Does teamwork really matter? Looking beyond the job posting to understand labour market demands

Vicinity Jobs’ data includes more than three million 2023 OJPs and thousands of skills. Most skills appear in less than 0.02% of job postings, so most postings rely on a small subset of commonly used terms, like teamwork.

Laura Adkins-Hackett, Economist, LMIC, and Sukriti Trehan, Data Scientist, LMIC, presented their research exploring trends in the skills listed in OJPs to develop a deeper understanding of in-demand skills. This research project uses pointwise mutual information and other methods to extract more information about common skills from the relationships between skills, occupations and regions.

1. Does teamwork really matter? 2024-05-31 Laura Adkins-Hackett & Sukriti Trehan Looking beyond the job posting to understand labour market demands

2. Project Background

3. Find connections in OJP to extract additional insights into the needs of the labour market Examine changes in skill connections across occupations and regions This research will show how we can enhance OJP as an analytical tool Show the importance of context when deciphering job postings A step in using OJP to identify skills shortage mismatches and demand Approach Impact Connecting skills Problem The same language is used across job postings, so skills in Online Job Postings (OJP) make it hard to identify what is required for the job. If skills in postings don’t represent reality, this limits the value of online job postings as a source of Labour Market Information (LMI).

4. The Data

5. 2023 OJP in data   Online job postings provide a rich source of big data, which presents both opportunities and challenges for accurately measuring job vacancies. LMIC has partnered with Vicinity Jobs to access data from job postings across the web. 3,078,987 20,482,226 4,550 710 Job postings Total requirements Unique requirements Requirements in at least 500 postings

6. Job postings by commonly requested skills NOC Legend: 0 -  Legislative and senior management occupations; 1 - Business, finance and administration occupations; 2 - Natural and applied sciences and related occupations; 3 – Health Occupations; 4 - Occupations in education, law and social, community and government services; 5 - Occupations in art, culture, recreation and sport; 6 - Sales and service occupations; 7 - Trades, transport and equipment operators and related occupations;  8 - Natural resources, agriculture and related production occupations; 9 - Occupations in manufacturing and utilities

7. Methodology

8. Methodology Frequency Measure the frequency of skills co-occurring to identify relationships between terms.​ Normalized pointwise mutual information The strength of skill relationships is measured using Pointwise Mutual Information (PMI), a metric to evaluate the frequency of terms co-occurrences compared to their independent occurrences.  We use a normalized version of PMI to restrict the range of values between -1 and 1.​ Directional relationship Measure a metric called confidence, which quantifies the conditional probability of observing one skill in a job posting when another skill is required, hence identifying most probable predictors for a skill. Utilize association rule mining with Frequent Pattern Growth algorithm to identify skill relationships with the highest confidence levels.

9. Normalized PMI Confidence for directional relationships   ⇒ PMI ( ) > 0   Confidence (x -> y)   The PMI value is normalized by dividing it by a factor of   Quality of association rules can be measured by certain metrics including support and confidence. Support (x -> y)   PMI (x, y)   For two skills x and y, Intuitively for two skills: communication and problem solving, if : Probability of events x and y occurring together and  

10. Results

11. Skill associations with Customer Service

12. PMI by NOC 6 vs others

13. Predictors of Customer Service Confidence Cash registers 75% Sales 70% Work under pressure 49% Fast-paced Setting 49% Multi-tasking 48% Work scheduling 46% Inventory Management 46% Goal Oriented 45% Occupational Health and Safety 44% Flexibility 43% Confidence POS software 79% POS systems 77% Cash registers 73% Interpersonal Skills 72% Problem-Solving 71% Occupational Health and Safety 70% Sales 70% Microsoft Outlook 70% Microsoft Office 67% Microsoft Suite 67% Confidence Multi-tasking 41% Work under pressure 40% Inventory Management 39% Office Administration 38% Microsoft Outlook 37% Fast-paced Setting 36% Ordering of supplies and equipment 35% Attention to Detail 35% Handling heavy loads 34% Scheduling 34% All Occupations  Sales and service occupations (NOC 6) All other occupations

14. Possible Paths Forward Understanding skill associations across additional modifiers, including TEER categories and remote work options Undertaking a qualitative review of the analysis, involving interpretation of findings using raw job posting text Exploring Modifiers Qualitative Review Temporal Analysis Conducting a longitudinal to examine how skill association evolves over time

15. Questions?

16. Appendix

17. Similar work Skill-driven recommendations for job transition pathways Developed a model to predict job transitions in Australia. The first stage focuses on skill similarities. Created “skill spaces” which identifies the likelihood of 2 skills appearing in the same job posting or being used interchangeably. Skill spaces are developed with a pairwise similarity and a skill-based revealed comparative advantage. An Open and Data-driven Taxonomy of Skills Extracted from Online Job Adverts Building a skills taxonomy based on the likelihood of 2 skills appearing in the same posting or being used interchangeably. Similarities are mapped into clusters to identify common themes, which become the basis of the new taxonomy. Removes the high traversal skills as they fall in multiple clusters. Skills for jobs 2022: Mapping skill requirements in occupations based on job postings data OECD study to identify skill shortages and surpluses. Builds a skills profile based on the revealed comparative advantage in different industries. To ensure the process is reasonable they do a statistical comparison of the results to the O*Net skill profiles. The skill imbalance is calculated based on the RCA, occupation imbalance, and occupation size.

18. Official data sources in Canada have limitations such as time lag, limited data granularity, small sample size, and exclusion of certain employer categories. Online job postings offer real-time insights into job trends with a high level of granularity and valuable information about the skills required. Challenges include inconsistencies between postings and actual vacancies, identifying and removing duplicate postings, limited transparency in algorithms, and technological barriers. There is also a bias in online job posting data for certain occupations: higher-skilled white-collar sectors have greater representation. Many other occupations are under-represented in online postings. Online job posting data as LMI Online job postings provide a rich source of big data, which presents both opportunities and challenges for accurately measuring job vacancies. LMIC has partnered with Vicinity Jobs to access data from job postings across the web.

19. "Teamwork" and networked skills

20. Overview of Skills Social-Emotional Skills Occupational Skills Technologies Tools and Equipment Number of skills in the group 51 302 1,733 2,464 Minimum number of job postings 35 2 1 1 Median job postings 58,089 3,055 29 14 Maximum number of job postings 1,322,765 921,645 329,208 60,250

21. Predictors of Teamwork Confidence Goal-Oriented  71%  Work under pressure  70%  Self-starter / Self-motivated  66%  Fast-paced setting  65%  Interpersonal Skills  64%  Decision-Making  64%  Key Performance Indicators  64%  Attention to Detail  64%  Analytical Skills  63%  Writing  63%  All Occupations     Top Predictor (1)  Top Predictor (2)  Top Predictor (3)  NOC 0  CRM software (75%)  Critical Thinking (72%)  Microsoft Access (72%)  NOC 1  Critical Thinking (70%)  Decision-Making (66%)  Coaching (66%)  NOC 2  Coaching (74%)  Work under pressure (73%)  Interpersonal Skills (73%)  NOC 3  Work under pressure (76%)  Dexterity (74%)  Writing (71%)  NOC 4  Dexterity (85%)  Critical Thinking (75%)  Reports preparation (72%)  NOC 5  Adobe Systems Adobe Creative Suite (77%)  Search Engine Optimization (73%)  Maya (73%)  NOC 6  Goal oriented (74%)  Work under pressure (74%)  Interpersonal Skills (69%)  NOC 7  Interpersonal Skills (72%)  Hand-eye coordination (71%)  Work under pressure (70%)  NOC 8  Interpersonal Skills (82%)  Dexterity (80%)  Team Building (80%)  NOC 9  Work under pressure (74%)  Dexterity (69%)  Self-starter / Self-motivated (69%)  By Occupations  Note: The values in brackets are the confidence, or likelihood that teamwork will be in a job posting based on the presence of the predictor skill 

22. Predictors of Communication skills Confidence Writing  83%  Analytical Skills  78%  Presentation Skills  78%  Negotiation Skills  78%  Conflict Management Skills  76%  Interpersonal Skills  75%  Research Skills  74%  Microsoft Access  73%  Multi-tasking  70%  Microsoft Outlook  70%  All Occupations     Top Predictor (1)  Top Predictor (2)  Top Predictor (3)  NOC 0  Writing (93%)  Microsoft Access (86%)  Microsoft Windows (83%)  NOC 1  Writing (86%)  Presentation Skills (83%)  Interpersonal Skills (80%)  NOC 2  Writing (85%)  Presentation Skills (84%)  Negotiation Skills (82%)  NOC 3  Analytical Skills (88%)  Writing (87%)  Computer Skills (85%)  NOC 4  Writing (83%)  Negotiation Skills (81%)  Multi-tasking (80%)  NOC 5  Interpersonal Skills (74%)  Goal Oriented (73%)  Analytical Skills (72%)  NOC 6  Writing (82%)  Analytical Skills (77%)  Presentation Skills (77%)  NOC 7  Writing (80%)  Analytical Skills (72%)  Microsoft Outlook (65%)  NOC 8  Writing (85%)  Analytical Skills (81%)  Lean Manufacturing (75%)  NOC 9  Writing (77%)  Microsoft Outlook (68%)  Analytical Skills (68%)  By Occupations  Note: The values in brackets are the confidence, or likelihood that teamwork will be in a job posting based on the presence of the predictor skill