Changing Attitudes and Provoking Action: Perspective-Taking Mobilizes White Americans for Prisoner Release

Abstract

Imagining oneself in another’s position can soften animus and promote empathy. When one’s loved ones have intense contact with carceral institutions, it can provoke a sense of injustice and political mobilization. Drawing on these insights, I design a survey experiment which assigns respondents to a no-treatment condition, an informational control, an egocentric perspective-taking exercise (imagining they are incarcerated), or a surrogate perspective-taking exercise (imagining someone close to them is incarcerated). I test the effects of the treatments on attitudes toward prisoner release and a semi-behavioral measure—whether respondents write a message to their sheriff in support of release. Relative to the no-treatment condition, the informational control doesn’t elicit changes. However, egocentric and surrogate perspective-taking can increase pro-release attitudes and mobilize respondents to write in support of release. These results push forward the literature on punitive attitudes by considering what forces might mobilize Americans against the carceral state.

After nearly six decades of government investment in carceral state institutions (Hinton 2016), the size and scope of incarceration and community surveillance in the United States are unprecedented (Gottschalk 2006). An estimated 1 in 40 adults are subject to carceral state control, which includes prison, jail, probation, parole, and other forms of community supervision, on any given day (Maruschak and Minton 2020).

Alongside the steady growth of incarceration, activists have organized in resistance (Davis 2003). In summer 2020, as the United States reeled from the COVID pandemic and Black Lives Matter protests swelled, abolitionist demands for decarceration gained new traction. Experts warned that prisoners, corrections officers, and the communities in which jails and prisons are located were at serious risk, due to chronic overcrowding and a lack of mitigation capabilities like sufficient ventilation (Carlisle and Bates 2020). As reports of COVID infections in lock-up spread, activists pushed for inmate release (Justice Collaborative 2020).

Why might people support decarceral efforts? What motivates everyday people to support prisoner release? Scholarship has largely posited that Americans are punitive, and typically support criminal justice policies that result in more incarceration, supervision, and surveillance (Beckett 1997; Gottschalk 2006; Enns 2016). Moreover, pursuing punitive policies is popular as evidenced by politically vulnerable Democratic governors’ choices to outspend and out-incarcerate Republicans in order to shore up their support (Gunderson 2022). And yet, at this moment there were campaigns to raise money for bail funds, petitions to release elderly prisoners, and some elected officials took steps to reduce prison and jail populations.

Drawing on proximal carceral contact research and literature from psychology, I argue that perspective-taking—or “the active contemplation of others’ psychological experiences” (Todd et al. 2011, 1027)—can encourage support for prisoner release, both in terms of policy attitudes and political mobilization. Moreover, I argue that perspective-taking need not be egocentric, imagining oneself in prison. Rather, surrogate perspective-taking is comparably powerful such that imagining one’s loved one in prison produces mobilization for prisoner release. Coalitions addressing criminal justice reform require a broad support base, making understanding white Americans’ attitudes critical. To that end, I test these expectations using a survey experiment fielded to a sample of white Americans and examine both policy attitudes and a semi-behavioral task—writing a letter to one’s sheriff about releasing jail inmates.

I make several key contributions to existing scholarship. First, I explore how white Americans think about policies that seek to shrink the size of the carceral state—a necessity to end mass incarceration. Second, I move beyond research showing surrogate perspective-taking can induce empathy to develop a theoretical argument about how it can induce political attitude and behavioral change. Third, qualitative analysis of the messages that respondents write to sheriffs shows that people draw on ideas about deservingness, threat, and public health to position prisoners as worthy or unworthy of release. These findings break new ground in our understanding of how Americans think about decarceration and how political mobilization in support of a hyper-marginalized population might be achieved.

Attitudes toward the Carceral State

Public opinion is a driver of policy change (Burstein 2003) and scholars argue that the expansion of the carceral state is a political response to the desires of a broadly punitive American public (Enns 2014, 2016; Miller 2016; Nicholson-Crotty, Peterson, and Ramirez 2009). The American public’s appetite for punishment has been on the rise since the 1960s (Ghandnoosh 2014; Enns 2014), and support for punishment is high across criminal justice institutions and practices (Ramirez 2013). Compared to other advanced democracies, the United States stands out for not only its exceptional rate of incarceration, but also for much longer prison sentences and public support for the death penalty (Mauer 2018; Unnever 2010). Wozniak (2014) shows that Americans perceive that prisons are unpleasant, difficult environments, and yet many desire still harsher conditions. Perhaps nothing exemplifies the punitive consensus so clearly as its bipartisan nature. At a time when many policy areas are notable for their partisan divide (Layman, Carsey, and Horowitz 2006), elected representatives of both parties have actively encouraged the reach and power of the criminal justice system nationally and subnationally (Alexander 2010; Beckett 1997; Eckhouse 2019; Hinton 2016; Murakawa 2014; Weaver 2007).

In recent years, however, the punitive consensus among Americans may have weakened. Some states and municipalities have taken steps to reduce crisis-level overcrowding by decreasing their prison populations; President Obama commuted sentences for 1,385 federal prisoners and some criminal justice reforms have passed into federal law with bipartisan support (American Bar Association 2018; Eggleston 2017; Malloy 2016). However, these piecemeal efforts have not substantially changed the scope of incarceration. Two million people remain incarcerated in the United States—many of whom have misdemeanor or drug convictions or are serving time for administrative reasons such as probation or parole violations (Natapoff 2018). Moreover, incarceration rose quickly after initial reductions during the beginning of the pandemic (Sawyer and Wagner 2022).

Despite this potential shift in attitudes toward the carceral state, political science scholarship has little to say about how the mass public thinks about shrinking the size or scope of the carceral state, let alone abolition (Davies, Jackson, and Streeter 2021). Instead, studies typically focus on support for punitive policies, crime spending, or being tough on crime (Gilliam and Iyengar 2000; Enns 2014, 2016). We might imagine these are simply two sides of the same coin, such that those who advocate for limited spending also support policies like prisoner release. However, given the inconsistencies and ambivalence in public opinion (Converse 1964; Alvarez and Brehm 2002), we should not assume this to be the case. For example, Chudy (2021) demonstrates that racial sympathy is not an opposite pole from racial antipathy, but is a different conceptual construct. Moreover, a respondent might say we shouldn’t increase spending on criminal justice either because they think we currently spend the right amount or because we spend too much. Without asking specifically about policies that retrench the carceral state, our understanding of the public’s appetite for decarceration will remain limited.

Race and the Carceral State

Though systems of racial control in the United States have varied over time such that the systems of slavery, Jim Crow, and mass incarceration are not points along a single historical trajectory (Gottschalk 2006), the roots of the American carceral state are inextricable from racial hierarchy (Alexander 2010). After emancipation, whites sought to reassert racial social control, and the expansion of criminal codes was a means to do so. This period brought about the Black codes and convict leasing, which served to continue slave labor albeit under a different economic system (Alexander 2010; Ayers 1984; Lichtenstein 1996; Mancini 1996; Oshinsky 1996). Enslaved labor was no longer predicated solely on race and birth, but on criminal status. And yet, the creation and enforcement of laws to label someone a criminal were tied to race. Thus, since the beginning of the American carceral state, it has been a racialized project, or an “effort to shape the ways in which human identities and social structures are racially signified, and the reciprocal ways that racial meaning becomes embedded in social structure” (Omi and Winant 2015, 13).

When the Civil Rights Movement eroded prior systems of racial control in the mid-twentieth century, the criminal justice system once again was a means to entrench the racial order (Eubank and Fresh 2022). As legalized discrimination and segregation ended, the carceral state expanded rapidly. Importantly, discrimination on felony status remained legal (Alexander 2010). Thus, just as many citizens gained new civil rights across the country, increasing numbers of people were labeled as felons thereby enabling discrimination against them.

The racialized roots of the American carceral system persist both in terms of who has contact with carceral institutions—Black, Latino, and Indigenous Americans are disproportionately likely to be subject to invasive and violent policing, arrest, incarceration, and longer sentences (Alexander 2010; Baumgartner, Epp, and Shoub 2018; Bobo and Johnson 2004; Bosick 2021; DeGue, Fowler, and Calkins 2016; Epp, Maynard-Moody, and Haider-Markel 2014; Mummolo 2018; Murakawa 2019; Soss and Weaver 2017)—and in terms of attitudes toward these institutions. White Americans are particularly punitive: they are the racial/ethnic group most supportive of criminal justice institutions and the death penalty (Hutchings 2015; MacDonald and Stokes 2006; Peffley and Hurwitz 2010; Tuch and Weitzer 1997); they tend to view victims of police violence as at fault (Israel-Trummel and Streeter 2022); the racial inequality of punishment can make them even more supportive of harsh punishments (Peffley and Hurwitz 2010; Hetey and Eberhardt 2014, but see Butler et al. 2018); and their racial attitudes are tightly correlated with their punitive attitudes and criminal justice policy preferences (Baranauskas 2022; Gilliam and Iyengar 2000; Wilson, Owens, and Davis 2015).

This is not to say that Americans of color are uniformly non-punitive in their attitudes (see Jefferson 2023). But the gulf in attitudes toward criminal justice institutions is wide, because the histories and experiences of these institutions are racialized. Because of these histories and experiences and the ways that power is distributed via racial hierarchy, it is imperative to consider how attitudes on racialized policies—like criminal justice—will depend on racial group membership (Anoll 2022; Pérez and Vicuña 2023). In their work on a different racialized policy (immigration), Masuoka and Junn (2013, 16) argue that “political attitudes are structured by the racial hierarchy, are formed at the individual level through the lens of group identity, and are the product of group interactions and historical memory.” This prism of racial attitude formation is apparent in the context of criminal justice. Anoll and Engelhardt (2023) demonstrate that recent negative interactions with carceral state actors produce greater changes in attitudes toward those institutions for white Americans than for Black Americans. They argue that this is because when it comes to the carceral state, Black Americans “carry deep knowledge reservoirs” which “anchor public opinion and attenuate the link between relevant experiences and attitudes” (Anoll and Engelhardt 2023, 1151). As I am interested in how perspective-taking might produce mobilization on behalf of incarcerated people, I focus on white Americans—a group that typically favors punishment and has less exposure to criminal justice systems.

Perspective-Taking

Attitudes toward others can be altered by encouraging people to imagine themselves in another’s position. This is referred to as perspective-taking and a bevy of studies show that it can reduce prejudice toward outgroups (Broockman and Kalla 2016; Galinsky and Moskowitz 2004; Kalla and Broockman 2020; Simonovits, Kézdi, and Kardos 2018; Todd et al. 2011; Vescio, Sechrist, and Paolucci 2003), reduce racial discrimination (Drwecki et al. 2011), and encourage helping behavior toward outgroups (Batson and Ahmad 2009; Shih et al. 2009). There are a range of interventions in these studies, but essentially, they all show that by asking people to consider the possibility that chance is the only reason they are not someone else, animus can soften. Psychology research shows that this largely happens by inducing empathy among those who perspective-take (Shih et al. 2009; Todd and Galinsky 2014).

Empathy is a complex process involving both cognitive and affective processes. For someone to feel empathy with another person, it requires “not only a minimal recognition and understanding of another’s emotional state, but also the affective experience of the other person’s actual or inferred emotional state” (Melloni, Lopez, and Ibanez 2014, 408). Research from psychology and neuroscience shows that empathy is responsive to both social context and priors. People express greater empathy for loved ones versus for strangers (Cheng et al. 2010), and perceptions of blame for one’s negative situation or outcome reduce feelings of empathy (Decety, Echols, and Correll 2010).

Empathy has political effects, with group empathy predicting a host of relevant political attitudes and policy positions (Sirin, Valentino, and Villalobos 2017). Beyond observational relationships, recent political science research indicates that empathy can be induced and lead to political effects. For example, perspective-taking can cause political mobilization to advocate for a disadvantaged group (Adida, Lo, and Platas 2018), and indirect forms of perspective-taking can shift policy attitudes in favor of outgroups (Williamson et al. 2021). In their study of support for refugees, Adida, Lo, and Platas (2018) show that asking Americans to imagine that they needed to flee their home made them significantly more likely to write to the president in favor of Syrian refugee resettlement.

There are exceptions to the power of perspective-taking however, and asking respondents to imagine they belong to a stigmatized group might fail. For example, Gloor and Puhl (2016) show null effects of perspective-taking on fatphobia and social distance from fat people, even though perspective-taking did elicit empathy. Perspective-taking is particularly likely to be resisted “if I have entrenched antipathy toward an out-group or if an out-group’s plight is the result of my actions or the actions of my group,” as “it may sensitize me to differences rather than similarities between myself and the target” (Batson and Ahmad 2009, 152). This may make perspective-taking challenging when it comes to prisoners as people who are in jail and prison are particularly stigmatized in the United States.

Crime and law-breaking are associated with “bad guys” (Barnes 2014; Claster 1992) and white Americans tend to see even those brutalized by police—let alone people who have been convicted and incarcerated—as responsible for their position (Israel-Trummel and Streeter 2022). Just world beliefs, which hold that bad outcomes are deserved because of bad actions (Lerner 1980), may pose particular challenges for perspective-taking with incarcerated people. Indeed system justification and beliefs about deservingness can limit the potential for taking action against injustices (Davis and Wilson 2022). Moreover, criminal justice policy is associated with people of color, particularly Black people (Gilliam and Iyengar 2000; Peffley and Hurwitz 2002), and white Americans tend to overestimate the rate at which crimes are committed by people of color (Ghandnoosh 2014). The connections between Blackness and criminality are so automatic that they are bi-directional: “Black faces and bodies can trigger thoughts of crime, thinking of crime can trigger thoughts of Black people” (Eberhardt et al. 2004, 876). Despite the end of de jure segregation, American society remains deeply racially segregated such that whites in particular have very white social networks (Anoll, Davenport, and Lienesch 2024; Bunyasi Lopez and Smith 2019), and the white-Black color line persists even though it has softened (Davenport 2018). If white people perceive that incarceration affects only or predominately Black people, and that those it affects are to blame for their own position, they may not engage in perspective-taking easily and may resist imagining themselves in an incarcerated person’s shoes.

There is reason to expect that despite these obstacles, perspective-taking could generate support for decarceration and prisoner release. First, this study takes place in the context of a global pandemic. While many white Americans may see people who are incarcerated as responsible for their incarceration, the pandemic was an external event that prisoners couldn’t be seen as causing. This is likely to limit any sense that those in prison were to blame for the historic health risks they faced, as deservingness beliefs reflect, in part, the idea that someone has caused an outcome (Feather 1999). Second, concurrent with the pandemic in summer 2020, the Black Lives Matter Movement organized historic protests that called attention to carceral violence and racial injustice. Though the public opinion effects may have been short-lived (Jefferson and Chudy 2021), there was meaningful change in racial attitudes among whites during this period (Reny and Newman 2021). Together, then, white Americans might have been more willing to imagine the perspectives of incarcerated people at this time. This leads to my first hypothesis: respondents assigned to an egocentric perspective-taking treatment will be more likely to support prisoner release and to write to their sheriff in favor of release.

Beyond imagining oneself as incarcerated, the literature suggests that imagining one’s loved ones in prison could produce political effects. Personal carceral contact and punitive carceral policies can degrade trust and affect political mobilization (Anoll, Epp, and Israel-Trummel 2022; Branton, Carey, and Martinez-Ebers 2023; Burch 2022; Lerman and Weaver 2014; Maltby 2017; White 2019b). By contrast, people whose loved ones have intense carceral state contact are typically more likely to participate in politics (Walker 2014; Walker and García-Castañon 2017; Walker 2020, cf. White 2022; 2019a). Anoll and Israel-Trummel (2019) draw on surrogate mobilization theory (Wilson and Ellis 2014) to argue that proximal contact might produce mobilization among people who care about others with limited access to the political process. They test this theory in the context of felony disenfranchisement policy, showing that having proximal carceral contact is linked to more political participation in the states with the harshest disenfranchisement policies. That is, people are more likely to participate politically when their loved ones are unable to do so.

The question here is whether simply imagining a loved one was incarcerated could produce changes in political attitudes toward incarcerated people and willingness to mobilize in their interest. Surrogate perspective-taking is certainly possible. Cheng et al. (2010) show that considering a loved one’s physical pain produces similar neural processes as one’s own pain. If surrogate perspective-taking induces empathy similarly to egocentric perspective-taking, we might expect it to also lead to the types of changes in political attitudes and behavior documented for egocentric perspective-taking (Adida, Lo, and Platas 2018). This is my second hypothesis: respondents assigned to a surrogate perspective-taking treatment will be more likely to support prisoner release and to write to their sheriff in favor of release.

Design

To test my hypotheses, I fielded a survey via Lucid from October 30 to November 20, 2020, to 2,001 American adults. ¹ This was prior to the FDA’s emergency approval of COVID vaccines and during a spike when daily deaths in the United States returned to 2,000 per day and the country surpassed 250,000 deaths. Thus, this was a time when the risk to prisoners was high and there were limited options to mitigate risk within corrections institutions (Carlisle and Bates 2020).

Lucid distributes surveys to respondents who have opted into various online survey platforms and quota samples with respect to age, gender, race/ethnicity, and region to achieve a diverse national sample. Though online opt-in panels are not nationally representative as with a probability-based sample, they can provide better estimates of causal effects in experimental designs than other convenience samples (Berinsky, Huber, and Lenz 2012). ² I analyze the 1,535 respondents in the sample who identified as white and who completed the survey. ³ The survey instrument included an experiment that randomized respondents into one of four different conditions: a no-treatment condition, an informational control, an egocentric perspective-taking exercise in which respondents imagined themselves to be incarcerated, and a surrogate perspective-taking exercise in which respondents imagined a loved one to be incarcerated. The no-treatment condition offered no information about COVID in prisons and jails before proceeding directly to the dependent variables. The three other conditions had identical introductions, which follow, and then had one of the three prompts in brackets:

As COVID-19 infections increase across the United States, some public health officials are particularly concerned about the risk of infection spreading among those incarcerated in prisons and jails. Several states have chosen to release some inmates who are particularly at risk of life-threatening complications from the disease, but jails and prisons remain overcrowded.

[Informational control: How does that make you feel?]

[Egocentric perspective-taking: Imagine that you were currently incarcerated. How would that make you feel?]

[Surrogate perspective-taking: Think of the person that you talk to the most about what’s happening in your life. Imagine that person was currently incarcerated. How would that make you feel?]

Write a few sentences. ⁴

In both perspective-taking conditions respondents were also asked what they might try to do to respond to the situation. This is a relatively brief intervention in which survey respondents spent a short amount of time considering this, rather than going through an intensive empathy-building exercise. However, the intervention mirrors those that prior studies indicate produce empathy (Todd and Galinsky 2014). The informational control provides all the relevant information about COVID risks in prisons and jails and asks respondents to think and write about their emotional response but does not ask them to imagine that they or a loved one was incarcerated. This allows me to compare whether providing this background information about the condition in prisons and prompting an emotional reflection changes respondents’ attitudes and behavior relative to the no-treatment condition, or whether it is the perspective-taking itself that does the work. ⁵

Following the treatments I captured two dependent variables. The first is an attitudinal scale of support for inmate release that captures support for releasing 1) those awaiting bail, 2) the elderly, 3) inmates who are immunocompromised, 4) nonviolent offenders, and 5) all offenders (full question text available in online appendix 1). I index these measures together to create a single scale of support for prisoner release ranging from 0 to 20 (Cronbach’s α=0.891 ⁶). The second outcome measure informed respondents that some sheriffs in the U.S. were considering releasing inmates to decrease the risk of COVID outbreaks and offered them a chance to share their thoughts on this policy idea with their sheriff. ⁷ Sheriffs are a relevant policy actor for this topic as they are able to exercise substantial local policy discretion (Farris and Holman 2024) and as primarily elected officials in the United States, may be responsive to public opinion. Moreover, this is a realistic choice as some sheriffs and local jail administrators did reduce their jail populations early in the pandemic through a variety of policy changes, including releasing some people (Widra 2021). This is a semi-behavioral measure, as they were offered the opportunity to provide a comment to a government official (a behavioral outcome) but within a survey environment rather than a real-world condition where they would have to send it themselves. However, this strategy has been previously used to demonstrate willingness to take action (Adida, Lo, and Platas 2018).

Six trained coders hand-coded these messages—without seeing the treatment condition to which respondents were assigned—as either pro-release (including those who advocated for release of only certain classes of inmates such as nonviolent or elderly offenders) or anti-release. The trained coders also coded respondents who were unsure of their opinion, respondents who stated that they had no opinion or didn’t want to share their opinion, responses that were unclear in their opinion, and those that were gibberish. ⁸ After coding all the responses, I examined all responses where there was any coding disagreement and reconciled those responses to then produce the semi-behavioral dependent variable where –1 indicates a message in opposition to release, 1 indicates a message in favor of release, and 0 indicates all other respondents. In total, 18.6% of respondents wrote in support of prisoner release and 19.2% wrote in opposition to release. Just over six in ten respondents chose not to write a message clearly stating an opinion on the issue. This indicates that among white Americans motivated to act on this issue, public opinion is closely divided.

What Do People Write?

Before analyzing what respondents wrote to their sheriff, I first explore the open-ended data from the treatment that asked them to reflect on their feelings. This offers a probe of whether the perspective-taking elicits empathy among respondents as expected. I did not directly measure empathy, but coded whether respondents in the informational control and perspective-taking treatments expressed concern, worry, or sympathy for incarcerated people when asked how the treatment intervention made them feel. Because those in the no-treatment condition were not asked about their feelings, I cannot examine empathy among this group.

Respondents in the egocentric and surrogate perspective-taking treatments expressed significantly more concern for incarcerated people than respondents in the informational control condition (refer to online appendix table A3). In the informational control, 14.3% of respondents expressed concern for incarcerated people after reading about overcrowding and health risks to prisoners. By contrast, nearly 60% of respondents in both perspective-taking conditions expressed concern. This suggests that the perspective-taking exercise was successful at provoking empathy. ⁹

Beyond empathy in the treatment tasks, what did respondents write in the messages to their sheriff, if they chose to send one? Qualitative statements within a survey environment offer researchers an opportunity to more fully explore how respondents think about particular issues and add important nuance to respondents’ attitudes that is difficult to glean from the causal analysis alone (Bates and Cole 2024; Bracic et al. 2023). Moreover, these statements help to situate prisoner release attitudes during this particular moment to help us understand how everyday people thought about what the country should do about people in lock-up during a public health crisis.

Reading respondents’ statements demonstrates the contingent nature of support for prisoner release and the sources of opposition. ¹⁰ Most respondents who wrote in favor of releasing prisoners noted the qualifications they wished to see around release. Some gave specific categories of prisoners they either supported or opposed being released. For example, one respondent wrote “I agree with this movement, however who gets released needs to be selected with care. Non violent offenders, inmates at risk, and petty charges for example,” while another wrote “Low risk inmates/crime should be the only people released.” ¹¹ Yet another wrote that “I respect the decision to release inmates and giving them their freedom in order to reduce the risks of COVID from spreading in incarceration systems. However, I believe that it should only be those with non-offensive crimes especially if they are of an older age.”

Respondents’ arguments in support of and opposed to prisoner release often draw on ideas about public health, deservingness, and danger. Some of these statements reflect the contextually specific conditions of the pandemic while others draw on broader claims about the purpose of prisons and punishment. Public health concerns tend to focus on the pandemic specifically, such as the respondent who wrote “I think it a great idea you don’t want a outbreak to happen and effect both guards and inmates.” By contrast, some of those who used public health to oppose release suggested that prison was a safer place during the pandemic. For example, one respondent wrote, “Even though they share cells, they are isolated. Why put them out where they for sure going to be inside their house at all times and most of them may dont wear masks and keep social distance. For me they are better off inside.” A different type of public health argument invoked the belief that Covid is not a serious disease (“I think that is ridiculous when you have a survival rate of 99.9%. It’s just barely more dangerous than the common cold”).

Across many of the statements both for and against release, respondents invoke deservingness: who matters and what should they get? People tend to think that when an outcome results from someone’s actions and choices it is deserved and when it results from outside forces it is not (Feather 1999). Research shows that beliefs about deservingness structure attitudes toward criminal justice (Israel-Trummel and Streeter 2022; Wilson, Owens, and Davis 2015; Wilson and Krysan 2022). In the case of imprisonment during the COVID pandemic, people often invoked desert either by noting that imprisonment results from individuals’ actions—and thus they deserve to be imprisoned—or that the pandemic was beyond individuals’ control—and thus they should not bear the costs of the health risks in prison. Examples of pro-release arguments drawing on concepts of desert included the respondent who wrote “I would have to agree that the non threatening inmates should be released. They deserve to live.” Another respondent suggested that the punishment for lawbreaking shouldn’t include possible death, writing “this is a good idea, the prison industrial complex is corrupt as is, people don’t deserve to die of covid for having maybe committed a crime.”

By contrast, some respondents indicated that prisoners deserve to get sick by virtue of their status as prisoners (“If they get sick, Oh well. Shouldn’t have ended up in jail in the first place”). One respondent wrote, “Thats some bullshit. I don’t give a fuck about this pandemic, if they’re in prison or jail there’s a good chance they deserve it. Fuck them, if they get sick them quarantine them, don’t fucking let them out.” Another simply wrote “Fuck that. Let them die.”

Still other respondents focused on the deservingness of the public, or the “good” people (“Criminals dont deserve your protection. Law-abiding citizens do.” and “The public deserves to be and feel safe. In some cases these are violent criminals. Don’t do it.”). These comments reflect the dichotomy perceived between criminals and law-abiding citizens, where criminals are by definition “bad” and unworthy (Claster 1992). Some wrote about what they, themselves, deserved and how no one was helping them (“Its ridiculous. I have underlying medical conditiobs [conditions]. I am not incarserated. I dont receive any perks from covid why should inmates? If you are so concerned send someone to do my spping [shopping]”).

Moral arguments against release tended to draw from the language of retributive justice where proportionate punishment is demanded for criminal acts (Tonry 2021). Many respondents suggested incarceration is required as just punishment for criminal behavior, such as the respondent who wrote “If they did the crime they should do the time. COVID should not be the grounds to release anyone from jail” or the angrier respondent who wrote simply, “BULLSHIT, THEY DID THE CRIME THEY DO THE TIME.” Another respondent wrote “As a law abiding citizen i am against allowing degenerates out just because of a sickness they should have thought about that before they did what they did to get locked up.” Each of these comments reflect what quantitative studies of public opinion toward punishment in the United States show—that a significant portion of the public supports harsh punishments and desires incarceration to be unpleasant and dangerous (Wozniak 2014).

The last theme that emerged from the qualitative reading of the statements is a focus on public safety. For example, one respondent wrote “Keep them locked up. They are a danger to us who obey the law,” while another argued “Releasing guilty people into the general population is ill advised. Your responsibility is to protect the general population, the good people.” Again, these arguments about public safety reflect ideas about who is morally deserving and is owed protection by the government. Another wrote more simply, “Protect our community and keep the bad guys locked up.” Some pointed specifically to crime rates as a reason to maintain current incarceration levels, including a respondent who wrote “Crime rates are already horrible. We don’t need criminals released because they may get covid. This should have been thought about way before this disater disease hit.”

Respondents’ arguments about prisoner release show that they are drawing on the values and frames that existing work suggests would be relevant to public opinion in this arena. In other words, people’s attitudes on prisoner release are structured and have meaning. Next, I turn to whether the treatments elicited changes in respondents’ attitudes toward prisoner release and in the likelihood they wrote to their sheriff.

Effects of Perspective-Taking

Figure 1(a) shows the average level of support for releasing prisoners in the sample of white respondents. Unsurprisingly, releasing all inmates garners the least support among the individual measures with an average score of 1.054 on the 0 to 4 scale. This indicates that the average response is just slightly above “somewhat disagree.” Releasing nonviolent offenders has the most support, but the average score is still just 2.097, or slightly above “neither agree nor disagree.” The other release conditions fall in between (those awaiting bail $ \overline{x} $ =1.850; the elderly $ \overline{x} $ =1.771; immunocompromised $ \overline{x} $ =1.641). Clearly, even amid a pandemic there is limited support for releasing prisoners. Figure 1(b) underlines this fact as it shows the frequency distribution of the 0 to 20 prisoner release attitudes scale. The modal score on the scale is 0, indicating strong opposition to release for each group of prisoners described. However, there is clearly variation available to analyze for treatment effects.

Figure 1 Attitudes Toward Prisoner Release

Notes: The scale for each of the individual items in panel A ranges from 0 to 4, with higher values indicating more support. The scale for the index measure ranges from 0 to 20.

Next, I examine whether there are differences in the level of support for prisoner release across the experimental conditions. Because there are multiple comparisons in the design, I report p-values from both naive t-tests and corrected for multiple comparisons using the Holm method for both dependent variables. Figure 2(a) plots the effect of each of the treatments compared to the no-treatment condition on the scale of support for prisoner release. ¹² The error bars show the 95% confidence interval around the point estimates. Figure 2(a) shows that the informational control and the no-treatment condition are statistically indistinguishable. This means that telling respondents the pandemic poses a public health threat in overcrowded jails and prisons and places prisoners at risk, and then asking them to think about how that makes them feel does not shift attitudes toward releasing prisoners relative to providing them with no information.

Figure 2 Perspective-taking increases support for prisoner release and mobilizes

Notes: Panel A shows the treatment effects on the release attitudes index. Panel B shows the treatment effects on writing a letter in support of release to the respondent’s sheriff. The treatment effects are compared to the no-treatment condition and include 95% confidence intervals around the estimates.

By contrast, perspective-taking increases support for prisoner release. Those who are asked to imagine that they were incarcerated score 0.842 points higher on the prisoner release scale compared to those who received no-treatment (p=0.042; p=0.126 with Holm correction) and 0.943 points higher relative to the informational control (p=0.023; p=0.103 with Holm correction). Respondents who imagined the person they are closest to was incarcerated were 1.539 points more supportive of prisoner release than those in the no-treatment condition (p=0.000; p=0.001 with Holm correction) and 1.640 points higher than the informational control (p=0.000; p=0.001 with Holm correction).

A comparison of the effect size to recent perspective-taking interventions is instructive. Williamson et al. (2021) report treatment effects in the range of 0.06 to 0.08 standard deviations. Here, relative to the informational control condition, support for prisoner release increased by 0.162 standard deviations in the egocentric perspective-taking condition and by 0.283 standard deviations in the surrogate perspective-taking condition. Compared to the no-treatment condition, support for releasing prisoners increased by 0.145 standard deviations in the egocentric perspective-taking condition and by 0.265 standard deviations in the surrogate condition. These findings indicate that even briefly imagining oneself or one’s loved one were incarcerated increases support for prisoner release. Moreover, Williamson et al. (2021) suggest that perspective-taking is best able to move affect toward outgroups, but often has limited effects on policy attitudes, which are less mutable. Here, however, I show that perspective-taking is able to significantly shift policy attitudes about prisoner release.

Online appendix table A5 disaggregates the scale into its component parts and shows quite consistent effects across all items except support for releasing all prisoners. For each of the subgroups of incarcerated people—immunocompromised, elderly, nonviolent, and awaiting bail—the egocentric and surrogate perspective-taking conditions increase support for release compared to the no-treatment condition. The effects are statistically significant for the egocentric condition for two of the categories (immunocompromised and elderly) and for the surrogate condition for all four of the categories. Due to the high proportion of zeros on the policy opinion scale, I also test the robustness of this result by creating a binned version of the variable in online appendix figure A1. This result replicates the finding here that perspective-taking increases support for prisoner release.

Changing policy attitudes is important politically, but motivating action might be harder. Therefore, the next question is whether the egocentric and surrogate perspective-taking treatments can elicit political mobilization on behalf of prisoners. I test this using the semi-behavioral measure of whether respondents chose to write a message to their sheriff about prisoner release.

Figure 2(b) shows the treatment effect on writing to their sheriff, where 1 indicates a message supporting release, –1 indicates a message opposing prisoner release, and 0 indicates any other respondent. As with policy opinion, there is no significant effect of the informational control on writing to one’s sheriff: on average, respondents in the no-treatment condition and the informational control condition were both more likely to write in opposition to release than in support. However, those who were in the perspective-taking conditions were more likely to write in support of release compared to those in the no-treatment condition. Compared to the no-treatment group’s mean score of -0.088, the average score for the egocentric perspective-taking treatment group is 0.060 (p=0.001; p=0.004 with Holm correction) and the average score for the surrogate perspective-taking group is 0.044 (p=0.002; p=0.013 with Holm correction). The differences between the perspective-taking treatments and the informational control miss statistical significance (egocentric: p=0.052 and p=0.188 with Holm correction; surrogate: p=0.113 and p=0.319 with Holm correction).

Another way to think about the effect of the perspective-taking treatments is in the proportions of each group who wrote a message for or against release. In the no-treatment group, 22% of respondents wrote an anti-release message and 14% of respondents wrote a pro-release message. In the perspective-taking groups, anti-release messages are less common (16% for egocentric and 16% for surrogate) and pro-release messages are more common (22% for egocentric and 20% for surrogate). As with the policy attitudes dependent variable, the effect sizes are substantial, with increases of 0.145 and 0.118 standard deviations for the egocentric and surrogate perspective-taking treatments, respectively, compared to the informational control, and an increase of 0.242 and 0.214 standard deviations compared to the no-treatment condition. Together these analyses indicate support for both hypotheses: egocentric and surrogate perspective-taking increase support for releasing prisoners and increase the likelihood of writing to one’s sheriff in support of prisoner release.

I test the robustness of the semi-behavioral finding both in terms of modeling and with two alternative specifications of the dependent variable. First, I test whether the findings are robust to an alternative model: ordered logistic regression. Online appendix table A6 shows that the results persist across model type. Using ordered logistic regression still produces statistically significant positive effects of both perspective-taking treatments on writing a letter in support of release. Second, I test two different alternative codings of the dependent variable. For the first alternative, instead of using the variable obtained from reconciling any disagreements among the coders, I use the modal code assigned by the six coders. The analysis in online appendix table A7 shows that the results persist. Second, I use the codes from the one coder who coded all of the open-ended statements. Again, online appendix table A7 shows nearly identical findings.

These results are striking. Asking white Americans to imagine that either they or their loved ones were incarcerated can shift attitudes to be more in favor of releasing prisoners and provoke mobilization in support of incarcerated people.

Who Responds?

While the average treatment effects show clearly that respondents are more supportive of prisoner release and more likely to mobilize in support after a perspective-taking exercise, it is possible that some groups of respondents are more or less responsive to the treatments. Moderation analysis is beyond the theoretical expectations I developed, but readers might be curious about some likely heterogeneous treatment effects. I argued that support should increase for prisoner release due to the empathy induced by perspective-taking; an implication of this argument is that the effects should be strongest among those whose priors make them less likely to support release ex ante. Predispositions fundamentally shape public opinion (Alvarez and Brehm 2002). In the context of an intervention designed to provoke empathy, this suggests that those who are most likely to respond are those who are least likely to already feel empathetic toward the target group. That is, the task is required to produce empathy because it is less likely to pre-exist.

Empathy requires both understanding and sharing a target’s emotional state (Melloni, Lopez and Ibanez 2014), and some people are predisposed to feel empathy without a treatment intervention. In the context of incarceration, we would anticipate that people with personal or proximal carceral contact would tend to feel empathy regardless of treatment assignment; simply being asked about prisoner release is likely to make them imagine those in their life (themselves or close social ties) who have had carceral contact, which might produce empathetic feelings. By contrast, people who have not personally or vicariously experienced incarceration would be more responsive to an empathy-inducing exercise relative to a no-treatment condition.

Similarly, we might expect partisan differences in treatment effects. The contemporary period has been marked by partisan polarization in American political behavior (Iyengar and Westwood 2015; Iyengar et al. 2019). Democrats and Republicans have divergent views on a variety of issues generally, and exhibited starkly different attitudes toward COVID, race, and criminal justice during this time (Anoll, Engelhardt, and Israel-Trummel 2022; Drakulich and Denver 2022; Pickup, Stecula, and van der Linden 2020; Reny and Newman 2021). More broadly, partisanship captures a bundle of pre-existing attitudes and values, such as authoritarianism and racial attitudes (Engelhardt 2021a; Engelhardt, Feldman, and Hetherington 2023; Jardina and Ollerenshaw 2022; Sides, Tausanovitch, and Vavreck 2022). These partisan differences are reinforced by partisan elite rhetoric, which differs sharply on questions of race (Engelhardt 2021b). The attitudes and values bundled into partisan identity are likely to indicate empathy with people who are incarcerated. Given these partisan differences, Republicans might respond more strongly to the perspective-taking treatments. Essentially, Democrats had already received messaging to move toward less punitive policy positions during 2020. In an experimental treatment that does not explicitly touch on partisan issues and only relies on perspective-taking, Republicans may therefore be more responsive.

I examine these potential heterogeneous effects in figures 3 and 4. ¹³ Figure 3 explores heterogeneous average treatment effects by prior carceral contact. I code respondents as having carceral contact if they indicate either that they have spent any time in jail or prison or if they have a close friend or family member who has spent time in jail or prison in the past five years. There are 592 respondents with carceral contact and 943 respondents without contact. The left panel shows that there are significant differences in policy attitudes between respondents with carceral contact and without carceral contact within each treatment group. Those with contact are consistently more supportive of prisoner release than those without contact. We can also see that both forms of perspective-taking have positive effects relative to the no-treatment and informational control conditions. However, for those with carceral contact, the confidence intervals on both perspective-taking treatments show that there is not a significant difference across treatment conditions. That is, among those with carceral contact, those who are assigned to engage in perspective-taking and those who are not exhibit similar attitudes toward the policy of releasing prisoners due to COVID.

Figure 3 Conditional treatment effects by carceral contact

Notes: The plots show the expected values from linear regression models interacting the treatment condition with an indicator for whether the respondent has carceral contact and include 95% confidence intervals around the estimates. Carceral contact is determined by whether respondents indicate that they 1) have spent any time in prison or jail or 2) have a close friend or family member who has spent time in prison or jail in the past 5 years. Anyone who answered yes to either of these questions is coded as having carceral contact, and all other respondents are coded as not having carceral contact.

Figure 4 Conditional treatment effects among partisans

Notes: The plots show the expected values from linear regression models interacting the treatment condition with an indicator for whether the respondent is a Democrat or Republican and include 95% confidence intervals around the estimates. Partisanship is coded from the standard 7-point partisan identity scale with leaners coded as partisans.

Among those without carceral contact, we see significant differences by treatment. Respondents assigned to the egocentric perspective-taking condition are significantly more supportive of prisoner release (effect of 1.220, p=0.014 compared to the no-treatment condition; effect of 1.113, p=0.026 compared to the informational control). Similarly, respondents assigned to the surrogate perspective-taking condition are also significantly more supportive of prisoner release (effect of 1.980, p=0.000 compared to the no-treatment condition; effect of 1.873, p=0.000 compared to the informational control).

The semi-behavioral outcome in figure 3(b) shows a similar pattern. The perspective-taking treatments do not increase the likelihood that someone with carceral contact writes to their sheriff, but they do increase the likelihood that someone without contact does so. Among those without carceral contact, the predicted likelihood of writing to one’s sheriff in the perspective-taking conditions is significantly higher than in either the no-treatment or informational control condition. ¹⁴ Most interestingly, respondents who didn’t have carceral contact and were assigned to one of the perspective-taking conditions wrote to their sheriff at statistically indistinguishable rates from respondents with carceral contact. Carceral contact, of course, is not randomly assigned. But this analysis shows that the treatment effects seem to work as I hypothesized they would among those who do not have carceral contact, and do not significantly change attitudes or behaviors among those with contact.

Finally, I turn to partisanship. I code so-called “leaners” as partisans, which yields 628 Democrats and 728 Republicans. Figure 4(a) shows significant gaps between partisans. Democrats are more supportive of prisoner release than Republicans across each treatment group. Among both Democrats and Republicans, the informational control has no effect on support for releasing prisoners relative to the no-treatment condition. Similarly, for Democrats the egocentric perspective-taking condition has no effect on support for prisoner release. The only statistically significant effect on policy opinion among Democrats is the surrogate perspective-taking condition compared to the no-treatment condition, where attitudes are expected to shift 1.386 points more in favor of release (p=0.024).

For Republicans, the positive treatment effect of egocentric perspective-taking on policy opinion just misses statistical significance compared to the no-treatment condition (p=0.062) but is significant compared to the informational control (p=0.024), where it boosts support for prisoner release by 1.450 points. The surrogate perspective-taking condition also appears to significantly boost support for prisoner release among Republicans. Republicans in this treatment are expected to score 2.036 points higher on the prisoner release scale compared to the no-treatment group (p=0.002) and 2.228 point higher than the informational control group (p=0.000).

When it comes to writing to their sheriff, we again see gaps between the partisans in figure 4(b). Across conditions, the mean score in each treatment group among Republicans is negative. This means that in each group, Republicans are more likely to write a message against prisoner release than in favor of release. By comparison, Democrats have positive scores across treatment conditions, though the estimate is nearly zero in the no-treatment condition. Further, it seems that Democrats react positively to the informational control. Democrats in this condition score 0.124 higher on the semi-behavioral outcome compared to the no-treatment condition (p=0.058). Democrats also respond positively to the perspective-taking conditions. Democrats in the egocentric condition are expected to score 0.132 points higher on the –1 to 1 scale compared to the no-treatment condition (p=0.033) and those in the surrogate condition are expected to score 0.153 points higher (p=0.019). This suggests that while Democrats’ attitudes toward prisoner release might not change as a result of either hearing about poor conditions in prisons or egocentric perspective-taking, these interventions can mobilize them to take action in support of inmates. Further, it seems that perspective taking is not necessary for Democrats: informing them about poor prison conditions is sufficient to induce mobilization.

Unlike Democrats, Republicans do not change their behavior in the informational control relative to the no-treatment condition. However, they do show positive treatment effects with the perspective-taking conditions. Republicans in the egocentric condition are expected to score 0.117 points higher on the –1 to 1 scale compared to the no-treatment condition (p=0.087) and those in the surrogate condition are expected to score 0.136 points higher (p=0.033). The confidence intervals around the estimates for Republicans in both perspective-taking groups encompass 0, indicating that for Republicans the perspective-taking treatments seem to move them away from voicing their opposition to prisoner release, and toward quiescence, on average. Republicans who perspective-take are also indistinguishable on the semi-behavioral measure from Democrats in the no-treatment condition.

The conditional treatment effects clarify who is most responsive to the perspective-taking interventions. Mentally stepping into the shoes of someone who is incarcerated or whose loved one is incarcerated promotes support for releasing prisoners and mobilization on their behalf, consistent with existing work that argues proximal contact can promote a type of surrogate political mobilization. The positive effects of the treatment conditions appear to be stronger among those whose priors make them least likely to support prisoner release. That is, people who are most likely to already feel empathy with incarcerated people don’t tend to change their attitudes as much when they go through a perspective-taking exercise. Rather, those who are least likely to already have empathy react most strongly to the exercise and become more likely to support release and to write to their sheriff in support of release. ¹⁵

Discussion

Over the past 70 years the United States has substantially expanded its carceral state such that it is now an international leader in human caging. Retreating from this would demand changes well beyond decreasing prison entries, and likely would require releasing people currently incarcerated (Ghandnoosh 2023). Political science has begun to document the harms associated with a sprawling carceral state (e.g., Epp, Maynard-Moody, and Haider-Markel 2014; Lerman and Weaver 2014) as well as the broad-based public support for these institutions (e.g., Enns 2016). This paper shows that white Americans’ support for releasing prisoners can be increased and mobilized using perspective-taking. In particular, I show that both egocentric and surrogate perspective-taking—imagining a loved one in jail—shift attitudes more in favor of prisoner release and increase the probability that respondents write to their sheriff to support release. This finding extends the scholarship from showing that surrogate perspective-taking can induce empathy to demonstrating that it can also provoke changes to political attitudes and behaviors. Moreover, this finding is politically meaningful as many more people are vicariously connected to the carceral state than immediately impacted; nearly half of Americans have had a family member incarcerated (Enns et al. 2019). It is, of course, important to note that attitudes do not shift when respondents are told that people in prison are at risk. Their attitudes only soften and they are only mobilized when they engage in perspective-taking. This suggests that people are more inclined to help people they are closer to, and indicates the challenges faced by any social movement for decarceration.

It is important to note that the data includes only those who completed the survey. It’s possible that the effects would not replicate among those who are less engaged survey-takers. However, the findings do suggest at least that among the more engaged public, perspective-taking can promote support for releasing prisoners. This matters because more engaged citizens may be better represented (Leighley and Oser 2018).

A skeptic might wonder if public opinion about incarceration matters for outcomes. But scholarship demonstrates convincingly that incarceration is not simply an inevitable result of crime but is a political choice (Eubank and Fresh 2022). The public has exerted incredible political power in the growth of the carceral state, and politicians respond to the public’s demand for punishment (Enns 2014; Nicholson-Crotty, Peterson, and Ramirez 2009). Public opinion will therefore be essential to any political efforts to dismantle punitive institutions. Moreover, public opinion isn’t static; it responds to elites, social movements, and political events. While historically the U.S. public has been relatively punitive, we may be witnessing a weakening of the punitive consensus.

Questions remain. This experiment was conducted at a unique moment due to both a worldwide pandemic and the largest American social movement to date, which drew attention to racial justice and disrupted whites’ attitudes and political practices (Anoll, Engelhardt, and Israel-Trummel 2024; Reny and Newman 2021). It’s possible that absent such disruptions to politics, it would not have been possible to shift respondents’ attitudes. We may even have returned to the earlier status quo already. Recent analysis indicates that 2020 was not a “racial reckoning” as backlash and counter-movements quickly followed. Initial steps to decrease police budgets did not last and white support for Black Lives Matter declined (Hoang and Benjamin 2024; Jefferson and Chudy 2021). This push-pull of movements and counter-movements reflects the longstanding history of competing racial orders: efforts to increase racial egalitarianism are frequently met with the reassertion of white racial supremacy (King and Smith 2005). We might view the results in this experiment pessimistically in this context; a single-shot intervention is never likely to last. However, we build our knowledge through the accumulation of small findings. Even a fleeting treatment effect can offer a glimpse into how attitudes might change or be mobilized. The finding here, that white Americans can become more supportive of decarceral efforts, suggests a potential political strategy for those who might seek to dismantle the carceral state.