The framework / research 0.2

Measurement framework

GDP measures production. Wealth is a stock of assets. Prosperity describes opportunities and living conditions. Life evaluation, daily feelings, meaning and resilience each answer other questions. A credible system must retain those distinctions.

What changed from the first proposal

We no longer recommend the early six-dimension harmonic score as the headline. Its normalization and weights create implicit tradeoffs, its zero rule is too consequential to hide, and a country average can conceal concentrated deprivation. It remains available as a documented early proposal.

1. A randomly situated life

Start with the experience of a person drawn from the population, rather than the average country statistic. A serious lifetime model would integrate age-specific survival, health, household resources, time and security. It would then account explicitly for unequal outcomes and uncertainty across people and future states.

Uᵢ = Eω [ Σₐ βᵃ Sᵢ(a,ω) u(cᵢ,a, tᵢ,a, hᵢ,a, fᵢ,a) ]

This is a research architecture, not a globally estimated equation. β is a disclosed time preference, S is survival to age a, and ω represents future states. Utility’s cardinal scale, treatment of death, baseline value of life, substitutions and ethical weights all require justification. Expected utility can miss tail risk; distributional aggregation needs additional assumptions.

A consumption-equivalent comparison can ask what fraction of reference-country consumption would make expected lifetime welfare equal, holding the reference profile fixed. Jones and Klenow provide an important economic precedent. We should not calibrate a worldwide version from national averages while pretending the necessary joint life-course data exist.

2. Rights and sufficiency remain visible

Publish deprivation at explicit thresholds alongside any welfare summary. Wealth cannot silently compensate for lack of bodily security or civil liberty. A veto threshold must be visible and contestable; it should produce an explanation rather than an unexplained zero.

3. Separate ethical choice, measurement error, and identification

Weights and inequality aversion are value choices. Sampling and source-model error are measurement uncertainties. Unknown household overlap is an identification problem. These require different displays and must not be mixed into one confidence badge.

Our essentials laboratory demonstrates partial identification. Given published access shares, all feasible patterns of joint deprivation are considered. The resulting range is conditional on the model and marginals, not a probability statement about true national prosperity.

4. Prefer robust comparisons to compulsory rankings

Country A is robustly above B for an explicitly defined set Θ only if the lowest admissible difference is positive: infθ∈Θ [W_A(θ) − W_B(θ)] > 0. When ranges overlap, we say the comparison is unresolved under the stated evidence. Failure to establish dominance does not prove equality.

Three weight scenarios are a finite sensitivity check, not proof across every possible preference. Randomly sampling weights measures a chosen distribution over assumptions, not the probability a country is truly better.

5. Keep stocks, outcomes and causes apart

Natural, produced and human capital support future possibilities, but their monetary sum can conceal ecological thresholds and their valuation can embed labor-market inequality. A longevity residual against GDP describes a pattern; it does not identify government effectiveness. Environmental footprints constrain the future and should remain visible even when today’s outcomes are strong.

6. Publish only what the data can support

Use independent source versions, explicit population/period joins and missingness. A 2026 publication can describe 2023–2025 opinions, a 2024 balance sheet release can stop in 2020, and a mobility dataset can describe people born decades earlier. Do not manufacture a common year or hidden joint distribution.

Research foundations

The mathematical tools are established. Pluraterra’s proposed contribution is an explicit, reproducible application that exposes assumptions and refuses to force conclusions the data cannot identify.