• Overview
• The Problem LMI Is Designed to Address
• Purpose and Goals of LMI
• How LMI Is Structured
• Scale and Outputs of the LMI Model
• Modelling Approach and Validation
• Perspective on Artificial Intelligence and the Workforce

Overview

Labor Market Intelligence (LMI) is an analytical approach focused on understanding, sizing, and forecasting the global labour force. At its core, LMI aims to measure the scale of the global workforce, analyse its current composition, and predict how it is likely to change over time. The labour force is treated as the driving engine of the global economy, and LMI seeks to provide structured insight into where it stands today and where it is heading in the coming years.

LMI is built on the belief that the global labour force is at a pivotal moment. Technological disruption and shifting socio-demographic trends are materially reshaping how people work across the world. In response to these changes, LMI focuses on understanding today’s workforce, projecting how it may evolve by 2030, and estimating the total labour cost to the global economy over the next five to seven years.

The Problem LMI Is Designed to Address

LMI was developed in response to challenges observed across strategic industries globally. Many countries have identified key sectors that are critical to future economic growth, particularly those that rely heavily on technical and STEM-based skills. Research into these sectors revealed a shared underlying issue: a growing gap in skilled labour.

This challenge often stems from a misalignment between educational institutions, governments, and industry needs. As a result, many organisations are struggling to find the skills they require, and this mismatch is expected to become a significant problem in the near future. LMI shifts the focus away from individual industries and instead examines the labour force and the skills landscape as a whole, using a single lens to better understand and anticipate these risks before they become unmanageable.

Purpose and Goals of LMI

The primary goal of LMI is to help institutions, companies, and governments anticipate and address structural workforce challenges. By providing forward-looking insight into labour market composition and occupational trends, LMI aims to support better planning and decision-making. Rather than reacting to shortages after they emerge, stakeholders can use LMI to identify emerging pressure points and evolving workforce dynamics, enabling earlier and more deliberate intervention.

LMI also seeks to create a common evidence base that connects education, policy, and industry. By grounding discussions in consistent occupational definitions, workforce distribution data, and sector-level analysis and skills, LMI helps reduce uncertainty and enables more coordinated and evidence-based responses to labour market pressures.

How LMI Is Structured

The LMI framework is built on a strong analytical foundation: a comprehensive occupation taxonomy. This taxonomy defines and organises roles across the global labour market, providing a consistent structure for analysis. It also enables the establishment of a clear baseline year, currently set at 2025, which serves as the reference point against which future labour market changes are measured.

From this foundation, the work is organised into three core components. The first component focuses on constructing and validating the baseline labour market, ensuring that the initial dataset is robust and reliable. The second component centres on forecasting, projecting how occupations and workforce dynamics are expected to evolve over time. The third component addresses compensation, recognising that understanding workforce distribution must be complemented by insights into wages, earnings, and the broader economic value generated by different occupations.

Scale and Outputs of the LMI Model

The LMI computational engine generates employment data for every country in the world at an occupation level. It covers more than 1,800 occupations and produces estimates for workforce size in 2025 as well as forecasts through to 2030. In addition to employment numbers, the model also produces compensation data for each country.

To improve accuracy, detailed country deep dives are also conducted, either to support specific client engagements or to validate the world’s largest economies. Strengthening the model for major economies improves the reliability of results across the entire global dataset.

Modelling Approach and Validation

LMI uses a dimensional modelling approach that applies guardrails, arithmetic operators, ceilings, and floors to ensure results remain realistic. For example, total employment figures are constrained so that occupation-level numbers always add up to verified national employment totals. These starting figures are validated using a combination of economic indicators, internal calculations, and external benchmarks.

This approach ensures that while the model is comprehensive and detailed, it remains grounded in real-world labour force constraints and does not produce implausible outcomes.

Perspective on Artificial Intelligence and the Workforce

LMI approaches the impact of artificial intelligence on the labour market with caution and data-driven realism. While AI is disrupting many industries, it is also creating new opportunities and roles. LMI emphasises that workforce change happens at a human pace, shaped by education timelines, skill acquisition, and career transitions.

Because occupations represent people, workforce numbers cannot change exponentially over short periods of time. Degrees and training take years to complete, and skills take time to develop. As a result, LMI does not assume sudden, large-scale job losses, but instead models gradual transitions that reflect how labour markets actually evolve.

LMI models AI’s labour market impact through a dual-dimension framework: augmentation (productivity enhancement) and automation (task displacement). The framework estimates the combined, occupation-specific net effect of these forces, moving beyond one-dimensional impact assessments.

• The Three Core Components: Inputs, Engine, and Outputs
• Labour Taxonomy: Organising the Global Population
• Geographic Taxonomy: Structuring Location Data
• Guardrails: Guiding the Model
• The Four-Step Computation Process
• Future Developments

The Labour Market Intelligence (LMI) computational model is a large-scale statistical and machine learning system designed to map and forecast the global labour market. It takes structured data from different geographic levels and labour segments and allocates values to occupations across countries. The model establishes a baseline year of 2025 and then forecasts forward to 2030, producing occupation-level estimates for every country in the world. Baseline years are periodically revised to reflect the most recent validated national statistics.

At its core, the model aims to classify the global population - approximately 8 billion people - into clearly defined roles and activities, providing both a snapshot of the present and projections for the future.

The Three Core Components: Inputs, Engine, and Outputs

The model is structured around three main components:

1. Inputs – including labour and geographic taxonomies, as well as model “guardrails.”
2. Engine – the computational system that processes the inputs.
3. Outputs – occupation-level numbers for each country from 2025 to 2030.

The taxonomies provide the structure. The guardrails guide the calculations. The engine performs the computations. The result is a consistent, global dataset of occupation numbers over time.

Labour Taxonomy: Organising the Global Population

The labour taxonomy is a hierarchical framework that categorises the entire global population based on what they do. On the employed side, the taxonomy follows a four-level structure built on the QS industrial taxonomy:

• Occupation
• Subsector
• Sector
• Industry

For example:

• Farmer (occupation)
• Crops (subsector)
• Agriculture (sector)
• Environment (industry)

This structure ensures that every occupation fits logically within broader economic categories.

On the unemployed side, the taxonomy goes beyond a single “unemployed” label. It includes categories such as:

• Stay-at-home caregivers
• Volunteers
• Primary and pre-K students
• Out-of-school students

Each of these groups is mapped across subsector, sector, and industry levels. This means the model does not only account for employed individuals, but for every person in the world and their current activity.

Geographic Taxonomy: Structuring Location Data

The geographic taxonomy follows the World Bank’s geographic structure and extends it into more granular levels such as:

• Cities
• States
• Countries/Territories
• Global aggregates

This hierarchy serves two key purposes. First, it ensures consistent geographic classification. Second, it helps fill data gaps. If data for a specific occupation in a country is missing, the model can move up the geographic hierarchy, such as to the state or regional level, to infer and allocate values. This improves completeness and reliability.

Guardrails: Guiding the Model

Guardrails are one of the most important inputs in the model. They act as informed constraints or estimates that guide how the model allocates occupation values.

There are two main types:

Absolute Guardrails. These are fixed numeric values. Example: “There are 120,000 nurses in France.”

Relative Guardrails. These are expressed as shares of a parent category. Example: “10% of the K–12 subsector consists of math teachers in the US.”

Guardrails also include operators, which determine how strictly they are enforced:

• Fixed – The input value must remain unchanged in the output.
• Approximate – The value can vary within a defined range (e.g., ±X%).
• Ceiling – An upper limit the model cannot exceed.
• Floor – A lower limit the model cannot go below.

These mechanisms provide flexibility while maintaining discipline in the calculations.

The Four-Step Computation Process

The model computation follows four main steps:

Step 1: Population Split

Each country’s total population is divided into:

• Employed
• Not Employed (including students)

This establishes the high-level structure before drilling down into industries, sectors, subsectors, and occupations.

Step 2: Baseline Year Calculations (2025)

Using the labour taxonomy and guardrails, the model calculates occupation-level values for 2025. During this process, the model respects the rules and constraints defined by the guardrails.

Step 3: Forecasting to 2030

Once baseline values are set, they are projected forward to 2030 using growth assumptions derived from a separate methodology. This produces forward-looking occupation estimates for each year.

Step 4: Rebalancing

The final phase ensures internal consistency. Occupation totals must sum correctly to their parent subsector, sector, and industry totals. For example, all occupations within a subsector must add up to that subsector’s total.

After rebalancing, the result is a complete dataset of occupation numbers for every country, for every year between 2025 and 2030.

Future Developments

The computational model continues to evolve. Planned enhancements include:

• A Wage Model – To calculate wages alongside occupation numbers, enabling estimates of total global labour costs.
  • The above is already completed, we have compensation data for every occupation by country
  • A wage forecasting model is in planning.
• AI-assisted analytical tools are being explored to support additional validation and anomaly detection processes.
• Greater Geographic Granularity – Expanding from country-level analysis to more detailed city- and state-level estimates.
• Supply Demand Model
• Labor and/or Skill Shortage Analysis
• Expansion of our total number of occupations

Together, these developments aim to increase accuracy, granularity, and economic insight—moving beyond counting occupations to understanding the full value and cost of global labour.

• Starting a New Country: Scanning and Validation
• Data Quality and Preparation
• Global Coverage and Scale
• Choosing Countries and Creating Benchmarks
• Core Outputs of the Model
• Industries, Strategic Sectors, and Custom Priorities
• Continuous Improvement and Model Evolution

This section of the Labour Market Intelligence (LMI) methodology focuses on the “country build” process at a high level. The country build represents a deeper, more detailed extension of the core model. While the computational engine already generates baseline and projection data for countries globally, the country deep dive is undertaken when a specific project or client engagement requires greater accuracy, validation, and confidence in the data.

The process is shaped by the purpose of the engagement. Before beginning work on a country, the team clarifies the client’s requirements to determine the scope, priorities, and level of detail needed.

Starting a New Country: Scanning and Validation

When work begins on a new country, the first step is a broad scan of available information. This includes identifying all relevant sources, projections, and datasets produced by governments, non-governmental organisations, and industry partners.

A crucial early task is validating the total employment number for the baseline year (2025) and assessing projections for 2030. This can be challenging because employment figures are calculated differently across sources. Some reports consider the entire population, while others include only individuals aged 15 and above or apply different retirement assumptions. Establishing a consistent and reliable starting point is essential for ensuring accuracy throughout the rest of the model.

Data Quality and Preparation

Beyond collecting data, ensuring its quality is equally important. The data must be up to date, comprehensive, and inclusive of all major sectors, including areas such as technology.

Once cleaned and verified, the data is structured so that it aligns with the QS occupation taxonomy. This taxonomy allows mapping of national data into a standardized framework. From there, an initial set of outputs is generated, forming the foundation for further analysis.

Global Coverage and Scale

The LMI model has already generated data for approximately 192 countries and territories. The country build process gradually expands and strengthens this global repository.

Given the size of the global labour force (estimated at around 3.3 to 3.5 billion people) certain countries have an outsized impact. For example, China and India together account for roughly 1.5 billion workers, making them significant outliers in global comparisons. Understanding such scale differences is critical when interpreting results and selecting benchmark countries.

Choosing Countries and Creating Benchmarks

The selection of countries for deep dives is strategic. Sometimes the real value of analysing one country lies in how it compares to similar economies. Countries may be grouped based on economic structure, industrial composition, or total employed population.

For example, work conducted in the Middle East—covering Saudi Arabia, Bahrain, and Oman—demonstrates how similar countries can serve as benchmarks for one another. Saudi Arabia, as the more developed economy in this group, provides insight into potential future trajectories for the others. Such comparisons help governments and institutions understand how policy decisions and sectoral focus areas may shape their future labour markets.

Core Outputs of the Model

Despite variations in engagement requirements, certain deliverables remain consistent across all countries. The model produces:

• Employment numbers for over 1,800 standardized occupations for 2025
• Median wage data for each occupation (in US dollars, 2025)
• Forecasts for 2030

These three core data points remain fixed. However, the way the data is presented or segmented can change depending on client needs. Custom scenarios or alternative cuts of the data may be produced to support specific strategic objectives.

Industries, Strategic Sectors, and Custom Priorities

The occupation taxonomy maps the total employed population of a country into 12 high-level industries. In some engagements, all 12 industries are validated and analysed. In others, clients may focus only on selected strategic sectors.

For example, in work conducted with the one government, attention was placed on 8–9 identified strategic sectors. After validating all industries, additional modelling was used to determine which occupations were most critical within those priority sectors. This approach helps stakeholders understand which jobs may be at risk from automation, which are vital for emerging industries such as advanced manufacturing, and where workforce planning efforts should be concentrated.

Continuous Improvement and Model Evolution

The country build process continues to evolve. Each new country provides valuable learning, not only in terms of processes and calculations but also in understanding how labour market patterns differ across economies.

The model operates using relative shares across occupations and industries. As more countries are analysed, these relative shares are refined and improved. This continuous expansion strengthens the engine, enhances its precision, and increases confidence in both baseline data and future projections.

In essence, the country build process does not just expand coverage, but also actively improves the quality and robustness of the entire Labour Market Intelligence model.

• What Is a Crosswalk?
• Why Country-Level Patterns Are Important
• The Four Phases of the Crosswalk Process

1. Finding the Best Source Data
2. One-to-One Mapping
3. One-to-Many Mapping
4. Splitting and Allocation

• From Crosswalks to Guardrails
• The Role of Crosswalks in the Bigger Picture

In the Labour Market Intelligence (LMI) work, crosswalks are one of the most crucial components. They sit at the centre of how country-level data is connected to the broader global model. While the process is technical, the purpose is simple: to ensure that labour market data from different countries can be accurately understood, compared, and integrated into a consistent framework.

Crosswalks allow the model to reflect both global structure and country-specific realities. They are a foundational step before forecasting, compensation analysis, and other downstream outputs.

What Is a Crosswalk?

At its core, a crosswalk is a method used to connect two occupational taxonomies. Each country typically has its own occupational or sector classification system. The LMI model, however, uses a structured labour taxonomy with four levels:

• Industry
• Sector
• Sub-sector
• Occupation

A crosswalk maps a country’s official classification system onto this four-level taxonomy. The goal is to align local labour data with the LMI structure as accurately as possible, ideally using the most granular (detailed) data available.

This mapping enables two key outcomes:

1. Estimating total employment for a country.
2. Understanding country-specific labour patterns and how that country structures its economy.

The crosswalk also becomes the base for future forecasting and compensation analysis.

Why Country-Level Patterns Are Important

Although the model operates globally, accuracy begins at the country level.

Governments publish labour data in different formats. Some report at the sector level (for example, “construction” or “real estate”), while others provide occupation-level data. The crosswalk dives deeper into these structures, translating them into the LMI taxonomy.

By validating and refining country-level mappings, the model improves iteratively. This country-by-country approach ensures that global outputs are supported by strong local foundations.

The Four Phases of the Crosswalk Process

The crosswalk process can be understood in four broad phases:

1. Finding the Best Source Data

The first step is identifying the most reliable and granular data source available. This may be an occupation-level taxonomy or a sector-level taxonomy, depending on what the country provides. The goal is to secure the highest-quality dataset possible before any mapping begins.

2. One-to-One Mapping

The next phase involves identifying direct matches between the country’s taxonomy and the LMI taxonomy. For example, if a country lists “cardio surgeon” and the LMI taxonomy also contains “cardio surgeon” with an identical definition, this is a one-to-one match.

Importantly, matches are not based on names alone. Definitions are carefully reviewed, as occupations may be defined differently across countries. Alignment in meaning is essential to maintain consistency in the model.

3. One-to-Many Mapping

Often, the LMI taxonomy is more detailed than national systems. With over 1,800 occupations, it is typically more granular than many government classifications, which may contain 600–1,000 occupations. In these cases, one occupation in a national dataset may correspond to multiple occupations in the LMI taxonomy.

For example, a country might list “secondary school teacher” as a single category, while the LMI taxonomy separates teachers by subject. This creates a one-to-many mapping. Again, definitions and sector alignment are reviewed to ensure that the split is conceptually sound.

4. Splitting and Allocation

When one occupation maps to multiple LMI occupations, the data must be split appropriately. An equal split would be the simplest approach, but it would not necessarily be accurate. Instead, additional research is conducted using industry reports, government data, and other reputable sources to determine realistic shares.

These splits determine how employment numbers are distributed across occupations. The results are expressed as shares of a parent sub-sector, which then feed into the next stage of the model.

From Crosswalks to Guardrails

Once mapping and splits are complete, the outputs are translated into guardrails. Guardrails are a structured set of rules that guide the model run. They can include:

• Absolute values (where exact employment figures are known)
• Relative shares (where proportions within a sub-sector are defined)

If detailed occupation-level data is available, guardrails can be set at the occupation level, improving accuracy. If only sub-sector data is available, guardrails are set at that level. The model then uses these country-level guardrails to rebalance and generate consistent occupational estimates.

The Role of Crosswalks in the Bigger Picture

Although highly technical, crosswalks are essential to the integrity of the LMI model. They ensure that:

• Local labour structures are accurately represented
• National employment totals are preserved
• Occupational detail is aligned with consistent definitions
• Forecasting and compensation analysis are built on a validated base

In short, crosswalks translate diverse national labour systems into a unified structure. They form the bridge between country-level data and the global labour market model, ensuring both precision and comparability.

• The 2030 Horizon
• The Core Drivers of Labour Force Forecasting
• A Top-Down and Bottom-Up Model
• Top-Down
• Assumptions and Limitations
• Common Misconceptions About Labour Market Growth
• Forecasting as Structured Discipline

Forecasting is one of the most important components of Labour Market Intelligence (LMI). The global labour market is shaped by significant uncertainty: some countries face rapidly growing populations, others are ageing; some are accelerating technological adoption, while others are still developing core industries. These shifts create complex questions about self-sufficiency, national specialisation, and long-term workforce planning.

Most labour market forecasts operate at a very high level, focusing on entire economies or broad sectors. However, there is limited forecasting at the occupation level. LMI addresses this gap by projecting employment across approximately 1,800 occupations through to 2030, offering a far more detailed view of where jobs are likely to grow or decline.

The 2030 Horizon

The year 2030 serves as a meaningful milestone. It marks the end of the current decade and provides a medium-term horizon that is far enough away to show structural change, yet close enough to remain grounded in current data and demographic realities.

The forecasting process begins with an established baseline: the total labour force in 2025. The key challenge is projecting that figure forward to 2030. Rather than relying on simple trend extensions, the approach incorporates both projections and structured modelling to estimate how the labour force will evolve.

The Core Drivers of Labour Force Forecasting

To estimate the future size of a country’s labour force, several high-level factors are considered:

• Population forecasts
• Ageing patterns and dependency ratios
• Labour force participation rates (including gender mix)
• Unemployment rates and expected changes
• Industry/Sector/Sub-Sector/Occupational trends and dynamics

For example, countries that experienced high birth rates 10–20 years ago are likely to see rapid labour force expansion as those cohorts reach working age. Conversely, countries with ageing populations may experience workforce contraction. By combining these four factors, it is possible to estimate the likely total labour force by 2030.

These high-level metrics act as an “aiming area” for the overall forecast, ensuring that occupation-level projections remain consistent with demographic and macroeconomic realities.

A Top-Down and Bottom-Up Model

The forecasting approach combines two perspectives:

Top-Down

At the macro level, total labour force projections set the overall direction. Growth rates are applied to occupations, and the results are checked to ensure they align with the projected 2030 labour force totals.

Bottom-Up

At the detailed level, occupation growth rates are shaped by sector, sub-sector, and industry dynamics. Historical data and external projections, including government forecasts extending to years such as 2025, 2030, 2033, and 2050, are used to establish realistic growth relationships between occupations.

For instance, if the education sector grows steadily, growth rates between related occupations (such as teachers and administrators) must remain proportionally consistent. The same applies in finance, technology, agriculture, and other sectors. This prevents unrealistic distortions, such as management roles doubling relative to analysts without justification. Because the model works both top-down and bottom-up, it must continuously reconcile macro labour force totals with occupation-level detail.

Assumptions and Limitations

One of the primary limitations in forecasting is data availability, especially at detailed sector and sub-sector levels. In many countries, long-term occupation-level projections are scarce. To address this, the model draws on a wide range of sources, including governmental and non-governmental datasets. It also uses causative drivers (such as past birth rates influencing future workforce size) rather than relying solely on extrapolation. It also looks at student data, to understand how enrolments, and graduate rates impact the labour force as well.

As more countries are analysed, a growing internal data repository strengthens the model, gradually reducing data constraints over time.

Common Misconceptions About Labour Market Growth

Working closely with the data reveals several common misunderstandings.

Growth Is Not the Same as Demand - Occupation growth rates (CAGRs) are often confused with demand. Even if demand for a role, such as AI-related occupations, is strong, growth may be constrained by supply. Many technical roles require 4–7 years of education and experience, limiting how quickly the workforce can expand.

Very High Growth Rates Are Rare - Historically, most occupations grow within moderate ranges. If a projected growth rate exceeds 4–5% annually over a five-year period, it requires careful review and strong justification. There are exceptions. For example, solar panel installers saw high growth as countries invested in green energy. However, even these surges tend to taper once installation capacity stabilises.

Growth Often Occurs in Traditional Sectors - Labour force expansion does not always occur in high-profile or technology-driven roles. In many countries with rapid population growth, sectors such as agriculture and mining may absorb large numbers of workers. Growth is often strongest in foundational industries rather than the most publicised occupations.

Some Labour Forces Will Shrink - Not all forecasts show expansion. In some cases, labour forces are projected to decline due to demographic structure. For example, in countries with ageing populations and sustained low birth rates, total workforce numbers may contract by 2030. These outcomes may be uncomfortable, but they are grounded in demographic data.

AI is unlikely to impose massive disruptions or augmentations in the workforce in a short timeframe, and neither will all countries realize these effects in the same time span or pace.

Forecasting as Structured Discipline

Forecasting in LMI is not about predicting headlines or amplifying trends. It is a structured process that balances demographic fundamentals, sectoral dynamics, historical patterns, and realistic supply constraints.

By combining macro labour force modelling with occupation-level analysis, the approach provides a disciplined and transparent framework for understanding where the global workforce is heading by 2030.

• Estimating Wages Across 1,800+ Occupations
• Adjusting Wages by Sector
• Integrating Employment Structure
• Multi-Layered Validation and Quality Control
• Occupational Tagging and Hierarchies
• Must-Check Occupations
• National Minimum Wage Baseline
• Sector-Based Adjustments
• Outlier Detection and External Cross-Checks
• Why Compensation Matters

Compensation is a core pillar of the Labour Market Intelligence (LMI) framework. While employment estimates and forecasts tell us how many people work in different occupations, compensation data explains the economic value of that work. By estimating wages, the LMI framework can segment the approximately $30 trillion global labour market enabling deeper analysis of economic trends.

Pairing wage data with employment figures allows for the calculation of total labour force costs at both country and global levels. This makes it possible to assess labour cost trends, project economic growth, and analyse large-scale shifts, such as the impact of artificial intelligence on job roles or the cost trajectory of high-growth sectors over time.

Estimating Wages Across 1,800+ Occupations

The wage model estimates median compensation for more than 1,800 occupations in every country. However, comparable and detailed occupational wage data is often incomplete, inconsistent, or overly aggregated, particularly across seniority levels.

To address this, compensation data is standardised using the QS occupational taxonomy. The process begins with sourcing reliable, granular wage data from reputable government sources. For example, in the United States, wage data is validated using multiple official sources, including the Bureau of Labor Statistics (BLS). This data is then mapped to the QS taxonomy and refined to produce a consistent and validated set of wage estimates.

A global repository of median wage data serves as a benchmark. Where country-specific data is missing or incomplete, benchmark wages are scaled using sector-specific wage ratios sourced from international datasets such as the International Labour Organization (ILO). This approach ensures that wage estimates reflect local economic realities while maintaining consistency across occupations and countries.

Adjusting Wages by Sector

Economic structures vary widely across countries. Some economies are heavily agriculture-based, while others are dominated by services or high-skilled industries. To reflect these differences, the model adjusts wages using sectoral data.

The World Bank’s sector framework divides the economy into three broad sectors: agriculture, industry, and services. The ILO provides median wage ratios for these sectors across roughly 100 countries. Each occupation in the QS taxonomy is mapped to one of these sectors, and sector-specific wage ratios are applied to reflect wage disparities between sectors in different countries.

When sector data is unavailable for a specific country, a clustering approach is used. Countries are grouped based on factors such as economic maturity, labour market compensation levels, and population size. A similar “proxy” country is then identified, and its sector wage ratios are used to scale the base wage data. This ensures robust estimates even where public data is limited.

Integrating Employment Structure

Wages alone do not define a labour market. How many people work in each occupation matters as well. To account for differences in labour market structure, employment figures for each occupation are used to weight wage estimates.

The model generates a calculated median wage based on both estimated wages and the country’s employment distribution across occupations. This model median wage is then compared to an officially published national median wage from trusted government sources. If there is a large gap, adjustments are made.

When the model median wage falls within a 10% margin of the published median wage (after multiple verification checks) it qualifies as a “silver level” estimate. This step ensures the results are realistic and aligned with national economic conditions.

Multi-Layered Validation and Quality Control

Validation is central to ensuring that compensation data is accurate, structured, and economically meaningful.

Occupational Tagging and Hierarchies

Occupations are assigned tags (such as directors, engineers, or analysts) to organise roles by hierarchy and sector. This makes it possible to compare wages across similar occupations and ensure logical progression across seniority levels. Consistent wage ratios, such as analyst-to-manager pay ratios, are maintained within defined bounds across occupational groups to preserve internal coherence.

Must-Check Occupations

Certain roles, such as heads of state or military officers, are flagged as “must-check” occupations. These positions often have publicly available compensation data and vary significantly across countries. Their wages are carefully cross-checked against external sources to ensure accuracy.

National Minimum Wage Baseline

The national minimum wage is used as a lower-bound reference point. This ensures that the lowest-paid occupations remain aligned with a country’s economic standards and prevents unrealistic wage estimates.

Sector-Based Adjustments

Sector-based wage adjustments account for country-specific economic conditions. For example, if U.S. healthcare wages are used as a base reference, they may be moderated when applied to another country to avoid distortion. This prevents high-wage benchmark countries from skewing results elsewhere.

Outlier Detection and External Cross-Checks

Unusually high or low wages compared to similar economies are flagged as outliers and reviewed. Once reviewed, they are cross-referenced within their occupational group and hierarchy to ensure internal consistency.

Benchmark wages also undergo external validation. Credible wage ranges are established through research and cross-checked against multiple external sources beyond the original government data. Finally, dataset extremes (the highest and lowest paid occupations) are re-evaluated and validated to confirm overall dataset accuracy.

Why Compensation Matters

Compensation data transforms labour market analysis from counting jobs to understanding economic value. By combining employment distribution, sector structures, and validated wage benchmarks, the LMI compensation model provides a structured and comparable view of how labour markets function.

This enables policymakers, businesses, and researchers to:

• Assess total labour costs at country and global levels
• Analyse sector wage disparities
• Understand income distribution patterns
• Forecast economic shifts in emerging industries
• Evaluate the financial impact of technological change

In short, compensation modelling ensures that Labour Market Intelligence does not just measure work but measures its economic weight and trajectory.

• Identifying Outliers in Large Datasets
• Applying Country-Specific Economic Logic
• Source Selection and Reliability
• Handling Data Gaps with Proxies and Ratios
• Classification System
• An Iterative and Evolving Process
• Conclusion

The Validation Model is the final stage of the Labour Market Intelligence (LMI) process at QS. It ensures that model outputs are accurate, logical, and consistent before being finalised. After the computational model generates workforce estimates for a country, the validation process reviews and refines those results to ensure they reflect real-world labour market conditions.

Validation is a structured, multi-layered process that plays a central role in maintaining the integrity and reliability of the LMI product.

Identifying Outliers in Large Datasets

The LMI taxonomy includes more than 1,800 occupations, making outlier detection a critical part of the process.

Validation begins at the sector and sub-sector level. If something appears unusual, analysts perform deeper dives into specific sections. Once sector-level numbers appear accurate, attention shifts to occupations.

Occupations are ranked by output level, and special focus is placed on:

• The top 100 occupations
• The bottom 100 occupations

Certain occupations are isolated for logical review. For example, it would be highly implausible for a country to have hundreds of astronauts or heads of state. These checks help identify obvious distortions that may not be statistically extreme but are logically unrealistic.

Applying Country-Specific Economic Logic

Validation does not rely on numbers alone. It also incorporates economic reasoning and country-specific context.

For example:

• In Saudi Arabia, oil and gas should dominate within the energy sector.
• In China, logistics should be dominant within the mobility sector.

At a high level, GDP sector composition is used as a reference point to assess whether industry outputs align with gross value added by sector. Beyond GDP data, unique legal and economic characteristics are considered. For instance:

• Bus conductors are rare in the UK.
• Distilling is legally restricted in Saudi Arabia.

These real-world constraints are embedded into the validation process to ensure outputs reflect structural realities.

Source Selection and Reliability

Maintaining accuracy also depends on reliable sources.  Data sources are assessed based on reliability, transparency, recency, and methodological rigour.

Certain datasets known for inconsistencies are avoided. Instead, higher-confidence sources are relied upon, and supplemented with:

• Business registration data
• Firm counts
• Workforce scaling patterns

This layered approach ensures that estimates are grounded in credible and relevant information.

Handling Data Gaps with Proxies and Ratios

Many countries lack complete labour market data. In fact, missing data is one of the most common challenges faced during validation. To address this, proxy country values are used. Countries with similar demographic and economic characteristics are clustered together, allowing regional benchmarks to guide estimation.

In addition, real-world ratios are applied, such as:

• Firefighters per fire station
• Police per 100,000 people
• Doctors per population
• Student-to-teacher ratios

These occupation-to-population relationships help generate logical workforce estimates when direct data is unavailable.

Internal research knowledge also supports this process. For example, if a country’s health system is known to be understaffed, health occupations should not appear disproportionately large in the output.

Classification System

Validation progresses through three classification levels:

Bronze Level. After the initial model run, baseline checks are applied to reach a preliminary “bronze” standard.

Silver Level – Hierarchy Checks. To move from bronze to silver, structural hierarchy checks are conducted. These ensure organisational logic holds across the taxonomy. For example, there must be more sales analysts than sales managers and there must be more managers than directors. These hierarchy checks are applied across all occupations to maintain logical workforce pyramids.

Gold Level – Relative and Cross-Country Checks. Moving from silver to gold involves deeper validation:

• Relative checks compare occupations against one another. For example, there must be more drivers than pilots.
• Cross-country comparisons use clustering methodology to compare countries with similar economic and labour market profiles.

Statistical benchmarking techniques are applied to identify unusual deviations from comparable economies. This allows analysts to investigate nuanced differences and refine outputs further. The gold stage is continuously evolving. It represents the most advanced and complex level of validation within LMI.

An Iterative and Evolving Process

Even at the gold level, validation remains iterative. Outputs are continuously compared against source datasets and benchmark thresholds to ensure confidence in the results. By the end of a two-week sprint, outputs are refined to a level that meets internal confidence standards. Once validated, they are integrated into the broader global LMI model.

Importantly, the validation model is not static. It evolves with each country reviewed, incorporating new rules, better benchmarks, and improved methodologies. This ensures that the LMI product becomes increasingly accurate and reliable over time.

Conclusion

The Validation Model is the final safeguard of the Labour Market Intelligence process at QS. It transforms raw model outputs into credible, real-world labour market insights.

Through structured sprint cycles, economic logic checks, hierarchy validation, cross-country benchmarking, and continuous iteration, the process ensures that LMI outputs are both statistically sound and practically plausible.

Ultimately, validation is not just about correcting numbers, but it is about embedding economic reasoning, structural logic, and global context into every workforce estimate produced.