Telling Canada’s story in numbers
Scott McLeish
Social and Aboriginal Statistics Division
UNECE Work Session on Migration Statistics
October 30-31, 2017
2016 Census of Population of Canada: Integration of
immigration administrative data
www.statcan.gc.ca
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA2
Presentation overview
• Census immigration data
• Administrative data
• 2016 Census developments
• Record linkages
• Improved processing
• New admission category variables
• Summary and next steps
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA3
The Census of Population is an important source of
information about immigrants in Canada
• The Canadian Census of Population collects several variables
related to immigration through the questionnaire:
• Immigrant status and year of immigration
• Place of birth
• Citizenship
• Generation status (based on place of birth of parents)
• In addition, several socio-economic topics are covered to
permit detailed analysis of immigrants in Canada
• Demographics
• Income, labour and education
• Housing, households and families
• Languages
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA4
Census processing
• Consistency edits and donor imputation are performed to
address missing, invalid or inconsistent values
• Missing, invalid or inconsistent values could be the result
of different factors:• Respondents may not understand the question (e.g. year of immigration vs. year
of arrival)
• Respondents may not remember their exact year of immigration
• Proxy responses (the person filling out the questionnaire may not know the true
values for everyone in the household)
• Ambiguous responses which could refer to multiple values (e.g. ‘Ireland’)
• Changing boundaries and place names
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA5
Donor imputation methods are used for
missing values• Goal: to find the nearest neighbour (donor)
• How do we define the “nearest neighbour”?
• Determine a list of variables that are correlated with the missing variable
(matching variables)• Languages
• Age
• Completed immigration questions
• Other ethnocultural variables
• Geography
• Determine the relative importance (weight) of the matching variables
• Calculate the dissimilarity between the record to be imputed and the
potential donor records using the matching variables
• For most census immigration questions, less than 2% of
records require imputation
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA6
Administrative data provides detailed information
on the flow of immigrants to Canada
• Immigration, Refugees, and Citizenship Canada (IRCC)
collects data on all immigrants when they are admitted to
Canada
• Detailed information on admissions since 1980
• Includes variables also collected on the census questionnaire:• Immigrant status and year of immigration
• Place of birth
• Includes additional details on conditions at time of admission• Immigrant admission category (e.g. economic immigrant, refugee, etc.)
• Intended occupation, knowledge of official languages, previous experience in Canada
• No updates are made and data does not account for deaths or
emigration
• This data is linked with annual tax files to create the
Longitudinal Immigration Database (IMDB)
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA7
2016 Census: Record linkage to combine both
data sources• Similar to linkages done in 2006 and 2011, the 2016
Census was linked with the administrative data:• Immigrants from 1980 to 2016
• Immigrants from 1952 to 1979• Less detailed records
• Temporary residents• Temporary foreign workers
• International students
• Asylum claimants
• About 90% of immigrants who landed from 1980 to 2016 were linked
• Estimated 0.3% false positives
2017-10-278
Administrative data illustrated effects of inconsistent
responses and corresponding consistency edits
7.6% of respondents who were linked to administrative immigration records responded “No” to the immigrant status question. This proportion increases the longer they have been in Canada. These cases are largely corrected during processing based on other responses (e.g. reporting going through the immigration process to become Canadian citizens.
0.0%
2.0%
4.0%
6.0%
8.0%
10.0%
12.0%
14.0%
36 34 32 30 28 26 24 22 20 18 16 14 12 10 8 6 4 2
Years since being admitted (from linked administrative records)
Percentage of respondents linked to IRCC immigration records who respond that they are not immigrants by years since landing
2016 Census 2011 NHS 2006 Census
2017-10-279
Year of immigration requires more imputation due to
consistency edit for immigrant status question
0
2
4
6
8
10
12
14
Place of birth Citizenship Immigrantstatus
Year ofimmigration
Place of birthof father
Place of birthof mother
Imputation rates for ethnocultural variables (%)
2006 Census 2011 NHS 2016 Census
Note: The use of IRCC data in donor selection necessarily ensures that the imputed values will be more similar to IRCC
values. However, these were not imputed deterministically - the linked IRCC values were only used to select a donor.
2017-10-2710
Administrative data improved donor selection by
introducing strongly correlated matching variables
63.9%
7.5%
14.7%
13.8%
14.8%
35.0%
6.5%
43.8%
0.0%
20.0%
40.0%
60.0%
80.0%
100.0%
120.0%
2016 Census 2011 NHS
Consistency of imputed years of immigration with linked IRCC values
Same 1 year different 2-5 year different More than 5 years different
63.9% of imputed 2016 Census year of immigration values matched their linked IRCC values. This compares with only 7.5% from the 2011 NHS.
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA11
Administrative data provided new analytical variables
on immigrant admission category
• In collaboration with IRCC, new variables were added to
the 2016 Census regarding admission category of
immigrants
• Immigrants are admitted into Canada for different reasons:
• Economic immigrants
• Sponsored family members
• Refugees
• Analysis can connect the conditions under which
immigrants were admitted with socio-economic outcomes
from the census
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA12
Economic immigrant categories
Economic immigrant
Federal investor
Quebec investor
Federal self-employed
Quebec self-employed
Federal skilled worker
Quebec skilled worker
Federal entrepreneur
Quebec entrepreneur
Provincial and territorial nominee
Business programs
Canadian experience class
Skilled trades worker
Caregiver
Entrepreneur
Investor
Self-employed
Worker programs
Skilled worker
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA13
Refugee categories
Blended visa office-referred refugee
Privately sponsored refugee
Government-assisted refugee
Dependant abroad of a protected person in Canada
Protected person in Canada
Resettled refugee
Refugee
Protected person in Canada
or dependant abroad
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA14
Similar to other census variables, consistency edits
and donor imputation methods were developed
• Imputation was stratified based on family characteristics
• Couples (especially with matching years of immigration)
• Children and parents
• ReliefF algorithm was used to determine the relative weight of
matching variables when selecting donors
• Non-parametric method using similar assumptions as nearest-
neighbour imputation: records with similar matching variables should
come from the same class
• More predictive variables are given more importance when selecting a
donor• Matching variables which are different between classes have weights increased
• Matching variables which are different within classes have weights decreased
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA15
Matching variables by predictive value (descending)
Adults in a couple
Adults not in a couple
Place of birth
Year of immigration
Spouse's admission cat.Year of immigration
Place of birth
Age at immigration
Province of residence
Sex
Subsidized housing
Spouse has same year of
immigration
Spouse has a different year
of immigration
Age at immigrationMarital status
Highest level education
Knowledge off. languages
Province of residence
Occupation
Sex
Spouse's admission cat.
Spouse's applicant type
Place of birth
Year of immigration
Age at immigration
Highest level education
Occupation
Province of residence
Knowledge off. languages
Spouse's year immigration
Highest level education
Spouse's applicant type
Sex
Knowledge off. languages
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA16
During testing, imputation results were in line with
original results even for small subpopulations
Source: Statistics Canada, Edit and imputation test data, 2011 National Household Survey and Immigrant landing file linkage
0%
10%
20%
30%
40%
50%
60%
70%
19
80
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
Year of immigration
Economic immigrant Immigrant sponsored by family Refugee Other immigrant
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA17
Final 2016 Census results were certified against
other data sources
0
50,000
100,000
150,000
200,000
250,000
300,000
19
80
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
20
15
20
16
Year of immigration
Number of immigrants by year of immigration and data source
2016 Census Administrative count (upper bound) 2014 tax filers (lower bound)
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA18
Admission category variables aligned with
other data sources after imputation
0
20,000
40,000
60,000
80,000
100,000
120,000
140,000
160,000
19
80
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
20
15
20
16
Year of immigration
Number of skilled workers by year of immigration and data source
2016 Census Administrative count (upper bound)
2014 tax filers (lower bound) Weighted linkage results
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA19
Summary and next steps
• 2016 Census immigration data were enhanced with
integrated administrative data
• Improvements made to processing
• New variables add analytical value to important data source for
immigrants in Canada (collaboration with IRCC)
• Immigration results were released October 25th, 2017• Including new admission category variables
• Reference material released providing context and explanations of
variables and comparability with past results
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA20
Acknowledgements
• Partners at Immigration, Refugees and Citizenship
Canada
• Lorna Jantzen and colleagues
• Social Survey Methods and Household Survey Methods
Divisions• Andrew Stelmack, Lyne Guertin, Sean Crowe, Piotr Biernot, Caroline
Pelletier, Paul Cascagnette, and Colin Babyak
• Social and Aboriginal Statistics Division• Kathryn Spence, Eric Mongrain, Laetitia Martin, Hélène Maheux,
Mireille Vézina, Chantal Poirier, Hyunji Lee, and Jarod Dobson
2017-10-27STATISTICS CANADA • STATISTIQUE CANADA21
Contact information
Scott McLeish
Chief, Census and Administrative Data
Social and Aboriginal Statistics Division
Statistics Canada
Ces diapositives sont aussi disponibles en français.