INTRODUCTION
At the end of 2022, 108 million people were living in forced displacement, having abandoned their homes due to violence or natural calamities (UNHCR 2023). The internally displaced persons (IDPs)—those who seek shelter within their countries’ borders—made up 71 million of this total; the highest number ever recorded (IDMC 2023a). More than two-thirds of all IDPs live in just ten countries, with Syria, Afghanistan, the Democratic Republic of the Congo (DRC), Ukraine, and Colombia among the worst affected. The scale of the internal displacement problem is large, and it has been growing in recent years.
Our aim in this study is to advance the understanding of why people volunteer to host the internally displaced. Following the recent psychological turn in this literature, we are especially interested in whether higher empathy might be correlated with a greater willingness to host. We also consider alternative explanations, given that hosting can be conceptualized not only as altruistic behavior but also as a type of cooperation if there is an expectation of reciprocity through, for example, labor or future offer of shelter.Footnote 1 Thus, we also consider the role of co-ethnicity, links to local authority figures, the wealth of the hosting family, security concerns, and religiosity of the potential hosts in informing hosting decisions.Footnote 2 The emphasis on empathy and consideration of the alternative factors were pre-registered as hypotheses.Footnote 3
The study took place in the context of internal displacement in the Democratic Republic of the Congo. The DRC has been experiencing prolonged conflict and is currently home to the third largest population of IDPs globally, with 5.7 million people—about 6% of the population—seeking refuge from violence within the country’s borders (IDMC 2023b). In a methodological advance on the existing literature, we measured the characteristics of potential hosting households before the arrival of IDPs. To do this, in 2019, we identified a research site in eastern DRC that was likely to receive an influx of displaced people in subsequent months and set out to survey all dwellings in 15 villages in that region. This allowed us to avoid the common trap of ex post rationalization of hosting decisions. Also, unlike most other studies in this literature, we measured empathy directly, modifying an empathy scale from psychology (Chatruc and Rozo Reference Chatruc and Rozo2021; Newman et al. Reference Newman, Hartman, Lown and Feldman2015). In another innovation, we measured hosting behavior not through self-reports of hosting or stated willingness to do so but through verifiable village chiefs’ reports. These reports covered a 10-month period after the initial survey. We believe our measure to be considerably more reliable than self-reporting, which is subject to a strong social desirability bias. Over the 10-month period, 24% of the 1,504 households surveyed started hosting IDPs. The majority of incoming IDPs are of the same ethnicity as most potential hosts, but there is much variation in ethnicity, both within the hosting villages and among the displaced.
The scholarly understanding of the multi-faceted problem of forced displacement remains limited. Much of the existing literature focuses on the causes of flight and the logic of where the displaced go. Early cross-national work argued that people flee conflict as the threat of violence increases (Moore and Shellman Reference Moore and Shellman2004). Other studies have shown that the displaced are more likely to stay within their country’s borders if neighboring countries are poor and undemocratic (Moore and Shellman Reference Moore and Shellman2006). Another important strand of the literature argues that the arrival of the forcibly displaced imports conflict into host communities by exacerbating ethnic and sectarian tensions, increasing competition over scarce jobs, or bringing weapons into the community (Salehyan Reference Salehyan2008; Salehyan and Gleditsch Reference Salehyan and Gleditsch2006; Tumen Reference Tumen2016). More recently, scholars have observed that while the short-term impact of the arrival of displaced persons might be net negative, in the long-term, the displaced can have a positive impact on local economies through the expansion of trade links, an increase in local human capital, and the influx of humanitarian assistance (Maystadt et al. Reference Maystadt, Hirvonen, Mabiso and Vandercasteelen2019; Verme and Schuettler Reference Verme and Schuettler2021; World Bank 2022; Zhou and Shaver Reference Zhou and Shaver2021).
Where the literature falls short is in addressing a fundamental question with regards to the dynamics of forced displacement, namely, in answering which factors explain the willingness of potential hosts to open their doors to the displaced and, from a policy perspective, in explaining how to encourage more hosting. This is particularly important for the internally displaced because the vast majority of IDPs are accommodated not in refugee camps, but in host communities, among other co-nationals (UN 2021). Having people stay within their own countries and in local communities, once forcibly displaced, might be desirable because this form of displacement minimizes the trauma of uprootedness.
In recent years, several pioneering studies have examined the correlates of attitudes and helping behaviors via-à-vis the forcibly displaced in host societies. In the context of refugees from Côte d’Ivoire fleeing to Liberia, Hartman and Morse (Reference Hartman and Morse2018) found that those in hosting communities who themselves had experienced violence were more likely to open their doors to the displaced. They dubbed this the “empathy born of violence” hypothesis and found additional supporting evidence for it in the context of Syrian IDPs being sheltered by other Syrians (Hartman, Morse, and Weber Reference Hartman, Morse and Weber2021). However, in a study among Lebanese hosts of Syrian refugees, Ghosn, Braithwaite, and Chu (Reference Ghosn, Braithwaite and Chu2019) uncovered no evidence that prior experience of violence increases positive predisposition toward the displaced. Thus, the specific pathway to more hosting remains in contention in this nascent literature.
Scholars have also studied attitudes toward refugees in Western societies. While most of this work focuses on relatively low-stake outcomes—like expressions of willingness to help and contributions in behavioral games—the findings do suggest that empathy might be one of the primary determinants of helping behavior (Adida, Lo, and Platas Reference Adida, Lo and Platas2018; Bansak, Hainmueller, and Hangartner Reference Bansak, Hainmueller and Hangartner2016; Williamson et al. Reference Williamson, Adida, Lo, Platas, Prather and Werfel2020).Footnote 4 However, a minority of the forcibly displaced are refugees, and only a fraction of these aim to and succeed in reaching Europe or the United States. Furthermore, the act of hosting strangers in one’s home is an extreme form of altruism or cooperation—it can be disruptive or even dangerous if guests prove violent or dishonest—and is therefore a different type of behavior from charitable contributions or support for petitions and requires scholarly attention in its own right.
In this study, linking pre-displacement characteristics of potential hosts to subsequent hosting behavior in the DRC, we find that empathy is the most important correlate of the willingness to open one’s doors to the displaced. There is a 20 percentage-point difference in the likelihood of hosting between the most and least empathic respondent in our sample. In exploring the correlates of empathy, we find limited support for the idea that experience of past violence makes individuals more empathetic, consistent with the empathy born of suffering hypothesis. Other factors that matter in explaining hosting decisions are security considerations—households headed by men are more likely to open their doors—and connection to authority figures, whereby those related to the village chief are more likely to accommodate IDPs. The effect magnitudes for these variables, however, are considerably smaller than for empathy. Contrary to existing work on other types of altruistic and cooperative giving, ethnicity, wealth, religiosity, and expectations of strategic benefits from IDPs are not correlated with hosting decisions.
To better understand the mechanism by which IDPs are matched with hosting families and to contextualize the findings, we conducted semi-structured interviews with five village chiefs and 150 randomly selected households in a random subset of five study villages. The interviews clarified that, in seeking shelter, IDPs approach household heads more or less at random, and that there is no formal matching process; the village head is informed of the newly arrived IDPs, but does not himself arrange hosting. The interviews also confirm that empathy was the main reason for hosting, whereas ethnicity did not play an important role.
From a policy perspective, it is important not only to understand the correlates of hosting but also to learn how to encourage people to host more. To get at this we designed an experiment where we primed cognitive empathy through a perspective-taking exercise, encouraging respondents to imagine what it would be like to be displaced by asking them where they would go and what they would take with them if forced to leave home (modeled after Adida, Lo, and Platas Reference Adida, Lo and Platas2018). To contextualize the importance of empathy relative to other factors we also separately primed religiosity and obedience to authority. Those randomized into the religious appeal intervention were visited by a community religious leader, who reminded them that it is a moral duty to help people in need. In the authority appeal, participants were visited by a local authority figure, the village chief, who stressed the importance of assisting potential IDPs who might arrive. To also test the importance of ethnicity to hosting decisions we implemented the experiment as a factorial design, whereby a random half of respondents were primed that the incoming IDPs would likely be of their own ethnic group, and others were told that the displaced would likely be of a different ethnicity.
Findings from the experiment suggest that it might be difficult to increase the willingness to host in the longer term. We found that neither the perspective-taking exercise nor the appeals to religion or authority affected hosting behavior relative to an untreated control in a setting where, on average, four months passed between the appeals being administered and the arrival of the displaced. This finding confirms the intuition from existing work on assistance to refugees in the U.S. that the effect of perspective-taking interventions might be short-lived (Adida, Lo, and Platas Reference Adida, Lo and Platas2018).
In terms of this article’s contributions, we are the first to measure the potential correlates of hosting before the arrival of the displaced, thus reducing the bias associated with the ex post rationalization of hosting decisions, and to measure hosting behavior in a way that minimizes misreporting due to social desirability bias. While our findings highlight the importance of empathy to extreme acts of helping like hosting, we are skeptical about policy makers’ ability to increase empathy levels in the population in the longer term using simple interventions. Given that the literature on refugee assistance in developed economies has already hinted at the relevance of empathy to helping decisions, we expect our findings to apply to the dynamics of refugee hosting in Western countries as well. It is a subject for future research how much empathy might affect other helping behaviors, including in everyday interactions outside of the context of assistance toward the forcibly displaced. Our findings also shed light on the ongoing debate whether past experience of violence results in pro-social behavior through post-traumatic growth (Blattman Reference Blattman2009) or parochial altruism in response to trauma (Bauer et al. Reference Bauer, Cassar, Chytilová and Henrich2014). The findings suggest that particularly empathetic individuals might reconceptualize who counts as an in-group member away from shared ethnicity toward shared victimhood.
LITERATURE AND HYPOTHESES
The role of empathy in explaining hosting is of particular theoretical interest in this study and something that we pre-registered as the primary correlate of hosting. Psychologists have long argued that there is a “giving type,” a person who is more empathetic and therefore more likely to engage in altruistic behavior (Batson Reference Batson2002; Hoffman Reference Hoffman2000), and noted that empathy reduces prejudice in interactions across ethnic and other group lines (Batson et al. Reference Batson, Chang, Orr and Rowland2002; Stephan and Finlay Reference Stephan and Finlay1999). This is because in empathizing with the suffering other, the affected individual develops an appreciation for how unjustly their interlocutor has been treated, and this reduces prejudice toward the other and enhances the desire to help. When the same emotions are shared by members of different groups, a bond of commonality might form between them. Consistent with common ingroup identity theory, we might expect that a new supraordinate identity would emerge between the person helping and the one who is receiving assistance (Gaertner and Dovidio Reference Gaertner and Dovidio2000).
Early studies on attitudes toward the displaced, primarily on refugees from developing countries in developed states, found that empathy is an important predictor of positive attitudes toward those in need. Newman et al. (Reference Newman, Hartman, Lown and Feldman2015) noted that respondents with high levels of empathy are less supportive of restrictive immigration policies. Bansak, Hainmueller, and Hangartner (Reference Bansak, Hainmueller and Hangartner2016) argued that Europeans are heavily motivated by humanitarian concerns, as they are most likely to accept vulnerable asylum seekers fleeing persecution. Subsequent studies have focused more on the cognitive dimension of empathy. The literature in psychology suggests that empathy has a cognitive and an affective component (Eisenberg, Fabes, and Spinrad Reference Eisenberg, Fabes and Spinrad2007). Whereas affective empathy is about a person feeling the same emotions as their interlocutor, cognitive empathy concerns an ability to understand what the other person is feeling without necessarily experiencing the same emotion. Studies have sought to prime the cognitive dimension of empathy through perspective-taking exercises by encouraging respondents to imagine refugees’ thoughts and feelings and by presenting participants with personal narratives of the displaced. Such exercises have been found to engender inclusionary behavior toward the displaced, heighten support for more liberal immigration policies, and reduce prejudice (Adida, Lo, and Platas Reference Adida, Lo and Platas2018; Alan et al. Reference Alan, Baysan, Gumren and Kubilay2021; Audette, Horowitz, and Michelitch Reference Audette, Horowitz and Michelitch2020; Chatruc and Rozo Reference Chatruc and Rozo2021; Simonovits, Kézdi, and Kardos Reference Simonovits, Kézdi and Kardos2018; Williamson et al. Reference Williamson, Adida, Lo, Platas, Prather and Werfel2020).
It remains an open question whether an extreme form of helping like hosting strangers in one’s home in the Global South is subject to the same set of determinants as less costly types of refugee assistance in Western countries. Encouragingly, the little work that exists on the correlates of hosting the forcibly displaced also attributes an important role to empathy. In their study in Syria, Hartman, Morse, and Weber (Reference Hartman, Morse and Weber2021) hypothesize that empathy is an important correlate of hosting, although they do not measure empathy directly. Hartman and co-authors argue that empathy is triggered by past experience of violence, and that it is those who had suffered violence who are most likely to host (see, similarly, Hartman and Morse Reference Hartman and Morse2018).Footnote 5 The empathy born of violence hypothesis is challenged by Ghosn, Braithwaite, and Chu (Reference Ghosn, Braithwaite and Chu2019), who in their study on the correlates of assistance toward Syrian refugees in Lebanon find no evidence that past exposure to violence leads to more positive attitudes toward the displaced. Instead, Ghosn and co-authors argue that it is past inter-group contact that facilitates positive predispositions toward those forcibly displaced. This claim does not challenge the relevance of empathy to hosting but rather stipulates a different pathway behind the effect.
In this study, we test whether empathetic individuals are more likely to host, including across ethnic group lines. Consistent with psychological theories about altruism, we hypothesize that those with higher empathy will be more likely to accommodate IDPs (H1a). Further, in line with the “empathy born of violence” argument we hypothesize a positive relationship between past experience of violence and empathy (H1b). Consistent with the literature, we expect cognitive empathy to be associated with a higher propensity for hosting.
While the theoretical focus of this study is on the role of empathy in hosting, we also consider alternative explanations. Hosting can be thought of as a form of cooperation if there is an expectation of reciprocity in the future. Dominant explanations for cooperative behavior focus on the cost-benefit calculations of social actors. Studies in this tradition tell us that ethnicity is an important determinant of cooperation: members of the same ethnic community are more likely to help each other because of natural affinities of language and taste, and because negative reputational effects of shirking might be stronger among closely networked co-ethnics (Habyarimana et al. Reference Habyarimana, Humphreys, Posner and Weinstein2009). Extending this logic, we might expect that those who stand to benefit strategically from giving, either because they expect something in return or because they are particularly visible in their community and therefore especially sensitive to reputation effects, will also cooperate and give more. This leads us to hypothesize that shared ethnicity should be one of the main correlates of the willingness to host (H2a). Likewise, we expect that those individuals who are most closely networked with authority figures with powers of punishment and reward in their communities will be more likely to host (H2b). We also hypothesize that those who think that hosting will bring economic rewards, either through the use of IDPs as cheap labor or by maximizing the chances of receiving aid from non-governmental organizations (NGOs), will be more likely to host (H2c).
If we consider hosting as an act of altruism it is reasonable to expect that one’s willingness to host might be constrained by the characteristics of one’s household. For instance, literature on charitable giving finds that the wealthy are more likely to donate and that people are less likely to help refugees as the cost of giving increases (Meer and Priday Reference Meer and Priday2021). Thus, we hypothesize that wealthier households will be more likely to host (H3a). Some authors have argued that those who have internalized the moral precepts of religious teachings become more altruistic (Brooks Reference Brooks2006; Putnam and Campbell Reference Putnam and Campbell2010), which leads us to conjecture that more religious households will be more likely to host (H3b).Footnote 6 Based on our field experience at the study site, we also expect that families with male household heads will feel more physically secure and will therefore be more likely to accommodate strangers (H3c).
CONTEXT: VIOLENCE AND DISPLACEMENT IN THE DRC
We set out to test these hypotheses in the context of the Democratic Republic of Congo. The DRC is a setting that is representative of a prolonged conflict leading to bouts of forced displacement. The DRC has been among the top three countries globally by the number of new displacements over the past five years (IDMC 2023b). There are over 250 ethnic groups in the country, and conflict is exacerbated by the multiplicity of ethnic cleavages. Fighting has continued for more than three decades, and though the Second Congo War ended in 2003, violence is still endemic, especially in the east. In 2022 alone, the number of IDPs due to conflict in the DRC increased by four million individuals, second only to Ukraine. The vast majority of IDPs do not reside in camps or larger cities but take refuge with host families in rural settlements (UNHCR 2021). The dynamics of displacement in the DRC make it similar on key dimensions to other countries with large displaced populations like Yemen, South Sudan, Northern Nigeria, and Afghanistan.
The study is set in eastern DRC, in the Kalehe region of the South Kivu province, which has seen sustained violence in the recent past. In 2019, the year when this study was in the field, South Kivu recorded some 400,000 new displacements (IDMC 2020). The Kalehe region is multi-ethnic. The Havu are the majority in the Buhavu chiefdom, and the smaller Buloho chiefdom is home to the Tembo. The study villages are located within the Mbingu South grouping in the Kalehe region (Figure 1). We selected this area because of a high likelihood of IDP inflows from the neighboring Kalehe highlands where armed combatants from the Conseil National pour la Restauration de la Démocratie (CNRD), a dissident wing of the Democratic Forces for the Liberation of Rwanda (FDLR), moved in large numbers shortly before fieldwork began. In November 2019, about two months after we conducted the household survey, the Congolese army launched a military offensive against the CNRD causing the displacement of thousands of civilians, including into our research sites. In Supplementary Appendix A, we provide a more detailed account of the conflict’s history and associated displacement patterns, as well as information on ethnic relations in the area and a description of what hosting arrangements entail in these communities. The process by which IDPs find hosts in a given community is described in Supplementary Appendix B.
Map of the Research Area
Note: Authors’ rendering. (A) The DRC, with South Kivu highlighted. (B) South Kivu, with Kalehe territory highlighted. The capital city of South Kivu—Bukavu—is also indicated. (C) Kalehe territory, the Buhavu and Buloho chiefdoms, the Kalehe highlands, and the study villages.
DATA AND EMPIRICAL STRATEGY
Village Selection and Data Collection
The project began in July 2019 with a visit to the prospective field site for an explorative survey. All 94 villages of the Mbinga South grouping were visited, and the field team met with village chiefs and collected data on village characteristics.Footnote 7 Twenty-one villages were found to fit the criteria for the study—there, cellphone reception was adequate and some displaced families had arrived in the preceding three months.Footnote 8 Of this total, fifteen villages were selected at random for inclusion in the project. Then, we met again with the village chief and the village council in the selected villages to explain the project in more detail and secure their approval to proceed. The field team also drew up a list of all households and dwellings in each village.
Next, the survey was fielded in September 2019 with the aim of interviewing every household head in all fifteen villages to collect pre-displacement information on household characteristics and to measure empathy levels among respondents. The experiment was embedded in the survey.Footnote 9 With the survey completed, we asked the village chiefs to keep the record of incoming IDPs and which households hosted them for the next 10 months. We told the chiefs that we would stay in regular contact and that the information they provided would be verified. To facilitate communications, we gave each chief a cellphone, which was theirs to keep, and also provided weekly top-up credit for the study’s duration.Footnote 10
The project’s field coordinator phoned the chiefs once every 2 weeks to discuss hosting dynamics and remind them to maintain records. The initial follow-up visit took place 4 months into the study. By then the conflict in the Kalehe highlands had flared up and IDPs were arriving in substantial numbers. The field team met with every village chief, discussed each of the entries, and digitized the records on incoming IDPs. We also visited five households selected at random from every chief’s list to check the accuracy of the hosting records; we found no discrepancy with the chief’s information. The final follow-up to collect the records on hosting dynamics—which was not pre-registered but became possible due to extra funding—was completed ten months into the study in July 2020; because of the Covid-19 pandemic, this was done over the phone.
While not pre-registered, we visited the field site once again in October 2021 for qualitative fieldwork to learn more about how the match between the IDPs and hosts took place and to contextualize the results of the quantitative analyses. We were especially interested to learn how the hosts and the hosted understood what motivated people to open their doors. Qualitative interviews with village chiefs and 150 household heads were completed in five randomly selected villages from the original sample of 15; we interviewed ten hosts, ten hosted, and ten non-hosting heads of households in each village. We describe highlights from these interviews in the results section and provide detailed information on the set-up and findings of qualitative fieldwork in Supplementary Appendix B. The timeline of field activities is presented in Supplementary Appendix C.
Sample
The fifteen study villages contain 1,660 dwellings. In the survey, we collected information from 1,504 dwellings.Footnote 11 We focus on the households that own the dwelling as they make the decision whether to host the IDPs. In the ten month-period following our survey, 1,274 new incoming displaced individuals were hosted among 386 of these households in the study villages, and 354 of these were captured in our survey.
The characteristics of potential hosts are summarized in Table 1.Footnote 12 A typical respondent is 43 years old. About half (49%) can read and write, and 60% of respondents are born in the village. Most respondents (71%) are Protestant, 19% are Catholic, and 10% follow other religions. A typical household consists of eight household members and has a dependency ratio of 54%.Footnote 13
Descriptive Information about Potential Hosting Households
Note: Components of the asset index: number of goats, poultry, houses, rooms, chairs, beds, foam mattresses, motorcycles, machetes, pots, cupboards, radios, and phones. For the strength of ethnic attachment and empathy measures, respondents were asked to score each item on a four-point Likert scale from “Strongly disagree” (0) to “Strongly agree” (3). Exposure to violence relates to the 12 months preceding the survey. Household size of 34 is an outlier; results hold when this observation is excluded from analyses.
Potential Correlates of Hosting
We hypothesized that individuals who are more empathic are more willing to host.Footnote 14 To measure empathy, we use a modified version of the Basic Empathy Scale. The full scale consists of 20 items (Jolliffe and Farrington Reference Jolliffe and Farrington2006). We designed a truncated six-item scale containing measures that in previous studies have been shown to correlate strongly with cognitive and affective empathy (see Supplementary Appendix D for details). Subsequently, we dropped one of the items because it was the only statement phrased in the negative and did not load in the same way as all the others in a confirmatory factor analysis.Footnote 15 The final scale that we use contains the five items reported in Table 1; the design of this scale allows us to separately measure affective and cognitive empathy. Higher scores denote higher levels of empathy. In the analyses, we use an additive index of the five-item scores. The average respondent is quite empathic with a score of 9 out of 15.
To allow us to test whether past exposure to violence might result in higher empathy we asked respondents how much and what type of violence they experienced during the preceding 12 months. The overall levels of violence exposure are high with almost everyone reporting that they had been afraid that their village would be attacked by an armed group (90%) and many stating that they had witnessed armed violence in their village (72%). These high levels of violence exposure are consistent with findings in other studies in this part of the DRC (Stoop and Verpoorten Reference Stoop and Verpoorten2021; van der Windt and Humphreys Reference Van der Windt and Humphreys2016). We will use the experience of having one’s home ransacked as our primary measure of violence exposure. There is meaningful variation on this measure (61% say that they experienced ransacking), and home ransacking, unlike targeted violence, might be plausibly exogenous to household characteristics.Footnote 16
When it comes to testing alternative explanations for hosting, there is variation on ethnicity in the sample. In the study villages, Havu are the largest ethnic group, comprising 69% of household heads. 14% of household heads are Tembo, and 10% are Shi (Table 1). Among the displaced that arrived during the 10-month study window, 31% are Havu, 23% are Kinyarwanda speaking (i.e., Tutsi or Hutu), 21% are Shi, and 17% are Tembo. To get at the relative importance of ethnicity among potential hosts we asked respondents how strongly they identify with their ethnic group. With a mean of 6.39 on a 0–9 scale, in-group bonds appear to be very strong. A related alternative hypothesis has to do with exposure to oversight from authority figures. We measure this by asking about the respondents’ connection to the village chief; these connections are defined broadly as family relations and friendship. 52% of household heads report some form of relation to the chief. In terms of strategic benefits from hosting, a little over a quarter of respondents think either that IDPs are a source of cheap labor or that hosting increases the likelihood that hosts might receive aid from NGOs.
The final set of alternative hypotheses has to do with the characteristics of hosting households: their wealth, religiosity, and physical security. We measure respondents’ wealth by recording the quality of construction materials of their house and via a factored index of the household’s possessions from farm animals to means of communication and transportation. An average household is quite poor with a decent roof but walls made of low-quality materials, like soil and straw, and no means of transportation. To measure religiosity we asked respondents about the importance of church in their daily lives and inquired how often they go to church. Consistent with expectations, respondents are generally very religious. To get at the underlying sense of security we recorded whether the head of the household is male—this is the case in 76% of households.
Empirical Strategy
To examine the correlates of hosting we estimate the following model:
$$ {Y}_{ij}={\beta}_0+{\varGamma X}_{ij}+\varLambda {W}_{ij}+{\alpha}_j+{\varepsilon}_{ij}, $$
where the indicator variable
$ {Y}_{ij} $
is equal to one if household i in village j started hosting in the 10-month period after the survey. Xij is a vector containing the study’s variables of interest: empathy, ethnicity, authority, perceived benefits, wealth, religiosity, security, and conflict exposure.
$ {W}_{ij} $
is a vector containing demographic controls. We include village fixed effects, αj, to control for differences in observable and unobservable predictors across villages. That is, we effectively control for any factor at the village level and higher that may explain hosting behaviors, such as the size of the IDP inflow and village-level governance dynamics. Standard errors are clustered at the village-level to account for within-village correlation of the residuals. The empirical model and control variables were pre-registered.
RESULTS
Correlates of Hosting
During the 10-month period after household surveys had been fielded, 24% of households started hosting newly arrived IDPs. A small number of IDPs (11%) already knew their hosts from before. We exclude hosting relationships based on prior acquaintance from the analyses because, substantively, we are interested in why people open their doors to strangers; this leaves us with 21% of households that started hosting strangers.Footnote 17
The main results exploring the correlates of hosting are reported in Table 2. Model 1 does not contain any controls, model 2 includes village fixed effects, and model 3—our preferred specification—includes both village fixed effects and demographic controls for age, literacy, being native to the village, religious denomination, household size, the household dependency ratio, as well as a control for whether the household was already hosting at the time of the survey. In model 4, we further explore the role of ethnicity in hosting decisions. We do that not at the level of households but in dyadic analyses. To do that, we constructed a dyad-level dataset that pairs each incoming IDP to all potential hosts in the village. This allows us to assess whether dyads where there is an ethnic match are more likely to initiate a hosting relationship.Footnote 18 Model 4 contains the same controls as model 3. We use IDP fixed effects, thus controlling for any differences in observable and unobservable predictors across incoming IDPs. The full set of results for Table 2 can be consulted in Supplementary Appendix F.Footnote 19 Coefficients are standardized for ease of comparison; we report magnitudes in standard deviations.
Correlates of Hosting
Note: Standard errors clustered at the village (models 1–3) and dwelling (model 4) level and reported in parentheses. Variables are standardized. The full set of results is in Supplementary Appendix F. *p < 0.10, **p < 0.05, ***p < 0.01.
First, we explore the role of empathy in hosting decisions. The host’s empathy level is the most important variable in the calculus of IDP hosting. The coefficient for empathy is positively signed and statistically significant across all four specifications. Following model 3, for one standard deviation increase on the 15-point empathy scale we find an associated increase in the likelihood of hosting by 0.07 of a standard deviation; moving from the lowest (1) to the highest (15) empathy score is associated with a 20 percentage-point increase in the likelihood of hosting.Footnote 20 When we separate the components of the empathy index into affective and cognitive empathy, we find that only the cognitive empathy coefficient is statistically significant, and that it is much larger than the one for affective empathy; see Supplementary Appendix F. Thus, consistent with recent arguments in this literature, we find suggestive evidence that understanding what one’s interlocutor feels is more important than feeling the same emotion as them even for an extreme act of helping like hosting.
The importance of empathy in hosting decisions is corroborated in the qualitative follow-up fieldwork. When asked in an open-ended question to explain why they decided to open their doors to IDPs, 66% of hosts said that they were motivated by compassion or pity, noting that they themselves had experienced displacement or could easily imagine their household being displaced. One of the hosts noted that “[the IDPs] were in difficulty, and I have gone through a similar situation.” Another, imagining what it might be like to be displaced, told us “I hosted them because I could find myself in the same situation and, in that case, I would need to rely on other people to receive me in their home.” The ability to imagine what displacement would feel like is illustrative of cognitive empathy in action. Our interlocutors also mentioned other reasons for hosting—notably, previous kinship relations, being asked by the village chief to help, a religious obligation to help those in need, and willingness to help “good” or respected families. In Figure 2, we summarize how often the various reasons for hosting were mentioned in the interviews, and more detailed information is available in Supplementary Appendix B. The interviews confirm that being empathetic toward the displaced is by far the most important reason for hosting.
Reasons for Hosting
Note: In qualitative follow-up interviews, we asked the open question “Why did you decide to host this IDP family?” to 50 randomly selected households that were hosting an IDP. This figure presents a categorization of their answers. Detailed information is available in Supplementary Appendix B.
Of the three cost-benefit centered explanations for hosting—ethnicity, relationship to the chief, and perceived economic benefits of hosting—only the respondents’ relationship to the village chief is statistically significant (Table 2). Those who self-report as having a connection to the chief are more likely to open their doors to IDPs. In the qualitative follow-up fieldwork we set out to ascertain what role village chiefs have in the hosting process. Interviews confirm that chiefs rarely match IDPs to specific families. Only 18% of hosts without a prior relationship with the IDPs said that it was the chief who made the initial introduction. In most cases, the IDPs knocked on doors at random (54%) or approached potential hosts in the street (20%) (see Supplementary Appendix B for details).
During our study period, most of the hosting households (72%) accommodated an IDP of a different ethnicity than the head of the hosting household. The role of ethnicity in hosting is explored at the household level in models 1–3 as the strength of ethnic attachment, and in the context of dyadic matches between potential hosts and IDPs in model 4 as a probability of co-ethnic dyads striking up a hosting relationship. Neither of the two variables is statistically significant, and both are consistently negatively signed. That ethnicity—an important predictor of cooperative behavior in everyday life and emphasized in the broader literature on the causes of cooperation (Habyarimana et al. Reference Habyarimana, Humphreys, Posner and Weinstein2009)—is not correlated with hosting decisions suggests that hosting by its nature is a type of behavior that is very different from more mundane cooperative interactions.Footnote 21
That ethnicity does not play a major role in hosting decisions is also confirmed in our qualitative fieldwork.Footnote 22 When asked an open-ended question about the logic of their hosting decisions, none of those who were hosting IDPs at the time of the interview mentioned ethnicity as a factor. Similarly, none of the IDP families referenced ethnicity when asked why they thought that the hosting family took them in. In a hypothetical hosting scenario—when asked to choose between IDP households of different ethnicities—only 12% of respondents said that they would prefer to host a family of co-ethnics. In contrast, 60% of respondents mentioned they would host any IDP household without discrimination. Supplementary Appendix B provides detailed information on these analyses.
We also hypothesized that household characteristics—factors like household wealth, religiosity, and the feeling of security—might be important in shaping hosting decisions. We find that neither wealth nor religiosity matter systematically. While having a high-quality roof—a common indicator for wealth in this setting—is positively and significantly correlated with hosting in models 1 and 2, this variable loses statistical significance once demographic controls are introduced in model 3. Higher religiosity appears to be consistently associated with a lower likelihood of hosting, although the coefficients never reach statistical significance. As a matter of empirics, the correlation between empathy and religiosity in our sample is low.Footnote 23 We do find that households with male heads are considerably more likely to accommodate IDPs. Qualitative interviews with female household heads suggest that they feel physically insecure relative to their male counterparts and are worried that male IDPs might assault them or refuse to leave. As a secondary factor, some female household heads mention that they are too poor to host IDPs and have smaller fields.Footnote 24
We check the robustness of the reported results in several ways. In the main specification, we study the hosting dynamics of households that own the dwelling. In a few cases, IDPs were received by households that were hosted themselves; in a robustness check, we include these additional households. Second, dwellings that were hosting at the time of the survey may be thought to be less likely to host additional households; to address this concern we drop these dwellings. Third, we drop households that left the village during the 10-month period after the survey. Fourth, to learn about the intensive margin of hosting, we look at the number of IDPs hosted, instead of a binary hosting variable. Fifth, we include hosting relationships based on kinship or prior acquaintance. Additionally, we address the issue that our model clusters standard errors for only a small number of clusters. Across all these robustness checks, the results do not change substantively (see Supplementary Appendix E).
Origins of Empathy
Having established that empathy is the strongest correlate of hosting, we now explore its origins. The literature in psychology and political science suggests that empathy might, at least in part, result from past exposure to hardship (Stephan and Finlay Reference Stephan and Finlay1999). We put the “empathy born of violence” hypothesis to the test in the context of displacement in the DRC. To do this we run an analysis with the empathy score as the outcome and measures of prior exposure to violence as independent variables. We measure past exposure to violence among potential hosts through a six-item index (model 1), by whether the respondent’s house had been ransacked (model 2), and by the number of times that a respondent had been displaced over her lifetime (model 3).Footnote 25 The analyses include controls for wealth, religiosity, strategic benefits, strength of ethnic attachment, and proximity to the village chief, as well as village fixed effects and demographic controls. As before, coefficients are standardized, and the reported effect is in standard deviations.
The results are reported in Table 3; full regression output is in Supplementary Appendix G. Past experience of violence is correlated with higher empathy across all specifications; however, the magnitude of the effect is small. A one standard deviation increase in the exposure to violence index is associated with an increase of 0.08 of a standard deviation in the empathy index. Moving from no exposure to violence to the highest level of exposure on the six-item index (i.e., from the minimum of 0 to the maximum of 6) is associated with a 0.27 standard deviation increase in empathy, corresponding to 0.6 of a unit increase on the 15-point empathy scale. Similarly, having one’s house ransacked or experiencing an additional instance of displacement are associated with positive, but small, increases on the empathy scale of 0.33 and 0.06 of a unit, respectively. These findings provide suggestive support for the hypothesis that past hardship might indeed make individuals more empathetic toward others who are suffering and therefore more willing to help.
Correlates of Empathy
Note: Standard errors clustered at the village level and reported in parentheses. Variables are standardized. The full set of results is in Supplementary Appendix G. *p < 0.10, **p < 0.05, ***p < 0.01.
The argument that past exposure to violence may be correlated via the empathy channel with a higher propensity among potential hosts to open their doors to the forcibly displaced, regardless of ethnicity, is an important addendum to the literature on the legacies of violence. There is an ongoing debate in that literature as to whether past exposure to violence leads to pro-social behavior through post-traumatic growth (Blattman Reference Blattman2009; Voors et al. Reference Voors, Eleonora, Verwimp, Bulte, Lensink and Van Soest2012) or to parochial altruism; that is, withdrawal from the public sphere and increased cooperation exclusively with one’s co-ethnics (Bauer et al. Reference Bauer, Cassar, Chytilová and Henrich2014; Lupu and Peisakhin Reference Lupu and Peisakhin2017). Our results help to make sense of this disparate set of findings. It seems that those with high levels of empathy—in part resultant from experiencing past hardships—might be more willing to view non-coethnics as similar and therefore act pro-socially toward them. In this sense, among particularly empathetic individuals prosocial behavior might be extended not only to coethnics but also to suffering others. This is consistent with the insights of common ingroup identity theory, which argues that meaningful shared experiences can activate supraordinate identity (Gaertner and Dovidio Reference Gaertner and Dovidio2000). That hosting communities in conflict-plagued societies appear to reconceptualize who counts as an in-group member away from shared ethnicity toward shared victimhood is something that thus deserves further attention (see, for instance, Kaufman Reference Kaufman2001; Wayne and Zhukov Reference Wayne and Zhukov2022).
Increasing the Willingness to Host
We have now established that empathy is the most important correlate of hosting. However, from a policy perspective, it is also important to know whether the feeling of empathy can be stimulated with a view to encouraging more potential hosts to open their doors to strangers in order to minimize the trauma of forced displacement. This is what we explore in the experiment that was embedded within the survey.
The experiment contained eight groups. Those randomly assigned to the empathy appeal participated in a perspective-taking exercise, where respondents were asked where they would go if displaced and what they would take with them. This type of intervention has been shown to activate cognitive empathy (Adida, Lo, and Platas Reference Adida, Lo and Platas2018). To contextualize the importance of empathy we also separately primed two other potential drivers of hosting. To test the sway of village elites over hosting decisions we asked the village chief to visit those respondents randomly assigned to the authority appeal to urge them to accommodate IDPs. Respondents assigned to the religiosity appeal received a visit from a local Catholic community leader, who reminded them that it is a religious duty to help those in need. Those in the control condition were read a brief message informing them that there might be displacement in the region; the same message preceded all other treatments. In addition, to learn about the role of ethnicity, half of the respondents were assigned at random to a prime that the incoming IDPs would likely be from their own ethnic group, whereas the other half were told that the displaced would likely be of a different ethnicity. Detailed information on the design of the experiment is in Supplementary Appendix H.
Results from the experiment are reported in Table 4. These specifications do not include any covariates other than village and enumerator fixed effects given the random nature of treatment assignment; full results with covariates are in Supplementary Appendix H.Footnote 26 We consider two outcomes related to hosting in the table: an expression of the willingness to host IDPs in the future as asked in the survey and actual hosting behavior over the 10-month period following the survey. The reason that we also report the self-reported willingness to host here is to show how different this attitudinal variable is from actual behavior. In the pre-analysis plan, we pre-registered additional outcomes unrelated to hosting: contributions to hypothetical future IDPs in an incentive-compatible dictator game, willingness to cultivate a field, proceeds from which would go to hypothetical future IDPs, and showing up to work in the field 2 weeks after survey completion.
Results of the Experiment
Note: Standard errors clustered at the village level and reported in parentheses. Variables are standardized. Details on the experiment are in Supplementary Appendix H. *p < 0.10, **p < 0.05, ***p < 0.01.
In the sample, 97% of respondents said that they were willing to host hypothetical IDPs. Only 24% actually started hosting. Likewise, 96% of respondents said they were willing to cultivate a field for future IDPs, but only 53% sent household members to do the work two weeks later. This discrepancy between attitudinal measures—commonly used in existing work on the correlates of refugee assistance—and actual helping suggests that self-reported attitudes are subject to social desirability bias, and that what the attitudinal measure captures, at least in this instance, is mostly cheap talk. Such high self-reported willingness to host and to cultivate the field made it very difficult to capture any treatment effects in the survey, and none of the treatments, including perspective-taking, had a statistically discernible effect on these outcome measures.
None of the treatments changed hosting behavior in the longer term (Table 4). These null findings are unlikely to be due to low statistical power, as there were about 1,500 participants in the experiment.Footnote 27 Instead, it seems that the treatments—including the perspective-taking intervention—did not leave a sufficiently long-lasting effect. On average, IDPs arrived four months after treatments had been administered. This was likely too long of a gap for a simple but scalable treatment such as perspective-taking to have an effect (on the short durability of perspective-taking effects see Adida, Lo, and Platas Reference Adida, Lo and Platas2018; Simonovits, Kézdi, and Kardos Reference Simonovits, Kézdi and Kardos2018). At the time of administering the surveys, we did not know how soon the displaced would arrive. It is possible that had they arrived within one or two months, treatment effects would have been discernible. Notably, humanitarian organizations like the UNHCR often find themselves in a similar situation of knowing that an influx of IDPs is likely but not knowing when these would arrive. The lesson from our study is that perspective-taking does not work 4+ months out from the time of the treatment.
When it comes to other helping behaviors—contributions to hypothetical IDPs in the dictator game and field cultivation—these are explored in Supplementary Appendix H. Authority and morality appeals both have a positive and statistically significant effect on donations in the behavioral game only in the sub-sample of respondents who think that IDPs are likely to arrive in the coming months and that future IDPs will reap benefits from respondents’ help.Footnote 28 Likewise, for this outcome, which is measured immediately following the treatments, the perspective-taking exercise has a positive effect (p = 0.11) in this subsample. Two weeks down the line, none of the experimental treatments have a positive effect on households sending members to cultivate the field. However, consistent with the study’s earlier correlational analyses on hosting behavior, respondents’ baseline level of empathy is positively and significantly associated with respondents or their family members showing up to work on the field in the subsample (p < 0.05).
Returning to the null effects of perspective-taking on longer-term costly helping behaviors like hosting or field cultivation, there is a plausible alternative explanation for them. Perspective-taking might not work in a population where most members have direct experience of or live in regular fear of displacement. 95% of respondents in our sample reported experiencing displacement at least once in their lifetime. It is possible that in a population where the possibility of having to flee from one’s home is a reality of everyday life that perspective-taking fails because imagining where to go and what to take when fleeing is a regular necessity. We are unable to adjudicate between the failure of the perspective-taking intervention as a matter of time lapsed versus saturation of experience or fear of displacement; this is something that requires further study.
CONCLUSION
We set out to explore what motivates individuals to open their doors to strangers and host the forcibly displaced. While the literature sheds light on the causes of displacement, the pathways that displaced individuals take, and on the economic and political impact of the displaced on host communities, it is largely silent about the factors that motivate potential hosts to accommodate them. And yet, if the trauma of displacement is to be minimized by facilitating the displaced to be hosted in communities closer to home and in private homes rather than camps, then we need to better understand what motivates hosting and how to encourage potential hosts to accommodate IDPs.
We found that empathy among potential hosts is the most important correlate of the decision to open one’s home to the displaced. As one goes from the minimum to the maximum value on the empathy scale, the associated likelihood of hosting increases by 20 percentage points. Empathy appears to matter considerably more in hosting decisions than any other factor. Exploring the origins of empathy, we also found, consistent with the altruism born from suffering hypothesis, that empathy levels appear to be higher among those who have experienced violence in the past. Feeling physically secure as a male household head or being connected to local authorities increases the likelihood of hosting too, but at a lower magnitude. Being of the same ethnicity as the IDP, wealthy, or religious does not appear to affect the decision to host, at least in the context of the Democratic Republic of the Congo.
The study featured an experiment wherein we encouraged respondents to host by priming cognitive empathy through a perspective-taking exercise. We also appealed to participants’ respect for authority and religiosity. The experiment produced largely null results indicating, among other things, that the effect of perspective-taking interventions is likely short-lived and cannot persist for multiple months. These results raise concerns with regards to policy makers’ ability to engender greater willingness to host those forcibly displaced in the longer term using simple interventions.
Our study is among the very first to explore the correlates of hosting (see also Hartman and Morse Reference Hartman and Morse2018; Hartman, Morse, and Weber Reference Hartman, Morse and Weber2021). It stands out from existing work because we were able to measure household characteristics—including empathy levels—prior to the arrival of IDPs; this allowed us to get around the problem of ex post rationalization of hosting decisions. In addition, our data come from a census of 15 villages, which means that the results are not a product of selective or biased sampling. Unlike most studies on assistance to migrants, we measured empathy via a scale that is used in psychology. Finally, the main measure of hosting in our study was not a self-reported willingness to accommodate IDPs or even a self-report of having hosted in the past but rather a verifiable record by the village head that a given household hosted IDPs in the 10-month period after the pre-displacement survey.
We expect the findings from this study to travel to other societies of primarily subsistence farmers living in chronic poverty and in an environment of high violence such as Yemen, South Sudan, Northern Nigeria, and Afghanistan. Notably, in the context of chronic violence, displacement is usually predictable, which primes the potential hosts to think of this as a possibility. Instances where displacement is a product of a natural calamity are different; there the dynamics of hosting, absent the expectation of IDP inflows, might plausibly be distinct. Therefore, the generalizability of our conclusions requires out-of-sample testing. In future work, it might also be useful to better measure exposure to past violence among respondents. The measure that we use captures recent experience of violence, whereas the psychology literature hypothesizes that hardships deeper in the past likely lead to more empathy. Our findings suggest that those more empathetic, and perhaps also with more past experience of hardships, might be more willing to consider victims as in-group members. This is something that could fruitfully be measured directly in future work. Finally, we expect that those with higher empathy will also be more likely to host refugees in economically advanced countries. Moving beyond the context of the forcibly displaced and hosting, in future work it would be interesting to explore how strongly empathy correlates with other forms of assistance, including in everyday interactions.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit http://doi.org/10.1017/S0003055424000923.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/IESQYC.
ACKNOWLEDGMENTS
We thank Han Il Chang, Jean-Francois Maystadt, Melina Platas, six anonymous reviewers, and seminar participants at Antwerp University, GLAC, NYU-AD, and EGAP for comments. Thanks to Emmanuel Kandate, Antoine Ciza, Ruilin Lai, RISD, and the late Jean Paul Zibika.
FUNDING STATEMENT
This research was funded by Research Foundation Flanders (12W8320N) and the Center of Behavioral Institutional Design and Tamkeen under the NYU Abu Dhabi Research Institute Award CG005.
CONFLICT OF INTEREST
The authors declare no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The authors declare the human subjects research in this article was reviewed and approved by New York University—Abu Dhabi (#040-2019). The authors affirm that this article adheres to the principles concerning research with human participants laid out in APSA’s Principles and Guidance for Human Subjects Research (2020).
Open access
Comments
No Comments have been published for this article.