Modelling Labour Supply Issues and Challenges Labour Supply

Modelling Labour Supply : Issues and Challenges Labour Supply Monitoring and Forecasting Workshop Vancouver, British Columbia October 17 -18, 2007

Context § § § Occupational supply modelling is currently done only at the national level All demand side components (expansion and replacement demands) are currently produced at both the national and provincial levels Interest has been expressed in creating provincial supply models 2

Purpose of this presentation § Describe the current labour supply modelling of COPS § Highlight the current issues and challenges (data, models) surrounding the production of national level labour supply projections § Highlight the additional challenges in undertaking labour supply projections at the provincial level 3

Current modelling structure of COPS 4

Uses and dissemination § The ability to do forward-looking analysis is considered essential in improving HRSDC’s capacity to deliver effective policy decision making. – Results from the projections are presented in “Looking-Ahead: a 10 -year Outlook for the Canadian Labour Market”. § Results from the projections are also used in the development of planning and decision-making tools. – HRSDC provides data and projections to Service Canada for the production of Job Futures. – Labour demand projections are also provided to the provincial governments and Service Canada’s regional offices, which use them to develop their own specific LMI products. 5

COPS modelling structure Aggregate Macroeconomic and Labour Market Indicators Demographics Employment by Industry Labour Force Expansion Demand Retirements School Leavers Immigrants Future Labour Market Imbalances by Broad Skill Level Future Labour Market Imbalances by Occupation Change in employment versus change in labour force by broad skill level Job openings versus new job seekers by occupation 6

School Leavers Model 1. Graduates by educational level 2. Graduates by major field of study 3. Allocate in or out of the labour force 4. School leavers allocated by occupation Graduates Drop Outs* FOS 1 In LF FOS 2 FOS 3 FOSi Out LF NOC 1 NOC 2 NOC 3 NOCn 7

1. Graduates by educational level -Graduates are projected by level of education -6 different levels of education are estimated, 2 are obtained residually: • Secondary • Trade and vocational • College • Bachelor • Master • Ph. D Drop outs Less than high school Some post-secondary The level of enrolments is projected based on source population, as well as economic factors. Graduates are then estimated using enrolments. 8

2. Graduates by major field of study Graduates are distributed by FOS according to the observed average distribution of the last 3 years (PSIS/USIS/CCSIS). • Fixed proportion over the projection period. FOS 1 Graduates FOS 2 FOS 3 FOSi No field of study for the following: • Less than high school • Secondary school graduates • Some post-secondary At the post-secondary level: • 49 FOS for Trade and vocational • 55 FOS for College • 58 FOS for the three university levels (Bachelor, Master and Doctorate). 9

3. Allocate in or out of the labour force § By field of study, the share of graduates in the labour market is determined using: FOSi – The Labour Force Survey for: • Less than high school (15 -24 years old); • High school grads (15 -24 years old); • Some post-secondary (20 -29 years old). – The National Graduate Survey for: In LF Out LF • Trade and vocational; • College; • University (Bachelor, Master and Doctorate). 10

4. Allocate by occupation Data sources: A) For the 3 non post-secondary levels: Labour Force Survey (LFS) B) For the 5 post-secondary levels: National Graduates Survey – In 2002, graduates from 2000 – In 1997, graduates from 1995 Other sources considered: § Job Futures; § Occupation descriptions from the NOC; § Matrix of Skills Transferability by Margaret Roberts (2003). In LF NOC 1 NOC 2 NOC 3 NOCn Ex-ante distribution: allocation not allowed if unrelated to FOS or skill level too low 11

4. Allocate by occupation It is assumed that each school leaver will offer his/her service to jobs related to his/her education level and field of study § E. g. : Bachelor graduates in nursing are constrained to occupy a job such as: – – – Nurse Supervisors and Registered Nurses (NOC 315); Other Technical Occupations in Health Care (NOC-323); Assisting Occupations in Support of Health Services (NOC-341); Manager in Health (NOC-031); Medical Technologists and Technicians (NOC-321). § and not occupy jobs like: – Managers in Retail Trade; – Cashier. 12

Immigration model 1. Projection of the number of immigrants 2. Allocate in or out of the labour force 3. Allocate by occupation* Immigrants In LF Out LF NOC 1 NOC 2 NOC 3 NOCn *Allocation based on information for recent immigrants (from the latest census) 13

Current issues and challenges surrounding the production of labour supply projections 14

Data challenges Issues § Timeliness § Sample size/Reliability Major data sources used in supply § Administrative data (enrolments, graduations) § National Graduates Survey § Labour Force Survey 15

Data issues - Timeliness A) Administrative data (enrolments, graduations) – Significant lags in data used to forecast enrolments and graduations : • • Data for high school is made available through the ESESP: latest data 2004/2005 Data for trade and vocational programs is acquired from TVOC survey: latest data 1999/2000 Data for community colleges comes from the CCSIS: latest data 1999/2000 University level data comes from the PSIS survey and is spliced with the (discontinued) USIS survey: latest data 2003/2004 – We thus need to forecast the past beforecasting the future! → Challenge: To speed up the release of data 16

Data issues - Timeliness B) National Graduates Survey – Produced occasionally (5 -year cycles): • • • Graduates of 1995 in 1997, follow-up in 2000 Graduates of 2000 in 2002, follow-up in 2005 Graduates of 2005 in 2007 (in process), follow-up in 2010 (planned) – Results depend on when the survey is conducted (economic cycles) : • E. g. : using 2000 NGS information to distribute computer graduates among occupations might be misleading given collapse in IT jobs in 2002. – Change in occupational coding every 10 years makes it difficult to use historical data (using many NGSs). 17

Data issues - Reliability A) Administrative data (enrolments, graduations) §The fields of study used in COPS are outdated: – e. g. 6 secretary programs at the TVOC level. → Challenge: Rework FOS according to CIP B) National Graduates Survey The NGS 2002 (class of 2000) has 38, 483 respondents. When building the distribution matrix of (5 post-secondary education levels) X (49 to 58 fields of study) X (140 occupations) → very small sample. § We use several NGSs to overcome small cell problems § Some fields of study are easier to allocate (e. g. nursing) than others (e. g. English literature). 18

Data issues - Reliability C) Labour Force Survey The sample size of the LFS is about 54, 000 households (accounts for approx. 83, 000 respondents). To build the school leaver distribution matrix for the 3 non post-secondary education levels (140 occupations), LFS data for youth is used. § Analysis of the coefficients of variation undertaken in 2003 shows that 3 -digit occupations data broken down by 33 industries is reliable at the national level. § We use the average distribution of the latest three years. 19

Supply modelling challenges School leavers model § Introduce dynamic in the distribution of graduates by field of study over time. § Redefine COPS fields of study (using CIP) § Re-examine the current model structure based on research currently undertaken § Introduce supply adjustment using substitution between broad occupational groups and skill level (ROA 2004) Immigration model § Re-examine the current model structure based on research currently undertaken (Cristiana Pescarus) 20

Other supply modelling challenges § Assessment of current labour market imbalances by occupation § Modelling of vertical occupational mobility (e. g. supply for managers) § Segmenting the labour market § Assessing work intensity (e. g. hours worked, full- versus part-time work) 21

Provincial supply modelling: additional challenges 22

Provincial supply challenges § Data timeliness: – The use of data collected by provincial Departments of Education can mitigate this problem. § Data reliability: – As we disaggregate further (in this case by province), the reliability of data declines: see next slide – Options: • Aggregate data (bundle FOS, educational levels or occupations) • Find alternate data source with larger sample size (Census, provincespecific NGS) • Seek to increase sample size through investment in data § Model: – Need to model inter-provincial migration 23

Data reliability § Data reliability is an issue at the provincial level. § It is difficult to produce projections when data have large standard deviations. Standard deviation of the annual growth in the number of enrolments and graduates by province (in % points) Type of university degree Enrolment (1985 -2004) Graduation (1976 -2004) Canada 2. 9% 2. 3% Newfoundland Labrador 9. 4% 7. 9% Prince Edward Island 5. 0% 13. 1% Nova Scotia 3. 2% 4. 0% New Brunswick 3. 1% 6. 1% Quebec 2. 0% 3. 9% Ontario 4. 7% 3. 0% Manitoba 7. 4% 4. 7% Saskatchewan 3. 0% 4. 6% Alberta 2. 7% 3. 8% British Columbia 3. 0% 4. 5% BA Bachelor's and other undergraduate degree Regions 24

Next Steps § At the national level – Continue to work on developing the supply model • Introduce dynamic in the projection of FOS • Redefine COPS FOS • Re-examine the current model structure based on research currently underway § At the provincial level – Provide technical expertise to provinces in building supply models • At the 2006 COPS Conference, supply models were presented and B. C. showed interest in modelling supply – Caveat: Provinces interested in building a supply model should first focus their attention on data (availability and reliability). 25