In the labour market, employers will demand different skills combinations: occupations will evolve, some will become obsolete and new ones will emerge. A growing number of studies attempt to provide guidance for employers and policymakers on reskilling priorities based on explicit predictions about the future (Nesta/Pearson, 2017; McKinsey, 2017; WEF/BCG, 2018).

Nesta’s collaboration with Pearson and Oxford Martin School predicts which occupations and skills will be in higher demand in the UK (and US) workforces in 2030. Panels of experts are presented with a subset of occupations and asked to predict whether they will experience growth (in both absolute and relative terms) in the workforce. Machine learning techniques are then applied to detailed data on the skills, knowledge and ability requirements of different jobs to extend the experts’ judgements to all other occupations in the workforce.

The Future of Skills suggests that a combination of strong social skills and higher-order cognitive skills, such as critical reasoning, originality and systems thinking, will grow in importance in the future.

But how do these job requirements of the future compare with the skills, knowledge and abilities our workforce currently has? Are workforce skills in the UK already moving to meet the needs of employers in the future, or is there a skills crisis brewing?

In this post, we attempt to answer these questions by comparing publicly available data on the workforce today with the predictions in The Future of Skills.

Summary

1. We determine the prevalence of a skill in the UK workforce on a scale of 1 to 5 using Office of National Statistics (ONS) Labour Force Survey data on the occupation composition of the workforce and O*NET data from the US Department of Labor on the skills, knowledge and ability requirements of different occupations.

The most prevalent features in the UK workforce turn out to be: oral comprehension, oral expression, customer and personal service, English language, active listening, speaking, problem sensitivity, speech clarity, written comprehension and speech recognition. The full set of results are available here.

The least important requirements include: dynamic flexibility, installation, explosive strength, night vision, sound localisation, peripheral vision, glare sensitivity, fine arts, repairing, equipment maintenance.

1. We establish how the prevalence of a job requirement has changed on a yearly basis since 2001. The requirements that have seen the greatest increase in importance in the last 16 years are: therapy and counselling, psychology, sociology and anthropology, medicine and dentistry, instructing, service orientation, social perceptiveness, learning strategies, customer and personal service, philosophy and theology. Those that have declined in importance the most are: manual dexterity, rate control, engineering and technology, repairing, operation and control, design, reaction time, production and processing, mechanical, control precision. The full set of results are available here.

3. Over time, skills, knowledge and abilities in the UK workforce have, on the whole, been steadily approaching what The Future of Skills predicts will be required. In particular, the requirements with the brightest outlook in the UK correlate well with the features that are currently prevalent (Kendall coefficient = 0.45, p-value=3.37e-13), and this correlation has been steadily increasing over time.

1. However, some ‘skills’ are not growing as quickly as the research suggests might be needed, and some are even shrinking. There appears to be a broadly linear relationship between the outlook for a job requirement, as determined by The Future of Skills, and the change in its importance in the UK workforce since 2001. Significant anomalies in this relationship are a simple way of identifying features that have grown historically by more or less than is warranted by future trends. It turns out, for example, that design, engineering and technology knowledge has markedly shrunk in importance over time - despite predictions it will be more highly valued in future - while clerical knowledge has become more prevalent. We offer different interpretations of these results.

Data

The web data portal O*NET, sponsored by the US Department of Labor, provides a list of occupations, coded to the US Bureau of Labor Statistics Standard Occupational Classification, along with their detailed job requirement profiles. In particular, we use data relating to how important each of 120 skills, knowledge and ability features are for a particular occupation, expressed on a scale of 1 to 5. (See The Future of Skills p30-31 for a more detailed description of these variables).

In order to use this resource for UK occupations, we follow The Future of Skills in employing a translation scheme that allows us to match a UK occupation (as defined by the ONS Standard Occupational Classification) to a US occupation. For this purpose, we use a mapping scheme (called a “crosswalk”) based on that in LMI For All, an online labour data resource that was developed by the UK Commission for Employment and Skills. (See The Future of Skills p103-111 for a detailed description of the crosswalk).

Finally, we use historical employment by occupation data, which is provided in the form of the “EMP04” dataset, available on the ONS website on a yearly basis since 2001. Note that the ONS switched from SOC2000 (the older version of the Standard Occupation Classification) to SOC2010 in 2010, so the change in classification system introduces discrete changes in that year.

Methodology

Our first goal is to obtain a measure of importance for each of the 120 O*NET skills, knowledge and ability features in the UK workforce. It reflects both how many people are employed in each occupation and the ‘skills’ currently required to do their job.

We have already mentioned how O*NET provides a requirements profile for each occupation. In order to derive an average requirement for the workforce as a whole, we compute a workforce average skill profile, weighted by how many people are employed in each particular occupation. This is intuitively equivalent to letting every worker vote on the importance of a given requirement in the workforce, assuming each of them votes in line with the O*NET ‘skill’ profile for their occupations, and then averaging across the votes. This means that occupations with more workers have more influence on the “average skillset” than less common occupations.

The average importance of each job requirement in the UK 2017 workforce so computed is available here. We check how this importance ranking compares with the ranking of requirements associated with occupations that are predicted to grow according to The Future of Skills (page 60), using the Kendall rank correlation coefficient. The result suggests that they strongly overlap (Kendall = 0.45, p-value=3.37e-13). While this finding serves as an independent sense check on the predictions in The Future of Skills, it says nothing about how job requirements in the workforce are changing and, if so, whether they are heading in the right direction. We can get a handle on this by repeating the calculation for every year for which we have data.

Figure 1 shows that this correlation has in fact been steadily rising year after year, suggesting that workforce skills, knowledge and abilities in the UK are indeed moving in the right direction: those features that employers will value more in the future are becoming more important in the workforce over time. Interestingly, we also see that the female workforce has a skills, knowledge and abilities profile that is markedly more correlated with what the research suggests will be in higher future demand.

Figure 2 explores which job requirements are driving this result. In particular, are we seeing growth in skills, knowledge and abilities that employers will value in the future and/or a decline in features that will become less valued? It does this by showing how the importance ranking of features has changed between 2001 and 2017. The figure suggests that the increasing correlation in Figure 1 is driven by both types of features.

Click here to view figure 2: Importance ranking of workforce skills over time, 2001-2017

Figure 3 plots this change over time for each job requirement (on the x-axis) against that job requirement’s correlation with occupations that are predicted to experience future growth in The Future of Skills (on the y-axis). The strong positive correlation confirms that those features that employers will value more in the future have already become more important over time. Note that both axes in the figure are dimensionless, as one is a correlation coefficient and the other is the absolute difference in importance between years 2017 and 2010 (measured on a scale from 1 to 5 as previously explained).

This figure is interesting, because any features that deviate significantly from the relationship may highlight areas where there are bottlenecks. For example, “design” (which O*NET defines as ‘knowledge of design techniques, tools and principles involved in production of precision technical plans, blueprints, drawings and models’) and “engineering and technology” (which O*NET defines as ‘knowledge of the practical application of engineering science and technology’) have decreased in importance, where the simple relationship might have expected them to be stable.

Meanwhile, the workforce importance of “explosive strength” and “dynamic flexibility”, two physical abilities involving quick and repetitive movements, has remained more or less unchanged on this measure, yet their outlook is one of decreased importance. In the right-hand side of the plot are features that have grown more strongly in importance than might have been expected given their future outlook, including “therapy and counselling”, “medicine and dentistry” and “customer and personal service” and “clerical”.

Limitations and future work

There are many limitations in both the data and methodology used in this post. Possibly the greatest is that we treat the requirements for a given occupation as unchanged over time. The O*NET data we have used are not longitudinal, nonetheless researchers in the US have used O*NET’s predecessors to create a time-series dimension. It will be important to investigate whether we can do this in our analysis too.

Another limitation reflects the relatively coarse 4-digit classification in the ONS’s Standard Occupational Classification for the UK compared with the 8-digit codes used in the US. For example, what is described as Medical Practitioners in the UK SOC (2211) corresponds to 13 different codes in the US (all of them starting with the same first 6 digits, 29-1069 and with slight differences in their skillsets). Because of this limitation, changes over time in the importance of many smaller occupations will not be reflected well in our analysis.

A third limitation reflects the fact that we estimate the average skill profile of the employed workforce, not the skills, knowledge and abilities of the unemployed. Nor do we account for variations across occupations and over time in skills gaps - gaps between what employers need their employees to do and what they can do - or in skills under-utilisation - when individuals have skills that are not deployed in their occupations.

Notwithstanding these limitations, we hope the analysis provides a useful diagnostic for identifying areas which may require attention from policymakers.