Homelessness during the Coronavirus Pandemic. Exploratory Study in Switzerland

ABSTRACT The coronavirus pandemic has negatively affected people of all social strata, and continues to do so, but its effect has been the most severe on members of the most precarious populations. In this exploratory study conducted in Switzerland, the specific situation of homeless people, a particularly vulnerable population, is examined from a criminological perspective. In total, we surveyed 32 homeless individuals: 14 during the first wave of the pandemic (March-September 2020) and 18 during the second wave (December 2020-March 2021). Results corroborate that the pandemic has had adverse effects on the respondents – both socioeconomic and psychological. Most of the participants do not use drugs and, overall, those who reported drug use did not report an increase during the epidemic. The occurrence of both victimization and offending is low among the participants. Ethical and methodological considerations such as the minimization of social desirability bias, satisficing, as well as the recruitment of difficult-to-reach participants and data collection more broadly during a pandemic are discussed.


Introduction
Exceptional situations with the potential to affect large swathes of a population, like biological threats, natural disasters or civil unrest, often have a large effect on the behavior of people, even more if they are unexpected, acute, and stressful (Hodgkinson and Andresen 2020). Since the beginning of 2020, the coronavirus disease (COVID-19 from now on) has been afflicting global health, forcing considerable changes upon the economy, social life, and social and cultural institutions in an attempt to contain the spread of the virus. Between 2020 and 2021, multiple countries have enacted lockdowns, states of emergency have been declared, borders have been closed, travel has been restricted, and for several months only essential businesses could remain open to the public. The stringency of public health measures varied between governments (Hale et al. 2021). For example, after the initial wave of lockdowns, many states imposed curfews, mandated testing for routine activities such as frequenting bars, nightclubs or traveling, and/or declared new national or regional lockdowns.
The COVID-19 pandemic has influenced multiple public spheres, criminality included (for a review, see Jaccoud, Burkhardt, and Caneppele 2021). For instance, Hodgkinson and Andresen (2020) analyzed Canadian crime trends during the first lockdown in 2020, concluding that the shift in crime trends during the pandemic is most consistent with the predictions of Routine Activities Theory (RAT) coined by Cohen and Felson (1979). Summarily, the pandemic changed the structure of opportunities for committing a crime during 2020 and therefore criminality. The largest study conducted thus far (Nivette et al. 2021) involving 27 cities belonging to 23 different countries found that the lockdowns were linked to an overall drop of 37% in urban crime. At the same time, some offenses have increased during the lockdowns, such as hate crimes against East Asians (Eisner and Nivette 2020;Gover, Harper, and Langton 2020;Tessler, Choi, and Kao 2020), non-lethal domestic violence (Arenas-Arroyo, Fernandez-Kranz, and Nollenberger 2021;Campbell 2020;Piquero et al. 2021) and cybercrime (Buil-Gil et al., 2021). However, other offenses, such as femicides, have not shown the same pattern and stayed stable or even decreased (Aebi, Molnar, and Baquerizas 2021;Hoehn-Velasco, Silverio-Murillo, and Balmori de La Miyar 2020).
Switzerland, a federal European republic with a population of approximately 8.4 million citizens (Confédération Suisse 2017) and the country in which this study has been conducted, was also negatively affected by the pandemic. Seeking to reduce the spread of the virus, Switzerland enforced in March 2020 a partial lockdown which required bars, clubs, and other non-essential businesses to be closed, public gatherings −including demonstrations− to be controlled, and strongly encouraged the facultative confinement of the general population. After the first semi-lockdown, Swiss restrictions varied according to the time of year: after a period of relaxation, a second semi-confinement was declared in December 2020 and lasted until February 2021 (Conseil fédéral 2021). Although the Swiss federal government appeared to have successfully managed the first wave of the pandemic, it has not been immune to criticism (Sager and Mavrot 2020), and the situation of precarious citizens −such as homeless individuals− has been a concern. Namely, for public authorities, the fact that homeless people infected with COVID-19 are difficult to detect, isolate, and quarantine has been a cause for concern both with respect to a) the health of homeless people themselves, i.e., their high risk of mortality and of otherwise developing life threatening health complications, and b) public health, because homeless individuals infected by COVID-19 could potentially infect a large proportion of the population on account of their outdoorsy lifestyles (Albon, Soper, and Haro 2020;Baggett et al. 2020;Banerjee and Bhattacharya 2021;Imbert et al. 2020;Kirby 2020;Lima et al. 2020;Morgan 2020;Tobolowsky et al. 2020;Tsai and Wilson 2020). In this regard, preexisting Swiss night shelters were already overcrowded before the pandemic and/or lacked resources to receive all homeless people during a contagion which required maximizing the number of collaborators while simultaneously dealing with social distancing and the implementation of the strictest hygienic measures. Furthermore, the governmental closure of public places forced the homeless to remain outdoors during the day and, therefore, the sanitary measures encouraged by the authorities were hardly accomplishable ("Face au coronavirus, la situation des sans-domicile fixe inquiète." 2020).
In the past, criminological studies have found a link between homelessness and both offending and victimization (see Baron 2004Baron , 2007Gaetz, 2004;Kipke et al., 1997;Tyler and Johnson 2004). For instance, in Canada, Baron (2004) tested Agnew's Strain Theory by analyzing the link between the offending rate of 400 young homeless and ten strains: emotional abuse, physical abuse, sexual abuse, homelessness, having been a victim of robbery, violence or theft, relative deprivation, monetary dissatisfaction, and unemployment. He found that all strains were predictive −either as main effects or in interaction with other variables− of offending. Gaetz (2004) argued that the victimization of the homeless youth in Canada was related to the social exclusion they experienced. Along the same lines, Kipke et al. (1997) studied 432 young homeless in the USA, finding that both homeless women and men manifested a high rate of exposure to violent victimization. Homelessness seems therefore a topic of interest for criminologists since the criminality and victimization of the homeless may constitute part of the dark figure of crime as well as due to their specific vulnerabilities and needs. However, to our knowledge, whilst criminological scholarship has carried out studies on the impact of COVID-19 on crime trends, the situation of the homeless has not been explored, except by Yakubovich and Maki (2021) and by Garriga (2021), whose objectives are different from ours: The former address women's homelessness during the pandemic as a consequence of intimate partner violence whilst the latter focus on the public discourses regarding homelessness during the pandemic.
This paper therefore aims to study the experiences of homeless people in the French part of Switzerland with respect to pandemic-related economic and emotional strains, victimization and offending, as well as drug use during the year 2020. Our goal is threefold: (1) to fill a gap in the existent literature on homelessness in the context of the COVID-19 pandemic; (2) to explore the relationship between strains and outcomes such as victimization, offending and drug use; (3) to illustrate the challenges (research design, respondent identification, and response validity) of surveying an understudied and difficult-to-reach group during a global pandemic.

The study
Thirty-two homeless individuals were recruited between May 2020 and August 2020 (first phase) and between December 2020 and March 2021 (second phase) (see Figure 1). The research began as a pilot study on the effect of the lockdown on the lives of homeless people in Switzerland (n = 14) and was extended while the pandemic persisted, monitoring the effects of its second wave as well (n = 18). It is to be noted that out of 32 participants, 28 were interviewed once, and two participated twice, i.e., once during the first round and once during the second one. However, it is seldom possible to know with certainty who participated twice in the questionnaire. Unfortunately, the homeless population is rather difficult to be studied in a longitudinal manner because of the challenges with providing codes to participants to identify them while preserving anonymity. Therefore, this study should be considered as consisting of two snapshots of the situation of the homeless during the first two waves of the coronavirus pandemic in Switzerland. Figure 1 illustrates the data collection procedure consisting of an iterative method in which we analyzed the data from the pilot before carrying on with the second phase. During the pilot, we contacted five night shelters situated in four different cities in Romandy, among which three accepted to assist in recruiting participants and/or collecting the data. These night shelters were reached either through networking (since the first author has been working for four years as a social worker) or with the help of two professors in social work who acted as gatekeepers. Because of the aforementioned social distancing and public health measures, we could not be present on the field to observe the functioning of these night shelters and to personally collect the data. Instead, we were obliged to create a short questionnaire (fillable on paper or online). The questionnaire (Annex 1) was originally written in French and it was then translated into English and Romanian based on the recommendations of the directors of the night shelters who informed us of the nationalities of their users. We sent the three versions of the questionnaire to the participating night shelters via e-mail, alongside a URL to access the online version. The social workers affiliated with the collaborating night shelters then distributed them among the homeless users and/or eventually conducted face-to-face interviews with interested participants, depending on their reading skills, the available time and space, and the resources available to the night shelter. During the second phase of the study (December 2020 to March 2021) we recontacted the already participating shelters, but we also sought new collaborators, among which a day shelter. Overall, we could count on the participation of structures operating in three cities. We used the same questionnaire with slight modifications to take into account the second wave and to ask the respondents whether they had previously participated in our study. This time, the first author of the manuscript was able to go on the field and observe and collect part of the data by herself: in total, the questionnaire was distributed to six persons by the staff of the shelters and the field researcher distributed other twelve. Moreover, she observed the field for roughly 15 hours, which allowed contextualizing, and therefore augment, the collected data evidently.
During the entire process, we maintained contact with our collaborators and remained available to satisfy any queries. In addition, in both rounds, the questionnaire was accompanied by a document containing instructions for the social workers (Annex 2), intended to increase the validity and reliability of the research. Information provided included the survey goals, its modes of administration (face-to-face or self-administered), as well as recommendations for decreasing potential biases. The recommendations were, among others, to administer the questionnaire individually and to avoid the possibility of other users overhearing, to clearly explain the survey goals and its voluntary and non-remunerated character, to encourage each user to answer honestly, to keep any information shared by the user confidential, and to read the questions as neutrally as possible. This latter was intended for minimizing the social desirability bias (Gaia 2020). It is also to note that we have experience conducting delinquency and victimization surveys with difficult-to-reach and vulnerable populations, and having been a social worker ourselves, we are trained for non-judgmental interactions with the users. Nevertheless, it is plausible that in spite of our efforts, social desirability was still present.
In total, seven questionnaires were filled online, and the rest on paper and sent to the research team, or completed by the research team. Overall, 21 questionnaires were filled on paper and sent to the research team by mail and seven were completed online. Eighteen surveys were filled through a face-to-face interview with the researcher or a social worker, seven participants completed the survey by themselves while asking questions to the administering social worker, and seven individuals completed the survey entirely by themselves. Twenty-two questionnaires were filled out in French, five in English, three in Romanian, and two in Spanish. We estimated the participation rate to around 50% although we do not possess reliable data since to document this aspect too time-consuming for the social workers. No participant dropped out once the questionnaire started but we were obliged to discard three questionnaires because the participants, users of the day shelter, were no homeless but just in a precarious situation. Those three questionnaires are not a part of the final sample (N = 32). Particular attention has been given to both ethical and methodological concerns 1 since both the population and the topic were judged sensitive. The data protection of the participants was fully insured by not collecting their personal data such as name, surname, and date of birth, and the research was both explained orally and on the questionnaire. We did also not discuss the specific answers given by the respondents with third parties, such as social workers, and as mentioned above, we requested the interviewers to likewise maintain the responses strictly confidential. We were also wary of both over-stigmatizing and over-victimizing a population living on the margins of society and facing daily struggles, and which may understandably be hesitant to discuss or to think about our 1 It is to note that our research did not need to count on the examination of an Ethical committee. At our university, this is a facultative procedure since our type of research does not enter the domain of application of the Swiss Federal Law related to research on the human being (Loi fédérale relative à la recherche sur l'être humain, 2014). research questions; in fact, it has been observed that surveys and interviews can be cathartic, but also traumatic (Birch and Miller 2000;Peritore 1990). We took special care in designing our survey questions, which were also approved by the directors of the shelters before being delivered to the participants. In addition, when faced with a person in a highly vulnerable situation, we oriented them toward the night shelter social workers for further assistance. In our case, comments provided by our respondents and feedback received from the social workers suggest that our target population received the questionnaire in a positive manner. For illustration, one social worker told us: Respondents perceived the questionnaire rather well. The explanation you gave me suited them and satisfied them. However, money matters are often difficult for them. Answering these questions made them reflect on how they experienced the confinement and sometimes led to great discussions. It should be noted that the questions asked addressed points that we rarely discuss with our users (Social worker).

Questionnaire
The questionnaire comprised the following sections of which the results are addressed in the findings section: (1) Sociodemographic data (basic): sex, gender, age, parental status, and civil status.
(2) Living conditions: regular place of sleeping, self-reported health, 'higher risk from coronavirus' status.
(3) Life during the pandemic (since March 2020): self-rated effect of the pandemic, spheres of life negatively affected by the pandemic, social interactions during the pandemic, and affect during the pandemic. (4) Substance use (tobacco, alcohol and drugs) during the pandemic (since March 2020).
(5) Victimization and offending during the pandemic (since March 2020): prevalence 2 of theft and physical assault, incidence 3 of theft and physical assault, place and time of day of victimization or offense. (6) Sociodemographic data (sensitive): monthly income and total savings, educational attainment, nationality, and residence status. Table 1 illustrates the sociodemographic characteristics of our sample, broken down by the first and second wave of the pandemic. Overall, twenty-six men, five women, and one intersex person participated in the study. Age distribution ranges from 18 years old to 71 years old (M = 43.6 years old, Mdn = 46.0 years old, SD = 15.1). Most participants were middle-aged or older, i.e., 45 years old or older. Most respondents were single (n = 23), with children (n = 20), and foreigners (n = 25). Eighteen had either an irregular or no residence status, and 15 had at most completed secondary education. Twenty-five participants sleep at a night shelter, two on the streets, one in an abandoned camping car, one in his own car, and seven at someone else's domicile. Most of the respondents (n = 17) have no monthly income and, among those who do, their income is markedly low (to provide a point of reference, the median salary in Switzerland is around 5,000 Swiss francs). In the same vein, most respondents (n = 21) possess little (less than 60 Swiss francs 4 ) to no savings.

2
The proportion of participants committing or suffering an offense during a given time period (Aebi 2006). 3 The number of offenses of the same type (e.g. theft) committed or suffered by an individual during a given time period (Aebi 2006   Most respondents self-rated as healthy or very healthy (n = 20), and only three persons self-reported as being at higher risk from coronavirus by virtue of being 65 years or more, or having health problems. It should, however, be noted that our sample includes seven respondents aged 65 or more.

Evaluation of the effects of the coronavirus pandemic
In general, the pandemic had a negative effect on the life of the homeless interviewees (Table 2). Among these, three women, and 19 men reported a negative or a very negative effect on their everyday life. Whereas respondents aged between 18 and 54 perceived the pandemic as having a negative effect on their lives (n = 18), those older than 60 were more heterogeneous in their judgment (n = 3 negative, n = 3 neutral, n = 1 positive).
Regarding the specific negative outcomes emerging from the pandemic, 17 homeless people lost their jobs and stopped meeting their friends because of social distancing, 15 individuals needed to spend most of their savings, twelve ate less, eleven lost their housing, five were furloughed, four felt their health worsen, nine had friends or relatives that were coronavirus positive, four people got infected with the coronavirus, and three had friends or relatives deceased because of the virus. In total, almost all have suffered at least one of the listed consequences. Moreover, 18 respondents have suffered three or more of these negative outcomes contemporaneously, seven have faced two negative consequences, and five respondents have endured only one of the aforementioned negative effects of the pandemic.

Effects of the Covid-19
Very negative n = 4 Negative n = 6 Neither a negative nor positive n = 1 Positive n = 1 2 missing values Very negative n = 3 Negative n = 9 Neither a negative nor positive n = 6 Positive n = 0

Effects: specifically (multiple answers possible)
Job loss n = 7 Stopped meeting friends n = 7 Spent most of the savings n = 6 Ate less n = 3 Lost housing n = 3 Furloughed n = 2 Health worsened n = 1 Friends or relatives caught COVID-19 n = 1 Caught COVID-19 n = 1 Job loss n = 10 Stopped meeting friends n = 10 Spent most of the savings n = 9 Ate less n = 9 Lost housing n = 8 Furloughed n = 3 Health worsened n = 3 Friends or relatives caught COVID-19 n = 8 Friends or relatives deceased due to COVID-19 n = 3 Caught COVID-19 n = 4 Number of negative outcomes during the pandemic 0 effects n = 1 1 effect n = 2 2 effects n = 5 3 effects n = 5 4 effects n = 1 0 effects n = 1 1 effect n = 3 2 effects n = 2 3 effects n = 2 4 effects n = 3 5 effects n = 2 6 effects n = 3 7 effects n = 2 5 The original query was an open question. In order to reduce the number and size of the intervals, we grouped these into clusters. Therefore, the scale is not continuous, nor are the categories exhaustive.
When comparing the responses at the first round of the questionnaire and at the second round, the difference between the number of participants that had lost their housing (three versus eight) or those who needed to eat less (three versus nine) is remarkable, and it might be symptomatic of the impoverishment of the respondents due to the continuation of pandemic times. Naturally, regarding their contagion with COVID-19, the more time passed, the higher the risks of exposure. Table 3 illustrates respondents' answers regarding their social interactions and emotions during the pandemic. Regarding the former, the participants spent most of the time either alone or in the company of strangers (n = 18). Despite the fact that most respondents lacked physical contact with friends or family (i.e., spent their days either alone or with strangers), only three participants never interacted with friends or family either online or by telephone, whereas 15 interviewees kept in touch weekly with friends or family.

Social interactions and emotions during the pandemic
In general, our respondents reported experiencing unhappiness (n = 17) during the pandemic rather than happiness (n = 4). The results are less straightforward regarding anxiety and irritation, the frequency of which is not skewed toward the extremes (never/always), while varying mostly between rarely and often. Nonetheless, when comparing the answers of the participants from the first wave with those of the second, more of the latter reported experiencing these two negative emotions rarely, if Table 3. Social interactions and emotions (N = 32).

First wave Second wave In-person interactions during pandemic
Spend the day alone n = 6 Spend the day with strangers n = 4 Spend the day with friends n = 3 Spend the day with family n = 1 Spend the day with the significant other n = 0 Spend the day alone n = 5 Spend the day with strangers n = 3 Spend the day with friends n = 6 Spend the day with family n = 2 Spend the day with the significant other n = 1 Online/telephone interactions with family/friends during pandemic Daily or almost daily n = 6 Several days each week n = 1 Sometimes n = 6 Never n = 1 Do not have close ones n = 2 Daily or almost daily n = 5 Several days each week n = 3 Sometimes n = 6 Never n = 2 Do not have close ones n = 2 Happiness during pandemic (Very) unhappy n = 6 A bit unhappy n = 2 Neither unhappy nor happy n = 3 A bit happy n = 0 Happy n = 0 Very happy n = 1 2 missing values (Very) unhappy n = 5 A bit unhappy n = 4 Neither unhappy nor happy n = 6 A bit happy n = 2 Happy n = 1 Very happy n = 0 2 missing values Anxiety during pandemic Always n = 2 Often n = 3 Sometimes n = 5 Rarely n = 0 Never n = 4 Always n = 1 Often n = 4 Sometimes n = 4 Rarely n = 4 Never n = 5

Irritation during pandemic
Always n = 1 Often n = 3 Sometimes n = 4 Never n = 6 Always n = 0 Often n = 4 Sometimes n = 2 Rarely n = 8 Never n = 4 Much better than during first wave n = 1 Better than first wave n = 8 Similarly than first wave n = 5 Worse than first wave n = 4 ever. This might be due to a habituation effect to the pandemic: among those participating in the second round of the questionnaire (n = 18), half indicated feeling much better during the second wave than during the first. Regarding the contemporaneous experience of all three negative emotions, only two respondents have often or always felt unhappy, anxious, and irritated.

Drug use, victimization and offending during the pandemic
In general, participants reported that they do not consume tobacco (n = 18), alcohol (n = 17), or illegal drugs (n = 26) (Table 4). Overall, 12 people reported not being consumers of any of the substances included in the questionnaire, ten reported being tobacco users, five reported consuming two of the drugs listed, and other five use all three substances. Among those who reported being consumers, seven reported increased usage of tobacco, and five of alcohol. Two reported using illegal drugs a bit more during the second wave. On the other hand, five respondents reported consuming less alcohol than before the pandemic and two reduced drug consumption. Regarding the distribution of drug users by sex, three (out of five) women and 18 (out of 28) men reported not being tobacco users; four women and 12 men reported not being alcohol users, and four women and 21 men reported not being illegal drug users. In addition, among those women who use any drugs, all of them reported neither A bit more n = 4 Neither more nor less n = 2 I do not consume n = 10 Much more n = 2 More n = 1 A bit more n = 1 Neither more nor less n = 3 Much less n = 1 Alcohol consumption I do not consume n = 9 Neither more nor less n = 2 Less n = 3 I do not consume n = 8 More n = 1 A bit more n = 4 Neither more nor less n = 3 Less n = 2 Illegal drug consumption I do not consume n = 12 Neither more nor less n = 1 Much less n = 1 I do not consume n = 14 A bit more n = 2 Neither more nor less n = 1 Much less n = 1 increasing nor decreasing their consumption compared to before the pandemic. In terms of age groups, senior respondents have the lowest prevalence of consumption of any of the three substance categories (n = 1 for tobacco, n = 2 for alcohol, n = 0 for illegal drugs). However, when comparing the first round with the second round, the findings suggest that more participants of the second round reported increasing their legal drug consumption since the pandemic started. Self-reported victimization and/or offending during the first year of pandemic is remarkably low among our sample: the overwhelming majority of participants declared having been neither a victim nor a perpetrator of either theft or physical assault. Regarding victimization, four people were a victim of a theft, and two of assault. Among these respondents, one person was a victim of both a theft and an assault. Therefore, within the entire sample, three people were victimized. These events occurred outside on the streets (n = 2) and at a night shelter (n = 1), at late hours (evening or night). It is worth mentioning that, although our sample is limited, the incidence of repeated victimization since March 2020 is noteworthy: with respect to assault, one victim reported two events, and another victim reported three events. Concerning victims of theft, two reported it happening once, one victim reported suffering it three times and two victims counted four times.

Prevalence of victimization*
Vis-à-vis offending during the first year of pandemic, two people disclosed committing theft and two people reported committing assaults. One person committed both a theft and a physical assault. These offenses were perpetrated in a supermarket (n = 1) and at a night shelter (n = 1) during both the afternoon and the evening. The profiles of victims and offenders do not overlap, that is none of the homeless respondents both committed and endured an offense during 2020.

Needs during the pandemic
When asked about their needs during the pandemic, most of the participants expressed material needs such as money (n = 7), employment (n = 9), housing (n = 10), internet access (n = 1), social aid (n = 1), and a table for writing (n = 1). Others communicated the need for general hygiene during the pandemic (n = 2) and to talk or see people (n = 3). For illustration, respondents expressed their desires in the following manners: "To restart normal life" (Participant 1, 1 st round).
"To stay positive and keep two meters, wash hands with soap, [and] all the time disinfectant" (Participant 11, 1 st round).
"See people around, find a place to rest, eat warm" (Participant 18, 2 nd round).

Had you a magic wand, what would you change?
Lastly, participants expressed what they would do if they had a magic wand. Five individuals shared community-oriented wishes, such as wishing everyone to be all right and happy, preventing the pandemic from killing people (n = 2), stopping people from destroying the planet (n = 1) and changing the behavior of the people (n = 1). Others expressed personal wishes such as to find a job (n = 9), to have a house (n = 8), to be happy (n = 4), to see their family (n = 4), to change their life (n = 2), to get a pension (n = 1) and to have some sort of small project and money to return to their home country (n = 1). "To do regular life and to avoid that the pandemic kills so many people" (Participant 1, 1 st round).
"Wrap the planet with peace, happiness and sense that there is a bright future for everyone" (Participant 2, 1 st round).
"To eliminate bureaucracy and activate my pension, they ask me a lot of documents and the social worker does not want to do anything [for me]" (Participant 7, 1 st round).
"A little project to earn enough money to go back to my country" (Participant 14, 1 st round).
"To find a job that I like and a place where to live" (Participant 17, 2 nd round).
"To be helped to go back to my family" (Participant 19, 2 nd round).
"Go to school and learn French, so I can learn a professional job" (Participant 23, 2 nd round). "To see my mother" (Participant 26, 2 nd round).

Discussion
This paper sought to describe the situation of homeless people, in particular night shelter users in French-speaking Switzerland, during the pandemic of the coronavirus over the course of 2020. By doing so, and given that scholars have mostly focused on the health issues afflicting homeless people (see Albon, Soper, and Haro 2020;Baggett et al. 2020;Culhane et al., 2020;Imbert et al. 2020;Kirby 2020;Lima et al. 2020;Morgan 2020;Tobolowsky et al. 2020;Tsai and Wilson 2020), we hope to fill a gap in the literature concerning the effects of the pandemic on the homeless population that is confronted by many physical, psychological, and emotional threats to their well-being. Our findings corroborate the observation that members of this population are highly precarious, socially isolated, and have been negatively affected by the coronavirus pandemic. Be that as it may, our respondents also reported feeling better during the second wave of the pandemic compared to the first, even though many more people lost their employment and housing during the former. This provocative result might be explained by some sort of habituation effect over time, and suggests a need to also explore resilience in homeless people alongside vulnerabilities.
The most unanticipated, and therefore surprising, finding in our research is that despite these risk factors, outcomes such as drug use, victimization and offending are appreciably low among our respondents. Although some people have consumed more legal drugs since the pandemic started, very few have increased the use of illegal drugs. In this regard, several hypotheses can be put forward. First, it might be that the homeless population in French-speaking Switzerland is precarious but has no drug consumption problems. However, it is also possible that our recruitment did not reach the most vulnerable population, i.e., the homeless drug addicts. Both hypotheses are plausible to some extent and unfortunately, the hardest-to-reach character of the homeless drug addicts challenges our access to them. In addition, we should consider the possibility that the presence of drug dealers on the streets was reduced during some months (see for example "Drogue," n.d.; Drogues et confinement, n.d.). If this is true, then a low rate of drug consumption would logically follow, all other things equal, due to a lack of suppliers. More broadly, since the pandemic has had economic costs for most of our respondents, it is also possible that the lack of economic resources partly explains the low drug use.
When comparing the first round with the second round of the study, more participants of the second round disclose an increased consumption of legal drugs. This finding is noteworthy and might indicate that the impact of the pandemic on substance use may increase as time passes and the pandemic persists. However, at this point this is a conjecture which requires empirical study.
During the first lockdown, the low rate of victimization and offending with respect to theft was not surprising under the light of Routine Activity Theory (Cohen and Felson 1979), since both victims and perpetrators were spending less time on the streets, which translates into reduced exposure to crime opportunities (as also stressed by multiple scholars, see, e.g., Aebi and Tiago 2020;Eisner and Nivette 2020;Hodgkinson and Andresen 2020;Nivette et al. 2021). However, given the need to share shelter with others, the low rate of physical assault is contrary to what was expected considering the results of research on domestic violence (Campbell 2020;Piquero et al. 2021). In principle, it would not have been surprising that the increase in night shelter occupation due to confinement policies combined with tensions due to a lack of resources had contributed to a noticeable number of violent interactions. A potential explanation for the contrary is the fact that the night shelters we studied are only open between 9 PM and 7 AM and therefore users were more likely to be asleep during occupation, which minimizes the chances for active violent interactions. However, the findings emerging from the second round also show low victimization and offending during the rest of the year 2020 and at the beginning of 2021, suggesting that the same pattern observed during the lockdown followed over the rest of the year. In this regard, it seems that even in a scenario in which there were more opportunities to commit an offense or to become a victim of a crime, participants' involvement in crime still remained low. Nonetheless, a factor to consider here is the age of the respondents, who were on average 44 years old. As research has consistently shown, the risk for not only offending but also victimization decreases with age (Hindelang, Gottfredson, and Garofalo 1978;Killias, Aebi, and Kuhn 2019), i.e., the older the person, the lower their criminal involvement (either as an offender or as a victim). Still, although it was a challenging exercise to gather information on the incidence of victimization and offending, the incidence of victimization among our respondents seems rather high. Therefore, our data suggest that although a stark minority of participants endured victimization, they endured it in a repeated manner. This also requires further research, since repeated victims and those who suffer multiple victimization are highly vulnerable and among the group of victims in greater need for intervention that decreases their risk exposure (Farrell 1992).
The survey we conducted provides added knowledge concerning a vulnerable and hard-to-reach population during an extraordinary situation, namely a pandemic, with respect to sensitive topics such as drug use, criminality, and victimization. The fact that we were able to collaborate with social workers who have pre-established relationships of trust with our population of interest has been invaluable, and paramount to the success of our study. This collaboration also allowed us to gather supplementary information following the data analysis (see further below). Furthermore, it is likely that lacking their input, including the recommendation of translating the questionnaire into English and Romanian, our participation rates would have been weaker. Nonetheless, we must acknowledge that a small sample size is among the main limitations of the study, alongside the fact that the research team could not directly instruct the collaborators on the field and monitor part of the data collection. Although the amount of data which could be collected with our questionnaire was limited by its length, this was also a design choice based on the existing literature on homeless people (Baron 2004;Whitbeck and Hoyt 1999), meant to minimize dropout rates and biases due to the response burden (e.g., satisficing). For instance, the latter could have emerged as a result of our participants lacking familiarity with questionnaires or experience reading long texts. Moreover, they could also have been motivated to answer without the required attention if they believed they could gain a material or immaterial reward from the research team or from the night shelter. To prevent response errors due to lack of ability or attention, we designed a short questionnaire comprising short and simple questions, we made efforts to minimize satisficing cues (e.g., leading questions), we disclosed from the beginning the lack of compensation for participating in the study, and we instructed the social workers to simply recruit respondents among their users without pressuring them to participate in any manner. In addition, we also asked the social workers to inform us whether a respondent answered dishonestly, and we were also prepared to discard filled questionnaires presenting internal inconsistencies, although neither of these two scenarios took place.
In terms of limitations, our findings are primarily descriptive in nature and cannot alone establish causal relationships. In that regard, it is unclear whether the pandemic decreased or increased the victimization and offending among the homeless because of lack of data from the prior period. It is important to stress the ethical and methodological pitfalls that we have previously acknowledged, namely the over-stigmatization and over-victimization of our respondents as well as the risk of social desirability bias. As explained, we aimed to be as transparent as possible with our interviewees, not to be judgmental and/or intrusive with our questions and we instructed the social workers to apply the same principles. Nevertheless, social desirability bias is a factor that cannot be discarded when interpreting the low rates of drug use, victimization and criminal behavior among our participants.
Along these lines, we argue that the collaboration between social workers and researchers is a valuable option −if not, the only one− in sensitive milieus in which obtaining access, gaining the trust of the participants, and observing them would require a great deal of efforts and time. Notably, the latter is a resource that many researchers may lack in the context of an ongoing global pandemic. At the same time, we have to acknowledge the challenges of these collaborations, such as researchers lacking total control of the data collection and the risk of gatekeepers instrumentalizing the research (Díaz Fernández 2019). Although compromise is unavoidable when entering the field to study the real world, these are issues which require careful deliberation. In our view, further research should also seek to reach homeless people that do not attend any shelter whatsoever, and whose situation might be (relatively) more vulnerable. As well, we believe that it would be insightful to study the situation of the homeless after the pandemic and to compare findings. However, the most important objective would be to design a self-reported victimization and offending survey to be administered among a more representative sample of homeless people in a periodical manner. Only by doing so we could be in measure to propose evidence-based interventions to policymakers to decrease victimization and offending regarding this group and to improve their well-being and social integration.

Conclusion
To summarize, this paper explored the situation of 32 homeless individuals in the French part of Switzerland during the first year of coronavirus pandemic which began in early 2020. Our findings seem to partially contradict the Routine Activities Theory (Cohen and Felson 1979), although we can only speculate on the virtual absence of physical assaults reported by our respondents and the lack of victimization and offending once the sanitary restrictions were relaxed. The advanced age of the participants as well as response biases (e.g., social desirability bias) might also have contributed to these low rates (i.e., skewed our results), despite our efforts to tackle both methodological and ethical challenges.
While we wish to emphasize the fact that we obtained access to a sensitive and vulnerable population and managed to investigate a sensitive topic in an extraordinary situation, we also have to acknowledge the low number of participants (N = 32) and the fact that our presence on the field during data collection was partial. Moreover, our study cannot establish causal relationships and its design does not allow us to address the counterfactual of whether, in an ordinary setting (e.g., prior to the pandemic), we would have observed comparable victimization and offending rates. At any rate, the survey seems to have been well received by both the respondents and the social workers, and we propose that future inquiries continue to capitalize on the valuable collaboration of practitioners in order to study the post-pandemic situation of the homeless. Other domains we believe would be valuable for further research are related to the study of the self-reported victimization and offending of a representative sample of homeless individuals, their resilience in the face of adversities, and their experiences with repeated victimization.