What if First Nations (and their poverty) were counted?

Kudos to the Globe and Mail for their front page story on Jan 23rd highlighting the fact that the official unemployment rate does not count First Nations reserves. You heard that right: First Nations reserves, some of the poorest places in the country, are not included in the official unemployment rate.

As unbelievable as that sounds, the reality is even worse. Reserves are regularly excluded from all of our regularly updated measures of poverty, wage growth, average incomes etc. The exception to this rule is during a Census, i.e. every four years (and as a result of legislation making the long form Census voluntary, concerns have been raised about the future reliability of these data). Otherwise, reserves—some of the poorest places in Canada–are statistic-free zones: out of sight…out of mind.

As someone who works regularly with Statscan data, this was hardly news to me. But I’m glad this issue has finally gotten the attention it deserves. How can we have an accurate picture of what’s happening in Canada when we’re actively and deliberately excluding some of the poorest parts of our country from our basic statistics?

But what might unemployment would look like in Canada if reserves were included? Since this data isn’t collected monthly, the only reliable figures are from the first week of May 2011, when the National Household Survey (NHS) was conducted. As you can see in Table 1, the seasonally unadjusted unemployment rate for Canada is 7.6% (close to the comparable Labour Force Survey (LFS) estimate of 7.5% in that month)—but on reserves it is a shocking 22%. Had reserves been included in the calculations, the Canadian unemployment rate would have been 7.8%, not the official 7.6%.

Table 1: May 2011 employment figures

May 2011 Unemployment figures

Note: non-seasonally adjusted, NHS Individual PUMF, 15 years and over

When reserves are included in the calculations, the employment rate (the proportion of the working age population that has a job) falls from 61.1% to 60.9%—pretty incredible, considering people on reserves make up only 1% of the Canadian population.

That’s for 2011, but what would it look like today? Figure 1 shows that, once reserves are included, the unemployment rate is a little worse than the “official” statistics indicate for Canada, Ontario and Quebec. But it is substantially worse for the Prairie provinces and BC. If everyone was counted, including folks on reserves, the unemployment rate in December 2014 would have jumped from 5.2% to 5.8% in Manitoba; and in Saskatchewan from 3.6% to 4.3%. BC would see its rate go from 5.4% to 5.7%.

Figure 1:

What if the jobless on reserves were counted?Excluding the poorest places in Canada from basic data collection may paint a rosier picture, but certainly not a truthful one. Canada has a responsibility to First Nations peoples who live on reserve, and they deserve to be counted. Yes, that will cost more money, but not including people on reserves in our data-gathering allows us to continue to ignore the appalling poverty that we’ve both facilitated and allowed to deepen for generations in our wealthy country.

Notes for Stats Nerds:
These calculations are approximations. The “on-reserve” designation was imputed by using band membership crossed with non-CMA locations from the NHS Individuals PUMF. This is clearly not perfect and should be treated as a proxy. This approach, while fast, overestimates the number of people on reserves. I’m applying differences in non-seasonally adjusted figures from 2011 to seasonally adjusted LFS data from Dec 2014. Those differences may not hold, although there is really no way of knowing since reserves aren’t included in the LFS.

Regular data collection should happen on reserves, but the responsibility for its non-collection shouldn’t be placed entirely at the feet of Statistics Canada. It’s more expensive to collect data on reserves, particularly if they are remote, and after  austerity-driven budget cuts, Statistics Canada’s resources have been significantly stretched.

David Macdonald is a Senior Economist with the CCPA. Follow David on Twitter @DavidMacCdn.


  1. I find articles such as this one only serves to misinform as there is much missing data from your report.
    Many Canadian provinces are missing, which have First Nations residents. Please update and resubmit your article with complete information this time.

    Many thanks. 🙂

  2. Good points, but there’s some misinformation.

    1) StatCan’s published 3-month moving average monthly unemployment rate estimates by Employment Insurance Economic Region (i.e. the ones used by ESDC for EI planning) include an adjustment to account for reserves. It’s just that the survey itself doesn’t cover them. And to be fair, most LFS estimates also do not. At least for EI planning they are taken into account, and StatCan errs on the side of overestimating their unemployment, which secures them better benefits.

    2) +/- 0.2 percentage points seems like a lot, and is a lot for NHS. But for monthly LFS, that’s much smaller than most of the confidence intervals. It would just register as noise. So despite radically different unemployment trends on reserves, the reserves don’t actually affect the national and provincial estimates a noticeable amount, although they are still an important demographic to consider.

    3) Contrary to the Globe’s article and other speculation, cost is not the primary barrier to covering reserves. Compliance is. LFS is a “mandatory” survey, which gives about 90% response rates and reasonable quality. But StatCan lacks the authority to enforce “mandatory” on reserves. Most reserves have also expressed a disinterest in being hassled by StatCan interviewers every month. The one (n=1) reserve used in the pilot study requested the study, so of course they were compliant, but in general most others are not. Without compliance or authority, it is very difficult for us to collect survey data there. Federal cuts are not helping.

    Get a bunch of reserves to sign a petition saying they want to be covered and then get an NGO to pay for the sample and it’ll be a different story…

  3. Great points Justin! Thanks for the clarifications.

    The figures I’ve got here are a bit rough as they aren’t explicitly those on reserve. I think what they point to is that there is likely little change nationally but that the unemployment figures could be affected in Saskatchewan and Manitoba and perhaps BC. In the other provinces it wouldn’t make much of a difference at the provincial level.

  4. True, the potential impact of missing reserves on Saskatchewan and Manitoba is much greater. I’d like to point out that LFS 95% CIs for December 2014 monthly unemployment rate in those provinces is about +/- 0.6 percentage points, higher than many other provinces. Taking that into account, the increase with reserves doesn’t seem as dramatic as it does on its own. I think it would be helpful if your graphs included confidence bands or some kind of measure of variability to frame the differences in some sort of context.

    That said, I don’t want to hide behind variability either. LFS estimates are slightly biased by missing those communities, and that is a meaningful part of the population that is being missed. It’s just that the impact on estimates for large areas is not as alarming as it appears on those graphs. And StatCan is looking at modelling approaches to get better projections for reserves in the mean time.

    Survey coverage of reserves would require establishing and maintaining treaties that allow consistent access for LFS interviewers onto reserve lands to not only collect but do significant nonresponse follow-up. Efficient collection would also require gaining access to some kind of telephone directories for reserve dwellings. And if those treaties change over time (e.g. they decide they don’t like being hassled by interviewers so often), suddenly removing them from the sample ad hoc has an impact on the survey design methodology and integrity. Comparability over time is affected. Trend estimates are affected. Total provincial sample size decreases (or gets re-allocated ad hoc), decreasing quality.

    In essence, the issue is far more complex than the Globe made it seem. It’s not simply a matter of the underdog being forgotten, as tantalizing a headline as that makes. Though it would be nice to see some workable solution in the near future.

Join the Discussion

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

Before commenting, please read our Comment Policy