CCL Home > Reports & Data > Systematic Reviews
Since its inception, labour market forecasting has been the subject of much debate among economists. The poor track record of forecasts and the complex task of “tuning” the labour market (Heijke, 1996) have led to the view that labour market planning is neither necessary nor useful (Borghans & Willems, 1998). The purpose of this review is to systematically and transparently gather, analyze and synthesize research devoted to discussing and determining whether it is possible to accurately predict labour market needs.
There are two major approaches to occupational forecasting: workforce projection and labour market analysis (signalling). Workforce projection produces longer-term federal and provincial forecasts, while labour market analyses (LMA) identify and continually adjust to current regional and short-term trends.
Workforce projections are commonly used at the federal level to provide long-term estimates (Van Adams, Middleton & Ziderman, 1992). However, estimates are only as good as the “plausible assumptions” on which they are based. Because there are many unforeseeable factors that may affect the economic growth of a country (and the labour needs this growth produces) the quality of forecasts tend to decrease as the length of the forecasting period increases (Campbell, 1997a). Projections in terms of sectoral change, on the other hand, are often reasonably accurate. The difficulty here appears to be translating the accompanying skills change into a profile that can be used to inform decisions and policies regarding looming training and education requirements (Psacharopoulos, 1991).
Labour Market analysis uses “signals” to forecast future labour and education requirements. These signals are available via newspaper job listings or from the provincial and public employment and social insurance services that collect information about job openings, placements and unemployment rates. These resources provide a wealth of data which, if analyzed, may offer insight into current shortages or surpluses of workers. This approach can be limited by the fact that the data gathered from newspaper listings and public employment services are not generally available in a form that can easily be analyzed, and are unlikely to be complete since many jobs are not posted and many unemployed do not register with public services. Other types of signals include employer and household surveys, enrolment data and tracer studies, all of which tend to be restricted by region and population.
In total, 38 studies were included in this review, 28 of those used empirical strategies to demonstrate the accuracy of a particular forecasting model. Our analysis and evaluation of these studies resulted in a two major conclusions. First, the quality of the literature devoted to labour market forecasting is inconsistent, meaning readers and reviewers of such material need to be proficient in econometric modelling and research design if they are to fully assess the value and or the flaws within the conclusions drawn by the authors. Second, forecasting research is very much source, location, and time specific. Consequently, it is not clear if the models will perform as well in other forecasting horizons. Our study suggests that there is no single forecasting model that can accurately forecast labour market needs in all situations. While some of the proposed models show an impressive level of accuracy in forecasting within a particular market, without controlled replicability the consistency of the forecast accuracy remains uncertain.
Ultimately, it appears that some forecasting models have the ability to estimate labour needs in very specific circumstances. What remains unknown is whether it is possible to develop a single model that will accurately forecast in a range of situations, under various conditions.
View the full report (630 KB)