Large Numbers, Life and Lives

We live in a world that is increasingly populated by rankings, graphs, maps and figures on an ever-broader array of topics, expanding their spheres of reference, especially with the advancement of digital technologies which facilitate the production and use of statistics. During the COVID-19 pandemic, this trend was intensified, bringing large numbers – on confirmed cases, deaths, vaccine efficacy and hospital occupancy – into the mainstream news, into the workings of government offices and even into everyday conversations. Their importance in shaping the pandemic experience meant numbers were also weaponised and politicised as actors with different ideologies challenged each other's positions.

During the most acute phase of the pandemic, graphs, numbers and percentages on infections and deaths were in constant circulation, serving as the basis for assessments of the situation and public policies, mainly in relation to ‘lockdowns’. In 2021, with the pandemic already one year old and vaccines being rolled out, numbers gradually began to occupy a different place, referring no longer to the present and the future (as had been the case of the predictive models that were all the rage in the early months), but to the past, serving to anchor narratives about what might have happened had certain measures (not) been taken (Camargo et al. 2021; Motta et al. 2022).

Numbers of confirmed cases and intensive care occupancy rates are metrics that make sense for the present and the future. However, the number of deaths from the disease is a cumulative figure – it only grows – and has come to be associated with collective suffering. The objective fact represented by the death of a person, its irreversibility and singularity, has made the figure that represents the sum total of those killed by COVID-19 a veritable legacy number.

The quantification of human life is and will always be a divisive and morally charged issue, since life is considered a higher value precisely because of its incommensurability and its understanding as a supreme, universal good which justifies the social contract. Yet during the pandemic, people's life and death – lives in the plural – were measured differently, albeit using metrics which have a long history.

As suggested by Georg Simmel (2004) in The Philosophy of Money, since its beginnings economic science has measured life in monetary terms. The calculation of compensation for the family of a murder victim, the value attributed to human labour and the calculation of risks of death are a few examples that cause little controversy. Similarly, the moral metric that attributes life expectancy unequally according to social class, skin colour, ethnicity or sexual orientation does not particularly offend our sensibilities (Neiburg 2020). Life insurance, which played a fundamental role in the development of the technical and conceptual instruments of modern statistics, operates precisely at the point of convergence between the probability of death and the monetary or monetisable losses that can be attributed to the life that was lost. Viviana Zelizer (1981), analysing historical changes in the cost of raising children in the United States, shows how investing in insurance or saving for education have been associated with different ideas about the sentimental and social value of children over time.

The pandemic expanded the scale and intensity of such quantifications, as well as the convention of equivalence between the value of a life and the price attributed to it, which was so widespread in social and economic practices. Calculations associating loss of life with economic loss were quickly consolidated after an initial phase marked by strong conflict between health and economics, as if the choice was dichotomous: either save lives by faithfully following the epidemiological model for ‘flattening the curve’ (Motta 2020) or give priority to the economy and avoid the catastrophic drop between 30 per cent and 50 per cent in the United States GDP forecast as a result of lockdown measures (Hansen 2020) – a prediction that was not confirmed. Gradually, graphs of confirmed cases and of economic output started to be presented together. Flattening the curve may have been the watchword of the first months of 2020, but other curves also had to be attended to, such as those indicating weakening economic activity, as argued by Pierre-Olivier Gourinchas (2020) and others.

Only later did this tension resolve itself in convergence, when econometric models indicated that in a situation of prolonged co-existence with the virus the cost of lives lost would be much greater than it would be had the most severe control measures been adopted along with the highly criticised practice of alternating periods of confinement and periods of easing measures (cf. The Guardian 2021). By explicitly calculating the cost of death, the pandemic forced us to confront extreme moral dilemmas on a daily basis while simultaneously fixing a reference for acting and making decisions about these questions: Who should be vaccinated first? What is the price in lives of having an economy functioning normally versus policies of social distancing and lockdown? With hospitals beyond capacity and shortages in healthcare supplies, who decides who dies so that others may live? Human lives lost something of their immaterial dimension while recognition of the cost of death grew, which revealed the sociological interactions at play in the attribution and equivalence of values related to distinct regimes of justification, to follow the thinking of Luc Boltanski and Laurent Thévenot (1991).

Denialism through Statistics: Public Numbers under the Bolsonaro Government

The rise of the new global far right has been accompanied by a rise in denialism, which also affects the ways statistics are produced and numbers are used to make arguments. A new phenomenon, its goal is to undermine public confidence in official statistics – something that was previously no more than a side effect of attempts by governments to window dress information on the economy and society and represent their own performance in a more favourable light. Statistics agencies, together with the press, science and academia, are coming under attack and having their ability to build consensuses around the truth undermined. The rhetoric of denialism relies heavily on continuously bombarding such institutions to make the fictions it wishes to portray seem more believable, fuelling conflicts and providing pretexts for exceptional measures (Camargo 2022; Kropf 2022).

As soon as he took office, President Jair Bolsonaro started taking actions to constrain the work of the country's research institutions and attempted to curtail the national census, potentially jeopardising the quality of the data and the Brazilian tradition of excellence in this field (Carazza 2021). On several occasions, the Bolsonaro government called into question the research methods used by the national statistics office, Instituto Brasileiro de Geografia e Estatística (IBGE; Brazilian Institute of Geography and Statistics), and the results, directly attacking its credibility. In 2019, when the national employment and unemployment figures compiled by the CAGED system showed that the formal employment rate had reached its lowest level in the time series, President Bolsonaro and the Ministry of Economy used information on the growth of the informal labour market, revealed in the national household sample survey, the Pesquisa Nacional por Amostra de Domicílios Contínua (PNAD; Continuous National Household Sample Survey), to argue for combining the two indicators, thereby overcoming the ‘backwardness’ of IBGE. In 2021, when the figures of the Cadastro Geral de Empregados e Desempregados (CAGED; General Register of Employed and Unemployed Persons) indicated an increase in formal job vacancies, it was the turn of PNAD data to be ignored and criticised for revealing a large increase in the number of people out of work (both seeking and not seeking employment) during the COVID-19 pandemic.

As far as the pandemic was concerned, the government denied the severity of the disease, denied the need for measures to restrict the circulation of people and denied the efficacy of the vaccines. The strategy also involved producing and publishing numbers about cases and deaths. In April 2020, the President's Secretariat for Social Communication presented a graphic, which they called ‘The Scoreboard of Life’, to show the number of Brazilians ‘saved’ and the number ‘in recovery’² but with no information on deaths. The executive secretary even saw fit to celebrate the fact that one million people had been cured, affirming that it made Brazil the ‘world record holder in the number of recoveries’.³ In June 2020, a number of reports came out in the national media about how they were being pressured by leaders from the Health Ministry to change the way they were presenting data on the pandemic (Marakawa 2020). The government wanted journalists to change the way they presented the number of deaths in a day, counting only deaths whose cause was confirmed on that day, which would exclude a large proportion of deaths prior to the date of publication, even if the cause was already confirmed (Megale 2020). Changing the criteria in this way would have brought down the death rate to less than one thousand a day, thereby meeting a direct demand of the president (Soares and Vargas 2020).

The federal government stopped publishing aggregate data, and access to the digital archives containing the tables was restricted for a few days. They stopped holding daily press briefings, which had been held regularly at 5:00 pm since early in the pandemic, to avoid the latest data from being publicised on the evening news. Faced with a potential crisis of confidence in the numbers produced by the federal government, the major media groups joined forces and began to gather data on the pandemic directly from state departments of health, creating a new federation-wide arrangement for access to information.⁴ Since then, federal bulletins have been systematically ignored by the press, which depend exclusively on the data compiled by the consortium of press and broadcasting companies. Paradoxically, the government itself fuelled mistrust in the reliability of the very statistics it produced, stimulating a controversy that only waned when the actors responsible for producing the numbers were removed from their positions and replaced.

At the same time, regular IBGE sample household surveys like PNAD, along with PNAD COVID-19, were not used either to estimate the number of people whose income was reduced significantly after the start of lockdown or to track their geographical distribution. Instead, an emergency benefit was designed and distributed without any prior recourse to evidence to correct for biases and distortions, resulting in billions being spent on people who were not in a financially vulnerable situation and did not meet the eligibility rules (Carazza 2021).

While statistics may have been disregarded when policies and funding decisions were made, numbers still retained their power as counterfactual evidence. The data on fires and deforestation produced by the Brazilian institute responsible for space research, the Instituto Nacional de Investigação Espacial do Brasil (INPE; National Institute for Space Research), were placed under suspicion several times when they revealed alarming increases in forest fires. The government repeatedly questioned INPE's competence, presenting fictitious numbers from unknown sources with the sole purpose of supporting the narrative that the Amazon was being well protected by the armed forces and agri-business. The idea was not to produce a demonstrable truth or an image of efficiency in public management. In short, ‘the successive attacks on statistics agencies and the use (and abuse) of fictitious numbers were designed to muddy the waters, making it harder to reconstruct the facts and reinstate objective procedures, thereby jeopardising the rational and methodical processes of proof and inference so key to the formation of a broad social consensus’ (Camargo 2022: 213).

The scenario presented above points to a general trend which is perhaps more visible in Brazil than it is elsewhere: the main conflicts during the pandemic were played out through statistics. That is why they are so important for observing and interpreting the construction of reality and the domestication of uncertainty. Along with more consolidated statistics, like COVID case numbers and deaths, hospital occupancy rates, unemployment rates, indicators of income and indicators demonstrating the unequal impact the pandemic had on vulnerable groups, three particular statistics emerged in the pandemic which largely framed how we lived through it. The first was the ‘moving average’, which the mainstream media created to calculate oscillations in case numbers, deaths and hospitalisations, at intervals of seven, fifteen and thirty days. The second was vaccination rates, which enabled the performances of different countries to be compared in the global bid to manage the pandemic. And the third, our focus here, consisted of estimates designed to quantify the political accountability of governments with a poor performance in containing transmission.

The Metrics of Avoidable Deaths

The Brazilian Ministry of Health has a definition of ‘avoidable death’, or, more precisely, ‘avoidable cause’, in its information system, DataSUS. It is one of many such categories that refer to the causes leading to each individual death that make up the immense statistical system that is DataSUS. It should be noted that this system is one branch of the Brazilian public health system, SUS (Sistema Único de Saúde; Unified Health System), which has a broad remit, ranging from direct care for patients to public health and prophylactic actions. According to the DataSUS guidelines, an ‘avoidable death’ is some ‘harm or situation [which is] preventable by the action of the health services, probably occurring when the health system is unable to meet the health needs and their determinants [and] are [sic] deficient in identifying and taking appropriate action’ (Ministério da Saúde n.d.).⁵ Therefore, even before the pandemic it was understood that some deaths may occur that could be avoided if the health service were to work properly.

The general concern about causes of death dates back to the seventeenth century (Drumond Junior 1998). But it was in 1976 that the concept of unnecessary deaths was systematised by David Rutstein and colleagues (1976) – the idea is that there are diseases and events that, if health services worked using all existing technical resources, would not exist or would not cause deaths. There is also a recognition that there are two types of causes of preventable deaths: those caused by improper prevention and those caused by improper treatment. The first type includes, for example, all diseases for which there are vaccines. And the second includes all those that are curable through therapies, medications and other forms of care.

This notion of the preventable death emerges and is consolidated as a way of measuring the efficiency and effectiveness of health services. It is, therefore, an index that one can use to assess the overall good or bad functioning of prevention and treatment policies. The category can be used, for example, to measure the impact of increased government investment in the area (Heijink et al. 2013). Brazilian literature on health mainly looks at preventable causes of deaths of babies, children and women in the postpartum period – known as sensitive indicators because their variation is evident and quite rapid – to look at changes in the quality of public services. Considering the origin and most common use by specialists, then, it is clear that the idea of responsibility is only coupled in very abstract terms and in relation to social values linked to the notion of what the public good is and not to particular individuals or incumbents of public functions.

The prominence of numbers in the management of the pandemic in Brazil increased the reach of the category ‘avoidable deaths’ beyond the confines of epidemiology. Through it, questions around the social underpinnings of health and the quality of public service delivery were asked. Through the deployment of this category – first by epidemiologists acting as scientific mediators, then by journalists and politicians, before and during the workings of the parliamentary enquiry committee – a number, that of deaths deemed as ‘excessive’, began to be linked to political decisions that were not, or should have not been, taken by state representatives.

As we shall see, the confluence of technical and political arguments transformed the category. It amplified its potential to reflect moral values and opened the possibility of attributing individual responsibility to political acts against the promotion of the public interest. As argued by Alain Desrosières (1993), the politics of large numbers shapes futures through the critique of reality. Here, the large numbers of politics – that is, the centrality that the category of avoidable deaths gained in public debate – wove together the moral and the legal and enabled a coordinated, collective response to the health emergency.

A report published in The Lancet on 20 February 2021 (Woolhandler et al. 2021) calculated the impact of the omissions of the former US president and his refusal to adopt harsher measures to prevent the spread of the virus. Considering the weighted average of COVID-19 deaths amongst G7 nations, the authors concluded that 40 per cent of US deaths could have been avoided had national plan been adopted, international cooperation not been eschewed, mass testing been conducted, schools been kept closed and large-scale events without mask wearing not been encouraged – that is, had the preventative measures and investments proposed, based on predictive models, been executed.

In the same period, The Lancet published a letter by the Brazilian researcher Pedro Hallal, in which he calculated the extent to which the Brazilian president, Jair Bolsonaro, was responsible for deaths during the pandemic. Pointing to factors similar to those raised in the US report, like the refusal to enforce non-pharmaceutical preventative measures, Hallal based his argument on the global average of deaths caused by the disease, attributing the high number of excess deaths – 156,000 at the time, or nearly 75 per cent of the total – to Bolsonaro's actions during the pandemic (2021: 374).

These and other subsequent initiatives did not use the same method or the same reference value. Indeed, we could ask: What should be taken as the basis for a normal curve: the average death rate among G7 nations, the regional average, the global average or some other figure? How can we measure intangible aspects, like the impact on a population of the words and actions of its president, who disregards public health protocols? Despite the risk of imprecision, these early studies largely failed to address this issue, given their completely justifiable strategy of mobilising international opinion and shining a light on acts that constituted a crime against humanity. What we have here is a pandemic-inspired mode of data activism: activism that harnesses statistics to criticise and confront institutionalised powers and thus promote greater social equity (Bruno et al. 2014: 5–30).

A crisis is characterised as a temporal construct that defines a certain period as abnormal (Koselleck and Richter 2006), which implies an attribution of normality to the periods immediately preceding and succeeding it. The calculation of deaths based on the size of each country's relative population in relation to the total world population – the mortality rate (per 100,000 inhabitants) – implies that even during a crisis parameters of normality are constructed around the notion of proportionality. The production of normality within abnormality was one of the fundamental operations for the subsequent construction of responsibility. It provided a framework of objectivity to measure the difference between the expected result in managing the pandemic and the actual result, which could be attributed to inaction and negligence. As Ian Hacking (1990) argues, the idea of normality is one of the most powerful notions of modernity, as it proposes the fusion of the ideas of recurrence and desirability.

One of the numbers that circulated most in the debate on the pandemic, and which helped justify the congressional enquiry, was the comparison of the relative mortality in each country. This was the number most used to assess the performance of the Brazilian government, especially when case numbers and deaths were on the rise. Denialists sustaining the government's defence tried to oppose this number with a counter-argument based on the number of people in Brazil who had been cured of the disease.⁶ Throughout 2021, this statistical counter-discourse also included the absolute number of people vaccinated, which was held to indicate that the country was on the world's vaccination leaderboard. In doing so, they blatantly disregarded the standard procedure of making comparisons by relative population size, not to mention the percentages of people who had had a second or third dose of the vaccine, which at the time was very low in Brazil.⁷

Between January and April 2021, estimates of responsibility for COVID deaths were largely restricted to the scientific milieu. The creation of the congressional enquiry, with its capacity to foster public debate, quickly changed this situation. With the media coverage it gained, other researchers set about estimating how many deaths could have been avoided had the federal government taken up offers of vaccines more quickly, and their findings started to appear in the press. One of the estimates that gained a lot of traction was that of the epidemiologist Pedro Hallal cited above. He claimed that 180,000 Brazilian deaths – three out of every four at the time – could be associated with the government's mismanagement of the pandemic. It is the same proportion that he had indicated in his January article in The Lancet, but the reference value was different. While in the first article he had used the average of global COVID deaths, in the more recent one he considered the number of vaccines that could have been purchased and administered had there been the political will to do so.

This study was followed by others which also used the number of vaccines available for purchase in addition to the vaccination rate, which had become enshrined as the new global benchmark for government accountability. One of these studies was by the sociologist Celso Rocha de Barros (2021) who in various opinion articles laid 100,000 deaths at Bolsonaro's door. This inspired one senator from the governing coalition to call for Barros to be investigated – which shows just how politicised numbers became in the pandemic, particularly numbers referring to ‘avoidable deaths’. As for the press, most outlets reported large round figures, such as 100,000 deaths. This was no coincidence, since such large numbers had the power to galvanise collective mourning and therefore public opinion (Camargo et al. 2021; Motta et al. 2022).

The testimony given in the enquiry by the current and past Ministers of Health and dozens of other people had the objective of collecting evidence about the policies implemented between March 2020 and October 2021, and thus to ascertain whether there had been any misconduct or potentially criminal conduct. The idea was to draw objective links between the actions and decisions taken (or not taken) by government leaders and the effects of the pandemic on the population. Taking the form of dates and quantities, statistical series, and sequences of events over time, numbers gained fundamental importance, embodying an objectivity evoked to accommodate moral and political positions. The two dimensions of law and statistics were once again combined to demonstrate links between legal duties and facts, between expected results and observed results.

Amongst the many issues debated in the enquiry over its months of work, two related numbers stood out, epitomising the government's inaction and its possible consequences: vaccines and COVID deaths. The number of vaccines offered in 2020 for advance purchase by Pfizer and Brazil's Instituto Butantan and subsequently rejected⁸ and the number of lives lost to the disease are the two numbers around which the government's responsibility was debated in the sessions of the congressional enquiry. These two numbers were also adopted as the core elements in the arguments used to attribute criminal responsibility in the enquiry committee's final report. In one section, the report expressly associates the crime of causing unacceptable prohibited risk (in Portuguese, geração de risco proibido) with the new category of ‘avoidable deaths’.

Following the rationale of the Brazilian researchers mentioned above, all of whom testified at the enquiry's hearings, the statistical model used in the report generated the finding that, given the country's vaccination capacity and the coverage of its public health system, had the vaccination campaign been intense and begun when Pfizer and Butantan had first offered the vaccine, 127,000 deaths could have been avoided. The novelty in relation to the estimates based on the global average of deaths, which was also cited, is that the enquiry's report refers to studies that draw an equivalence between lives and their economic value, like the study entitled ‘Modelling the Impact of Delaying Vaccination against SARS-CoV-2 Assuming Unlimited Vaccine Supplies’ (Amaku et al. 2021). The same model was applied to Brazil in an article written by researchers from the University of São Paulo, Fundação Getulio Vargas, Butantan and the London School of Economics and was also used to sustain the report's findings (Pereira et al. 2021). By applying the Value of a Statistical Life (VSL) to the optimal number of vaccinations that could have been delivered, a loss of R$418 billion was estimated – a value that exceeds many times over the cost of purchasing the vaccines at the initial offering price, at around ten dollars per dose, plus potential claims for indemnification and hospital costs due to side effects (Senado Federal 2001: 1015–1017).

The report mentions another group of studies such as that of Fábio Santos and colleagues (2021) which use a similar methodology to estimate the number of people over eighty, seventy and sixty years of age who would still be alive had they received the second dose of the Pfizer vaccine in March, April and May 2021, respectively. The calculation considers the capacity of Brazil's vaccination system, the efficacy of the vaccine in preventing death, the delay in the delivery of the vaccine and potential wastage, taking as a basis for the calculation the seventy million doses that Pfizer offered Brazil in August 2020 and which the federal government rejected. The enquiry's report replicates the conclusion of the articles, namely, that 12,663 people over sixty years of age would not have died in the three months in question, attributing these deaths to the federal government's decision (Senado Federal 2021: 1019–1020).

Along with a mathematical model that estimates the impact of the delay in vaccination on the number of cases and deaths, the report cites calculations from other studies, such as one by Jurema Werneck (2021), which aimed to measure the impact of government decisions on deaths amongst the population. Using a methodology similar to the initial argument used by Hallal, these studies calculate deaths that could have been avoided by actions such as mass testing, contact tracing, mask wearing and restrictions on movements, as well as healthcare services, including human resources, hospital beds, medications, oxygen and ventilators. The results indicate that 120,000 lives could have been saved by the end of March 2021, or 40 per cent of the COVID deaths until then.⁹

Another study, published in The Lancet (Xavier et al. 2022) by researchers from Fiocruz, the University of Brasília and the Federal University of Rio de Janeiro, analysed differences between the determinants of the waves of COVID-19 infection in Brazil. They found that while inequality of income and healthcare infrastructure were the main determinants during the first wave, in the second, the administrative districts where Bolsonaro won a majority of votes in the 2018 presidential race had higher mortality rates, indicating that this wave ‘was explicitly shaped by the partisan choice of municipalities’. This association held true even after taking into consideration structural inequalities between these districts in terms of income, human development and health services. This was the first study to present a ‘political’ map of the pandemic by considering the geographic distribution of avoidable deaths based on votes for Bolsonaro, the Human Development Index (HDI), the Gini Index and mortality per 100,000 residents.

The delays in the purchase, distribution and administration of the vaccines, as well as the failure to adopt non-pharmaceutical measures to prevent the spread of the virus, served as the basis for a lawsuit to hold the federal government accountable for the comparatively high mortality from COVID-19 in Brazil, calculated as being 89 per cent above the global average. The report identified ‘reckless management by the government in the fight against the pandemic’, involving a ‘strategy of propagation of the virus conducted systematically by the federal government, followed by attempts at resistance by the other federal powers and entities’ (Senado 2021: 1021).

Amongst the crimes attributed to the president, causing unacceptable (‘prohibited’) risk is based entirely on the emerging category of ‘avoidable deaths’. Scientific and journalistic research into the impact of the president's actions and inactions, such as variations in hospital bed occupancy rates before and after large meetings and gatherings in different cities attended by the president and without the use of face masks,¹⁰ was also used as supporting evidence to establish culpability and charge the head of state with crimes related to the failure to observe preventative sanitary measures (Senado 2021: 1049).

The conclusion of the work of the congressional enquiry committee, as set forth in its final report, demonstrates the effect of using vaccination rates as a benchmark in framing the public debate. The comparison of the performances of national governments following the development and acquisition of COVID-19 vaccines served as an instrument of proof that preceded any debate, in the meaning formulated by Desrosières (1993), providing a stable parameter for evaluation and judgement – a parameter that did not exist in the first year of the pandemic.

Conclusion

While the politics of large numbers shapes our possible futures and the way criticisms can be formulated, the large numbers of politics during the pandemic brought to the fore the contingencies around the attribution of moral values based on quantitative information. The category of ‘avoidable deaths’ was evoked and transformed at the interface between the politics of large numbers and the large numbers of politics with reciprocal influence between politicians, political actions, experts and technical arguments.

We show in this article that what was at stake in the transformation of this category expresses a particular political and institutional dynamic in which the occupants of state power acted to weaken common spaces and language while their opponents made technical and logical arguments. What we see from this is that the actions of authoritarian groups in the state, therefore, must also be seen to shift the analytical agenda of social studies on quantification.

While battles over numbers and their political use have been a constant in the history of statistics, the combination of the COVID-19 pandemic and the rise of far-right governments has made statistics-based denialism – along with denialism in other spheres, especially science, climate and history – a threat to well-established social consensuses, such as the reliability of population censuses, indicators of forest fires, and vaccine efficacy. In particular, this attempt to displace and dismantle public arenas has focussed its efforts, both in party politics and everyday politics, on radicalising the debate on major national issues, fuelling heightened animosity and targeting imaginary enemies. In this scenario, the terms expressed through the various metrics about death do not address so much the efficiency of public administration as life itself as an all-encompassing, supreme value, as well as the role of the state and the roles of various democratic institutions in its protection.

The displacement and transformation of statistical categories, avoidable deaths in our particular case, provide us with two main analytical insights. First, their investigation makes it possible to understand the different but interconnected articulations of law, morality, institutional politics and science at a particular historical moment. And second, in showing historical inflections in these various articulations, they also challenge current theoretical framings.

The black box that usually hides the technical procedures behind the production of numbers has been opened to the eyes of the public, who are more aware than ever that consensus is a fragile political construct. The mainstreaming of technical and political processes of quantification, with politicians, journalists and experts from various fields publicly debating which numbers matter most, where they come from and how big they are, has two possible results. One is an awareness about the centrality of statistics in people's daily lives – quite literally a matter of life and death in times of crisis. Another, more dangerous result is a decline in trust in numbers, to use Theodore Porter's (1995) term. If large numbers are that political, how can we believe in the legitimacy of any expert argument?

Where the enquiry committee's report will lead is a moot point. We do not know whether or how figures on avoidable deaths and inferences made from such figures will be included in any formal charges against the former president by the Attorney General's Office or the Federal Supreme Court. In June 2022, the Attorney General of the Republic, a close ally of Bolsonaro, asked the Supreme Court to reject the committee's final report and its findings.¹¹ His argument against accepting the charge of the crime of epidemic resulting in death, for example, hinged on a lack of causal links between the then president's acts and the contracting of the disease by specific people. It is an objective, individualising causal link that ignores the broader notion of responsibility, based on the relationship between the resources available to a public official and his or her legal obligations. The court has yet to reach a decision.

Desrosières (1993) called ‘large numbers’ the public statistics produced by the state. But during the pandemic, there was an inflection in this sense, since the moral grandeur expressed by the large numbers framed the public debate. The moral force of large numbers and large quantities has spawned categories, such as ‘genocide’, which have been widely used as a numerical, moral, political and legal nexus. Although the formal application of this term has been widely debated, it is understood to encapsulate technical public indignation by referring to other historical events and thus proposing a place of memory for the Brazilian pandemic experience. From our analytical perspective, the issue is how the sheer magnitude of the number of apparently avoidable deaths triggers the moral sensibilities associated with it.

The controversies over pandemic statistics in general, and over the concept of avoidable deaths in particular, show a change in the public life of numbers. On the one hand, numbers come out of the pandemic with their importance in collective life bolstered, their role in solving problems involving moral dilemmas confirmed. On the other hand, the profusion of estimates and the discrepancy between their reference values, while perhaps not jeopardising numbers as a common language, exacerbate their political use, bringing to the fore the conventional nature of quantification, which is normally encapsulated in state and social routines. This paradoxical situation stems from the struggles for democracy being played out in Brazil, leading to a hyper-politicisation of numbers, central both in establishing criminal responsibility and in the interweaving of legal and moral categories to construct a past for the pandemic.

Our analysis foregrounded the category of avoidable deaths as a lens through which to reconstruct the public life of numbers during the pandemic and to observe the various translations operated in the processes of objectification and decision-making during the health crisis. At the same time, we approached avoidable deaths as an ethnographic category by debating the statistical value assigned to the lives lost, in the plural – but also in the singular – by offering a collective response to trauma in defence of society. More than discussing what the numbers of avoidable deaths are, we tried to show what they do through an empirical observation of the genesis of a quantification infrastructure, which continues to be under construction at the time of writing.