📄 earth_science

A country-sector dataset of environmental intensity and geographic variability across industrial activities

This paper presents a harmonized dataset of direct, supply-chain, and systemic environmental intensity for 6,781 sectors across 44 countries and five Rest-of-World regions, utilizing thirteen indicators to demonstrate that while both sector characteristics and geographic context significantly influence environmental performance, sector effects account for a larger share of the observed variation.

Original authors: Osama Diab

Published 2026-07-15

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Osama Diab

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the global economy as a massive, bustling kitchen where 44 different countries are cooking up 6,781 different types of dishes (industrial sectors). Some dishes are just a simple salad (services), while others are complex, multi-layered lasagnas (manufacturing) or smoky, fire-heavy grills (mining).

For a long time, chefs and critics tried to figure out which dishes were the "messiest" in terms of environmental impact. But they were speaking different languages. One chef measured the mess in "spilled milk," another in "burnt toast," and a third in "smoke in the air." It was impossible to compare them fairly.

Enter Osama Diab, a researcher from KU Leuven, who decided to translate all these messy measurements into a single, universal language: a ranking system. Instead of saying "this sector spilled 500 liters of water," the new dataset says, "this sector is in the top 90% of water messiness compared to everyone else."

The Three Layers of the Mess

To get the full picture, the study looked at the environmental "mess" in three distinct ways, like peeling an onion:

The Direct Mess (Direct Intensity): This is the smoke and spills happening right inside the kitchen while the chef is cooking. It's the pollution generated by the sector's own operations.
The Hidden Mess (Supply-Chain Intensity): This is the mess created before the chef even starts. Did the flour grow using dirty water? Was the steel melted using coal? This layer captures the pollution hidden in the ingredients and the delivery trucks.
The Total Mess (Systemic Intensity): This is the final score, combining the kitchen smoke and the hidden ingredient mess. The author used a special math trick (a geometric mean) to ensure that a sector can't hide a huge problem in one area just because it's clean in the other. If you have a smoky kitchen and dirty ingredients, your score goes way up.

The Great Discovery: It's About the Recipe, Not the Chef

One of the most exciting findings is answering a big question: Does the country matter more, or does the type of work matter more?

Imagine two bakers: one in Belgium and one in Brazil. If they both bake the same type of cake, is the mess different?

The Paper's Finding: The study ran a statistical test (like a referee blowing a whistle) and found that the type of activity (the recipe) explains more of the difference in messiness than the location (the country).
The Nuance: While the recipe is the biggest factor, the location still matters a lot. Some sectors, like mining and electricity generation, show huge differences in messiness depending on where they are. This suggests that for these specific jobs, the local technology and rules can make a massive difference. But for others, like manufacturing, the process itself is so inherently "messy" that it doesn't matter much where you do it.

The "Top 20" Most Variable Jobs

The study identified the top 20 sectors where the environmental performance varies the most from country to country. These are mostly mining, extraction, recycling, and electricity generation. It's like finding out that while a "frying pan" is always a frying pan, the quality of the oil used to fry in it varies wildly depending on who is holding the pan.

What This Dataset Is (and Isn't)

This dataset is a powerful tool for ranking and comparing. It helps policymakers and companies see who is doing better or worse relative to others.

However, the paper is very clear about what this tool cannot do:

It's not a ruler for absolute weight: A high score doesn't mean a sector is "heavy" with pollution in absolute tons; it just means it's "heavier" than most others.
It's not a measure of total pollution: Because the scores are based on money (economic output), a high-tech, expensive industry might look "clean" per dollar earned, even if it produces a lot of total pollution. It's like saying a luxury car is "efficient" because it costs a lot, even if it burns a lot of gas.
It's not a magic bullet: The paper suggests these scores are great for benchmarking and spotting trends, but they don't solve the problems themselves.

The Bottom Line

By turning 13 different types of environmental data (from greenhouse gases to water use) into a single, fair ranking system, this dataset gives us a clear map of the industrial world. It tells us that while what we produce is the biggest driver of environmental impact, where and how we produce it can still make a significant difference. It's a new lens that helps us see which industries are the most sensitive to change and where we might get the biggest bang for our buck in making the world cleaner.

Technical Summary: A Country-Sector Dataset of Environmental Intensity and Geographic Variability

Problem Statement
Existing environmentally extended multi-regional input-output (EEMRIO) databases, such as EXIOBASE, are critical tools for quantifying environmental impacts but suffer from significant heterogeneity. Environmental indicators are typically reported in disparate units and scales, rendering cross-indicator, cross-sector, and cross-country comparisons difficult. Furthermore, while these databases contain the necessary data to explore whether environmental pressures stem from inherent sector characteristics or geographic contexts (e.g., technology, regulations, resource quality), they lack harmonized metrics to facilitate such systematic analysis. Consequently, there is a gap in understanding the magnitude and geographic dependence of environmental pressures across industrial systems.

Methodology
The study constructs a harmonized dataset covering 6,781 economic sectors across 44 countries and five Rest-of-World (RoW) regions, derived from the EXIOBASE 3.8.1 database. The methodology involves three primary stages:

Indicator Selection and Transformation:
- Thirteen environmental indicators were selected to cover climate change, air pollution, acidification, eutrophication, ecotoxicity, human toxicity, photochemical ozone formation, and water use. These were chosen for their availability, conceptual distinctiveness, and representation of midpoint impact categories.
- To address unit heterogeneity and skewed distributions, all indicator values were transformed into percentile ranks across all country-sector observations. This approach preserves relative ordering while minimizing the influence of outliers and extreme values. Structural zeros (non-applicable combinations) were excluded prior to normalization to prevent conflating non-applicability with low environmental burdens.
Calculation of Environmental Intensity Dimensions:
- Direct Environmental Intensity ( $D_i$ ): Calculated as the arithmetic mean of percentile ranks for indicators generated within a sector's own operations.
- Supply-Chain Environmental Intensity ( $P_i$ ): Estimated using Leontief inverse analysis ( $L = (I - A)^{-1}$ ) applied to the technical coefficient matrix ( $A$ ) and direct environmental coefficients ( $e$ ). This captures pressures embodied in upstream production networks. The resulting values were also converted to percentile ranks and averaged.
- Systemic Environmental Intensity ( $S_i$ ): Integrated into a single metric using the geometric mean of direct and supply-chain scores ( $S_i = \sqrt{D_i \times P_i}$ ). This method was chosen to emphasize activities that are environmentally significant across both operational and supply-chain perspectives, reducing the weight of observations that score highly on only one dimension.
Geographic Variability Quantification:
- Geographic variability was defined as the standard deviation of systemic environmental intensity scores across countries for each specific sector. This metric assesses the dispersion of environmental performance for identical activities under different national contexts.
Data Filtering and Validation:
- Observations with valid data for fewer than 50% of the selected indicators were excluded. Robustness checks confirmed that increasing this threshold to 75% yielded identical results, indicating high data coverage.
- Technical validation included Spearman rank correlation analyses to test the stability of rankings against alternative aggregation methods (e.g., arithmetic vs. geometric mean, mean vs. median percentiles) and indicator coverage thresholds.

Key Results

Dataset Composition: The final dataset comprises 6,781 country-sector observations with harmonized scores for direct, supply-chain, and systemic environmental intensity, alongside geographic variability metrics.
Sectoral Patterns:
- Manufacturing and Construction: These sectors exhibit moderate direct intensity but significantly higher supply-chain intensity, indicating that their environmental burdens are heavily driven by upstream networks.
- Primary Sectors (Agriculture, Forestry, Mining, Utilities): These sectors display the opposite pattern, with high direct intensity and lower supply-chain intensity, suggesting their impacts are intrinsic to their operational boundaries.
- Services: Service sectors generally show the lowest systemic environmental intensity scores.
Geographic Variability:
- A two-way analysis of variance (ANOVA) revealed that sector effects explain a larger proportion of variance ( $\eta^2 = 0.414$ ) than country effects ( $\eta^2 = 0.224$ ). This suggests that environmental intensity is more strongly associated with the nature of the production activity than with the geographic context.
- However, significant geographic variability exists in specific sectors, particularly mining, extraction, recycling, electricity generation, and agriculture. These sectors are characterized by high environmental intensity and high cross-country dispersion, making them prime candidates for sustainability upgrades and technology transfer.
Robustness: The study found very strong agreement between rankings derived from geometric and arithmetic means (Spearman's $\rho = 0.974$ ) and between mean and median percentile scores ( $\rho > 0.98$ ). The correlation between direct and supply-chain scores was positive but weak ( $\rho = 0.337$ ), validating the necessity of measuring both dimensions separately.

Significance and Claims
The paper claims that this dataset provides a transparent, reproducible framework for multidimensional benchmarking of environmental performance without requiring users to process raw EEMRIO data or perform complex input-output calculations. Its primary significance lies in:

Harmonization: Enabling comparison across heterogeneous environmental dimensions, countries, and sectors by converting diverse indicators into a common ordinal scale.
Differentiation: Distinguishing between direct operational pressures and upstream supply-chain pressures, which often follow different patterns across sectors.
Policy Utility: Identifying sectors where environmental performance is highly sensitive to geographic conditions versus those where it is intrinsic to the production process. This distinction supports targeted industrial policy, technology transfer analysis, and sustainable supply-chain management.
Relative Benchmarking: The authors explicitly state that the scores represent relative environmental intensity rather than absolute environmental burdens. Consequently, the dataset is most appropriate for ranking, comparative assessment, and exploratory analysis, rather than the quantification of absolute physical impacts.

The work is presented as a foundational resource for sustainability strategies, allowing researchers and policymakers to assess whether environmental intensity is driven more by sector characteristics or by the geographic context in which those sectors operate.