Browse Catapults
Transport System Catapult - Logo Home

Modelling and mapping job skills to improve JAM prospects with Essex County Council

“The Data Science Fellowship with Essex County Council (ECC) sought to identify the demand of job skills within Essex at a district level, using publicly-available jobs data from online jobs listings. This geographic understanding of job skills demand across Essex would enable the ECC to re-position their existing skills training workshops, or co-design new training, to the near-term needs of Essex-based employers. In turn, the ECC would be better equipped to alleviate employment-related deprivation within “Just About Managing” (JAM) households, by encouraging participation in targeted upskilling.

Importantly, this Fellowship project also sought to establish a methodology for longer-term monitoring of job skills trends across each Essex district. These job skills trends can provide valuable labour market insight, helping to guide the provision of ECC services and strategic policy for education and economic development. 


A JAM household is characterised by a complex combination of social, economic and cognitive factors. Whilst a JAM household is not solely defined by a job role or an income level, a JAM household income ranges between 75 – 100% of the Minimum Income Standard (MIS). This income level can result in “limited mental bandwidth” due to limited month-to-month financial stability and a “scarcity mindset” from their greater vulnerability to exogenous financial shocks. The MIS, as maintained by the Centre for Research in Social Policy (CRSP) at Loughborough University, is published as an annual benchmark for a minimum standard of living within the UK, as determined by members of the public. Below this household income level, there may be an impact on a household’s ability to meet material requirements and participate in society.

Figure 1: Fellowship project outline

Prior to the Data Science Fellowship, the ECC commissioned an external consultant to identify areas of intervention to alleviate the drivers of deprivation affecting JAM households. The external consultant highlighted the improvement of careers, skills and training opportunities within Essex, with a recommended focus on “increased re/upskilling” and providing “career role models” to JAM households.

To maximise the impact of the 12-week Fellowship project, the project scope progressed with a fundamental grounding in the conclusions and recommendations established within ECC’s prior work.

In order to capture a holistic representation of the job skills demand within Essex, two analyses workstreams were executed in parallel.

“The quantitative analysis workstream applied Natural Language Processing (NLP) topic modelling techniques to extract job skills from publicly-available online job data.”

Online job listings for each district of Essex were collated from well-known online jobs boards. The detailed job descriptions of these job listings contained unstructured “free-text”, which varied substantially in terms of written structure and text content. This inherent variation within the job descriptions meant the text required extensive cleaning before use within the topic model. The cleaned job descriptions also underwent preprocessing, where the descriptions’ long-form sentences were converted into word “tokens” ready for input into the Latent Dirichlet Allocation (LDA) topic model.

Figure 2: Quantitative analysis outline

After tuning the model’s hyperparameters, a final set of 10 unlabelled topics were presented to the ECC domain experts for interpretation. Each unlabelled topic comprised of a group of 10 keywords, which were understood to have some common semantic meaning and coherence by the LDA model. The ECC domain experts had to interpret these unlabelled groups of keywords and label them as job skills; interesting interpreted job skills included “ability to work under pressure” and “customer service”.

“The qualitative analysis workstream extracted consistently-occurring, underlying job skills from themes discussed during semi-structured interviews with Essex-based employers.”

Figure 3: Qualitative analysis outline

For the qualitative analysis workstream, the Fellowship project intended to collect employers’ insights by interviewing a sample of employers in proportion to Essex’s local economy. However, due to COVID-19 severely disrupting businesses and the project’s 12-week time constraint, only a limited number of employers were successfully interviewed. These interview transcripts were inspected for common job skills themes discussed across all interviews.

Whilst not statistically significant, valuable employer insight was gathered and a selection of job skills from the qualitative analysis were presented to ECC domain experts for impact analysis. The qualitative analysis added complementary real-world context to the project’s investigation of job skills and enriched the impact analysis by capturing valuable insight directly from Essex-based employers.

Figure 4: Impact analysis

“The Fellowship project leveraged the multi-disciplinary domain expertise within the ECC by engaging key colleagues from Data & Analytics, Research & Insight, JAM policy, Skills & Economic Growth, Adult Community Learning and Education departments.”

An interactive ‘impact analysis’ workshop was hosted by a Catapult service designer, to facilitate the domain experts’ evaluation of the job skills produced by the quantitative and qualitative analyses. By directly evaluating the job skills on their expected 12-month impact to the JAM group, the domain experts were quickly able to define a shortlist of “Quick Wins” job skills. These “Quick Wins” job skills represent an opportunity for the ECC to offer “easy-to-teach” yet high impact skills training to JAM individuals in “finding a job” or “performing a job” contexts.

Notably, the optimal format and level of the “Quick Wins” skills training will need to be defined in future work. The Fellowship project recommends the ECC continue to deepen collaboration efforts with Essex-based employers to co-design skills training opportunities. Additionally, increasing JAM participation in skills training opportunities requires careful consideration of their specific needs, including “targeted advertisement through JAM channels”, “time commitment flexibility” and “Adult Community Learning (ACL) Ambassadors within the JAM community.

Figure 5: Heatmap and radar plots displaying district-level insight of job skills demand

The ECC is now equipped with an evidence-based methodology to gather job skills insight from the demand-side of the Essex labour market. The project’s district-level insight into job skills demand enables the ECC’s Adult Community Learning department to re-position the provision of skills training to align with the near-term needs of Essex-based employers. Additionally, the ECC is now better equipped to target and alleviate employment-related deprivation for JAM households.

Collaborating with ECC’s domain experts at the project scoping, results interpretation and project recommendation stages was critical. This collaboration enabled the Fellowship project to deliver relevant insights to senior ECC stakeholders, whilst offering project recommendations suitably aligned with existing ECC strategic goals and policies.

The identification of job skills can be further refined by interviewing a larger number of Essex-based employers and collecting a larger number of Essex-based online jobs. The Fellowship project used online job data which was available mid-way through the coronavirus (COVID-19) lockdown. It would be prudent to monitor changes in job skills demand as the Essex economy normalises. Lastly, including interviews with Essex-based employees would enable future work to corroborate the relevance of the job skills identified from the employers’ perspective.

 

By Darren Ko – Data Science Fellow, Essex County Council