Explaining the Gender Wage Gap: The role of worker and
Download
Report
Transcript Explaining the Gender Wage Gap: The role of worker and
Workplace and Employee Survey
Marie Drolet
WES Research Manager
Business and Labour Market Analysis Division
Statistics Canada
[email protected]
Outline of today’s presentation
General survey overview
•
•
•
•
•
•
Why a linked survey?
Survey content
Potential research questions
Collection
Methodology
Data Access
Comparative Research:
• Can the workplace explain gender pay differentials?
Goals of WES
To develop an ongoing survey that will
link workers and workplaces at the micro level
provide information from both the demand and
supply sides of the labour market
==> enriching research studies
provide longitudinal information allowing
researchers to control for both individual and
workplace effects that are not possible in other
data sets
improve survey infrastructure
Survey Content
Employee outcomes:
Hours polarization;
Wages;
Training received;
Workplace characteristics:
Technology implemented;
Operating revenues,
expenditures, payroll,
Employment; hiring, vacancies
Business strategies;
Work organization
Compensation schemes;
Training provided;
Occupation mix
Organizational change;
Subjective measures of
productivity, profitability
Type of competition
Worker/job characteristics;
Education;
Age/gender;
Occupation, management
responsibilities;
Work history, tenure;
Family characteristics;
Unionization;
Use of technology;
Participation in decision
making;
Wages and fringe benefits;
Work schedule/arrangements;
Training taken
Workplace outcomes;
Employment, revenue growth;
Organizational change;
Implementation of technologies.
Potential research questions
Are unionized workers more actively involved in
workplace decision-making and employee
participation programs?
Are there industrial sectors that are replacing less
skilled workers with higher skilled workers?
What are the characteristics of workers leaving
their jobs, thereby creating job vacancies?
Do alternative work practices such as job rotation
and participation in work groups reduce quit rates?
Methodology:
Target Population
All business locations in Canada that have
paid employees EXCEPT employers in
•
•
•
•
•
Yukon, Northwest Territories
agricultural, fishing, hunting, trapping,
public administration
religious organizations
military
Employee content
• receives a T4 slip from Revenue Canada
Methodology:
Sampling frame
Stratified 2 stage design
Workplace component
• physical location of business with paid workers
• frame stratified by industry, region, size
• size cut-offs are different for industry / region
combinations ==> model based approach
• sampling weights assigned to each unit = the
inverse probability of selection with adjustment*
Worker
Component
• lists of employee made available by employers
Methodology:
Longitudinal strategy for workplaces
*** up to 6 years
Methodology:
Longitudinal strategy for workers
*** 2 years
Methodology:
Weighting and Estimation
To produce estimates from a sample that
relates to a population of interest requires
the use of survey weights
Weights are adjusted for
• complete non-response
• stratum jumpers
• calibration or “benchmark” to known totals from the
Survey of Employment, Payroll and Hours
• linked analysis weights adjusted for live units with
no responding employees
Methodology:
How to compute variances that take into
account the complex survey design of WES?
Not done in most statistical packages
Use bootstrap weights provided
• (100 in total for each observation, each based on 50
iterations)
• idea: re-sampling technique to capture variability
To calculate the variance
• produce an estimate based on each set of bootstrap
weights
• compute the variance estimate
• apply an adjustment to make the variance design
consistent (50/100)
Methodology:
Why is taking account of the complex survey
design of WES SOOOO important?
NOT taking into account of the design effect
results in an UNDERESTIMATION of the
variance
Hypothesis testing and constructing confidence
intervals requires accurate standard errors
May erroneously report that a statistic is
significantly different from zero when it is not.
Rules of Thumb: NO
Response Rates
Employer
Employee
1999
96.5%
83.3%
2000
95.8%
86.8%
2001
92.7%
87.8%
Data Access
Research Data Centres
• SSRHC application
• evaluated by 3 member peer committee
• judged on scientific merit, viability of methods,
appropriateness of data
• deemed employees of STC
• research falls within the mandate of STC
Remote Access
•
•
•
•
proposal sent to WES research manager
dummy data set sent to researcher
programs sent to Statistics Canada
designed primarily for multivariate analyses
Important websites
Statistics Canada: www.statcan.ca
WES site
• General Information:
– http://www.statcan.ca/english/survey/business/workpl
ace/workplace.htm (English)
– http://www.statcan.ca/francais/survey/business/work
place/workplace_f.htm (French)
• Questionnaires (1999-2002)
– http://www.statcan.ca/english/concepts/wes.htm
– http://www.statcan.ca/francais/concepts/wes_f.htm
• Workplace Evolving Series
– http://www.statcan.ca/english/freepub/71-584MIE/free.htm
– http://www.statcan.ca/francais/freepub/71-584MIF/free_f.htm
Can the workplace explain Canadian
gender pay differentials?
Marie Drolet
Canadian Public Policy, Summer 2002
Evolving Workplace Series, No
Objectives
To move beyond ‘traditional’ analyses by
incorporating workplace characteristics in
the wage outcomes of men and women
household data = focus on worker
To determine the contribution of the
workplace in ‘explaining’ the gender
wage gap
linked employee-employer data = focus on
gender segregation
Determinants of wages
Usual suspects
• human capital, demographic, job
New WES variables
• high performance workplace practices
» self directed workgroups
» performance pay
•
•
•
•
•
•
•
foreign ownership
non profit organizations
quantity and timing of labour demanded and supplied
required & actual education match
training costs per employee
workplace rate of part-time employment
industry / occupation / firm size
Do wages differ by workplace
characteristics?
expected robust association between wages
and ‘established’ variables
impact of workplace variables
(-) percent working part-time
(+) training expenditures per employee
(+) foreign ownership
(+) self directed workgroups
(+) workers receiving performance-based pay
(+) quantity and timing of labour
(+) undereducated
(-) overeducated
(NS) non-profit organizations
Are the workplaces of men and
women different?
* significantly different
Men
Women
Average workplace rate
of part-time employment*
19.7%
33.7%
Self-directed workgroup*
36.3%
29.0%
Non-profit*
13.9%
28.6%
Foreign held*
11.0%
5.5%
$257
$247
Over-educated
36.4%
38.4%
Under-educated
16.3%
14.5%
Training expenditures
Main Finding #1:
Women are concentrated in low wage
workplaces
When usual suspects are taken into
account ….
Pooled OLS model:
• women earn 15% less men when there are NO
controls for the workplace
Workplace fixed effects model*
• women earn 8% less than men when controls for
the workplace are included
Main Finding #2:
Industry measure in WES ‘explains’ more of
gender pay differentials
Using ONLY Worker characteristics +
industry & occupation
– 58% of gender wage gap ‘explained’
– industry ‘explains’ 34% of gender wage gap
– significantly different from other studies
• Using Survey of Labour and Income Dynamics 1997:
– 50% is explained
– industry accounts for 15%
Main Finding #3:
The workplace accounts for more of the gap
than the worker
Excludes industry & Includes industry &
occupation
occupation
Total
100.0
100.0
Worker
characteristics
10.7
18.7
Workplace
characteristics
25.8
41.4
“Unexplained”
63.5
39.9
Main Finding #4:
The contribution of specific workplace
characteristics in ‘explaining’ the gap
based on Oaxaca decomposition method, base = male pay structure
Excludes industry Includes industry
& occupation
& occupation
Industry
19.7
Workplace part-time rate
17.7
10.9*
Self-directed workgroups
2.5
2.1
Performance-based pay
2.3
2.2
Foreign held
2.4
2.2
Non-profit
-1.0
1.3
Timing of labour
2.2
1.4
Training expenditures
0.3
0.2
Job-education mismatch
1.8
1.0
Main Finding #5:
Despite addition of new variables, a
substantial portion of gap is unexplained
In most detailed specification about 40% of the gap
is unexplained
Adjusted: women earn 92% of male average hourly
wage rate
Differs from analyses using SLID
• unexplained: 51%
• adjusted: 89%
Important from a public policy
perspective
Current research necessary since policies tend
to address different components of the gap
Knowledge of contribution of workplace to
gender pay differentials is essential since
workplace contributes more to explaining
differentials than the worker