Methodology

The data within the Skills-First Workforce Initiative is based on publicly-available nationwide labor market information across thousands of employers.

The Burning Glass Institute (BGI) research team conducted an analysis of the skills across nine distinct roles. These roles include both Bureau of Labor Statistics Standard Occupational Classification (SOC)-level occupations and some roles that fell outside of SOCs that we then categorized based on Lightcast suboccupation-level roles:

  • First-Line Supervisors of Retail Sales Workers
  • Sales Managers
  • Software Developers
  • Financial and Investment Analysts
  • Customer Service Representatives
  • Retail Salespersons
  • Transportation, Storage, and Distribution Managers
  • Customer Service Managers (Suboccupation-level)
  • Product Managers (Suboccupation-level)

For each of these roles, BGI identified and assessed dozens of skills based on nationwide market demand signals, creating a detailed skills taxonomy. No proprietary data from any participating firm was used in conducting these analyses, nor are the analyses focused on the participating firms, the wages they pay, or their hiring activities. This taxonomy provides a foundational understanding of the key proficiencies consistently sought by employers for each role and how these skills may vary across the labor market. Additionally, BGI calculated a series of market-level metrics to evaluate each role and skill together (skill-role pair) for insights into labor market demand and trends.

Identifying Top Skills for Each Role

To systematically determine the most relevant skills for each role, BGI employed a weighted methodology leveraging skill rate, how often a skill appears across a nationwide universe of job postings, and Term Frequency-Inverse Document Frequency (TF-IDF) analysis, which helps identify skills that are uniquely important to the role. BGI uses this weighted methodology to balance between skills that are too generic across all roles and skills that are too unique to a particular role to get a comprehensive list of skills important to a role. BGI generated an initial list of approximately 70 skills per role using a weighted combination of skill rate (30% weight) and TF-IDF (70% weight), and then manually reviewed the top 70 skills per role to eliminate outliers and anomalies, ultimately finalizing a curated list of 60-65 skills per role. The following calculations used Lightcast’s 2023 job postings data to calculate skill metrics by role.

Skill Rate Calculation

Skill rate represents the prevalence of a skill in job postings for a given role. For example, the skill rate of "Communication" for Customer Service Managers was calculated as the number of job postings mentioning the skill "Communication" divided by the total job postings for Customer Service Managers.

TF-IDF Calculation

TF-IDF helps determine the relative importance of a skill to a specific role compared to all roles. This metric ensures that highly distinctive skills for a role are prioritized over more generic, widely applicable skills. This metric is calculated as the Term Frequency (TF) of a skill within job postings for a specific role multiplied by the Inverse Document Frequency (IDF), which downweights skills that appear frequently across many roles. If a skill is mentioned often in job postings for a particular role but rarely in postings for other roles, it will have a high TF-IDF score, indicating its significance to that specific role.

Calculating Market Metrics for Each Skill-Role Pair

To provide a data-driven perspective on skill demand and which skills help workers command a higher wage, BGI calculated four key market metrics:

  • Frequency in Role: the normalized calculation (adjusting the data to a common scale so values can be fairly compared within a role) of the skill rate metric mentioned above, indicating the proportion of job postings for a role that mentions a specific skill.
  • Specificity in Role: the normalized calculation of the TF-IDF metric mentioned above, which identifies how uniquely important a skill is to a given role relative to other roles.
  • Growth Rate in Role: the percentage change in skill rate from 2019 to 2023, highlighting skills that are becoming more or less prevalent over time.
  • Skill Specific Wage Impact: the relative impact of a specific skill on securing a wage premium within a role. This was estimated using a fixed-effects regression model (isolates the effect of one variable by controlling the other variables that don’t change over time) that compares job postings that require a given skill versus those that do not, controlling for education level, experience requirements, occupations, and year-over-year trends.

These metrics serve as the foundation for categorizing skill-role pairs into four key labels: High Growth, High Value, Low Demand, and Durable.