Survey of Labour and Income Dynamics SLID All Congress Symposium The riches of the Research Data Cen

1
Survey of Labour and Income Dynamics (SLID) All
Congress SymposiumThe riches of the Research
Data CentresJune 2, 2004University of
Manitoba, Winnipeg Heather LatheIncome
Statistics Division
2
Points to cover

• Survey design highlights, content, analytical
  uses
• New content as of 2002
• Getting started SLID retrieval system (SLIDRET)
• Also available in photocopies
• Variance calculation bootstrap
• Sources for more information

3
Original objectives of SLID

• Labour market and income flows
• Determinants of change
• Impacts on the family
• Also
• Main source for cross-sectional income data

4
Design highlights

• Longitudinal data, among the first at Statistics
  Canada
• Labour market and income data together
• A wide variety of additional explanatory
  variables
• Family make-up and changes are key
• Timeframe
• Individuals followed for 6 year period
• Annual contact to update information

5
Sample design

• Up to 2 interviews per year
• Labour in January and Income in May if no T1
  access permission
• Preliminary information asked during first
  contact

6
Who is interviewed?

• Longitudinal respondents selected at start of
  panel
• Cohabitants also interviewed (every member of the
  household)
• Movers followed (longitudinal)
• Labour, education and income collected for
  persons 16, labour not collected 70
• Housing variables collected for every in-scope
  household, disability for every person
• 10 provinces, non-institutional, off-reserve

7
2001 sample sizes
8
Sample size by CMA 2001 households

• Toronto 1 546
• Montreal 1 326
• Winnipeg 1 117
• Vancouver 957
• Ottawa-Gatineau 850
• Edmonton 737
• Calgary 653
• Kitchener
  501
• Halifax 555
• St.Catharines-Niagara 455
• Quebec City 523
• London 418
• Hamilton 405
• Victoria 321

9
SLID content

• Over 1000 variables
• Database organized into logically related groups
  of variables
• Database key variables
• person
• job
• job-absence
• economic family
• household
• household relationships
• monthly receipt of EI / WC /SA
• jobless spell
• census family
• marital status
• education certificates

10
Organization of content
11
Employment and unemployment dynamics research

• Gross changes in employment, unemployment and
  inactivity between months or years
• Gross flow data of persons or jobs by industry,
  occupation, worker characteristics and job
  characteristics
• Duration of spells (jobs or unemployment)
• Examples
• To what extent are long spells of unemployment
  experienced by the same individuals?
• Why do people withdraw from the labour market
• What precedes a transition into self-employment?

12
Life-cycle labour market transitions

• Labour market transitions associated with
  particular stages of life cycle
• Examples
• Transitions from school to work
• Transitions from work to retirement
• Work absences taken to raise children
• What are typical life-cycle patterns in Canada
  today?
• What are the subsequent activities of high school
  drop-outs, and what precedes a return to school?

13
Job quality

• Examples
• Wage differences between men and women
• Job benefits
• Underemployment
• Career change
• Career advancement and earnings
• Job polarization, e.g., in terms of wages and
  hours worked

14
Family economic mobility

• Examples
• How stable is family income?
• What proportion of families experience a
  significant improvement or deterioration in
  income between two points in time?
• What are the determinants of these changes?
• How important are changes in family composition
  (divorce, remarriage) in explaining a change in
  financial well-being?

15
Dynamics of low income

• Incidence and duration of spells of low income
• Factors associated with moves into or out of low
  income
• Examples
• Identifying the existence or otherwise of a
  persistently poor sub-population, and
  associated characteristics
• Studying flows into and out of low income in
  relation to government transfer programs and
  taxation

16
Life events and family changes

• Identifying the various types of families and how
  families change, including step-families and
  multi-generational families
• What are the determinants or impacts of life
  events?
• Example
• - What are the familys economic circumstances
  preceding a marriage break-up, and what are they
  for each spouse and any children following a
  separation?

17
Educational advancement and combining school and
work

• Possible to view educational activity in the
  context of an individuals other activities or
  family circumstances
• Examples
• Financial well-being, or family responsibilities,
  of people pursuing post-secondary education
• To what degree do high school or post-secondary
  students combine work and school
• What is the labour market involvement of high
  school drop-outs and what precedes a return to
  school

18
New content

• Starting with 2002
• Housing
• Education/training
• Geography new base

19
Housing content

• From 1994
• dwelling tenure (rented/owned) and type
• From 1999
• with/without mortgage number of bedrooms
• From 2002
• the need for repairs
• farm or home business operated from the property
• Homeowners shelter costs
• mortgage payments, utility costs
• property taxes, condominium fees
• Renters shelter costs
• monthly rent, utility costs
• amenities included in the rent (parking,
  appliances, etc.)
• whether the rent is subsidised by government or
  an employer

20
New housing research examples

• Do changes in employment trigger a housing move?
• Is a new job or promotion associated with a move
  to more suitable, affordable or adequate housing?
• What proportion of families experience a
  significant improvement or deterioration in
  housing between two points in time?
• How important are changes in family composition
  (divorce, remarriage) in explaining a change in
  suitability, affordability or adequacy of housing?

21
Education coverage

• More complete in 2002
• secondary school
• community college
• business school
• vocational school
• university
• courses, workshops, seminars
• training related to employment

22
Education collected characteristics

• Main subject / subjects of courses
• Reason for choosing area of study
• Goals related to current or future employment
• Employer support
• Number of months
• Number of weeks
• Number of hours per week
• Total number of hours
• Full time / part time
• Area of studies - code
• Receipt of certificate or diploma

23
New education-related research

• Impact of programs offered by elementary and high
  schools, trade schools, colleges, CEGEPs and
  universities
• gender, level of education, age, job tenure,
  labour market experience and occupation.
• Life-long learning
• Does training have impact on career progress or
  income?

24
Geography

• Census 2001 geography introduced
• 1992-1998 keeps 1991 geography
• From 1999 new geography
• Some variables needs to be recoded back to 1992
• Old 1991 geography
• Old information for recoded variables and years
  1999-2001 stored in new old geography variables

25
Access to confidential microdata

• Custom data retrievals
• On-premises access
• Research Data Centers
• Remote access
• SLIDRET in RDCs

26
SLID database retrieval software (SLIDRET)

• Produces a flat rectangular file in Text (ASCII)
  format to eliminate the requirement to understand
  the database structure
• By producing this study file, you get started on
  some of the main choices for your study

27
SLIDRET basic inputs

• Type of analysis longitudinal or
  cross-sectional, and years of study.
• Unit of analysis Person, Person-job, Family,
  Various spells , etc. (11 choices in all).
• Browse the variable descriptions
• Loads the appropriate weight field according to
  other specifications.

28
SLIDRET variable choice

• Pick variables of interest.
• Can add weights (on top of default one)
• If desired, possible to restrict the population
  and sort the file.
• Note SLIDRET prevents the user from making
  inconsistent choices.

29
SLIDRET outputs

• Output file in text format
• Output file in dbf format
• Record layout
• Data dictionary
• Variable labels
• Code sets
• Univariate statistics
• Query - saved for future use or modifications

30
Variance estimation

• SLID uses bootstrap variance estimation
• Bootstrap - replication method (process done 1000
  times)
• n is the number of PSUs in a stratum
• (n 1) PSUs selected by Simple Random Sampling
  with replacement in each stratum to obtain a
  bootstrap sub-sample
• Set of bootstrap weights calculated

31
Bootstrap weights

• Bootstrap weights are available for SLID
  reference years 1996-2001
• For all types of weights
• Cross-sectional weight (ICSWT26)
• Labour weight (ILBWT26)
• Longitudinal weight for each panel (ILGWT26)
• Combined panels longitudinal weight (ILWCP26)
• 1000 sets of weights

32
When to use bootstrap

• Bootstrap variance estimation can be done for all
  direct estimates
• totals, means, proportions, medians, percentiles,
  parameters from a regression, etc
• Works for estimation of fixed effects does not
  for random effects
• technical reason - few PSUs in first stage of
  sampling
• E.g.,
• multilevel modeling
• random effects models

33
Why to use bootstrap

• Bootstrap variance estimation accounts for
  complex survey design, specifically
• cluster sample
• non-uniform response mechanism
• Use it for population models
• to account for sample design

34
How to estimate the variance

• Calculate regular estimate ?
• Execute the estimation program using each set of
  bootstrap weights (1000 sets) and keep each
  resulting estimate ?i, i1, 2, , 1000.
• The following formula estimates variance

35
Which bootstrap weights to use

• Always use the bootstrap weights that correspond
  to the weight used to produce the estimate
• if an estimate is produced using the
  cross-sectional weight (ICSWT26) for 2001, then
  the bootstrap cross-sectional weights for 2001
  should be used for variance estimation
• In order to maintain the maximum precision, all
  1000 weights should be used, especially for
  estimations on small domains

36
More info available

• A document on how to calculate variance
  estimation is available in RDCs
• Some SAS macros and related documentation, as
  well as examples of programs doing bootstrap
  estimation with SLID data, will be available in a
  near future
• More information from Longitudinal Data Analysis
  Group at Statscan

37
Survey overview

• Like internet home page just for SLID
• www.statcan.ca Our products services
  Free Personal finance household finance
• Overview
• Products services
• Notes, definitions, methodology
• Questionnaires
• Research papers
• Data dictionary

38
For more information on SLID
Contact the Client Services Section at
income_at_statcan.ca or toll-free at
1-888-297-7355 or 951-7355