The State’s Tobacco Business

The Examination’s estimate attributes at least 57 million deaths to China Tobacco across 1990–2023. Its scale demands attention. Its meaning demands equal care: this is a modeled cumulative attribution, not an annual toll or a company-specific register of individually proven deaths.

The method combines population mortality estimates with a company allocation based on market share. That can make institutional scale visible where no single administrative record exists. It also introduces an assumption: a share of the market is used to allocate a share of harm. The calculation is not, by itself, a complete account of what would have happened if the firm had not existed.

The cross-industry comparison adds another distinction. Tobacco’s market share is measured through product volume, while the method uses revenue for ultraprocessed food companies. Different proxies may be necessary, but they make the ranking conditional on methodological choices. The interactive explains those choices without pretending to reproduce the calculation.

The company-specific death total is a modeled attribution, not a register of individually adjudicated deaths.

Inspect each stage: population burden, company allocation, historical accumulation and the resulting estimate. The stages are an explanation of the published method. They are not independent measurements that can be adjusted freely. A slider with invented parameter ranges would create the appearance of sensitivity analysis without the underlying data needed to justify it.

The institutional question is how commercial interests meet public-health responsibilities. Ownership is part of the setting, but it does not establish the motive for a specific decision. A convincing explanation would trace authority, incentives and the ability to challenge a decision. The estimate supplies a reason to scrutinize those arrangements; it cannot supply the decision record itself.

There is a strong objection to methodological caution: uncertainty can become a way to postpone responsibility indefinitely. But precision need not have that effect. A bounded estimate can still describe a grave burden. Separating uncertainty about the exact allocation from the governance question makes scrutiny more specific, rather than making the harm disappear.

The reverse argument also matters. Public ownership could, in principle, create power to change commercial behavior as well as incentives to preserve it. Whether that capacity is used is an empirical question. The Notebook should examine decisions rather than assume that the ownership label alone answers how responsibility is exercised.

A replication could test whether the published arithmetic is reproducible; alternative inputs and assumptions could test how robust the result is. Those are different exercises, and neither has been performed here. The figure therefore keeps the published estimate, its period and its assumptions attached. Readers can examine the basis of the claim without being asked to mistake numerical precision for certainty about every dimension of responsibility.