Is compassion the flip side of punitiveness? Incorporating COVID-19 crisis in experimental vignettes to examine support for visitation and vaccination in prison.

Objectives: The public hold both punitive and pragmatic attitudes toward prison policy. Yet it is unclear whether the public supports compassionate efforts that do not directly relate to recidivism. This study explores the role of exclusionary symbolic aims (prioritizing non-prisoner groups), inclusionary symbolic aims (minimizing health risk for the vulnerable), and cost (taxes).
Methods: Using a quota-based national sample fielded in spring 2021 (N=1260), we embedded two experimental vignettes to assess support for vaccination priorities and personal protective equipment (PPE) for in-person visitation. We also examine respondent experiences (e.g., exposure to COVID-19, vaccine status, personal or vicarious arrest) and beliefs (e.g., political ideology, racial resentment, stigma).
Results: Consistent with dominant exclusionary symbolic aims, respondents showed strong preferences for non-prisoner groups in facilitating safe in-person visits (in long-term care facilities) and vaccine priorities (to prison guards). Inclusionary symbolic aims were less clear when examining risk from vaccine side effects or helping vulnerable populations (the elderly). High cost reduced support for compassionate policy. Conclusions: Public support for policies aimed at maintaining the health of individuals who are incarcerated may be motivated by similar factors as punishment preferences. Supplementary Information: The online version contains supplementary material available at 10.1007/s11292-022-09523-z.

The COVID-19 pandemic has presented a unique opportunity for society to reevaluate priorities of social relationships, health, and inequality. One area that has been given more attention—which overlaps with each of these—is the impact of COVID-19 on those who are incarcerated. As the pandemic forced lockdowns and social distancing, coverage of the pandemic behind bars proliferated.1 Mainstream news highlighted exacerbated social isolation as personal visits were abruptly halted, while the risk for disease remained high due to the porous nature of facilities (guards, transfers, and new admissions), along with an inability to socially distance and maintain hygiene (e.g., Goudie et al., 2020; Klein, 2021; Williams et al., 2020). Prisons were tracked as key outbreak clusters over the first year of the pandemic (New York Times, 2020). Awareness of the impact of COVID-19 on the penal system may lead to growing support for the well-being of justice-involved individuals (Giorgis, 2020; Stillman, 2020).

Public opinion on these matters is consequential as policymakers are responsive to changes in public sentiment—and correctional policy is driven by the public’s punitive mood (Pickett, 2019). While the public can hold ambivalent or even contradictory opinions (Cullen et al., 2000; Unnever et al., 2010), when aggregate sensibilities are overwhelmingly harsh, treatment of those in the criminal justice system will match (Enns, 2016; Garland, 2001; Tonry, 2004). Since the peak in the 1990s, punitiveness has been waning (Enns, 2016; Johnston & Wozniak, 2021; Pickett, 2019; Ramirez, 2013). Greater awareness of the severity of conditions in the system during COVID-19 may elicit more measured beliefs about appropriate criminal justice conditions (Tonry, 2004). This could include support for compassionate policies, such as vaccination priorities and personal protective equipment (PPE) for in-person visits.

On the other hand, growing awareness of challenges faced by individuals in confinement may not translate to support for compassionate policies. Considerable research suggests that the public holds stigmatized views toward the (formerly) incarcerated as evidenced by lack of support for opportunity in varied domains, including housing, employment, and public assistance (Bandara et al., 2020; Denver et al., 2017; Johnston & Wozniak, 2021; Lehmann et al., 2020; Ouellette et al., 2017; Socia et al., 2019). Further, support may depend upon (saving) costs (Bandara et al., 2020; Garland et al., 2013; Johnston & Wozniak, 2021; Socia et al., 2019; Thielo et al., 2016; Vuk et al., 2020) or concerns that individuals in prison are given some benefit that the public is not (Garland et al., 2013).

We found four public opinion studies examining support for strategies to address COVID-19 among incarcerated populations that were fielded in the first several months of the pandemic, many months before COVID-19 vaccines became available (Denney & Garibaldo Valdez, 2021; Eno Louden et al., 2021; Ivanov et al., 2021; Miranda et al., 2021).2 Across the studies, support for release options was moderate at best. In the two that examined support for hygiene and healthcare measures, the public endorsed those options more strongly (Denney & Garibaldo Valdez, 2021; Eno Louden et al., 2021).

Building on this work, we use two experimental vignettes embedded in a national online survey conducted in late May to early June 2021 to assess the public’s support for vaccine priorities and taxes for PPE for in-person visits. Given the relatively low support for release options, we explore policies that would protect individuals while incarcerated. We assess the public’s support for those strategies around the time vaccines were becoming widely available in the US.3 We randomly assigned conditions that may influence endorsement of these measures, including affected party (prisoners vs. non-prisoners), health risks, and financial costs. We also consider the roles of respondent experiences (e.g., exposure to COVID-19, vaccine status; personal or vicarious arrest) and beliefs (e.g., political ideology, racial resentment, and stigma). Identifying which measures the public supports for addressing COVID-19 among individuals who are incarcerated can set priority areas. Further elucidation of which characteristics influence support could inform strategies to garner the public’s backing for these measures. More broadly, the study will contribute to our understanding of public opinion toward compassionate treatment for incarcerated individuals when the interventions do not directly address criminal risk or behavior.

Public attitudes about compassionate criminal justice policies

Punitive attitudes have long dominated the US public, but this masks a more nuanced view (Cullen et al., 2000). While global punitive attitudes are persistent, asking about specific policy solutions can garner support for rehabilitation that is similarly stable (see Cullen et al., (2000) or Jonson et al., (2013) for summaries). This concurrent support for punishment and rehabilitation characterizes the “pragmatic” American who is open to a “balanced approach to crime control that include[s] more progressive initiatives” (Unnever et al., 2010; p. 453).4

What is not clear is whether this pragmatism extends to more compassionate treatment that is not directly aimed at reforming criminal behavior. Often research showing endorsement of progressive approaches frames it in terms of helping those individuals refrain from crime (e.g., rehabilitation) (Thielo et al., 2016; Vuk et al., 2020). Support for health-related priorities, if not viewed as directly impacting criminal behavior, may be lower on the list. For example, support for substance abuse (92%) and mental health (90%) treatment during reentry was exceptionally high among Missouri residents (Garland et al., 2013), likely because the public perceives these conditions as related to criminal behavior. Meanwhile, far fewer supported health care for the formerly incarcerated (46%) (see Ouellette et al. (2017) for similar trends in South Carolina).

The public health crisis of COVID-19 in correctional facilities provides an opportunity to gauge public support for various approaches to promoting health and well-being for individuals in prison that can apply beyond the pandemic. Policies aimed at protecting incarcerated people from COVID-19 have been a source of public debate (Lighty, 2021; Spicuzza, 2021), and research has begun to tease out the sources of public support for such policies. In contrast to policies centered on punishing or rehabilitating incarcerated people, policies centered on reducing COVID-19 risks have no explicit crime reduction goals. Rather, they constitute compassionate criminal justice policies for which the primary aim is promoting the wellbeing of individuals who are incarcerated.

Existing research suggests that public support for health/hygiene measures is higher than endorsement of release options to address COVID-19 in correctional facilities. Denney and Garibaldo Valdez’s (2021) quota-based national sample (n = 1040) supported sanitation/PPE (76%) and preventative care/medical treatment (61%) more than releasing those already eligible (56%) or as many people as possible (25%). In a small convenience sample (n = 124), Eno Louden and colleagues (2021) similarly showed average support higher for non-release conditions (making testing available [4.1 out of 1–5 scale], eliminating co-pays [4.0], and making hand sanitizer [3.9] and soap [3.9] available) than release ones (those with less than a year to serve [3.6], low risk [3.5], elderly [3.4], or had chronic health conditions [3.3]). When only release options were presented, respondents in a national online convenience sample (n = 650) preferred to release individuals with nonviolent charges (elderly/jail 64%; elderly/prison 61%; jail/age not specified 57%) than those with violent offenses (elderly/prison 16%; jail/age not specified 14%) (Ivanov et al., 2021). This suggests that perceived public safety risk may be more important for endorsing release than COVID-19 risk (e.g., elderly, in poor health).

Factors predicting support for compassionate COVID-19 policies included racial attitudes (Denney & Garibaldo Valdez, 2021; Ivanov et al., 2021; Miranda et al., 2021), stigma (Eno Louden et al., 2021) and empathy (Ivanov et al.) toward prisoners, “hierarchy enhancing ideologies” such as meritocracy and anti-egalitarianism (Miranda et al., 2021), and perceptions of COVID-19 (e.g., risk, fear) (Denney & Garibaldo Valdez, 2021; Eno Louden et al., 2021; Ivanov et al., 2021). Experimental conditions presenting respondents with additional information about current events show mixed results. Information on the increased risk of COVID-19 in correctional facilities did not significantly increase support for release options (Ivanov et al., 2021), but information on racial bias in the criminal justice system did (Denney & Garibaldo Valdez, 2021). None of the studies explicitly examined whether symbolic aims or financial cost influenced support for COVID-19 policies.

In the current study, we extend this literature by exploring two theoretical explanations for public support for compassionate COVID-19 prison policy, which may also help to explain compassionate criminal justice policy support more broadly. Our specific focus is on explanations that map onto the symbolic and instrumental considerations that have been found to underlie public support for other types of criminal justice policy (e.g., Gerber & Jackson, 2016; Lehmann et al., 2020; Silver & Pickett, 2015; Tyler & Weber, 1982). We focus specifically on support based in symbolic aims and perceived costs of compassionate criminal justice policies.

Understanding support for compassionate COVID-19 policies

Research on public support for criminal justice policies has largely centered on “symbolic” concerns about maintaining social boundaries that are challenged by crime or “instrumental” concerns about public safety (Geber & Jackson, 2016; Lehmann et al., 2020; Silver & Pickett, 2015; Tyler & Weber, 1982; Tyler & Boeckmann, 1997). In contrast to policies centered on punishing or rehabilitating, policies that center on the health and wellbeing of incarcerated people do not have explicit crime reduction goals. Symbolic and instrumental support for compassionate policies may take on additional forms as a result.

First, compassionate criminal justice policy support may be influenced by the desire to maintain social boundaries in much the same way as that symbolic concern motivates support for punitive policies. By supporting punitive policies—or rejecting compassionate ones—the public can express disapproval of behaviors that break social norms and threaten the social solidarity of “law-abiding” society (Silver & Silver, 2017; Tyler & Boeckmann, 1997; Unnever et al., 2010). Such preferences are thought to be rooted in desires to maintain social boundaries between “us” and “them” and corresponds with the concept of stigma—often characterized by wanting to keep greater social distance from (i.e., not associate with) or “othering” those in prison (e.g., seeing as different, lesser, or more dangerous) (e.g., Bandara et al., 2020; Hirschfield & Piquero, 2010; see also Maruna, 2011). In the case of compassionate criminal justice policies, public stigma toward those in prisons is influential to the concept of “least eligibility,” which is giving individuals in prison benefits that are not freely available to the public (see Garland et al., 2013). Least eligibility may be especially salient when allocating resources to help incarcerated people in ways that are not linked to crime reduction, which may mean diverting resources away from the public, without the purported justification of enhancing public safety.

Broadly, research indicates that members of the public are often reluctant to support policies that may be described as “compassionate” that benefit (formerly) incarcerated people, such as those related to housing, public benefits (e.g., food stamps), and employment (Bandara et al., 2020; Johnston & Wozniak, 2021; Ouellette et al., 2017; Socia et al., 2019). One study that explicitly asked about prioritizing individuals in prison for COVID-19 interventions found that the largest group of survey respondents (43%) thought that “protecting those outside prison should take priority” (Denney & Garibaldo Valdez, 2021).5 Similarly, less positive views about incarcerated individuals are associated with lower support for hygiene, medical, and release options (Eno Louden et al., 2021). Thus, we anticipate that public support for policies that allocate protective resources to incarcerated people will lag behind support for policies that allocate similar resources to non-incarcerated groups (those in care facilities or prison guards). This expectation leads to our first hypothesis regarding exclusionary symbolic aims:

Public support for compassionate COVID-19 policies will be lower for incarcerated individuals than non-prison groups, with higher support for vaccines for prison guards (Experiment 1) and PPE for individuals in care facilities (Experiment 2) (H1).

However, as compassionate criminal justice policies are not explicitly aimed at expressing disapproval of crime, these policies may promote social solidarity through an additional, and underexplored, symbolic concern—inclusionary symbolic aims. Rather than serving the purpose of symbolically dividing “us” and “them,” compassionate criminal justice policies blur this line: under such policies, people in prisons are treated as individuals who are deserving of the full care and consideration afforded to non-incarcerated populations. As a result, compassionate criminal justice policies that provide aid to some of the most marginalized and vulnerable members of society (i.e., incarcerated individuals) may serve to promote and reify universalistic moral values centered on the notion that all individuals—regardless of social status—are deserving of care and fairness. Importantly, such inclusionary moral concerns are widely endorsed (Graham et al., 2018; Haidt, 2012) and are thought to organize people into social and moral communities (Haidt & Graham, 2007; Janoff-Bulman & Carnes, 2013), just as expressive punishment may organize people around the exclusion of rule-breakers. As a result, compassionate criminal justice policies that explicitly and unambiguously promote the wellbeing, health, and humanity of incarcerated, vulnerable individuals may be perceived as better symbolizing widely shared universalistic moral values and will be more strongly supported than policies that less clearly demonstrate inclusionary symbolic aims.

Research on support for compassionate criminal justice policies in the context of COVID-19 is limited, and the existing research on the role of criminal justice policy as demonstrating universalistic moral values is equivocal. Two studies suggest that instrumental concerns (public safety) may outweigh inclusionary symbolic aims, as the public were more likely support release for those viewed as lower risk of recidivism than those who were more vulnerable to disease: greater support for non-violent than elderly or with medical conditions (Ivanov et al., 2021); greater support for low recidivism risk or near release than elderly or with chronic health conditions (Eno Louden et al., 2021). As these studies focus on release, for which public safety is a relevant concern, there is a need to examine compassionate policies that provide care solely within prisons, which may more closely align with inclusionary symbolic aims centered on the public demonstration of care for vulnerable and marginalized individuals. In line with this inclusionary symbolic purpose of compassionate policies, we include the second hypothesis:

Public support for compassionate COVID-19 policies will be greater when the policies have clear universalistic moral values, like causing the least harm (Experiment 1: vaccines with minimal vs. major side effects) and helping the vulnerable (Experiment 2: greater support for PPE for in-person visitation for elderly vs. non-elderly) (H2).

Finally, in the context of compassionate criminal justice policy support, instrumental concerns may center on direct costs to the public in the form of taxpayer burden, rather than downstream costs that are typically the focus of punitive/rehabilitative policies, such as future crime and victimization costs (e.g., Gerber & Jackson, 2016). In previous research, support for substance abuse, mental health, and health care for individuals in reentry were each about halved when respondents were told it would mean higher taxes (Garland et al., 2013), indicating that support for progressive policies in the criminal justice system is tempered when those policies are perceived as higher cost. To date, however, research has largely not explored the extent to which direct costs to taxpayers may shape support for compassionate criminal justice policies in general, or COVID-19 policies specifically. We expect that perceived higher upfront costs will reduce the public’s support for compassionate measures to address COVID-19 in prisons. Specific to these instrumental concerns, we hypothesize that:

Public support for compassionate COVID-19 policies will be greater when it is presented as lower cost (Experiment 2: $20 vs. $100 for PPE for in-person visitation) (H3).

Data and methods

Respondents were recruited via Qualtrics Panel (QP) in May 2021. QP collaborates with online sample providers to deliver responses from opt-in panelists to researchers. QP validates respondents’ profiles via IP address and digital fingerprinting technology (Guillory et al., 2016). Respondents receive electronic invitations to participate in research surveys and complete surveys in exchange for small payments (e.g., gift cards, donations to charity, airline miles). QP uses quota sampling to populate the sample to be representative of the demographic composition in the population on gender, race, age, and region. Our survey began with six demographic questions6 for the quota sampling prior to the experimental vignettes. Studies suggest that QP panels are diverse, representative demographically and politically (Boas et al., 2020) and have higher data quality than crowdsourcing platforms such as MTurk (Zack et al., 2019).

As the current study incorporates experimental designs, it is important that respondents pay attention to experimental stimuli. Thus, we placed an attention check question after the experiments, with scale (e.g., stigma) and additional demographic items at the end of the survey. Our attention check question was adapted from Berinsky et al. (2014) (see online Appendix A for wording). The main analysis includes all respondents as the findings are intent-to-treat effects (Aronow et al., 2019), which is in accordance with recent criminological research (Nix et al., 2020). Of the 1260 respondents, 863 (69%) of the respondents passed the attention check question. We provide supplemental analyses using only the subset of the sample that passed the attention check (see online Appendix B). The experiment results were not substantively different than those obtained from the full sample.

Regarding demographic composition of the full sample, 71% of the respondents are white, 49% are male, and 41% have a 4-year college degree or higher education. The demographic distribution is roughly similar to the population. According to 2019 US Census, 79% of the population are white, 49% are male, and 32% have a bachelor’s degree or above. We use listwise deletion for each experiment separately, resulting in different analytic samples after restricting to those who had responses on the experimental items and control variables.7

Experimental vignettes

Experiment 1: vaccination priorities

Experiment 1 uses a 2 × 2 between-subject factorial survey design, which has two two-level factors and four different hypothetical scenarios. It describes a scenario in which a proposed vaccination program would prioritize individuals in prisons over non-prisoners. Each respondent was randomly assigned to one of the four scenarios. Experimental manipulations included the prioritization of prisoners or prison guards (H1: exclusionary symbolic aims) and the severity of side effects of the vaccination (H2: inclusionary symbolic aims). After the vignette, respondents answered one question about their support for the vaccination plan (coded from 1 = strongly oppose to 5 = strongly support). The text for the experiment is:

“The government is rolling out a vaccination program to address widespread COVID-19 infection in prisons. In rare cases, the vaccination may result in [severe side effects, including death/minor side effects, including muscle pain]. The government is prioritizing the vaccination of [prisoners/prison guards] over the general public.”

Experiment 2: prison visitation

Experiment 2 uses a 2 × 2 × 2 between-subject factorial survey design and describes a proposed tax increase that would support the provision of PPE to institutionalized individuals for family visitation. There are three two-level factors and eight hypothetical vignettes in total. Each respondent was randomly assigned to one of the eight hypothetical vignettes. Experimental manipulations include prison/non-prison setting (H1: exclusionary symbolic aims), elderly/non-elderly recipients of PPE (H2: inclusionary symbolic aims) and low/high cost to taxpayers (H3: instrumental concerns). After the vignette, respondents answered a follow-up question gauging support for the specific policy proposal of raising tax dollars to facilitate family visitation (coded from 1 = strongly oppose to 5 = strongly support). The Experiment 2 text is as follows:

“Your local government is proposing to raise property taxes in the current year by [$20/$100] per household. The increased tax dollars would help provide masks and hand sanitizer to family members who want to visit their [elderly parents/adult children] in [long term care facilities/prisons].”

Additional variables

In addition to the experimental manipulations, we included several variables as possible predictors of support for the COVID-19 policies proposed across the three vignettes. The full wordings of all survey measures are provided in online Appendix A. Like prior work (e.g., Lehmann et al., 2020), we measure political conservatism with a single item ranging from very liberal (= 1) to very conservative (= 5). To measure racial resentment, we follow prior work adapting items from the Henry and Sears (2002) Symbolic Racism 2000 scale (e.g., Pickett & Baker, 2014; Unnever & Cullen, 2010), averaging the responses across four items to create the scale (alpha = 0.751). Our scale for stigma was created by averaging across four items adapted from the Hirschfield and Piquero (2010) stigma scale (e.g., danger, dishonesty, social distance; alpha = 0.880). A single item assessed “how worried” respondents were about the COVID-19 outbreak (1 = not worried at all; 4 = very worried). Experience with COVID-19 was measured through three items: personally tested positive (0 = no, 1 = yes), family members or close friends tested positive (0 = no, 1 = yes), and you, family members, or close friends hospitalized (0 = no, 1 = yes). Respondents were also asked, “When, if at all, do you plan to get the COVID-19 vaccine after it is made available to you?” Responses were already vaccinated (= 1), plan to get vaccinated8 (= 2), and no plans to get vaccinated (= 3). Respondents indicated if they had ever been arrested (0 = no, 1 = yes), as well as if any immediate family members (e.g., spouse, parents, siblings, or children) had ever been arrested (0 = no, 1 = yes).

We also measured demographic controls: age (in whole years, calculated from birth year), gender (1 = male; 0 = female and two non-binary/other respondents), race (1 = white, 2 = Black, 3 = other), ethnicity (1 = Hispanic/Latino; 0 = else), region (coded from state of residence; 1 = Northeast, 2 = Midwest, 3 = South, 4 = West), education (1 = less than high school; 7 = doctorate), and income (1 = less than $10,000; 12 = greater than $150,000; see Table 1). All covariates were balanced across the experimental conditions.9

Table 1

Descriptive statistics

	N	Mean	SD	Min	Max
Dependent variables
Exp 1: Support Vaccine Priorities to Prisons	1260	3.33	1.36	1	5
Exp 2: Support PPE Tax for Visits	1256	2.94	1.48	1	5
Respondent characteristics and beliefs
COVID-19 Worry	1260	2.73	1.09	1	4
Not at all (%)	234	18.57
Slightly worried (%)	255	20.24
Somewhat worried (%)	386	30.63
Very worried (%)	385	30.56
COVID-19 Positive Test-Self (%)	1259	13.82
COVID-19 Positive Test-Family (%)	1257	44.55
COVID-19 Hospitalization-Self/Family (%)	1259	23.67
COVID-19 Vaccine status	1259
Already have (%)	565	44.88
Plan to (%)	454	36.06
No plan (%)	240	19.06
Past Arrest-Self (%)	1250	22.00
Past Arrest-Family (%)	1257	34.69
Conservative	1258	3.00	1.21	1	5
Racism	1260	3.35	1.00	1	5
Stigma	1259	2.99	1.02	1	5
Demographic controls
Age	1259	46.39	17.73	18	94
Gender (% Male)	1256	48.65
Race	1257
White (%)	892	70.96
Black (%)	184	14.64
Other (%)	181	14.40
Ethnicity (% Latino)	1256	17.36
Education1	1258	3.88	1.56	1	7
Income1	1256	6.03	3.62	1	12
Region	1251
Northeast (%)	227	18.15
Midwest (%)	269	21.50
South (%)	469	37.49
West (%)	286	22.86

1See Appendix A for frequencies across all response categories on these control variables.

Analytic strategy

We estimate linear models using ordinary least squares with robust standard errors.10 Multicollinearity was not an issue with variance inflation factor (VIF) values below 1.55 on all covariates (Mn = 1.30). We present results of the experiments with and without controls (Haner et al., 2021; Lin, 2013). Finally, both experiments include conditions for non-prisoner groups; therefore, we provide a supplemental analysis replicating the models in the prisoner-only conditions (see Appendix C). We examine different situational factors in the vignettes that correspond with exclusionary (H1) and inclusionary (H2) symbolic aims, as well as instrumental concerns (H3). In Experiment 1, we examine the impact of prison rather than non-prison groups (H1) as well as harmful vaccine side effects (H2), on support for vaccine prioritization. In Experiment 2 (support for tax providing PPE), we again examine the impact of prison rather than non-prison groups (H1), as well as harms, this time comparing elderly to non-elderly individuals (H2). Experiment 2 also includes the impact of costliness to taxpayers (H3). Although we do not have specific hypotheses regarding interactions, we conduct exploratory analyses for each vignette examining interaction effects of experimental manipulations.

Results

Table 2 shows the results from regression models predicting support for a vaccination program prioritizing prisons (Experiment 1; models 1–2) and support for a tax providing PPE for in-person visitation (Experiment 2; models 3–4).11 Concerning experimental manipulations, we find a significant “prisoner” effect, where respondents prefer to prioritize the non-prisoner group for both policy interventions (support for H1). In line with exclusionary symbolic aims, support for vaccination priorities over the public was higher when the experimental condition was for prison guards (Experiment 1; models 1–2). The coefficient suggests that respondents assigned to the prisoner condition had average support for vaccine priorities for that group that was about 0.37 lower than those assigned to the prison guard condition (on a 1–5 scale) in the full model (model 2). Similarly, support was higher for taxes to purchase PPE for in-person visitation when the experimental condition was for residents of care facilities (Experiment 2; models 3–4). The expected difference in support for taxes for PPEs for in-person visitation was about 0.18 lower (on a 1–5 scale) in the prisoner condition (model 4). Figure 1 provides a visual depiction of the expected size of difference in policy support across the experimental conditions.

Table 2

OLS regressions predicting support for strategies to address COVID-19

	Experiment 1 (vaccine priorities)						Experiment 2 (PPE taxes)
	Model 1			Model 2			Model 3			Model 4
	Coef	SE	Sig	Coef	SE	Sig	Coef	SE	Sig	Coef	SE	Sig
Experimental conditions
Prisoner (ref = guard)	− 0.33	0.08	***	− 0.37	0.07	***
Severe Side Effects (ref = mild)	− 0.21	0.08	**	− 0.19	0.07	**
Prisoner (ref = care facility)							− 0.18	0.08	*	− 0.18	0.07	*
Elderly parent (ref = adult child)							0.11	0.08		0.10	0.07
Tax = $100 (ref = $20)							− 0.32	0.08	***	− 0.24	0.07	***
Respondent characteristics and beliefs
COVID-19 Worry				0.19	0.04	***				0.28	0.04	***
COVID-19 Positive Test-Self				0.17	0.10					0.09	0.11
COVID-19 Positive Test-Family				− 0.04	0.08					0.01	0.09
COVID-19 Hospitalization-Self/Family				0.11	0.09					0.06	0.10
COVID-19 Vaccine status (ref = already have)
Plan to				0.13	0.09					0.24	0.10	*
No plan				− 0.66	0.11	***				− 0.18	0.11
Past Arrest-Self				0.07	0.10					0.23	0.10	*
Past Arrest-Family				0.06	0.09					0.19	0.09	*
Conservative				− 0.12	0.04	***				− 0.22	0.04	***
Racism				− 0.08	0.05					− 0.11	0.05	*
Stigma				0.10	0.04	*				0.13	0.04	**
Demographic controls
Age				− 0.01	0.00	*				− 0.02	0.00	***
Male				0.09	0.08					0.02	0.08
Race (ref = White)
Black				− 0.15	0.11					− 0.01	0.12
Other				− 0.21	0.11					− 0.17	0.12
Latino				− 0.01	0.10					− 0.01	0.11
Education				0.03	0.03					0.08	0.03	**
Income				0.01	0.01					− 0.01	0.01
Region (ref = Northeast)
Midwest				0.06	0.12					− 0.15	0.12
South				0.07	0.11					− 0.06	0.11
West				0.10	0.12					− 0.09	0.12
Constant	3.59	0.06	***	3.48	0.29	***	3.13	0.08	***	3.43	0.30	***
R2	0.020			0.199			0.016			0.289
N	1219			1219			1216			1216

Coef regression coefficient, SE standard error, sig. statistical significance.

*p < 0.05; **p < 0.01; ***p < 0.001.

Fig. 1

Predicted policy support with 95% confidence intervals

Moving to inclusionary symbolic aims as evidenced by concerns about harms and health risks, respondents indicated more support for vaccination priorities when side effects were presented as mild (Experiment 1; models 1–2), but there was no difference in support for taxes for PPE when the affected group were elderly (Experiment 2; models 3–4) (partial support for H2). Finally, instrumental concerns were predictive of policy support: respondents supported taxes for PPE more when the cost was lower ($20 vs. $100) (support for H3). We also estimated supplemental models to test interaction effects for the experimental conditions (available from the authors upon request). None of the interactions of experimental manipulations were statistically significant, but we do observe some small differences among the sub-set of respondents who received the prisoner conditions in each experiment which is described in the next section.

Based on the full models in Table 2 (models 2 and 4), we predicted policy support across the experimental conditions12 with 95% confidence intervals (see Fig. 1). Predicted support was above the mean on the scales (greater than 3) for all the vaccine priority conditions in Experiment 1, but only above the mean for three of the four low tax ($20) conditions in Experiment 2 (all except the low tax for visiting adult children in prison). In Experiment 1, the 95% confidence intervals for predicted probabilities of support do not overlap between the prisoner and guard conditions and very slightly for the side effect severity conditions (such as severe effects prisoner condition support upper limit at 3.17 and mild side effects prisoner condition support lower limit at 3.11). In Experiment 2, the 95% confidence intervals overlap across the pairs of conditions, with the least overlap across the tax manipulation pairs (about 0.04) and the most for the age pairs (about 0.19). Across the eight conditions in Experiment 2, the least overlap in predicted support was between the $100 tax/prisoner/adult child condition (2.78–3.06) and the $20 tax/care facility/elderly parent condition (3.05–3.34). Figure 1 visually displays the respondent preference of greater support for non-prisoner groups and lower cost interventions.

Supplemental analysis: prisoner-only conditions

Certain factors may only predict policy support among respondents receiving the prisoner condition. For the most part, the results were consistent with those in the main analyses, with a few exceptions (see online Appendix C for model results). Severity of vaccine side effects no longer predicted policy support among those within the prisoner condition in Experiment 1 (failure to support H2); however, respondents in the prisoner condition in Experiment 2 were more likely to support taxes for PPE for in-person visitation when the prisoner was elderly (support for H2). The amount of taxes was not predictive in the prisoner-only subgroup (failure to support H3). As noted previously, however, there were no significant interactions among the experimental conditions in the full sample, indicating that these differences in significance in the sub-group do not represent meaningful differences across the prison/non-prison conditions in the full sample.

Discussion

COVID-19 is an unexpected hit on the US prison system, which is already strained by overcrowding and limited medical resources (Cohen, 2015). Resolving the public health crisis requires a rational public that agrees punishment should be humanitarian and proportional, a knowledgeable public that understands the severity of the situation, and a sympathetic public that is willing to allocate more resources to the correction system. We examined support for compassionate policies related to incarcerated individuals during a global pandemic using two survey experiments. Although COVID-19 is a unique contemporary event, the problems revealed in this crisis are deeply ingrained in the society and shed light on important factors shaping policies that affect incarcerated people more generally.

We examined the role of symbolic and instrumental aims in fostering support for compassionate responses to the impact of COVID-19 on incarcerated individuals. Because compassionate criminal justice policies center on promoting the health and wellbeing of justice-involved individuals rather than on crime control, we argue that the relevant symbolic and instrumental aims may go beyond those underlying support for punitive policies. We found that exclusionary symbolic aims centered on “othering” incarcerated people appeared important in predicting compassionate policy support as respondents were less likely to prioritize “prisoners” over non-prison groups (H1). However, we found little evidence that additional inclusionary symbolic aims centered on demonstrating esteem for vulnerable incarcerated individuals (H2) were relevant to the public’s support for compassionate COVID-19 policies. Instrumental concerns (i.e., costs to the public; H3) predicted support for compassionate COVID-19 policies.

Across both vignettes, respondents expressed a preference for prioritizing non-prisoners in vaccination priorities and taxes for PPE—confirming prior work on “least eligibility” for generous services to individuals leaving prison (Garland et al., 2013). This finding was also consistent with prior research examining if the public wants to prioritize COVID-19 responses to individuals outside of prison (Denney & Garibaldo Valdez, 2021; see also Géa et al., 2022). In a public health crisis, such attitudes may impact mortality—and at the very least inflict additional harm on individuals who are incarcerated. While restrained with their support for COVID-19 mitigation policies that impact prisoners, respondents appear aware of the hazardous environments in prisons as evidenced by a slight majority of respondents who received the prison guard condition (53%) supporting vaccination priorities for them over the public (vs. 45% supporting prioritizing prisoners). Similarly, the highest predicted policy support was for prison guard vaccine policies (Fig. 1). Thus, as expected, respondents appeared to perceive incarcerated individuals as less deserving of care and concern—consistent with exclusionary symbolic aims.

In contrast, our expectations regarding support for policies that express inclusionary symbolic aims—particularly universalistic ideals of care for even the most marginalized members of society—were largely unsupported by the data. In Experiment 1, we found that respondents were less likely to support vaccine priorities when side effects were severe, as expected. However, the impact of the severity manipulation—while still in the expected negative direction—was not significant among the subset of respondents who received the prisoner condition. We also saw no main effect for the age condition in Experiment 2 (manipulation of “elderly parents”), nor an age by prisoner condition interaction (although among the subset of respondents who received the prisoner condition, support for taxes for PPE was higher for elderly parents). That is, our results suggested that the public does not reliably prioritize the medically vulnerable for non-release policy options, consistent with research on release from custody for those at highest risk of health problems (Eno Louden et al., 2021; Ivanov et al., 2021). One explanation is that adult children in long-term care facilities were viewed as similarly vulnerable to elderly parents in facilities, given their need for long-term care. However, taken together with the findings regarding vaccine side effects, our results also suggest that the public is not more supportive of policies that demonstrate universalistic moral values centered on care and fairness for vulnerable individuals.

We also predicted that, reflecting instrumental concerns, respondents would be more supportive of policies that were less costly to taxpayers. Consistent with other research showing sensitivity to cost for prison and reentry services (Garland et al., 2013; Johnston & Wozniak, 2021; Vuk et al., 2020), we found that respondents were more supportive of PPE for institutionalized individuals when the cost was lower (although the results were non-significant in the prisoner condition they were still in the expected direction). We interpret this finding as suggesting that, when crime prevention is not a policy goal, direct monetary costs are an important instrumental concern.

Although not central to our analyses, we also found relationships among policy support and respondent characteristics worth noting. Consistent with prior work, conservative respondents were less likely to support either compassionate policy (e.g., Denney & Garibaldo Valdez, 2021; Unnever et al., 2010), while respondents scoring high on racial resentment were less supportive of PPE for visitation (e.g., Ivanov et al., 2021; Jackl, 2021; Miranda et al., 2021; Pickett & Baker, 2014; Unnever & Cullen, 2010). Worry about COVID-19 was associated with greater support for compassionate policies in both experiments; however, having personal or vicarious experience with COVID-19 (positive test, hospitalization) was not, consistent with mixed results in prior work (Eno Louden et al., 2021; Ivanov et al., 2021). Examining a new factor, vaccine status, we found that compared to those already inoculated (as of late May/early June 2021), those who planned to get the vaccine were more supportive of PPE taxes and those who had no plans to get vaccinated were less supportive of vaccine priorities. It appears that vaccine hesitancy corresponds with lower general support for vaccine priorities for others as well. We also found that personal and vicarious experience with the justice system predicted PPE tax support—a result consistent with broader public opinion work on CJ experience corresponding with more inclusive policies (Lehmann et al., 2020; Rade et al., 2018)—but not support for vaccine priorities. In all, the effects of the control variables on compassionate COVID-19 policy largely echoed existing literature on punitive and compassionate criminal justice policies.

Our findings regarding stigma measured at the individual level, however, are unexpected and contrary to our findings regarding the prison/non-prison manipulation as well as prior work (Eno Louden et al., 2021). In our survey, those who stigmatize prisoners endorsed policies that would keep them safer (PPE for visits; vaccines) while they are separated from society. Stigmatizing attitudes are associated with preferences for social avoidance (Snider & Reysen, 2014), perhaps leading people to prefer sanitary measures that would reduce the need for releases in response to COVID-19. Future research will need to disentangle and explore the dimensions of stigma that may link to increased support for compassionate COVID-19 practices.

Altogether, the current study may provide theoretical insight into the sources of support for compassionate policies more broadly. Whereas a large body of research focuses on public punitiveness relating to sentencing (e.g., Bandara et al., 2020) or post-release policies (e.g., Garland et al., 2013; Johnston & Wozniak, 2021), less research has examined public support for policies that impact individuals who are currently incarcerated (but see Vuk et al., 2020). Our results indicate that, for the most part, the sources of public support for compassionate prison policy were similar to those driving support for punitive policies in prior literature: both exclusionary symbolic aims and instrumental concerns appeared relevant. It is also worth noting, however, that most of our findings were similar across the prison-/non-prison conditions in the supplemental analyses,13 suggesting that many of our results may reflect broader orientations toward the provision of care during a health crisis.

The present study has several limitations. First, we use an online nonprobability sample. Although previous studies show the quality of data collected via Qualtrics Panel is high (Boas et al., 2020; Zack et al., 2019) and the current sample mirrors the US population composition, it is not a probability sample and, therefore, comes with those concerns about generalizability. For example, while the overall demographics are comparable to the US population composition, there may be issues with the joint distribution (Mullinix et al., 2015). Second, the study uses cross-sectional data. The circumstances of COVID-19 evolve and the public’s attitudes concerning this public health crisis are likely to change. As such, the relationship between time sensitive covariates, such as COVID-19 worries and vaccination status, and the dependent variables should be interpreted as a snapshot of the survey period (May/June 2021). Third, the order of the experimental vignettes was not randomly assigned; therefore, answers to the second experiment may have been primed by the first (see online Appendix A for survey items ordering). Fourth, experimental manipulations were limited. For example, we indicated the high risk/vulnerable population as “elderly parents” in Experiment 2. However, prior work has used age with medical or health conditions as indicators of vulnerable groups (Eno Louden et al., 2021; Ivanov et al., 2021). Fifth, our measurement of experience and attitudinal factors may have impacted our results. While we used scales measures based on prior work (e.g., racial resentment (Pickett & Baker, 2014; Unnever & Cullen, 2010) and stigma (Hirschfield & Piquero, 2010)), our item for COVID-19 exposure was unique to this survey (“tested positive”). Items measuring extent of prior illness (e.g., asymptomatic to hospitalization) could be more predictive of support for COVID-19 mitigation policies. Additional work in this area could further elucidate factors that impact the public’s endorsement of health and safety measures to assist those in prison.

To prevent the corrections system from becoming the epicenter in the next pandemic, it is important to understand the public’s attitudes toward individuals who are incarcerated, as well as how to gain their support for policies that help improve health and welfare conditions in facilities. The study presents experimental evidence on public opinion about vaccination and safe visitation practices during a public health crisis and represents an important step toward better understanding public support for compassionate criminal justice policy more generally. While the study provides important evidence in this area, research exploring qualitative and longitudinal data is still needed to fully understand the problem and strategies for implementing possible solutions.

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 521 KB)