|
SLIDRet 4.0 User’s Manual
SLIDRET is
the SLID (Survey of Labour and Income Dynamics) data RETrieval system. It was
developed to simplify access to SLID data; users are able to create a dataset
according to their own specifications, without having to understand the
internal SLID database structure. Everything is contained in one screen.
Windows open to allow users to clarify various selections.
Users are
able to create, save and modify specifications to create cross-sectional and
longitudinal datasets containing only the variables selected. The database can
be created in two formats: a text (ASCII) flat file or a Foxpro table. SLIDRET
can also generate a data dictionary, a record layout, a file containing the
variable labels, variable codesets, and univariate statistics for each of the
variables selected. The output data file can then be read into the analytical
software preferred by the user. An additional feature is that the software
suggests a weight variable based on the specifications (if desired, this
default can be changed by the user).
You can
also specify the order in which the data are stored (e.g. in ascending order of
an identification variable) and put restrictions on the content of your
database (i.e. persons with a total income greater than 100 000 dollars).
To use all
features of SLIDRET, one must have access to the internal database (for which
access is restricted). However one can create the “query” files from the public
version of SLIDRET which contains an empty database structure. These query
files can subsequently be used for remote access (more information on remote
access in another section of this manual) or when the user has access to the
internal database.
Depending
on the user’s selections and the type of computer used, the extraction may take
some time. We anticipate that this will not be viewed as problematic since
SLIDRET is not used frequently. The primary purpose is to create a file which
will meet the needs of a specific analytical or research project.
Version 4.0.0 is a complete redesign of the SLIDRet application. The application interface is now programmed in C#.Net (rather than FoxPro). Data are kept in SQL Server 2005 databases (rather than FoxPro). The net results of this redesign are:
·
A
more user friendly interface.
·
Decreased
processing times for queries.
· Enhanced functionality
Creates a new query.
Loads
a query which has been saved from a previous session.
Saves your query.
Closes SLIDRet.
Sets user specific SLIDRet options.
o General Options
§ Preferred language:
Sets the application language. SLIDRET will remember your language preference
for your next session.
§ Check variables on first mouse
click:
Sets whether a single or a double click is required to select a variable in the
Variable Selection tab.
§ Default queries location:
Specifies the default location where queries will be saved to. The user is not
limited to this location and may specify a different save location at any time
§ Show welcome screen:
Specifies whether the SLIDRet Option welcome screen should be displayed when
the SLIDRet application starts up.
o Data sources Manager
§ Add data source
·
Data
server name:
The name of the SQL Server instance where the data reside.
The Refresh button updates the neighbouring dropdown list so that it contains
all visible SQL Server instances.
·
Data
database name:
The name of the database to be used for this data source.
The Refresh button updates the neighbouring dropdown list so that it contains
all databases on the specified SQL Server instance.
·
Dmms.xml
location:
The full path and filename of the Dmms (Data Dictionary) file to be used in
conjunction with the specified database.
Dmms files should be named using the following format: Dmms_yyyy.xml (where
yyyy represents the four digit year corresponding to the Dmms file in
question).
Only a Dmms file corresponding to the same year of the data should be used
(i.e. data for 2007 should be using the Dmms_2007.xml file).
The neighbouring folder icon permits navigation to the desired file.
·
The
Add button saves the entered data source information. Exiting the Options window,
even by using the Save and close button, without Adding a data source will
cause the data source information to be lost.
§ Current data sources
·
Displays
information for all saved data sources.
·
Data
sources must be added through the Add data source section.
·
A
data source can be deleted by first selecting it, then pressing the to the right of the column headings for this
section.
§ Save and close
·
Saves
changes made in the General Options section.
·
Does
NOT add or save data source information
· Closes the form
: Creates a new query.
: Loads a query which has been saved from a
previous session.
: Saves your query.
: Navigates to the previous tab. Is
inactive (greyed out) when unavailable (as when the current tab is the first
tab).
: Navigates to the next tab. Is
inactive (greyed out) when unavailable (as when the current tab is the last
tab).
Search is used to find variables meeting the
search criteria
:
Where search term is entered.
: Identifies parts of the variable
to search:
: Displays the results of the search
:
: Opens
Data Source options.
The user must choose
between cross-sectional and longitudinal. In general, cross-sectional analysis
refers to analysis covering only one reference year (or a series of one
reference year analysis) whereas longitudinal analysis spans over several
years.
SLID (Survey of
Labour and Income Dynamics) data cover years from 1993 to present.
SCF (Survey of
Consumer Finance) data cover years 1976 to 1997.
SLID + SCF data cover 1976 to present.
The unit of analysis identifies the unit of observation for one record on the
output file. Not all choices are available for both cross-sectional and
longitudinal analysis. The choices are:
·
Person: one
record for every person
·
Person-job:
one record for every job held (Persons with no jobs are not included in this
file.)
·
Job
absence: one record for every job absence (Persons with no jobs or with jobs
from which no absence was taken are not included in this file.)
·
Jobless
spell: one record for every period of joblessness (Persons with no such periods
are not included in this file.)
·
Marital
status spell: one for every marital state held by each person during his/her
lifetime (Complete marital history is not collected by SLID for some persons
with multiple marriages.)
·
Postsecondary
degree, diploma or certificate: one record for every postsecondary degree,
diploma or certificate received during each person’s lifetime, with some
exceptions for persons with multiple degrees. (Persons without postsecondary
educational attainment are not included in this file.)
·
Household:
one record for every household
·
Household
relationships: one record for every pair of persons in every household
(Households with only one person are not included in this file. In general, a
household with n persons will have n*(n-1) records on this file.)
·
Economic
family: one record for every economic family
·
Census
family: one record for every census family
· Monthly receipt of EI/WC/SA: one record for every type of benefit received by all persons. Benefits covered are Employment Insurance, Workers’ Compensation and Social Assistance. (Persons without receipt of any of these three benefits are not included on this file).
For Longitudinal queries
chose the panel(s)
While all combinations
of years are accepted, users requesting longitudinal (year) analysis should pay
attention to the survey design and panel rotation. Starting with 1993, a panel
of respondents is introduced every three years and stays in sample for a
six-year period. So it is possible to have one panel for analysis for six years
with half the total survey sample or two panels for analysis for three years.
So the application will only let you choose from among the following time
periods:
Three years: 1993 to
1995, 1996 to 1998, 1999 to 2001, etc.
Six years: 1993 to 1998, 1996 to 2001, 1999 to 2004,
etc.
In addition, one can
also select the year prior to these ranges (i.e., 1992) to expand the data
available for analysis. Situational data such as province of residence, marital
status and family composition are available corresponding to the end of the
year prior to the start of the panel.
The Select all years
option will check/uncheck all of the currently available reference years
options
Since the SLID database contains
many variables, the location and selection of those variables of interest may
take some time. SLIDRET uses the same variable classification into themes and
sub-themes as the SLID Electronic Data Dictionary (http://www.statcan.ca/bsolc/english/bsolc?catno=75F0026X). So, if you are familiar with the data dictionary,
you will be able to follow the same path in SLIDRET. SLIDRET also allows you to
search for a variable, and to select it for your query at the same time. To
select what you want in SLIDRET, use the computer mouse to point and click on
themes and variables of interest.
The Variables pane title is that of the
selected theme/sub-theme. The pane’s contents are variables applicable to the
theme/sub-theme selected in the themes pane. If no theme is selected the
Analysis pane will display all variables applicable to the chosen Analysis unit
Variables can be selected, one at a time, by
either single or double clicking (depending on how the corresponding option is
set in the options page) the checkbox next to a variable name. Single or double
clicking (depending on how the corresponding option is set in the options page)
a checked variable un-checks (de-selects) a variable.
Checking the Select all visible variables
checkbox will check all variables that are currently viewable (whether they are
currently visible or not), that is to say all variables belonging to the
currently selected theme are selected. Conversely un-checking the Select all
visible variables checkbox will un-check all variables that are currently
viewable (whether they are currently visible or not), that is to say all
variables belonging to the currently selected theme are de-selected.
The Selected Variables pane displays the entire
list of currently selected variables. As they are selected from within the
variables pane, variables will appear in the Selected Variables pane.
Conversely, as variables are de-selected from the variables pane they are
removed from the Selected Variables pane.
A variable can be removed from the Selected
Variables pane by first selecting it and by then press the button.
Pressing the Clear variables button will remove
all variables from the Selected Variables pane
The pane which spans the lower right of the
Variable Selection tab displays details regarding the last selected variable
(the last variable name that was clicked on in the Variables or the Selected
Variables panes. The contents of this pane are for information purposes only
and as such are read-only.
Selected Variables displays a summary listing
of all selected variables in the Variable Selection tab. Further details on the
selected/highlighted variable can be found below the list. Contents of this tab
cannot be modified. Any changes must be made at the original data entry
location.
Analysis Type as selected in the Required
Information tab. Contents of this tab cannot be modified. Any changes must be
made at the original data entry location.
Analysis Unit as selected in the Required
Information tab. Contents of this tab cannot be modified. Any changes must be
made at the original data entry location.
Themes and Entities displays a list of all
themes and entities covered by current query. Contents of this tab cannot be
modified. Any changes must be made at the original data entry location.
A listing of all the years as specified in the
Required Information tab. Contents of this tab cannot be modified. Any changes
must be made at the original data entry location.
Displays the restrictions (if any) that have
been applied to this query. Its contents are managed through the Line editor.
To modify or remove a restriction, select it and perform the corresponding action
(modify or remove) in the Line editor.
Allows for the following management of queries.
·
Add:
Specify the required fields and press the
Add/Modify button.
·
Modify:
Select the desired restriction
from the Restrictions table, make desired changes in the Line editor, and press
the Add/Modify button.
·
Remove: Select the desired restriction from the
Restrictions table and press the Remove button
Provides a detailed description for the
variable selected in the Line editor.
Allows specification of the desired file output
formats from among:
·
Univariate statistics (.ust):
Produces statistics on the distribution of the data in the resulting dataset. For numerical variables, ranges will be established based
on the weighted mean of all valid, non‑zero
values of the variable. The distribution of valid, positive, non‑zero
values of numerical variables follows this pattern:
( 0, B1 = mean divided by 2.00]
(B1, B2 = mean divided
by 1.50]
(B2, B3 = mean divided
by 1.25]
(B3, B4 = mean divided
by 1.10]
(B4, B5 = mean]
(B5, B6 = mean
multiplied by 1.10]
(B6, B7 = mean
multiplied by 1.25]
(B7, B8 = mean
multiplied by 1.50]
(B8, B9 = mean
multiplied by 2.00]
(B9, Maximum valid value
of the variable]
NOTE: Univariate
Statistics are not available for Longitudinal data
NOTE: Univariate Statistics can greatly increase the time required to complete
a query.
NOTE: All files once created do not maintain any link to the actual data in
SLIDRet. The files are a temporary snapshot of the SLIDRet data represented by
the query at the time they were created. Changes made to SLIDRet data will NOT
be automatically reflected in the files you have created.
Specifies the language used in the output file
for application generated headings or messages. This does not affect the
language of such items as variable names or variable values.
Specifies the weight to be used with the query.
Only weights applicable to your chosen Type of Analysis, Survey, Analysis Units
and Reference Years will be available. Weights are covered further in the Weights
section later in this document.
Once the query is run information regarding its
progress will be displayed here.
Once the query is run information regarding its
progress will be displayed here.
Once the query is run information regarding its
progress will be displayed here.
The information in this
section has been added as a result of questions from users.
·
The name
given to a query must begin with an alphabetic character.
·
Do not use
hyphens (-) or spaces in the name of a query.
·
The weight
used to select the sample appears in the record layout with the letter AA@ as prefix (e.g. AICSWT26 instead of ICSWT26 or
AILWCP26 instead of ILWCP26, etc.). That automatic or Asystem@ weight will not appear in the data dictionary,
unless you add it to your list of selected variables.
·
Never
delete your queries using Windows Explorer. Use instead the Delete
option from the SLIDRET Query menu. The reason is that SLIDRET
creates a list of all the names of queries you created for its internal use,
and only by deleting your queries with the Delete option will it erase
the name of those queries from its list. If you delete queries using something
else, problems may pop up later.
·
SLIDRET
gives the query name to all files created for this query. Only the file
extensions will change: *.dbf, *.cod, *.dat, etc (exception: *z.cdx).
·
To improve
the performance of SLIDRET, close as many other applications as possible.
·
You can
stop a submitted query at any time by pressing the escape key. This may stop the query or crash
SLIDRET. You then just need to reopen it again. Since your query has been
saved, you can resubmit it.
·
There are
reserved codes in SLID. The pattern followed by the reserved codes is:
...6 => Interim
Processing Code
...7 => Don’t know
...8 => Refusal
...9 => Not
applicable
You can find the codeset
associated to a particular variable in the codeset file (*.cod) produced by
SLIDRET, and in the data dictionary.
·
The year,
appropriate weight and identifiers are automatically inserted into the query
results, and do not appear in the list of variables available.
·
For the reference
year(s) selected, SLIDRET automatically inserts the cross-sectional weight
associated with the unit of analysis: for PERSON, the associated weight is
ICSWT26 (exception: if you select any
variable from the theme LABOUR, the labour weight ILBWT26 is automatically
selected); for PERSON-JOB, JOB ABSENCE and JOBLESS SPELL, the associated weight
is ILBWT26. For all other units of analysis, the default weight is ICSWT26.
·
As a consequence of the
integration of the SCF and SLID samples for the 1993-1997 period, new cross-sectional weight variables were created
this year to ease the use of weights for data users. Cross-sectional weights
are now available if one wants to use data from the SCF sample exclusively (WTCSCF26)
or if one wants to use data from the SLID sample exclusively (WTCSLD26) or if
one wants to use data from the combined sample of SCF and SLID for the years
that they overlap (ICSWT26).
Note that the labour weight ILBWT26 is a person weight, not a household integrated weight as are the cross-sectional weights ICSWT26, WTCSCF26 and WTCSLD26: be careful when you use it in estimates of families or households.
·
SLIDRET
extracts the cross-sectional sample for weights greater than zero for selected
year(s).
·
After
selecting PERSON as the unit of analysis, it is possible to select job
characteristics. For persons with more than one job during a particular year,
the job characteristics on the output file are those of the main job held
during the year (i.e., the one at which the person worked the most hours).
Those without a job would have a value of ANot applicable@ for all job variables.
·
Similarly,
for queries with units of analysis HOUSEHOLD, ECONOMIC FAMILY and CENSUS
FAMILY, it is possible to select person-level characteristics. These
characteristics on the output file pertain to the Major Income Earner (MIE) in
that group of persons for the year of interest. The MIE is the person in that
group of persons who received the greatest income among all group members. Job
characteristics of the MIE’s main job may also be included.
Here are the three possibilities:
Please, use the following statement
in the Restrict Option (see section at the end of the manual)
RESP99 equals 90
Please, use the following statement
in the Restrict Option (see section at the end of the manual)
RESP99 not equals
90
Please, use the following statement
in the Restrict Option (see section at the end of the manual)
(YEAR99 more than 1989
AND YEAR99 less than
1996
AND RESP99 equals 90)
OR (YEAR99 more than
1995
AND YEAR99 less than
1998
AND RESP99 not equals 90)
·
The units
of analysis Marital status and Education certificates
automatically give you all data in the database up to the latest year selected,
regardless of which other years you selected.
·
For
Longitudinal (year) type of analysis SLIDRET extracts the longitudinal sample
for weights greater than zero and response code RESP99 = “01” (longitudinal
person in scope) for the latest year selected in the query.
·
For
Longitudinal (panel) type of analysis SLIDRET extracts data for all the
longitudinal persons in the panel(s) selected for the years indicated
regardless of the weight values and response code.
·
SLIDRET
will automatically choose the weight that will allow you to have the largest
sample available. ILGWT26 will be used for extractions over 4, 5, or 6 years,
or when the panels do not overlap. ILWCP26 will be used to combine two panels
and study the behaviour of their population for the years they do overlap.
·
SLIDRET
users who make longitudinal (year) queries should take account of the panel
rotation when selecting the reference years covered by the query. Panel 1 started with reference year 1993, and
subsequent panels start every 3 years.
The life of each panel is 6 years.
In addition to the six full years of data, some baseline information is
available for the year prior to the start of the panel (for example, 1992 for
panel 1). Those users who select more
than 7 years or those users who select a range of years which do not correspond
to the sample rotation will get results which may not be very useful. Variables corresponding to years in which the
person was not in the SLID sample will
have a value of "not applicable" on the data file produced by
SLIDRET.
The following chart
identifies the years covered by each panel:
Panel |
Full data |
Baseline data |
1 |
1993 to 1998 |
1992 |
2 |
1996 to 2001 |
1995 |
3 |
1999 to 2004 |
1998 |
4 |
2002 to 2007 |
2001 |
5 |
2005 to 2010 |
2005 |
... |
... |
... |
And the next one
illustrates how panels overlap:
1993 |
1994 |
1995 |
1996 |
1997 |
1998 |
1999 |
2000 |
2001 |
2002 |
2003 |
2004 |
Panel 1 |
|
|
|
|
|
|
|||||
|
|
|
Panel 2 |
|
|
|
|||||
|
|
|
|
|
|
Panel 3 |
Here is a description of
the seven different units of analysis for longitudinal (year) queries.
·
PERSON: all
persons selected as a member of the active longitudinal panels who were
respondent in the last year included in the query and who were in scope in the
last year included in the query. The Ain scope@ part means that those who die, are
institutionalized or emigrate would be excluded. There would be one record per
person on the file. On a longitudinal extraction for 1993 to 1995, there are 26,875,830
estimated persons in Canada.
Notice that for all the
subsequent files the persons included are subsets of the persons on the PERSON
file.
·
PERSON-JOB: one record for every job (up to 6 jobs per year) held by
every person on the PERSON file during the years included in the query. A
person with no job in any of the years would not be included in the file. On a
longitudinal extraction for 1993 to 1995, there are 27,392,244 estimated
person-jobs in Canada.
·
JOB ABSENCE: one record for every absence (up to 2 absences per job per
year) taken from every job included on the PERSON-JOB file during the years
included in the query. A person with no job would not be included in the file.
On a longitudinal extraction for 1993 to 1995, there are 5,477,868 estimated
job absences in Canada.
·
JOBLESS SPELL: one record for every jobless spell experienced by each
person included in the PERSON longitudinal file during the years included in
the query (up to 7 jobless spells per year). A person without a jobless spell
in all years would not be included in the file. On a longitudinal extraction
for 1993 to 1995, there are 16,154,065 estimated jobless spells in Canada.
·
MONTHLY
RECEIPT OF EI/WC/SA: one record for each person (up to 3 record per year) included
in the PERSON longitudinal file for each program from which benefits were
received during the years included in the query. A person who does not receive
any of these benefits during any of the years would not be included in the
file. On a longitudinal extraction for 1993 to 1995, there are 8,590,762
estimated monthly receipts in Canada.
·
MARITAL
STATUS: one record for every marital spell experienced by every person included
in the PERSON longitudinal file. Every person will have at least one marital spell
(i.e., single). On a longitudinal extraction for 1992 to 1995, there are
49,544,035 estimated marital status in Canada.
·
EDUCATIONAL
CERTIFICATE: one record for every post-secondary degree, diploma or certificate
received by every person in the PERSON longitudinal file. A person without such
a certificate would not be included on the file. On a longitudinal extraction
for 1992 to 1995, there are 14, 076,174 estimated education certificates in
Canada.
For Longitudinal (panel) type of analysis files
have the same content as described above but for all the longitudinal persons
in the panel(s) selected for the years indicated regardless of the weight
values and response code.
Combined SCF/SLID sample integrated
cross-sectional weight. This weight takes into account the respondents of the
two surveys combined (Survey of Consumer Finances and Survey of Labour and
Income Dynamics) for the overlapping years of 1993 to 1997. For the years prior
to 1993, this weight is equivalent to the cross-sectional weight applicable
when using only the SCF sample (see variable WTCSCF26). For the years after
1997, this weight is equivalent to the cross-sectional weight applicable when
using only the SLID sample (see variable WTCSLD26). This is an integrated
weight, which is to say that all persons in the same household receive the same
weight. For estimates of persons, the weight is counted once for each person in
the household. For estimates of families or households, the weight is counted
once for each family or household.
Most recent revisions: With the release of data
for 2003, the weights were revised for 1990 to 2002. With this revision a new
calibration strategy was implemented which included adding the number of
earners as an additional dimension of the calibration. The number of earners
was obtained from the Canada Revenue Agency T4 files. In addition, population
control totals used in the calibration were updated based on the 2001 census.
With the release of data for 2004, the weights were revised for 2003. The
calibration of the original weights for 2003 used modeled T4 data because the
final T4 data was not yet available. The revision of the 2003 weights the
following year (with the release of data for 2004), used the final T4 data in
the calibration step. With the release of data for 2005, the weights were
revised for 2004. The calibration of the original weights for 2004 used modeled
T4 data because the final T4 data was not yet available. The revision of the
2004 weights the following year (with the release of data for 2005), used the
final T4 data in the calibration step. With the release of data for 2007, the
weights for years 1993 to 1997 were revised to take into account the combining
of the SLID and SCF samples.
Regular integrated cross-sectional weight for
SCF. This weight is valid for cross-sectional estimates using observations from
the Survey of Consumer Finances only and is available for all years of that
survey, i.e., years 1976 to 1997. See also variable ICSWT26. This is an
integrated weight, which is to say that all persons in the same household
receive the same weight. For estimates of persons, the weight is counted once
for each person in the household. For estimates of families or households, the
weight is counted once for each family or household. Most recent revisions:
With the release of data for 2003, the weights were revised for 1990 to 1997.
Regular integrated cross-sectional weight for
SLID. This weight is valid for cross-sectional estimates using the Survey of
Labour and Income Dynamics only and is available for all years of that survey,
i.e., 1993 onwards. See also variable ICSWT26. Note that the SLID sample consisted
of only one panel of respondents from 1993 to 1995, while since 1996 it
consists of two panels of respondents; the weight takes this into account. This
is an integrated weight, which is to say that all persons in the same household
receive the same weight. For estimates of families or households, the weight is
counted once for each family or household.
Most recent revisions: With the release of data
for 2003, the weights were revised for 1993 to 2002. With this revision a new
calibration strategy was implemented which included adding the number of
earners as an additional dimension of the calibration. The number of earners
was obtained from the Canada Revenue Agency T4 files. In addition, population
control totals used in the calibration were updated based on the 2001 census.
With the release of data for 2004, the weights were revised for 2003. The
calibration of the original weights for 2003 used modeled T4 data because the
final T4 data was not yet available. The revision of the 2003 weights the following
year (with the release of data for 2004), used the final T4 data in the
calibration step. With the release of data for 2005, the weights were revised
for 2004. The calibration of the original weights for 2004 used modeled T4 data
because the final T4 data was not yet available. The revision of the 2004
weights the following year (with the release of data for 2005), used the final
T4 data in the calibration step.
Cross-sectional weight for the person, adjusted for
non-response to questions on non-income variables such as labour market
activity and education. It is adjusted for more cases of non-response than is
the regular integrated cross-sectional weight (ICSWT26), thereby minimizing the
percentage of persons with missing values ("Don't know") after
weighting. Nonetheless, there are still missing values since only minimal
imputation is performed for non-income variables.
Most recent revisions: With the release of data for 2003,
the weights were revised for 1993 to 2002. With this revision a new calibration
strategy was implemented which included adding the number of earners as an
additional dimension of the calibration. The number of earners was obtained
from the Canada Revenue Agency T4 files. In addition, population control totals
used in the calibration were updated based on the 2001 census. With the release
of data for 2004, the weights were revised for 2003. The calibration of the
original weights for 2003 used modeled T4 data because the final T4 data was
not yet available. The revision of the 2003 weights the following year (with
the release of data for 2004), used the final T4 data in the calibration step.
With the release of data for 2005, the weights were revised for years 1993 to
2004 to incorporate a new strategy to deal with non-response.
Longitudinal weight for the person. Applicable to one
panel of respondents.
Most recent revisions: With the release of data for 2003,
the weights were revised for 1993 to 2002. With this revision a new calibration
strategy was implemented which included adding the number of earners as an
additional dimension of the calibration. The number of earners was obtained
from the Canada Revenue Agency T4 files. In addition, population control totals
used in the calibration were updated based on the 2001 census.
Longitudinal weight for the person
which takes into account the respondents of the two panels combined.
Most recent revisions: With the release of data for 2003, the weights were revised for 1993 to 2002. With this revision a new calibration strategy was implemented which included adding the number of earners as an additional dimension of the calibration. The number of earners was obtained from the Canada Revenue Agency T4 files. In addition, population control totals used in the calibration were updated based on the 2001 census. With the release of data for 2006, the weights were revised for years 1993 to 2005 to correct a problem with calibration totals used. For information on combined panel weighting, see the paper called " Combined-panel longitudinal weighting, Survey of Labour and Income Dynamics" by Jean-François Naud. Statistics Canada Catalogue no. 75F0002MIE, No. 008.