Whose Values Decide? A Global Survey of 3,198 People Sets New Weighting Factors for LCA

In short: Every time a sustainability assessment produces one number, someone has decided how much a year of human health is worth relative to a species at risk of extinction. Usually that decision is invisible. A UNEP-hosted study asked 3,198 people across four income groups to make it explicitly — and the answers differ systematically by where respondents live.

The problem weighting solves, and creates

Life cycle assessment produces results across multiple impact categories. Often they point in different directions: one option is better for climate, worse for water; one reduces toxicity but increases land use. At that point the decision-maker faces an unavoidable question — which matters more?

Under ISO 14044, weighting is an optional element, defined as converting and possibly aggregating indicator results across impact categories using numerical factors based on value-choices. That last phrase is the crux. Weighting is where value judgement enters the calculation openly.

The study's authors make a sharp observation about what happens when weights are unavailable: decision-makers weight the categories anyway — consciously or subconsciously, often equally — and remain exposed to criticism that their weighting favours conclusions they already preferred.

The GLAM context

The Global Guidance for Life Cycle Impact Assessment Indicators and Methods (GLAM) project, supported by the Life Cycle Initiative hosted by UNEP, aims to establish consensus on LCIA indicators and methods. Its methodology rests on three endpoint categories, called Areas of Protection:

Human health (HH) — measured in DALYs, days of healthy life lost per person per year
Ecosystem quality (EQ) — percentage of global terrestrial species put at risk of extinction
Natural resources and ecosystem services (NRandES) — USD lost per person per year

No existing weighting set fit these specific categories. So the project's weighting subtask built one.

The methodological groundwork is worth noting: the subtask developed WEMSS, a freely available software assessing 35 weighting methods against 50 key decision-making features, to help analysts select an appropriate method. Running GLAM's own requirements through WEMSS narrowed 35 methods to a shortlist of 10, from which a discrete choice experiment was selected.

How the survey worked

A discrete choice experiment (DCE) presents respondents with competing scenarios and infers their preferences from the choices they make. Each scenario described the three Areas of Protection at one of nine levels — ranging from a 100% reduction to a doubling of the current impact.

Respondents saw nine choice tasks, each with a reference scenario (no change) and two hypothetical alternatives, and picked the one they preferred.

The reference scenarios were anchored to real normalization values: 19 days and 55 days of healthy life lost per person per year (two values, to test sensitivity); 12% of terrestrial species at risk of extinction per the IUCN Red List; and 6,480 billion USD in annual global losses of natural resources and ecosystem services — equivalent to 7.5% of global GDP.

That last figure was translated to per-capita terms by income group, producing very different absolute stakes: 3,398 USD for high-income countries, 711 for upper-middle, 170 for lower-middle, and 51 USD for low-income countries.

Sampling was deliberately global. Face-to-face interviews were conducted in Uganda (five districts, five villages each, five randomly selected households per village) and Burkina Faso (eight of 13 regions, every 10th household from a random urban starting point). Survey companies recruited quota samples in Tokyo, Shanghai and Mumbai. Türkiye combined face-to-face and snowball sampling. The survey ran in 11 languages.

The authors are direct about why: internet penetration is 89% in high-income countries but 19% in low-income countries — so web administration alone would have excluded exactly the populations most often missing from this literature.

The headline weights

Two independent calculation approaches were applied to the same choice data — an econometric approach grounded in random utility theory, and an MCDA disaggregation approach grounded in deterministic value theory. Using two methods was a deliberate robustness test.

Across all income groups:

| Area of Protection | Econometric | MCDA linear | |---|---|---| | Human health | 0.42 | 0.41 | | Ecosystem quality | 0.31 | 0.32 | | Natural resources & ecosystem services | 0.26 | 0.27 |

Two fundamentally different methods, applied to the same preferences, agreed to within 0.01. That convergence is the study's strongest methodological result.

The confidence intervals also confirm these weights differ from equal weighting (1/3, 1/3, 1/3) in a statistically significant way — meaning the common default of weighting everything equally does not reflect what people actually prefer.

Where it gets interesting: income changes the ranking

Disaggregated by World Bank income group, the picture shifts:

| Income group | HH | EQ | NRandES | |---|---|---|---| | High | 0.34 | 0.41 | 0.25 | | Upper-middle | 0.36 | 0.36 | 0.28 | | Lower-middle | 0.36 | 0.32 | 0.32 | | Low | 0.54 | 0.24 | 0.22 |

High-income respondents are the only group that ranks ecosystem quality above human health. Every other group prioritises human health — and in low-income countries the weight reaches 0.54, rising to 0.58 when the higher health reference scenario was shown.

The authors note that the low-income group's human health weight, even at its lowest (0.49), exceeds the highest human health weight recorded in any other income group.

This is not a minor calibration difference. It means an LCA using high-income weights and one using low-income weights could reach opposite conclusions about the same product — and both would be faithfully representing real human preferences.

The authors connect this to Eco-indicator 99's cultural theory framework: high-income respondents display egalitarian traits (prioritising ecosystems), while lower-income groups are more individualist (prioritising human health).

The population adjustment

Here the study does something unusually careful. The survey sample doesn't match the world's population distribution: low-income respondents were 29% of the sample but 9.14% of world population, while lower-middle-income countries are 43.23% of world population but only 16% of responses.

Recalculating with population shares produces different global weights:

| AoP | Survey-based | Population-adjusted | |---|---|---| | Human health | 0.42 | 0.37 | | Ecosystem quality | 0.31 | 0.34 | | Natural resources | 0.26 | 0.29 |

The adjusted weights sit noticeably closer to equal weighting — a reminder that a global average depends heavily on who gets counted.

A worked example, and why it matters

The study includes an application to canned yellowfin tuna in brine (1 kg functional unit). After normalising the three impacts and applying the weights, the single score breaks down as:

Ecosystem quality: 95.4% of total impact
Human health: 4.6%
Natural resources: 0.001%

The weights ranged only from 0.42 to 0.26 — barely a factor of two. Yet one category dominates the result almost entirely. The normalization values, not the weights, drove that outcome.

The authors flag this dependency explicitly: the weights are likely dependent on the normalization values used in the survey, and they specify tiered recommendation levels — high confidence when used with the normalization values they were developed with, medium confidence within tested upper boundaries (110 days per person per year for health, 24% of species at risk, 6,796 USD lost per person), and low confidence outside them.

Three limitations worth carrying forward

ISO forbids public single scores in comparative claims. The authors state plainly that these weights — like any weights — cannot be used where an ISO-compliant study needs to compare product systems and disclose weighted single-score results publicly.

These weights are global, not local. They are unsuitable for assessments focused on local or regional scale impacts, where weighting should come from the affected population.

Separating humans from nature is itself a value choice. The authors raise this candidly: splitting Areas of Protection into human health, ecosystem quality and natural resources can imply that human beings are separate from nature, risks underplaying the interactions between social and natural systems, and may neglect long-term effects such as future generations affected by ecosystem damage.

They also note a splitting bias risk: dividing an objective into two sub-objectives tends to increase the combined weight — meaning comparisons across methods with different category structures are not straightforward.

What a business should take from this

The transferable lesson isn't the specific numbers. It's a governance principle.

Any aggregated sustainability score contains a value judgement. If an organisation reports a single ESG score, an eco-score, or a composite sustainability index, weights were applied — whether stated, defaulted to equal, or embedded invisibly in a vendor's methodology.

Weights should come from somewhere defensible. The three legitimate options, per this study's framing, are: elicited from the actual decision-makers, elicited from affected stakeholders, or drawn from a documented external set like this one — used precisely when decision-makers cannot or prefer not to impose their own preferences, or when no decision-maker is involved.

Whose preferences are embedded is a strategic question. For a company operating across Latin America, Europe and the United States, this study demonstrates empirically that stakeholders in different markets weight these trade-offs differently. A single global weighting scheme is a choice about whose values take precedence — one worth making deliberately rather than by default.

Frequently asked questions

What is weighting in life cycle assessment? Under ISO 14044, weighting is an optional element defined as converting and possibly aggregating indicator results across impact categories using numerical factors based on value-choices. It is where value judgement enters the calculation openly.

What weighting factors did the UNEP GLAM study produce? Across all income groups: 0.42 for human health, 0.31 for ecosystem quality, and 0.26 for natural resources and ecosystem services using the econometric approach — and 0.41, 0.32, 0.27 respectively using the MCDA approach. Two different methods agreeing to within 0.01.

Do weighting preferences differ by country income level? Substantially. High-income respondents are the only group ranking ecosystem quality above human health, at 0.41 versus 0.34. Low-income respondents weight human health at 0.54, rising to 0.58 with the higher health reference scenario.

What are the three Areas of Protection in GLAM? Human health, measured in DALYs; ecosystem quality, measured as percentage of global terrestrial species at risk of extinction; and natural resources and ecosystem services, measured in USD lost per person per year.

Can these weights be used in an ISO-compliant public comparison? No. The authors state plainly that these weights — like any weights — cannot be used where an ISO-compliant study needs to compare product systems and disclose weighted single-score results publicly.

Is equal weighting a neutral default? No. The confidence intervals confirm these weights differ from equal weighting (1/3, 1/3, 1/3) in a statistically significant way. Weighting everything equally is itself a value choice, not the absence of one.

Why is normalization as important as weighting? In the study's tuna case, ecosystem quality accounted for 95.4% of the single score while the weights ranged only from 0.42 to 0.26 — barely a factor of two. The normalization values, not the weights, drove the outcome.

How Sustek.co makes value choices explicit

If your sustainability score aggregates multiple dimensions into one number, weights were applied — and someone should be able to say whose. Sustek.co structures that decision openly:

Sustainability Pulse — Determines which impact categories are decision-relevant for your sector and markets, before any index is built on top of them. Sustrategize™ baseline diagnostic: ESG maturity assessment, data gap analysis, and circular economy potential mapping.

Sustainability Navigator — Structures materiality and prioritisation within the transformation roadmap with value choices made explicit rather than buried in an aggregate index. Double materiality assessment, circular value-at-stake in CFO-legible figures, 4IR technology sequencing, and a board-ready Transformation Blueprint aligned to CSRD, GRI and TCFD.

Sustainability Command — Reports against those priorities continuously through Quarterly Executive Impact Dashboards and investor-grade disclosure, with social value quantified through ValueFlow™ — our SROI platform powered by the MeasureUp proxy bank, built precisely so the social dimension carries real numbers rather than narrative.

For organisations operating across LATAM, Europe, Brazil and the US — where, as this study demonstrates, stakeholder priorities genuinely differ — Iconet® provides on-demand access to 93+ vetted sustainability consultants with regional and sector-specific expertise, without fixed structural cost. The Iconet® x AWS reStart program carries a verified SROI of 6.41:1.

Sustek.co is an S.A.S. BIC certified under Sistema B Colombia, and part of Google for Startups and AWS Activate.

Whose values are embedded in your sustainability score right now? If nobody can answer that, the diagnostic is where to start: sustek.co/services

Source: Bayazıt Subaşı, A., Askham, C., Dancke Sandorf, E., Dias, L.C., Campbell, D., Taş, E.F., Itsubo, N., Nagawa, C.B., Kyarimpa, C.M., Djerma, M., Bazie, B.S.R. & Cinelli, M. (2024). "Weighting factors for LCA—a new set from a global survey." The International Journal of Life Cycle Assessment, 29, 2107–2136. ISSN 0948-3349. Open access under CC BY 4.0. Developed within the GLAM project of the Life Cycle Initiative hosted by UNEP.