This work contributes to the “Jobs and Skills” module of the Going Digital horizontal project and to the Skills Outlook 2019 on Skills and Digitalisation. It results from the cooperation between the Directorate for Education and Skills (EDU) and the Directorate for Science, Technology and Innovation (STI).
It proposes an experimental methodology and first time estimates of the monetary cost of training needed for occupational transitions to occur. It builds on Bechichi et al.’s (2018a,b) analyses of the quantity and type of training needed to facilitate occupational transitions, and of the education and training policies enabling such transitions.
The training effort required to move individuals to a different occupation is considered to be a function of the difference in cognitive skills (namely, literacy and numeracy and task- based skills requirements characterising the occupations of origin and of destination (e.g. management and communication and Information and Communication Technology skills). Occupational transitions are considered “acceptable” if they are welfare and well-being enhancing. Such conditions are met when the occupations that can be reached after a predetermined training effort (of up to 6 months, 1 year or 3 years, respectively) do not entail an excessive wage cut for the worker, nor too large skills excess relative to the occupation of origin, if any at all.
The cost of training is assumed to encompass both a direct and an indirect component. The direct cost of training is the monetary cost of an education and training programme of a given length. It therefore reflects the differences that may exist, along a number of dimensions, in education and training programmes both within and across countries. The indirect cost, or opportunity cost, is calculated in terms of foregone workers’ wages, for the duration of the training, as it is assumed that workers do not work while on (re)training. The study also implicitly assumes that training is effective and that workers can (at least in theory) successfully complete the training spell considered and acquire the skills they need to transit to the identified occupations.
The analysis first proposes estimates of the cost of moving any worker in the workforce to another occupation, i.e. focuses on the whole population of workers. It then narrows the focus and estimates the cost of moving only those workers in occupations at high risk of automation to another job. The concept of “safe haven” is proposed in the case of workers in occupations at high risk of automation (henceforth high ROA, estimated to be 14% at the lower bound). “Safe haven” occupations are here defined as acceptable occupations of destination that are not at high risk of automation. While characterised by a relatively lower risk of automation, “safe haven” occupations do not shield workers from unemployment altogether, nor necessarily imply a high(er) job quality, and may be exposed to changes if their tasks change in the future.
As not all workers in high ROA occupations may be able to reach “safe haven” occupations upon small training spells of at most 6 months (either contiguously or periodic), the study further proposes estimates of “minimum training costs”. These correspond to the cost of moving workers in occupations at high risk of automation to the “safe haven” occupations that can be reached within the narrower training effort possible, whether small (of at most 6 months), moderate (of at most 1 year) or important (of at most 3 years).
This work relies on a number of assumptions and operational choices mostly dictated by data availability and existing evidence, which may have first order effects over the estimates proposed (Section 8 of the paper and Annex B discuss and synthesise the key assumption made). Among them: the estimated cost relates to upgrading cognitive skills, and only partly task-based skills; the direct cost of training is derived using information related to the education system; individuals are assumed not to work while training; training is presumed to be effective in allowing occupational transitions to occur; on average all individuals are assumed to be able to bridge the same cognitive skill gaps within a certain training spell; on average, countries’ education systems have the same effectiveness; and wage decrease of more than 10% are considered unacceptable.
Some of these assumptions stem from the impossibility to account for a number of individual and education-related characteristics known to shape learning, including workers’ attitude towards learning, stock of human capital, and quality of education and the efficiency of the education sector. Also, little is known about how adults learn, how effective lifelong training can be or about the effect that technology may have on both the efficiency and the effectiveness of training or education – all aspects left to future research.
The use of education expenditure to proxy the direct cost of retraining workers implicitly assumes that individuals retrain through formal or non-formal education and training programmes, but not (at least, not solely) by learning at work or at home. It further assumes that the cost of different types of training is similar to the cost of formal training in an education institution.
This work further offers a static picture of occupational transitions, as it assumes that workers can seamlessly transit from work to training and back to working again, and leaves aside considerations related to labour market frictions and unemployment dynamics, as well as labour demand conditions.
The main results are:
Training Costs Associated with All Workers:
On average across all countries considered, when both direct and indirect costs are factored in, depending on the occupation of origin, lower bound estimates show the average total per worker costs of training, by scenario, to be:
o in the training effort scenario entailing at most six months of training (“small”): between USD 1 000 and 15 000 (average: USD 7 000)
o in the scenario entailing a training effort of at most one year (“moderate”): between USD 2 000 and 35 000 (average: USD 13 000)
o inthescenarioentailingtrainingofuptothreeyears(“important”):betweenUSD 24 000 and 98 000 (average: USD 50 000).
These figures are subject to approximation in light of the methodology chosen and its caveats, as detailed in the body of the study and in Annex B. Therefore, they should be considered as first estimates, which may change in case more and better data become available and some of the assumptions can be relaxed.
The per-worker average total cost increases with the amount of the training needed to move from one occupation to another, but this progression is non-linear, meaning that twice as long training periods will generate more than twice as high costs.
Total estimated costs vary depending on the type of costs considered, i.e. direct cost, indirect cost or both. The average direct (re)training costs per worker is USD 1 600 in the case of small training needs, doubles in the moderate training scenario, and rises sharply in the important training scenario, to USD 12 600. The average per-person opportunity cost (i.e. the indirect cost) is about USD 4 300 in the first scenario, USD 9 900 in the moderate one, and rises sharply to USD 37 000 in the case of the important training scenario. Generally, the opportunity cost of training represents approximately 75% of the per-person total cost. This is true on average across occupations and training scenarios.
The total cost is generally larger for highly skilled than for low-skilled occupations, as highly skilled workers face larger opportunity costs (given that their salaries are generally higher than those of low-skilled workers). This result underlines the importance of allowing individuals to both work and learn, to mitigate the overall cost of education and training policies aimed at facilitating occupational mobility.
The Cost of Moving Workers in Occupations at High Risk of Automation:
When a small training scenario is considered, 57 out of 123, or equivalently 46% of occupations, have no acceptable transitions. This is especially the case for professionals and technicians among high skilled occupations, and plant and machine operators and assemblers among low-skilled ones. When moderate training efforts are undertaken, the outcome is much more positive: 87% of the occupations find an acceptable transition.
Workers in some high-ROA occupations of origin would need to undergo at least a moderate (re)training (up to 1 year) to find an acceptable occupation of destination at low or medium risk of automation. Reassuringly, though, no one needs being left behind, as “safe haven” occupations can be identified for all high-ROA occupations in the important training scenario.
Nevertheless, bringing workers in high ROA occupations to safe havens will be a non-trivial undertaking. It will cost between USD 3 400 to 7 400 in direct training costs per worker, depending on the country, and USD 9 200 to 21 100 in indirect (foregone wages) costs. In total, the lowest estimated cost is USD 12 600 per worker to move from an occupation at a high-risk of automation to one that at medium or low risk.
In the countries considered, per-worker average figures vary substantially, as do the cost of education and the salary and income distributions. The country-specific minimum total cost of moving workers to a safe haven is found to be positively related to the share of high-ROA employment in manufacturing, and negatively related to the share employed in market services. Also, the relatively higher density of high-ROA workers in manufacturing is likely to make the sector more exposed to future investment in automation and employment-reducing technologies, and may require larger expenditure in workers’ training in the foreseeable future.
Constraining transitions so that workers can move to occupations of destination that are not at high ROA increases the average cost of such occupational transitions. Transitions towards acceptable occupations at lower risk of automation for workers in high ROA jobs are generally characterised by relatively higher cognitive skill requirements. The per- person cost of training an average high-ROA worker to move to the nearest safe-haven set of acceptable occupations thus appears to be costlier than moving the worker to an average acceptable occupation of destination.
The range of rough or “ballpark” estimates of the country-level minimum cost (direct + indirect) of helping workers at high ROA move to “safe haven” occupationsreaches between 1-5% of a single year’s GDP on average across the countries considered, or 6-26% of a single year’s welfare expenditure. On the lower bound, such estimated costs are lowest for Norway (0.32% of GDP) and highest for Chile (2.26% of GDP); on the upper bound, they range between 1.35% of GDP in the case of Singapore and 9.9% of GDP for Chile. Estimates reflect a number of structural features, including the occupational and skills distribution of the population and the industrial structure of countries.
These resources may not be sufficient to move all high-ROA workers to “safe haven” if training programmes are not effective in providing all trained workers with the necessary skills. A first attempt to address this issue relies on completion rates of educational programmes and suggests that up to 25% of workers may fail to transit to a “safe haven” occupation.
With the caveat of needing to rely on a number of important assumption, the study also proposes first experimental figures of the country-wide training cost that may be needed to bridge differences in task-based skills between occupations. This cost component is added on the top of the country-level cost needed to train workers and move them away from high-ROA occupations. Estimates based on 2015 Eurostat’s Continuing Vocational Training Survey (CVTS) data suggest such a cost to range between 0.06% – 0.3% of GDP.
The total cost of up-skilling or re-skilling workers is lower for occupations seeing a relatively higher presence of young people, even when focusing on occupations at high risk of automation only.
Occupational transitions need not entail salary losses. In the case of skilled workers, while they may find it more costly to move away from occupations at high risk of automation, they may also enjoy hourly wage gains from the transition which are increasing in their cognitive skills, that is, the more they are skilled, the more they can gain.
The analysis shows the urgency of a renewed approach to workforce training and lifelong learning in the digital era. This could result from concerted efforts between businesses, governments and other stakeholders – including workers. Such actions should encompass the design of effective training programmes enabling workers to acquire the skills needed to change jobs; facilitate learning on the job, for all workers. This along with better preparing future generations of workers for occupational mobility, for instance through fostering multidisciplinary knowledge and by means of enhancing readiness to learn. Also, it will be important to agree on how to best share these costs and to identify or design incentive mechanisms able to make occupational transitions-related policies both efficient and effective.
Chosen excerpts by Job Market Monitor. Read the whole story at OECD iLibrary | Occupational transitions: The cost of moving to a “safe haven”