2.1The timing of measures in 1918
The two studies of 1918 that the platform record and this article lean on share a design and differ in scope. Hatchett, Mecher and Lipsitch take seventeen cities in the autumn of 1918 and ask whether the moment at which a city introduced nonpharmaceutical interventions, measured against the course of its own epidemic, is associated with the height of its peak death rate. Their answer is that early action went with lower peaks. Philadelphia and St Louis serve as the paired illustration: in Philadelphia, first civilian cases were reported in mid-September and measures came only in early October; St Louis acted within days (Hatchett, Mecher and Lipsitch 2007). The study compares cities by timing and does not isolate the effect of any single measure.
Markel and colleagues widen the frame to forty-three cities and to the whole of 1918 and 1919. They find that the combination of school closure with bans on public gatherings, introduced early and kept longer, went with lower excess death rates, and that cities which lifted measures early often saw mortality rise again. From these historical, observational data they argue for interventions that are layered, early and sustained (Markel and others 2007).
Read side by side, the two studies make a timing thesis: when a city acted, measured from the first signs of its own outbreak, mattered. The thesis needs, for each city, two dated points, the start of the local epidemic and the coming of the measures, and a mortality series fixed by a common observation period and denominator. None of these is a fact that a city's officials, in the week they acted, could have held in the form the later study uses. The start of an epidemic is a date fixed afterwards, from a series not complete until the season ended.
Alfred Crosby's history of the 1918 influenza in the United States, first published in 1976, offers a different kind of account of the same city. It narrates the autumn wave in American cities, Philadelphia among them, where the epidemic overwhelmed hospitals, doctors, nurses and the means of burial; and it asks why so deadly an epidemic left so little mark on American memory (Crosby 1989). Crosby's question is about memory, not timing, and it marks what the comparative design leaves aside: a single city's experience of a season that could not be compared while it was happening.
2.2Counting while the outbreak runs
The second line of argument concerns the counts on which any comparison of timing depends. Lipsitch and colleagues set out three sources of bias in estimates, made during an outbreak, of the share of cases who die. Deaths lag behind cases, so early ratios of deaths to cases understate it. Mild cases are missed, so ratios overstate it. Case definitions, testing and reporting differ between places and over time, so comparisons across places and periods need these differences taken into account (Lipsitch and others 2015). The work is about the arithmetic of a modern outbreak, not about 1918. Its relevance here is that the weekly reports of 1918 and the report of 2020 both print counts made while an outbreak ran, and both print them with some mark of their incompleteness.
The timing thesis reads a pattern across cities from mortality figures fixed after the fact. The counting line warns that figures fixed during the fact differ by place and period in ways that are not random. A later study can correct for such differences if it has the means; a report of the hour cannot, and the question here is whether the reports of the hour say so.
2.3Precaution and its critics
Sunstein's Laws of Fear makes two claims that bear on the cases. The first is that a strong precautionary principle is incoherent, since precaution against one risk creates others and risks lie on all sides of a decision. The second is that the better course is to weigh costs and benefits with margins of safety; public fear he explains through availability and probability neglect (Sunstein 2005).
The Late Lessons editors work from the other end. Their cases, among them asbestos, benzene, PCBs and BSE, are hazards for which warnings were available long before action was taken. The lessons they draw include acknowledging ignorance as well as uncertainty, keeping up long-term monitoring of early warnings, and looking for blind spots in scientific knowledge (Harremoës and others 2001). Their cases are environmental and slow; the cases here are epidemic and fast. That difference is a fit to be watched, not ignored. A warning about asbestos could wait decades for action. A warning about influenza in September 1918 could wait, at most, days.
Side by side, the two works agree that decisions are taken under uncertainty and disagree about where the larger error lies: for Sunstein, in a principle that gives no guidance because action and inaction both carry harm; for the Late Lessons editors, in warnings left unheeded. Neither says what an epidemic report should contain. Read against the two reports, both suggest what to look for: whether a report states risks on more than one side, and whether it states what it does not know as well as what it is unsure of.
2.4Who may declare
The record of 11 March 2020 concerns a statement by the World Health Organization; the report of February cites a declaration by the same body on 30 January. Three works frame such statements. David Fidler reads the SARS outbreak of 2002 and 2003 as the first post-Westphalian public health crisis: the World Health Organization used information from non-state sources and issued travel advisories without the consent of states, and China's delay in reporting showed how far the older regime depended on states notifying; this shaped the revised International Health Regulations (Fidler 2004). Lawrence Gostin's treatise on global health law sets out the Regulations of 2005 as they stand: states must notify the Organization of events that may constitute a public health emergency of international concern; the Director-General, advised by an emergency committee, decides whether to declare one and issues temporary recommendations, which do not bind; and the Organization depends on states' reporting and has few powers to compel (Gostin 2014). Annelies Wilder-Smith and Sarah Osman review the declarations of such emergencies from H1N1 influenza in 2009 to COVID-19 in January 2020, the criteria and committees, and the debates over declaring too late or too early; they note that the Regulations provide for the emergency of international concern, not for a declaration of a pandemic (Wilder-Smith and Osman 2020).
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.
A declaration under the Regulations is a legal act with a defined consequence; a characterisation outside them is a statement. Anyone who would date the moment the world was warned has to say which is meant.
2.5What would change the decision
The last line of argument supplies the second half of this article's question. Richard Neustadt and Harvey Fineberg's study of the 1976 swine influenza programme traces overconfidence built on thin evidence, the momentum of a programme once announced, the merging of separate decisions, and the absence of planned points at which the decision would be looked at again in the light of new evidence; it recommends asking what would change the decision and building such points in (Neustadt and Fineberg 1978). Neustadt and Ernest May propose that decision makers separate what is known, what is unclear and what is presumed about a situation (Neustadt and May 1986). Charles Rosenberg's essay "What Is an Epidemic?" gives an epidemic a dramaturgic form, unfolding as in acts, the first of which is a slow, often reluctant recognition he calls progressive revelation, and the last a subsidence followed by retrospection (Rosenberg 1992). Barry Turner holds that disasters incubate: warning signs accumulate, unnoticed, misread or discounted, and after the event an inquiry reconstructs what should have been seen (Turner 1978). Sheila Jasanoff holds that the science used for regulation differs from research science, since it must give answers under time pressure and uncertainty (Jasanoff 1990).
These works do not form one side. Neustadt and his co-authors prescribe what decision makers should write down; Rosenberg and Turner describe how recognition happens and how hindsight rebuilds it; Jasanoff describes the kind of science a report written under pressure contains. Together they supply the reading tools used below.