Consultation on Census Data
March 9, 1998
recorded by L. Simich
Summary
On Monday, March 9, 1998, CERIS (Joint Centre of Excellence for Research on Immigration
and Settlement -- Toronto) and the Canadian Race Relations Foundation (CRRF) co-hosted a
public consultation on census data at the CERIS offices. Moy Tam, Executive Director, CRRF
introduced the session and Valerie Preston, Associate Director of CERIS and Professor of
Geography, York University, chaired the meeting. Panelists included:
- Korina Besednik, Statistics Canada, Ontario;
- Tim Rees, Access and Equity Centre, City of Toronto;
- Professor Michael Ornstein, Institute for Social Research, York University.
Contents
Background
On Monday, March 9, 1998, CERIS (Joint Centre of Excellence for Research on
Immigration and Settlement -- Toronto) and the Canadian Race Relations Foundation
(CRRF) co-hosted a public consultation on census data at the CERIS offices. Moy Tam,
Executive Director, CRRF introduced the session and Valerie Preston, Associate Director of
CERIS and Professor of Geography, York University, chaired the meeting. Panelists
included:
- Korina Besednik, Statistics Canada, Ontario;
- Tim Rees, Access and Equity Centre, City of Toronto;
- Professor Michael Ornstein, Institute for Social Research, York University.
Ms Tam welcomed the participants and introduced the chair and panelists, explaining
that CRRF was established in October 1996 (and opened officially in November, 1997) as a
result of government efforts at redress to rectify the kind of racist policies which led
to discrimination against the Japanese in Canada during WWII. She said that CRRF
activities are aimed at eliminating racism in concert with other Canadian groups and
citizens. Census data may be used to advance equity, especially for minorities and
immigrants. Now that over 11% of the Canadian population is self-identified as a visible
minority, how can research using census data best be done?
Presentation by Statistics
Canada:
Korina Besednik of Statistics Canada, who works with clients requiring customized data
at the Toronto office, reviewed the types of 1996 census data which are available to the
public. She explained that the data is available at a variety of geographic levels, from
provincial to the smaller Enumeration Areas (EA’s). She particularly recommended
consulting the two census products that provide data on a small scale area, the Profiles
and the Basic Summary Tabulations.
Generally, the Profiles (for 1986 and 1991) were available in public libraries two
years after their production. The 1996 Profile data will be available in March 1999,
although it will also be available on CD-ROM at a cost of $5000 in October of 1998. Ms
Besednik noted that the census data is currently being released in stages. For census
products, such as catalogues, she urged consultation participants to call the Stats Can
information line.
Questions to Ms Besednik were answered as follows (bullets indicate comments from
audience):
- What definition of "immigrant" is used by Statistic Canada? Anyone not born in
Canada.
- Can census data be broken down and made available by subject? Special requests for data
by subject, age or other categories can be made.
- Is there a way of linking data, for example, age with health care needs? While Stats Can
doesn’t do this, it was suggested that other existing surveys might be used to link
data categories, for example, some health surveys may have immigration codes.
Dr. Morton Beiser noted, however, that the National Population Health Status Survey is
not administered to non-English or French speakers, lessening its utility for research
among minority groups. (The issue of linking data for research purposes came up throughout
the consultation, and seemed to indicate a felt need for Stats Can to work more closely
with researchers to identify ways to make census data more useful.)
- Are the census boundaries geo-coded? Yes, as are the Enumeration Areas.
- What is the meaning and purpose of "visible minority?" Besednik replied that
it is not a category created by Stats Can, but by Employment Equity.
- What determines the category of "other", e.g., "other languages"?
Relative size? Yes, only the largest 50 are specified in order to avoid low cell counts
when cross-tabulated with small geographic areas. (One way to preserve confidentiality.)
Ms Besednik acknowledged the changes in 1986, 1991 and 1996 Census questions make
comparisons of data difficult. Previously there were check-off boxes concerning identity,
including one for "black", which were replaced in 1996 with write-in spaces. As
a result many minority members chose to write in "Canadian," or provided
specific national origin information.
- Why does Stats Can use confusing categories such as "Arab/West Asian" which
may not seem culturally or geographically accurate to respondents? Furthermore, it is not
clear how this category relates to the "visible minority" designation. Ms
Besednik answered that several census questions do specify place of birth, even if other
questions cluster identities. Questions about the appropriateness of the "visible
minority" phrase have to be referred to Employment Equity.
- Can educational institutions receive the census data sooner than the public? Prof.
Preston noted that the only way to secure data in advance of public availability now is by
paying extra. Statistics Canada is required to do cost recovery.
- With regard to a particular minority language group, one audience member noted that,
based on his personal knowledge of the community, the census surely underestimates the
number of speakers. It was acknowledged that the census is imperfect and that such an
undercount could be a mistake.
Presentation by Tim Rees and
Michael Ornstein:
Mr. Rees outlined the context for the analysis of the 1991 Census ("Ethno-Racial
Inequality in Metropolitan Toronto: Analysis of the 1991 Census), which Prof. Ornstein had
undertaken for the Access and Equity Centre. As Mr. Rees noted, until then their work of
promoting equity was based primarily on qualitative research, but more concrete
information was needed for policy development and program delivery. For that reason, Prof.
Ornstein was asked what data was available and how the Access and Equity office could use
it to get a sense of actual disparities among different communities and to set priorities.
Prof. Ornstein reported that he had examined the Census data to determine 1) what
groups to compare, and 2) how to compare them. He found that useful comparisons could only
be made with costly special tabulations, which were difficult to design in the early
stages of the project.
He noted that "small" findings proved to be revealing. For example, Iranian
refugees were not, as a whole, well-off in Canada; whereas Egyptian immigrants, even
though from a relatively poor country, were relatively well-off . In other words, further
research on conditions of immigration and socio-economic background is needed in analysing
the data. Unfortunately, the data are not sufficiently stratified now for such detailed
comparisons.
Prof. Ornstein explained that there are other complexities in the data, such as
multiple ethnic origins, e.g. Black and British both might be selected by a respondent, so
he chose to work with the more "visible" designation for the purposes of the
research, resulting in a more "fine-grained" picture. He also noted that it will
be impossible to replicate his particular research with the next census because the
questions and categories will have changed again.
Prof. Ornstein reported on some limitations on the manipulation of the census data. For
example, he was unable to divide Central and South American immigrants into groups, except
for Mexico, even though it is common knowledge that the different waves of immigration
resulted in differences among the current groups, e.g. earlier Latin American arrivals
were richer and better educated, whereas later Latin Americans were poorer. Therefore, the
Latin American category was not culturally or demographically cohesive from the beginning.
He discovered that the strongest link with socio-economic status is unemployment, not
low-wage employment nor total household income (since single-earner households are
generally poorer). If he could do the research again, he would focus on a) economic data,
and b) the distribution of young and old within groups.
Ornstein made three points in conclusion:
1) It is important to look at variance, e.g. in incomes, because concentrating on means
can conceal the real needs.
2) One has to think in terms of incidence and numbers both, because the larger groups
do have more impact.
3) More individualized data is needed to answer some research questions. For example,
groups with lower education have lower incomes, but do individuals?
One overall problem with the way Statistics Canada makes the data available for
research is that these types of questions are hard to answer with standard tabulation
data. Part of the problem is that the researcher has to request and pay for specialized
data before the research problem can be thought through.
In the discussion that followed it was noted that Statisticss Canada is moving in the
direction of making special tabulations possible, but special permission for reasons of
confidentiality is required of researchers. Although products like the Profiles on CD-ROM
are useful, it is not possible to do cross-tabulations, e.g. for the university educated
and visible minority categories. As Prof. Ornstein said, it is hard to develop the
foresight necessary for getting $5000 worth of data at the beginning of a research
project. And as one graduate student added, how can a Ph.D. candidate pay that much for
data anyway? Discussants noted that the Data Liberation Initiative is a step toward
addressing some of the issues of availability at least at univeristies.
- Are the relevant categories dealt with in the Analysis of the 1991Census
"ethnic?" Prof. Ornstein replied that he stuck with self-described ethnicity and
wasn’t able to pull out relationships other than by year of arrival.
Prof. Preston said that there is a proposal to remove the ethnic origin question
entirely from the census, but most academics are opposed.
- Did Prof. Ornstein discuss gender differences within groups? No, but every table could
have been profitably separated by gender.
- For what reasons were Iranian immigrants poor? Perhaps because they were students?
A discussion followed which touched on the relevance of stages and conditions of
immigration, and particularly language barriers. Again the limitations of standard
tabulation data were noted, as well as the imprecision of constructed and self-reported
categories like language, ethnicity and visible minority status.
- Sheine Mankovsky, City of Toronto, commented that, given the limitations of the census
data, we should still rely on complementary qualitative research. She asked whether
Statistics Canada might be able to suggest more useful ways to analyze the data for the
purposes of setting policy. Could a special program be set up to focus on relevant
problems and issues not currently addressed, such as women and/or refugees? She noted that
we shouldn’t lose sight of the purpose, which is understanding the causes of and ways
to alleviate inequities.
Prof. Preston agreed that such issues should be addressed. Dr. Beiser returned to the
example of
accessibility of health care utilization data as a relevant issue, and the question of
the feasibility of linking sets of data was raised again. In principle, it is feasible to
link data sets. Economists use employment insurance data in this way.
The problem with the approach of linking sets of data drawn from the census is that
people answer census questions believing that confidentiality will be preserved. If
individual data sets were made available, there would be privacy concerns to be resolved.
- It was pointed out that census data can still be used to suggest links between
variables.
- The question was asked whether Stats Can collects data other than at census time. There
are apparently other large-scale surveys done, but not necessarily in a way useful for
research about minorities and immigrants.
Ted Richmond pointed out that Quebec, having more flexibility in immigration
procedures, has linked more data sets and integrated more files, making individual data
available to researchers. But there are, as a result, more confidentiality concerns. He
also noted that using census data is not the only possible route, e.g. health care
providers could be asking more questions about ethnic status. Prof. Ornstein observed that
agencies such as OHIP don’t keep records for the benefit of researchers.
- The comment was made that it is difficult to see how existing local data bases can be
used for research purposes when collection standards vary so much. Social services users,
being in need, may not make good informants, and the service providers taking down the
information may not be reliable informants either (for methodological purposes).
Dr. Beiser added that, even while researchers are looking for alternative sources of
data, Statistics Canada should not be allowed to continue to design massive longitudinal
studies without better stratifying the data and getting larger samples. Prof. Preston
observed that the policy concerns of Stats Can’s main clients, the provinces, must be
reflected in the way the samples are constructed.
Recommendations for the 2001
Census
The primary recommendations included:
1) Reconsideration of the category "visible minority"
2) Making available more individual data sets
3) Lowering the cost to researchers and others
Other recommendations were as follows:
- Those responsible for constructing the categories (such as Arab/West Asian) should be
more familiar with the cultures of those geographic regions.
- Whatever changes are made in the census questions should be "collapsible
backward" so that comparability is possible over the years. Besednik noted that some
questions, such as what is a person’s "mother tongue," are not usually
subject to change.
- Given the sometimes negative connotations of "ethnic" (i.e. non-European), the
wording could be reconsidered.
- It was suggested that "spatial" categories would be useful to include, e.g.
"the 905 region."
- For some respondents, national categories are preferable to "visible minority"
status, which is confusing and even objectionable. On the other hand, as it was pointed
out, national categories alone (such as South African) can also obscure important
differences.
Consensus was that one inadequate term tends to replace another, but that the
alternative of removing questions from the census is untenable. The political question is
why the visibility of "visible minorities" is important, and to whom.
- Gender differences in data could be usefully highlighted.
- Occupational data (by other than outdated industry categories) would be useful.
- Data on refugee status might be useful, but the difficulties of getting accurate
information and not offending respondents were acknowledged.
