COVID-19 in prisons: a study of habeas corpus decisions by the São Paulo Court of Justice

Abstract Brazil has become the epicenter of the COVID-19 pandemic in the Global South-a pandemic that disproportionately affects vulnerable populations, especially those detained and imprisoned. Legal institutions are struggling to respond. In this paper, we focus on the National Council of Justice’s Recommendation 62, issued March 17, 2020, which recommends that judges take several measures to reduce the risk of COVID-19 infection in prisons. We test this recommendation’s impact by looking at habeas corpus decisions in the São Paulo Court of Justice. The exploratory findings presented here indicate that Recommendation 62 has little impact on habeas decisions. In general, citing the recommendation does not lead the Court to grant early release or house arrest to those detained, and most habeas actions are decided against petitioners. This is true even when petitioners claim to be part of a risk group, or their alleged offense did not involve violence or serious threat-factors that should favor habeas relief under Recommendation 62.


INTRODUCTION
Brazil has become the epicenter of the COVID-19 pandemic in the Global South-and indeed one of the epicenters worldwide. As of May 31, the pandemic had killed over 28,000 people, with a toll of nearly 1,000 deaths every day (Ministério da Saúde, 2020). The pandemic disproportionately affects vulnerable populations, especially those detained and imprisoned (Rodrigues, 2020;Incarceration and Law, 2020).
Over the last quarter century, Brazil's prison and jail population grew from 73,000 people in 1995 to over 748,000 in 2019 (more than one-third of whom are under pre-trial detention). The country has the third largest prison population in the world. There is also severe overcrowding in the prison system. On aggregate, the occupancy in Brazil's prisons and jails is over twice its designed capacity (Conselho Nacional do Ministério Público [CNMP], 2019; Departamento Penitenciário Nacional [DEPEN], 2019;Walmsley, 2018).
The state of São Paulo, the country's richest state, also holds its largest prison population. According to the latest data from June 2017, the state detained 233,089 people in jails and prisons, a total that is three times higher than the state with the second largest prison population (Minas Gerais, with 74,712 individuals). Overcrowding is also a problem in São Paulo, where 176 prison and jails have a total capacity for only 147,942 prisoners, resulting in a 150% occupancy level (DEPEN, 2019).
Overcrowding causes or aggravates several health risks in prison: unsanitary and unhygienic living conditions, insufficient provision of essential items (including food), spread of infectious diseases (tuberculosis and HIV/Aids are more prevalent inside than outside prison), and violence caused by inmates and correctional officers. The healthcare conditions in São Paulo's detention facilities are also far from ideal. In 2018, only 37.7% of prisons and jails had a doctor present daily and 20% of them provided no healthcare service.
These poor health and healthcare conditions clash with the rights and guarantees in Brazil's constitution and correctional law. The constitution protects human dignity and forbids torture or inhumane treatment. It also establishes a constitutional right to health that guarantees access to comprehensive healthcare on an equal basis to all. Moreover, the country's correctional law determines that the government must provide prisoners with healthcare and allow them to be treated outside of prison if necessary.
The dreadful conditions in Brazilian prisons have been challenged in thousands of lawsuits seeking early release of prisoners or structural reforms in prisons. In 2015, the Federal Supreme Court (STF) declared in a preliminary decision the "unconstitutional state of affairs" in the Brazilian prison system, concluding that the "incarceration culture" and public authorities' systematic failings and inertia systematically violate prisoners' rights. Then, in 2017, the STF ruled that prisoners have the right to financial compensation for damages resulting from incarceration under inadequate conditions. And in 2018, it mandated the de-incarceration of mothers under pre-trial detention. Meanwhile, the Nacional Council of Justice (CNJ)-the body responsible for the strategic planning and administration of the judiciary but with no power to review judicial decisions-has also been promoting policies to reduce incarceration through research projects and recommendations to judges.
Despite these measures, Brazil's prison population keeps rising and prison conditions have barely improved. This is particularly concerning in light of the COVID-19 pandemic when the combination of overcrowded prisons, poor healthcare services, and a highly infectious disease can lead to tragic consequences. Recent reports show that the virus's lethality within prisons is five times higher than in the general population (Pauluze, 2020). As of May 29, prisons in the state of São Paulo acknowledged 100 suspected COVID-19 cases, 76 confirmed cases, and 12 inmate deaths (DEPEN, 2020). The actual infection and mortality rates are surely higher because only 4,873 people-about 2% of the prison population-have been tested (Brazilroad, 2020). Moreover, recent reports show that not only prisoners are in danger, with 147 prison officers and staff reporting the infection and 12 dying (Infovirus, 2020).
Several institutions have responded to this health crisis. In this paper, we consider National Council of Justice Recommendation 62, issued March 17, 2020, which recommends that judges take several measures to reduce the risk of COVID-19 infection in prison facilities. We focus on those measures aimed at reducing prison overcrowding: the reduction of pre-trial detentions and the early release and house arrest of convicted prisoners, especially those in risk groups or vulnerable detainees whose alleged offenses do not involve violence or serious threat (Box 1).
These are fundamental measures to mitigate COVID-19 infections because prison overcrowding presents an unsurmountable obstacle to enabling social distancing and improving healthcare and sanitary conditions in prisons. Yet it has been unclear whether lower courts would follow Recommendation 62, particularly because the CNJ's guidance has been controversial. Then-Minister of Justice Sergio Moro and Supreme Court Justice Luiz Fux have made public declarations cautioning against mass de-incarceration due to the pandemic (Fux, 2020;Ministério da Justiça, 2020). And while a March 18 STF decision advised judges to consider Recommendation 62, the Court did not order them to follow it, as two dissenting opinions proposed. The actual impact of Recommendation 62, then, has been unknown. This paper offers preliminary evidence of its effects by analyzing habeas decisions by the São Paulo Court of Justice-a particularly significant court because it reviews habeas petitions in the state that has the largest prison population, reports the highest number of COVID-19 infections and deaths, and was the country's first epicenter of the disease.

BOX 1 NATIONAL COUNCIL OF JUSTICE RECOMMENDATION 62/2020
Article 4 -To recommend investigating judges to consider the following measures in order to reduce the epidemiological risks and in light of the virus spread at each location: I -To reconsider pre-trial detention, in accordance with Article 316 of the Criminal Procedure Code, with preference to: a) pregnant women, lactating women, mothers or people caring for under-12 child or disabled person, elderly people, native Brazilians, people with disability, people in risk group*; b) People detained in institutions running at above full capacity, without dedicated health care professionals, under an interdiction order, with cautionary measures determined by an international jurisdictional body, or with facilities conducive to the new coronavirus spread; c) Preventive detentions that surpassed the 90-day period or that are related to crimes that involve violence or the threat of violence against a person […] Article 5 -To recommend correctional judges to consider the following measures in order to reduce the epidemiological risks and in light of the virus spread at each location: I -Early release from closed and semi-closed prison following the Supreme Court binding precedent 56, particularly in relation to: a) pregnant women, lactating women, mothers or people caring for under-12 child or disabled person, elderly people, native Brazilians, people with disability, people in risk group*; b) People detained in institutions running at above full capacity, without dedicated health care professionals, under an interdiction order, with cautionary measures determined by an international jurisdictional body, or with facilities conducive to the new coronavirus spread; […] III -House arrest for people serving sentence in open and semi-closed prison, under the conditions to be defined by the correctional judge. IV -House arrest for people with confirmed or suspected Covid-19 diagnosis according to medical report, if adequate space for isolation is not available at the facility […] Note: According to Recommendation 62, the COVID-19 "risk group" includes elderly people; pregnant people; people with chronic diseases; people who are immunocompromised; and people with respiratory diseases and other preexisting comorbidities that may cause a deterioration in their health condition when infected, with special attention to diabetes, tuberculosis, renal diseases, HIV, and co-infections. Source: Recomendação Nº 62, de 17 março de 2020 (CNJ, 2020).

METHOD AND GENERAL RESULTS
We collected every one of the 6,771 habeas decisions made by the São Paulo Court of Justice, extracted from the Court's official gazettes ("diário oficial") where its decisions are published on a daily basis. We read the 185 official gazettes published between December 1, 2019 1 and May 4, 2020, searching for habeas decisions that mention "covid, " "pandemia, " or "coronav. " We analyzed (a) all of the collected decisions through "regex" or the mapping of regular expressions using the software Python (version 3.7) 2 ; (b) a statistically significant random sample of 371 decisions 3 ; and (c) a purposeful sample of decisions regarding class action habeas petitions (what we call "collective" habeas petitions).

Universe analysis
The earliest decision meeting our search criteria is from March 18. The number grew in the following weeks (Graph 1), 4 following a similar trend found in other Brazilian courts (Balthazar & Mariane, 2020 The peek of habeas decisions between March 28 and April 4 th suggests that many habeas petitions were filed as soon as Recommendation 62 was issued and decided by the São Paulo Court of Justice in a surge over the following weeks. Most habeas decisions in our universe involves individual petitions. Through a regular expressions' analysis, we found only 25 collective habeas decisions. The overwhelming number of decisions (88%) went against petitioners (Graph 2).

Random sample analysis
We examined the random sample of 371 decisions by mapping key variables, such as whether (1) the petitioner was represented by a public defender or private lawyer; (2) the habeas petition was individual or collective; (3) there were motions for preliminary injunctions or relief; (4) the petition was successful and led to release; and (5) if the judge explicitly considered the petitioners' circumstances or alleged offense, as recommended in Recommendation 62. In almost 30% of the sample the petitioners were represented by the state public defender's office (Defensoria Pública do Estado de São Paulo). It is not possible to attest that the remaining 70% of cases comprise only paid legal counseling since the public defender's office can and often does delegate legal representation to pro bono lawyers. Within the sample, we found only 2 of the 25 collective habeas, all of which will be analyzed in the next section.
As shown above, the Court denied habeas relief in nearly 90% of the decisions in our sample (see Graph 3). In 67% of the decisions, the Court explicitly mentioned Recommendation 62. There appears to be no association between mentioning Recommendation 62 and either granting or rejecting the habeas petition. There is information regarding the offense the petitioners were convicted or accused of in 248 decisions, 5 although the decisions rarely contain details about the facts and circumstances of each offense, charge, or conviction. The details matter because for most cases we are unable to differentiate acts that fall under the same offense but that are substantially different in many ways. For instance, robbery can be committed by a person who waits at a bakery door while his unarmed partner orders the cashier to give them money (Case 3) and one who hits the victim several times before taking their property (Case 51).
Recommendation 62 calls on judges to reconsider pre-trial detentions lasting more than 90 days for those charged with crimes that do not involve violence or serious threat against a person (VST) (article 4, I, c) 6 . Graph 4 separates the crimes identified in the 248 cases between those involving VST or not. Source: Elaborated by authors. 5 The other 123 refer to 121 decisions in which the offense was reported as "not informed" and 2 collective habeas petitions. 6 The distribution of habeas petitions involving VST and non-VST crimes does not necessarily follow the distribution of alleged crimes at pre-trial detention hearings. However, figures from before the pandemic suggest that the same three prevalent crimes found in our sample -robbery, drug trafficking, and theft also formed the majority of crimes brought before judges for pre-trial detention hearings. For more information, see Fórum Brasileiro de Segurança Pública (2017).
We also reviewed every decision for information about whether the petitioner's alleged health condition would put them in a risk group, finding that less than a third of the decisions within the sample mentioned the risk group at all. Graph 5, Graph 6, and Table 1  Total 137 * A prisoner may mention more than one cause of vulnerability. ** Recommendation 62 does not define "elderly, " but Brazilian law defines elderly as someone who is 60 or older. Source: Elaborated by the authors.
Recommendation 62 does not offer an exhaustive list of those in a risk group. It is, therefore, difficult to assess with accuracy if the alleged condition of each petitioner falls under Recommendation 62's scope. Most of the conditions mentioned, however, involve claims of respiratory diseases or age, conditions that are explicitly described by Recommendation 62 as important risk determinants that justify early release from detention during the COVID-19 pandemic. A judge's choice to mention the risk group condition, however, does not seem to affect their ultimate decision-in only five of those decisions did the Court rule in the petitioner's favor.

Successful habeas petitions
We singled out the decisions in which the Court granted the habeas petition or ordered further analysis or action. Petitioners may raise more than one reason for the illegality of their detention, and hence it is important to understand how often reasons related to COVID-19 affected the decision. In our sample, we found 21 successful habeas petitions and 6 partially successful petitions 7 .
Among these 27 decisions, in 9 we found no information regarding the alleged offense that led to the prisoner's conviction or detention; in 5 the offense involved violence or serious threat (robbery (n=2), domestic violence (n=2), homicide (n=1)); and in 13 the offense did not involve violence or serious threat (drug trafficking (n=10), theft (n=2), drinking and driving (n=1)).
The 21 successful petitions can be divided in four groups. First, 5 decisions in which the main ground for the decision was that the petitioners were in a COVID-19 risk group. All five petitioners were serving their sentences following a criminal conviction. Four of them were transferred to house arrest and one was released on parole. Second, 9 decisions questioning the legality and necessity of the pre-trial detention. Neither Recommendation 62 nor COVID-19 seemed determinant to the outcome but might have affected how the judge used their discretionary power. Third, 3 decisions in which the illegality of the pre-trial detention was so evident that COVID-19 was irrelevant to the decision. Fourth, 3 decisions in which the petitioners were convicted and serving time in a closed facility despite of their right to move to a less strict regime according to the law. In sum, in only 14 cases (3% of the sample) was there evidence that COVID-19 and Recommendation 62 played a role in the Court's decision to grant habeas relief.

Collective habeas petitions
Apart from the random sample of decisions, we also drew a purposeful sample containing every collective habeas decision. Collective habeas petitions are filed on behalf of a group rather than specific individuals. This writ is a relatively recent innovation in Brazilian law and was first used in 2018 to benefit all provisionally detained women who are pregnant or caring for a small child. Collective habeas petitions can, of course, have a potentially larger impact because they can lead to the release of large numbers of people and to structural changes.
Using regular expressions, we found 25 decisions involving collective habeas petitions (see note 1). Three of them were excluded because they were filed on behalf of a group of identified prisoners and thus do not fit the concept of collective habeas petitions used in this paper. Three other decisions were excluded because the Court made no decision on the habeas petition.
Among the 19 remaining collective habeas petitions, 14 sought early release or home prison for risk group prisoners in specific facilities (7 were on behalf of elderly prisoners, 5 on behalf of prisoners with health conditions, and 2 did not specify the risk group). Two collective habeas petitions were filed to benefit all prisoners in a specific facility (i.e., it was not limited to a risk group), while 3 habeas were filed for all prisoners in risk groups (without specifying a particular facility). None of the collective habeas petitions were granted. The most common justification was that habeas relief cannot be granted to an undetermined group because the individual circumstances of each individual need to be assessed (n = 16). All but three decisions mention Recommendation 62.

DISCUSSION
Our exploratory findings indicate that Recommendation 62 has had little impact on habeas decisions. In general, most habeas petitions are denied, and citing Recommendation 62 does not lead judges to grant early release or house arrest to those detained. This is true even in decisions when the judge mentions a risk group or when the petitioner's alleged offense does not involve violence or serious threat (such as drug trafficking-the offense that appears in most habeas decisions).
Our data suggests that the São Paulo Court of Justice underestimates the risks of contagion strongly correlated with confinement in precarious and overcrowded facilities like Brazilian prisons. In only 4 of the 371 cases in our sample was it explicitly acknowledged that the risks of contagion were higher inside prison and jail facilities, while in many decisions judges explicitly declared otherwise. The CNJ and other experts around the world have raised concerns about the high risks of COVID-19 contagion in prison populations (Kinner et al., 2020;Okano & Blower, 2020), but these warnings do not seem to resonate with judges in the São Paulo Court of Justice.
By failing to appreciate the higher likelihood of contagion within prisons, the São Paulo Court of Justice appears to ignore the reality of prison life and the increased risks to detainees, especially those who are part of risk groups 8 . Moreover, this wrongful conception of the relative risks inside and outside of prisons is exacerbated by judicial balancing of public safety and public health, whereby some judges even argue that keeping the rest of society safer has more value than protecting the life and health of detainees-even those accused of offenses that do not involve violence or serious threat. As one judge wrote: "In both situations, what one aims to protect is public health, the wellbeing of all, so there is no reason to expect a different treatment. The virus free is dangerous, so, if we cannot arrest it, we confine ourselves. A free drug dealer is also dangerous, but we can get rid of him, so let's arrest him or keep him locked up, at least for some time so he can reflect on the seriousness of what he has done" (Habeas Corpus n. 2053292-65.2020.8.26.0000).
The resistance of most judges to implement Recommendation 62 is unsurprising given earlier studies. Other scholars have shown not only that the São Paulo Court of Justice resists granting habeas relief in general (Freitas, 2020), but that lower courts rarely follow higher courts' de-incarceration recommendations (Machado, Barros, Guaranha, & Passos, 2018) 9 . Taking the effects of the pandemic seriously requires acknowledging that, in a context of mass incarceration, mass release is the only solution to the uncontained spread of the virus and the loss of human lives 10 .

CONCLUSION
This paper is the first step in a larger effort to understand how the COVID-19 pandemic is affecting Brazil's judicial system and prison system. The findings cause serious concerns about the health of the Brazilian prison population-concerns that could have implications for the general population and on broader policies for containing the virus. They also vindicate the perception that the right to health of prisoners is not taken seriously, not even in the context of the most serious public health crisis in generations.
Our findings are robust but have important limitations. We analyzed habeas decisions by the appeals court of only one state. We cannot affirm that similar results will be found in other courts and instances (first-tier judges, higher courts, and appeals courts in other states). In theory, the results might also differ if we analyze other procedural avenues for seeking early release or house arrest if one has an open case under review. However, based on the high number of habeas decisions published at the São Paulo Court of Justice official gazettes, we have reasons to believe that the habeas writ has been used as a preferable procedure. Moreover, given the challenges of accessing the full court records for each case, we analyzed the judicial decisions only (many of the court records are filed under seal and cannot be accessed by the public). This means we can have only an approximate understanding of petitioners' claims and circumstances. Lastly, 9 The National Council of Justice does not have judicial powers, making recommendations like recommendation 62 one of its more controversial activities. However, Sao Paulo State Court's lack of compliance with this more substantive rule exemplifies a tendency found also when CNJ is performing an administrative function like establishing policies to standardize information systems throughout the Judiciary. A recent study has found that lower courts resist even these procedural regulations. For more, see Oliveira and Cunha (2020). 10 See Lenin (2020) discussing how in the U.S. context this exceptional situation and crisis offer "an opportunity to recognize the cruelty, inhumanity and destructiveness" that mark criminal policies in "normal times" (Lenin, 2020, p. 2). we did not analyze the impact of other institutional responses, such as those from the Federal Department of Prisons or from each state administration, nor did examine further elements of Recommendation 62. This paper is the first effort to understand a topical issue with serious implications for criminal justice and for public health. The gravity of the COVID-19 crisis calls for a robust and prolonged research effort, of which this paper is a fair start.