A strong negative cross-sectional association.
Open discussions OPEN DISCUSSION DRAFT · VERSION 1.0 · 28 JULY 2026
AIR POLLUTION · LIFE EXPECTANCY · CAUSAL IDENTIFICATION
When a strong association
does not license an effect claim.
A dual-track reanalysis of PM2.5 and life expectancy across countries, designed to preserve the observed pattern while refusing a causal conclusion that the data cannot identify.
当显著相关不足以支持“效应”结论：PM2.5 与期望寿命跨国数据的双轨再分析。
WHAT THIS PAPER DOESIt keeps the signal.
It keeps the signal.
It lowers the claim.
The first GHEAS paper is preserved as a frozen baseline. This new discussion draft begins again from the full evidence trail: one track stress-tests the country panel; the other searches for stronger policy-based designs.
原论文仍作为冻结范本保留。这一版不掩盖显著相关，也不把模型中的符号变化解释成真实健康效应；它只保留当前设计能够承担的结论。
01 — THE ESTIMATE PATHThe answer changes
The answer changes
as assumptions change.
94.7% attenuation after measured development context.
A sign reversal—not a credible protective effect.
The estimate becomes small and inconclusive.
A future value predicts the present outcome: the design fails.
Estimates are years of life expectancy per 10 µg/m³ higher PM2.5.
02 — TRACK AFour gates tested.
Four gates tested.
Four gates failed.
Stable direction
FAILThe coefficient changes sign across reasonable models.
Core intervals
FAILIntervals do not sustain one stable substantive effect.
Placebo logic
FAILFuture exposure predicts current life expectancy.
Bias scenario
FAILResidual confounding remains a quantitatively plausible explanation.
The licensed Track A conclusion is descriptive association—E1—not an independent or causal PM2.5 effect.
03 — TRACK BA valid stop is
A valid stop is
a scientific result.
The causal-design pipeline worked on synthetic data. In three real policy candidates, the exposure stage failed before a health estimate could be justified.
US PM2.5 standard
The apparent first stage disappears after county-specific trends. The threshold is too weak for health estimation.
STOP BEFORE HEALTH →US nitrogen-budget policy
Ozone changes are preceded by strong differential trends. The design stops before a health outcome is analysed.
STOP BEFORE HEALTH →India catalytic-converter policy
The published arithmetic reproduces, but alternative comparison sets and city trends erase precision. The design stops.
STOP BEFORE HEALTH →What the evidence supports
Countries with higher PM2.5 often have lower life expectancy in the observed data. That relationship is real as a description.
What it does not support
These national panel data do not identify the direction or magnitude of a causal PM2.5 effect on life expectancy.
05 — READ THE FULL DRAFTTwo languages.
Two languages.
One auditable argument.
当显著相关不足以支持“效应”结论
完整的双轨分析、数值结果、证据门槛、局限和参考文献。
When a Strong Association Does Not License an Effect Claim
The complete methods, numerical results, evidence gates, limitations, and references.
The original paper remains unchanged as the frozen baseline ↗
06 — OPEN REVIEWChallenge the boundary,
Challenge the boundary,
not just the coefficient.
We especially welcome criticism of:
- the choice of national life expectancy as the outcome;
- the Track A model set and future-exposure placebo;
- the stop rules applied to the three Track B candidates;
- whether any sentence exceeds the evidence level E1.
Comments remain attached to the paper so that criticism and later revisions share one public record.
Loading comments…