In the last days of September 1918 a reader of the federal Public Health Service's weekly journal could learn that influenza had reached Massachusetts in August, that hospitals in North Carolina were crowded, and that Pennsylvania had so far reported relatively little of it among civilians. A week later the same reader would find Philadelphia named for the first time, in a sentence of a few words. In the issue of 18 October the journal gave statewide totals of registered deaths for Pennsylvania and said that the larger share had fallen in Philadelphia. Between those issues, by the account of a modern platform record, the city held a large parade on 28 September and introduced school closures and limits on assembly on 3 October. The weekly reports mention neither.
In early February 2020 the Centers for Disease Control and Prevention published, first online and then in its weekly report, an account of what it and other agencies had done since Chinese health officials reported a cluster of respiratory illness on 31 December 2019. It listed eleven confirmed cases in the United States and described screening at airports, a new laboratory test, guidance for clinicians and a presidential proclamation on entry. It said what was known, what was unclear, and on what assumption the measures rested. On 11 March 2020, after its publication, the World Health Organization's Director-General said that the spread of COVID-19 could be characterised as a pandemic. The report says nothing of that, since it was written before.
Skeptron's records hold one case on COVID-19, the World Health Organization's statement of 11 March 2020; lockdowns, vaccines and national responses are not in them, and this study does not use them as evidence.
This article asks a plain question of both texts. When authorities acted, what did each report say it knew, and what, in its own words, would make the action change? The question is narrower than whether the actions were right, and it can be answered from the page. A report written during an epidemic carries the knowledge of its hour, and carries or omits some statement of what would count as reason to act otherwise.
The question matters for two debates. The first concerns the influenza of 1918 and the timing of public measures. Richard Hatchett, Carter Mecher and Marc Lipsitch compared seventeen American cities and found that those which acted early in their local epidemic had lower peak death rates; they set Philadelphia, where measures came only in early October, beside St Louis, which acted within days (Hatchett, Mecher and Lipsitch 2007). Howard Markel and colleagues, across forty-three cities, found that school closure and bans on gatherings put in place together, early and for longer, went with lower excess death rates, and that cities which lifted measures early often saw mortality rise again (Markel and others 2007). Both studies are observational. Against the confidence that timing can be read from such comparisons stands work on how counts made during an outbreak mislead, by place and over time (Lipsitch and others 2015). The two lines share an author and do not contradict each other, but they pull in different directions on the same evidence: one toward the conclusion that acting early mattered, the other toward caution about what the numbers of the time could show.
The second debate concerns acting before the evidence is complete. Cass Sunstein argues that a strong precautionary principle is incoherent, since precaution against one risk creates others, and recommends weighing costs and benefits with margins of safety (Sunstein 2005). The European Environment Agency report edited by Poul Harremoës and colleagues collects hazards on which early warnings came long before action, and draws lessons that include acknowledging ignorance as well as uncertainty (Harremoës and others 2001). Sunstein asks what a precaution costs; the Late Lessons editors ask what a delay cost. Both bear on what to do with signs that are not yet proof.
The contribution of this article is to give both debates a reading of the documents that recorded the signs at the time. These reports are not the evidence on which the later comparisons of cities, or the arguments about precaution, chiefly rest. They are the federal government's own weekly account of what it had heard, written before anyone knew how the season would end. Beside each stands the modern summary Skeptron wrote for its platform. The comparison shows what each layer keeps. The weekly reports of 1918 keep prevalence and counts, with their gaps marked, and no orders at all for Philadelphia. The report of 2020 keeps a sorted account of knowledge and an explicit assumption, and names revision only as a general promise. The modern summaries keep the dates the later argument needs, and take them from the later argument.
The placement of the secondary works in these debates, and their application to the two cases, are this article's own. The works are cited only for what they argue.
Section 2 sets out the debates; Section 3 the texts, records and ledger; Sections 4 and 5 read the two cases; Section 6 brings them together; Section 7 concludes.