Literature Review onEstimating Inter-occupational Labour Mobility ina Federal, Provincial and Territorial Environment
Prepared for:
Forum of Labour Market Ministers
Labour Market Information Working Group
C/O Human Resources and Skills Development Canada
Prepared by:
The Centre for Spatial Economics
15 Martin Street, Suite 203
Milton, ON L9T 2R1
March31, 2005
The Centre for Spatial Economics
Abstract
Labour mobility is critically important in the healthy functioning of a modern economy. A properly functioning labour market permits the economy to adjust to economic shocks and helps to match the skills of workers with the requirements of their jobs, which boosts productivity and therefore living standards. Labour mobility is particularly important today considering the phenomenon of globalization and technological advancement could be changing the type of occupations and skills needed in the Canadian economy.
Two key aspects of labour mobility are occupational and geographic mobility. This report examines the theoretical and empirical literature on occupational and geographic mobility and also examines current international practices in occupational models and forecasting. A set of best practices in occupational forecasting and modelling are also identified, the application of which could improve the accuracy of Canadian occupational modelling and forecasting. An improvement in the accuracy of Canadian occupational forecasting and modelling would pay dividends by providing better advice to policymakers, and better information to employers and employees regarding future occupational trends.
About this Report
This report was prepared by The Centre for Spatial Economics, a consulting organisation created to improve the quality of spatial economic and demographic research in Canada. The report was sponsored by the Forum of Labour Market Ministers (FLMM).
The authors accept all responsibility for any remaining errors or omissions. The views in this report reflect those of the authors and do not necessarily reflect those of the FLMM or its member agencies.
Questions or comments about this report can be sent to:
The Centre for Spatial Economics
Mail and Courier Address
The Centre for Spatial Economics
15 Martin Street, Suite 203
Milton, Ontario L9T 2R1
Canada
Phone Numberse-mail addresses
Robert Fairholm(416) 346-2739
Robin Somerville(905)
Ernie Stokes(905)
Fax Number(905) 878-8502
Table of Contents
Executive Summary
Introduction
Occupational Mobility
Human Capital
Screening, Signalling, Filtering Hypothesis
Labour Queue or Job Competition Theory
Job Search
Job Matching
Segmentation Theory
Occupational Mobility: Summary
Geographic Mobility
Regional and International Migrants
Causes of Migration: Why Move
Who Migrates
Adjustment of Migrants to Host Region
Impact on Host Region
Geographic Mobility: Summary
Impediments to Regional Labour Mobility
Integrated Interregional-Occupational Model
Econometric Issues
Demand Side Analysis
Supply Side Analysis
Observations
Occupational Modelling and Forecasting
Step One: Historical Labour Force Data
Step Two: Demographic and Labour Force Projection
Step Three: Industrial Employment Forecast
Step Four: Occupational Employment Forecast
Step Five: Qualifications Data
Step Six: Qualification Forecast
Replacement Demand
School Leavers
Best Practice in Occupational Modelling and Forecasting
Conclusion
References
Appendix I: International Occupational Models
The Netherlands
United Kingdom
Ireland
Germany
France
United States
Canada
Australia
Appendix II: International Occupational Models Features and Drawbacks
The Centre for Spatial Economics
Literature Review on Inter-occupational Mobility in Federal/Provincial/Territorial Environment Page 1
Executive Summary
Labour mobility is a critical ingredient in the functioning of a modern economy. A properly functioning labour market permits the economy to adjust to economic shocks and helps to match the skills of workers with the requirements of their jobs, which boosts productivity, lowers costs and, therefore,improves living standards. Labour mobility is particularly important today considering the phenomenon of globalization and technological advancement could be changing the type of occupations and skills needed in the Canadian economy.
Two key aspects of labour mobility are occupational and geographic mobility. Of the two, occupational mobility is significantly larger, with almost 40% of employed workers changing their occupation between 1994 and 1997. The size of these flows are much larger than interprovincial migration or international migration, which in the 2001 census were both estimated at roughly 1% of total population over the previous year.
There are many similarities between occupational and geographic mobility, and some researchers suggest that job seekers consider both together. Most of the economic literature, however, tends to examine one of these aspects of labour mobility, although many of the underlying theories and empirical results are similar.
For occupational mobility, the economic literature provides a number of competing models – human capital, signalling, job search, job competition, job matching, and segmentation. The literature is less clear if one model is superior to the others. There is evidence supporting at least some aspects of each of these approaches. From the perspective of occupational forecasting, one of the most pertinent aspects of the empirical literature is that there is a fairly consistent list of factors that affect occupational mobility – financial and non-financial costs of changing occupation, age, formal education, training, gender and perhaps ethnicity. The literature also indicates that there are likely substitution effects and crowding out of less qualified employees by more qualified employees at least when labour demand is reduced and perhaps on a continuing basis that could lead to a persistent mismatch between skills needed to perform jobs and the qualifications of employees.
In the economic literature, it is hypothesized that part of the reason for the occupational mismatch is because jobs are location-specific, and there could be (financial or non-financial) barriers that prevent workers from moving to where the jobs are located. This insight means that the spatial element must also be considered when trying to model occupational demand and supply. Even at the provincial/territorial level, however, the geographic size of the jurisdiction might be too large.
Most research of geographic mobility assumes that people will move when the discounted flow of monetary and non-monetary benefits of moving exceeds the costs. The real question of the research then becomes what are the benefits and costs that influence migration. Also some of the literature directly tries to examine the role of uncertainty and expectations in these decisions. The dominant financial factors that the literature points to includes: wage differences, probability of being employed (unemployed) in the new region, government income support payments, living cost differences (particularly for housing), and the financial costs of the move, which tends to increase with distance. There is a long list of possible non-financial factors that affect migration including social networks, psychological costs and location specific human capital. Since many of these non-financial factors are unobservable, the personal characteristics of migrants are used as proxies. Personal characteristics that are linked to migration include: age, formal education, occupation, income level, gender and perhaps ethnicity.
The discussion in the chapter on impediments to regional migration shows that there is likely to be a significant structural break in the geographic mobility of Canadians after 1995, and again after 2001 because of the agreement to harmonize national occupational standards across Canada. This means that any time series analysis needs to take these potential changes into account. Furthermore, probit, logit or cross-sectional analysis of the factors of mobility (or immobility) that use data prior to 2001 may not be representative of the current mobility of workers, particularly for those occupations that were affected by the change in mutual recognition.
Occupational models tend to use, explicitly or implicitly, human capital and/or matching theories to model occupational mobility. For geographic mobility, occupational models tend to use at best only a few of the financial aspects, such as regional wage and unemployment rate differences. Furthermore, regional migration is typically exogenous to these models and, therefore, migration does not directly respond to changes in financial or non-financial factors. This is an area where the application of some basic research will improve the dynamics of the occupational modelling system. And given the fact that the economic literature clearly shows that regional migration is the dominant equilibrating mechanism in the labour market, the lack of an endogenous regional migration component in these models is a major failing.
If the researcher is interested in studying both occupational and geographic mobility, then we must ask whether the decision to change location and the decision to change occupation are independent. The evidence indicates that these two decisions are not independent and, as a result, ordinary least squares estimation techniques cannot be used. Fortunately, there is a vast literature on possible corrections for these issues. These range from simple single equation instrumental variables techniques to more complicated systems of equations. The application of co-integration theory to panel data sets has led to the development of dynamic panel data estimation techniques. This is an enormously powerful tool for the analysis of the large cross-sectional databases collected over time that characterize detailed labour market statistics. Finally, qualitative and limited dependent variables techniques (examples include tobit and logit models) can also be designed to correct for any potential bias and are frequently employed to model individual behaviour or choice typical in labour supply models.
Drawing best practices from among established occupational modelling systems is a moving target because modelling groups are continually revising their models and methodologies. Best practice also depends on many factors, such as data availability and quality, the amount of resources that are to be applied to the project, and the ultimate need that the model/forecast fulfills. A set of best practices for a modelling system that can produce occupational projections and policy simulation work can be identified, however, no system is always better than the others, so the following selects from various approaches.
Data: The Netherlands is generally acknowledged to have the most complete set of demographic accounts in the world. Development of the type of data in the Netherlands would help to improve the modelling and forecasting of Canadian occupations.
For Canada there is a trade-off between census data and other data sources like the LFS and SLID. The census provides very detailed data on occupations by age, sex, ethnicity, migration status and educational attainment in an internally consistent manner. Census data, however, can not be used to answer some basic questions regarding mobility, and the fact that it is collected once every five years limits the type of analysis that can be done. In contrast, the LFS data provide time-series estimates by occupations, which at the national level are fairly detailed, but these data will not be available at the same level of detail for provinces and territories. SLID also suffers from a small sample, which means that its occupational detail is very aggregate.One approach that is often used to get around this common trade-off is to model a dynamic model at a more aggregate level, and then to provide more disaggregated information by using an extrapolative trend approach for component shares. This is done in Australia and the Netherlands, for example.
The development of a complete model of occupational and geographic mobility requires a tremendous amount of data. In particular, flow data are needed that represent the movement of people from one labour force state to another or from one occupation to another, for example. This type of data is generally not available. To inform their models, researchers have used the cohort component method to derive flow data from the available stock data. This is done in the Netherlands, Canada, US and Australia, for example, and represents the best way to derive the flow data that are needed for detailed occupational models.[1]
Demographics: Demographic characteristics are an important factor that influences occupational and geographic mobility. This means that, at a minimum, the demographic information should include: age, education, gender and perhaps ethnicity. International migration is treated exogenously in all models. This seem appropriate given the barriers to international migration, such as border controls. In terms of regional migration, most models do not include endogenous geographic labour mobility. The Australian model includes regional labour mobility to equilibrate the regional labour markets, which should make impact analysis more realistic and is consistent with the empirical evidence that shows that regional labour markets tend to be equilibrated via inter-regional migration.
Aggregate GDP and Final Demand: It has been found in the literature on economic forecasting thataconsensus forecast tends to be more accurate than individual forecasts. This means that ideally a consensus economic outlook should be used to provide the underlying foundation for the occupation forecast. An important aspect of this stage is that sufficient expenditure detail is available in the model in order to provide the needed final demand information that will be used in the input/output matrix transformation to produce industry gross output and therefore industry value added. The US uses the most extensive final demand outlook, by expanding the available macro forecast with final demand bridge equations to provide additional final demand detail. This could be replicated in Canada.
Industrial Output and Employment Projection:It is generally acknowledged in the literature that some form of macroeconometric (or CGE) model is needed in order to produce a consistent industrial employment outlook. Given the impact of technological change and relative prices on industry output, the input/output coefficients should ideally be projected into the future. The Australian approach accomplishes these tasks, by combining information from a macro model and a CGE model with an explicit projection of input/output coefficients into the future.
A number of approaches have been used to generate industrial employment; however, it is not clear what approach would produce the best results in the Canadian situation. This is a case where empirical analysis will be needed to determine the most appropriate functional form and econometric technique.
Occupational employment:A variety of different approaches are taken toward occupational employment outlooks. No one system dominates in all aspects. Most tend to model the share of industrial employment comprised of different occupations. The more sophisticated approaches try to model changes in the occupational structure by including various influences. For example, the Canadian model includes a variable that reflects the business cycle, which in turn ensures that occupational demand reflects the position in the business cycle. The Dutch system, takes a further step by using a random coefficient approach that weights the coefficients between single equation estimation for that occupation in that industry and a pooled dataset for that occupation in all industries. This permits more reliable parameter estimation without affecting the specification of the separate occupation equations.
The literature indicates that technological change must also be accounted for in determining occupational trends. This is done in the US model, which includes an extrapolative staffing trend, TFP and government policy in the occupational employment equations. The best practice in this area would seem to combine the random rounding approach used in the Netherlands with a functional form that includes both cyclical and structural factors like technological change.
Qualifications Data: As with the basic demographic data, the Netherlands is widely acknowledged as having the best set of consistent demographic accounts in the world, inclusive of education. Ideally, Canada should develop more complete demographic accounts. This is unlikely to happen any time soon. Current Canadian educational data are problematic, and will likely severely constrain any attempt to improve the forecasting of skills and qualifications in the near future. Over time more Canadian education data using Classification of Instructional Programs (CIP) will become available, but it will be several years before the 2006 census data are available using the same classification system. And it will likely be a couple of years before an official concordance between Major Field of Study (MFS) and CIP will be available from Statistics Canada.
Qualifications Projection:Canada and the Netherlands both have education sub-models. The Canadian model produces an estimate of school leavers, which adds to supply. However, there is no interaction between supply and demand in the Canadian modelling system. The Netherlands is generally acknowledged to have the most sophisticated system in the world to project qualifications, which permits the interaction between supply and demand by education and occupation. It has a very sophisticated approach to modelling demand substitution effects and supply related crowding out in the labour market. If the modelling exercise aims to determine the likely interaction – and therefore outcomes – of school leavers (and others in the labour force), the ROA approach is the most suitable.
The Canadian system, explicitly takes into account the outcome that immigrants experience based on their historic pattern. Since the literature indicates that there is a dilution of immigrants’ human capital, this step is essential to correctly forecasting their occupational outcome. The Canadian approach does not, however, go far enough, since immigrants are also known to experience significant occupational change as they stay in their host country longer.