The tax data looks really promising as well and is probably a lot more accurate than data collected from surveys asking people to self-report their income over the last year. With the shift to providing most low-income benefits through tax credits made in the 1990s, the tax filing rate has increased dramatically and undercoverage is extremely low. Perhaps Statistics Canada should be using the tax filer data to calculate the proportion of people living in low income instead of the SLID?
But then, tax filers data is limited because it doesn’t provide any background info – level of education, etc. – that can help us explain differences in low income among the population.
I suppose this is a long-winded way of saying that there are limitations of surveys and we only get estimates, which can fluctuate year to year because of sampling issues, but I still think that they are useful to keep track of and measure progress by, especially over the medium and longer term.
I have worked with the tax filer data as well. What I would like to do is have more effort put into trying to reach these people or at least quantify the amount of bias. I should have a look at the research you mention. I think I recall the report you noted.
I guess the best one can do is acknowledge the data gap and the bias. I have worked on both the Survey of Consumer finance (now cancelled) and Famex and I do agree that the surveys have a much harder time capturing the lower and upper ends than the admin sources, However, the admin sources have their own problems, which as an example, like you mention, have little in terms of content.
Potentially there are some alternative strategies to combat this under coverage. However, one sure method to combat it is to ensure user’s of the number are well aware of the under coverage, which is rarely mentioned and in many cases even understood by the users of the data or for that matter some of the researchers.
I do want to come to the defense of relative poverty measures. If you believe (a la Spirit Level and similar arguments) that income inequality itself imposes costs on society, then a relative measure of poverty is quite appropriate. Also, relative poverty measures take into account evolving social norms about what constitutes a “decent standard of living.” Absolute measures (like the official U.S. poverty threshold measure) assume there’s a fixed standard that never changes (the U.S. measure was set in the 1960s); they virtually predetermine that poverty rates will “fall” over time simply because average incomes tend to rise relative to the cost of that fixed consumption basket.
I hope nobody is taking “The Spirit Level” as the received truth. Entirely apart from the way that it gradually slides from correlation to causation, there are many problems with the stats. It looks as if the conclusion (inequality is bad) was decided ahead of time and that a little data-mining occurred. See:
I agree that there is merit in relative measures of low income, and Statistics Canada seems to be strongly pushing the LIM-AT, which is a purely relative measure. The real benefit here is the ability to compare internationally, as it’s easier to calculate median incomes in every country and much harder to compare definitions of “decent standard of living”.
I also agree that income inequality is an important issue to keep track of, but I think it goes beyond poverty and low income. This is why I’m not convinced that a single indicator can provide an adequate picture of both poverty and inequality.
Purely relative measures like LIM only look at the bottom half of the income spectrum so by definition they only provide half the picture. Given that the major driver of income inequality in the last two decades has been the incredible rise of incomes at the top, relative measures of poverty will not be able to pick that up.
We’re now living in a country where the middle class is barely able to maintain the real income levels their parents’ generation enjoyed, which is why I think that given our particular circumstances, it’s better to measure poverty with a consumption-based measure and track inequality trends separately with other indicators.
I prefer calling it consumption-based, rather than absolute measure because I think we all agree that the definition of a “minimally appropriate standard of living” changes over time and we need to update the consumption basket accordingly. Which is why I like the MBM as a measure of poverty. Its main limitation is that it’s expensive/time-consuming to produce and that it’s not suited for international comparisons.
That said, threshold-based measures only provide part of the picture – if we managed to get everyone just a tad above the threshold then how much have we really achieved?
This is why I wish Statistics Canada published not just the low income threshold and the share of people below this threshold, but actually plotted the adjusted income distribution with the threshold marked. I think looking at the entire distribution and how it changed over time would be very telling.
I still remember my first econometrics prof who kept reminding us to look at our data and not just at the summary stats. I miss being a grad student and having access to micro-level data that would allow me to plot the distribution myself.
I hope Statistics Canada plots the distributions in their forthcoming multi-index report on low income.
team up with a professor, and put a project through the Regional Data Center, and you may be able to get access to the micro data. As long as it is Academic based, you should be good to go. Let me know and I can give you a hand with all the paper work and such. It does take quite a long time to get the final access straightened out, but there is a route.
Of course that is not feasible for a small project, it would have to be something larger.