Cedefop's medium-term skills forecasts have proven very popular. But how does the skills project approach the topic and draw conclusions? This publication provides an overview of the methods underpinning the project. Cedefop’s forecast is not intended to replace forecasting efforts in individual countries, but to share the knowledge acquired during the development of different systems and methods, and to highlight the results. This shared knowledge can help to improve the methods used in each country and to resolve outstanding issues.
Cedefop’s forecast can also inspire new forecasting initiatives.
The feedback provided by countries can in turn help make the European forecast even more precise. The more solid the method, the more reliable the results.
Download Skills supply and demand in Europe.
Introduction
1.1. Background and rationale Equipping the labour force with the right skills is one of the key policy focuses of the European Union’s (EU) strategy for smart, sustainable and inclusive growth. Anticipation of skill needs has received more attention in the EU, as illustrated by several policy documents such as the Integrated guidelines for growth and jobs (2008-10) (European Commission, 2007), the Council Resolution on new skills for new jobs, and the Spring 2008 Council Conclusions (Council of the European Union, 2007; 2008), in which the EU Member States asked the European Commission to report on future skills requirements in Europe up to 2020. New skills for new jobs: action now (European Commission, 2010a), a report prepared by an expert group set up by the European Commission, also emphasised the need for a coordinated approach to improve Europe’s capacity to anticipate change. After a wide consultation among stakeholders, the Commission included the New skills and jobs agenda as a flagship initiative in the EU 2020 strategy (
European Commission, 2010b).
It is in this context that Cedefop conducts regular, coherent and systematic skill demand and supply forecasts. In 2008, Cedefop released the first pan-European skill needs forecast, i.e. employment projections by sector, occupation and qualification level across Europe up to 2015. In 2009, Cedefop forecast the supply of skills by gender, age group and qualification level. Finally, in 2010 the first parallel forecast of skill supply and demand up to 2020 was presented.
A variety of forecasting methods are used. The accuracy of methods must be tested and compared with available alternatives to increase the quality of results. At the same time statistical authorities publish new data regularly. Finally, forecasting is an ongoing exercise, affected by changing reality, which means it is important to use the most up-to-date information and to reflect trends and changes to achieve the most reliable results.
This publication presents the complex methodological framework used by Cedefop to forecast skills supply and demand and some current attempts to improve it. It does not promote Cedefop’s methodology as the only correct methodology. Moreover, Cedefop’s forecast does not replace those conducted at national level. Instead, this publication presents the problems that we have encountered and the solutions we have adopted to produce a unique pan-European skills supply and demand forecast
1.2. General overview of the methodological framework Europe’s pan-European forecast of skill needs requires complex methods, relying on long-term research and drawing on the expertise of several high-level European research institutions. The modelling tools have been designed to enable further development and customisation. The general framework consists of methods developed in two pilot studies on Future skills needs in Europe (Cedefop, 2008) and Future skill supply in Europe (Cedefop, 2009). These were combined to produce the first pan-European forecast of skills supply and demand in Europe (Cedefop, 2010). Forecasting is a dynamic process, and important developments took place in 2010 made possible by the modular approach adopted, which enables the different parts of the system to be improved independently. As shown in Figure 1, the model breaks down into different building blocks and into several interrelated components.
Even though the modelling framework has proven to be rather robust, a dialogue must be established with experts from European countries, who are likely to have much greater knowledge of employment trends and data sources within their own countries. By making it easy to incorporate new data and alternative or additional assumptions, the modelling framework provides an opportunity for knowledge and input of experts to be built in efficiently and transparently.
The project involved developing consistent databases and related tools to produce a comprehensive and consistent set of skill projections for all EU Member States plus Norway and Switzerland (EU-27+). The system, models and modules rely upon official data sources, drawing primarily on Eurostat, in particular on Eurostat demographic data, national accounts (NA), the EU labour force survey (EU-LFS), as well as additional data on flows of qualifications. Compilation and harmonisation of the best possible data available for measuring employment was a major achievement of the project. Historically, most countries have invested considerable resources in developing data for their NA. In many respects estimates of employment on this basis are to be preferred as they are consistent with other key economic indicators, such as output and productivity. On the other hand, the EU-LFS has the advantage of providing measures of employment structured by skills (occupation and qualification), as well as by gender and age, which are not available from NA-based estimates.
1.2.1. Supply of skills The skill supply projections produce consistent pan-European projections broken down by age, gender and formal qualification. The results indicate the future skill supply by highest qualification held as well as by age groups and gender for the population and labour force aged 15 and over. The skill supply projections are compatible with the skills demand projections (when focusing on qualifications).
The historical analysis and projections of overall labour supply by age and gender are provided by an extended version of the existing pan-European macroeconomic model E3ME developed by Cambridge econometrics, which incorporates a new demographic and labour-supply module. E3ME models labour supply as a function of economic activity, real wage rates, unemployment and other benefit rates. At present, the model parameters are estimated for labour market participation in each country by gender and separately for different age groups. This is of key importance for modelling educational participation and attainment since these are known to be gender and age specific. This expanded model framework is then used to create a detailed set of baseline projections for labour supply, disaggregated by country, age groups and gender and covering a 10-15 year period. This model forms a key input for the analysis of the supply of qualifications and provides the link between economic activity and labour market supply. Finally, this link can be used to provide a range of projections of available skills through scenario-based analysis around the baseline forecast, indicating areas that are most sensitive to the economic climate and change.
Modelling and forecasting the supply of qualifications ideally requires a detailed and comprehensive stock-flow model, with behavioural links which can be used to predict the distribution of people in the total population and labour force (employed and unemployed people) by highest qualification held. In practice, this ideal is hard to realise, as a detailed demographic or educational and labour market accounting system is still lacking at EU level.
The methods currently used for modelling forecasts are less ambitious. They range from rather simple models, based on fitting trends of aggregate qualification patterns among the population and/or labour force, to more sophisticated approaches based on econometric analysis of microdata on individuals, mainly using LFS data. All focus on overall stocks rather than flows.
1.2.2. Demand for skills
The demand side involves four main elements or modules. Each module contains its own database and models. The results focus on future demand trends at a pan-European level (EU-27+): by sector (up to 41 industries based on NACE classification); by occupation (up to 27 occupations based on ISCO classification); by qualification (three broad levels based on the ISCED classification); plus replacement demands by occupation and qualification. Together these produce estimates of the numbers of job openings (net employment change plus replacement demand) by skill (as measured by occupation and by qualification). The detailed classifications and aggregations used are provided in Annex 2.
The forecast of employment by economic sector is provided by a module which is based on results from the existing pan-European multisectoral macroeconomic model (E3ME). This model delivers a set of consistent sectoral employment projections, which are transparent in terms of the assumptions made about the main external influences on the various countries (including technological change and the impact of global competition).
E3ME combines the features of an annual short- and medium-term sectoral model, estimated by formal econometric methods, with the detail and some of the methods of the computable general equilibrium models that provide analysis of the movement of the long-term outcomes. It can also be used for dynamic policy simulation and for forecasting and projecting over the medium and long term.
The LFS conducted in all countries provides a source of information for the construction of occupation-industry matrices of employment. These surveys have the advantage of being conducted regularly. They also adopt standardised sets of questions and systems of classification. While there are still some differences among countries, LFS provide a broadly consistent set of data which can be used for producing occupational employment projections within the industries identified in macroeconomic models such as E3ME. The forecasting module designed to calculate changes in employment (expansion demand) by occupation (EDMOD) based on these data works out the implications for occupational employment.
Occupational employment patterns are only one way of measuring skills. An occupational category can be understood as broadly describing a particular job (related tasks, requirements, position, etc.). Qualifications represent the characteristics of people filling these jobs as well as one of the selection criteria for filling a particular job. From the education and training policy and planning point of view, the types of qualifications typically required are important. Even with only weak data for (formal) qualifications, it has been possible to develop the module (QMOD) which allows inferences to be made about implications for qualifications.
In addition to changes in overall occupational employment levels, it is important to consider replacement demand arising from outflows from a job/occupation, such as retirements and deaths, transition to non-employment, net migration and inter-occupational mobility. Estimating replacement demand is not straightforward and is quite sensitive to the data sources used. Ideally, detailed data on labour market outflows and transitions (mainly retirements and occupational mobility) would be required to analyse replacement demand more accurately. However, these are not currently available and therefore this forecast relies on a methodology that is based on stocks of age-cohorts by occupation and qualification, and excludes transitions from one occupation to another.
From the LFS, it is possible to analyse the demographic composition of each occupation. This allows specific rates of retirement to be estimated for each occupational class (but still not taking account of inter-occupational mobility). LFS data can also be used to estimate rates of outflow. The replacement demand model (RDMOD) has been developed on the basis of data sources that are similar to the occupational model (EDMOD). The model is driven in part by the occupational and qualification employment levels projected from EDMOD and QMOD, combined with models and information on the probability of leaving employment owing to retirement or migration and for other reasons (e.g. transition to inactivity).
1.2.3. Comparing skill supply and demand To provide information on possible labour market imbalances and skill mismatches, a further module (BALMOD) has been added. This module compares the skill demand and skill supply projections (focusing on qualifications) and attempts to reconcile the two.
The possibility to analyse potential skill imbalances in the labour market is important from a policy and individual point of view. Such information can, in conjunction with corresponding demand estimates, shed light on possible future developments in European labour markets, highlighting potential mismatches and thus helping to inform decisions on investments in skills (especially in formal qualifications) made by individuals, organisations and policy-makers.
However, simply comparing current demand and supply projections is problematic for both practical and theoretical reasons. Although the two sets of results are based on common data and are carried out simultaneously, they do not incorporate direct interactions between supply and demand and, therefore, they cannot be directly compared. Cedefop has started to work on modelling interactions between supply and demand, but due to the complexity of the task these interactions might be incorporated only in the medium to long term. There are various other conceptual and methodological issues regarding imbalances that need to be considered to avoid misleading inferences and interpretations.
A final adjustment has been made to the estimates of employment by qualification (demand side) to take account of the labour market accounts residual. This residual measures the difference between employment as measured for the NA estimates (workplace based, jobs) and the corresponding LFS estimates (heads, residence based). Both measures are used in the project (5). The difference between the NA and LFS can be quite significant and needs to be considered, especially when comparing demand and supply.
Differences between skill demand and supply can include:
(a) double jobbing (some people have more than one job) or one full-time job is shared by two or more people;
(b) distinction between residence and workplace (many people do not live in the same country as they work; this is especially significant for some small countries such as Luxembourg);
(c) participants in training and related schemes who are also working in parallel (they may be included in the labour force and in education statistics – double counting);
(d) different definitions of unemployment (e.g. ILO definition versus limited to unemployment beneficiaries);
(e) statistical errors (in measures of employment, unemployment and related indicators, including sampling and measurements errors);
(f) other differences due to the use of different data sources such as treatment of nationals working abroad.
The Council on Higher Education Accreditation (CHEA) Board of Directors has appointed 19 individuals to the Advisory Council of the CHEA International Quality Group (CIQG).
/https%3A%2F%2Fprofilepics.canalblog.com%2Fprofilepics%2F1%2F0%2F1076071.jpg)