Who Helps Refugees in South Africa? An Examination of Cultural and Social Factors.

Pro-refugee philanthropy is beneficial for refugee integration and an important contributor to intergroup relations in a South African context. No study, however, has provided an in-depth quantitative analysis of what factors drive this type of behaviour. This article investigates what cultural, social and religious factors may predict participation in charitable behaviour towards refugees in South Africa. Three types of behaviours were examined: (i) material donations; (ii) volunteer activities; and (iii) information sharing. Using data from a 2019 Ipsos Migration Survey, the study found that intergroup threat was a robust predictor of charitable behaviour of all kinds. Friendship contact with a pro-refugee volunteer was also positively associated with philanthropic behaviour. Institutional trust and religious orientation were correlated with volunteer activities and information sharing but not donations. Study findings can be utilised to design interventions that increase public participation in actions that help refugees in an African context. © International Society for Third-Sector Research 2022.

Introduction

There were more than a quarter of a million refugees (including asylum seekers) living in South Africa in 2020. The country does not encamp refugees and this group is required to self-settle in host communities (Handmaker, 2011). There is limited government support provisioned for refugees, and many face severe challenges to integration including accessing basic services, acquiring appropriate documentation and finding employment (Human Rights Watch, 2020). In addition, this group is frequently the target of xenophobic prejudice and, in some cases, violence (also see Dinbabo et al., 2021). Stepping in to fill the gap left by the state, charitable organisations as well as altruistic individuals, have helped provide support (including shelter, medical aid and legal protection) for refugees in the country. Although most of this philanthropic behaviour is informal and non-registered, it has become crucial to refugee integration and an important foundation for intergroup harmony. Given the societal value of this kind of behaviour, the question becomes how can it be encouraged?

Quantitative scholarship on the decision to volunteer or to give to charity has grown substantially in the last few decades (for a review of this research, see Bekkers & Wiepking, 2011; Brooks, 2007; Wilson, 2012). However, the scope of the available work on charitable behaviour targeted at refugees is limited. Moreover, the African context is under-represented in the contemporary empirical scholarship on volunteering and non-profit organizations (Compion, 2017). There are only a handful of large-sample studies that seek to identify the determinants of African philanthropic behaviour. As a result of these factors, little is known about the general characteristics of those who participate in South African pro-refugee philanthropy. Few scholars have examined the profile of these participants, and none have provided an in-depth quantitative analysis of what factors drive individual engagement in this type of intergroup help. Given the societal benefits of this sort of philanthropy, this knowledge gap is troubling.

The article examines what drives participation in pro-refugee philanthropy in South Africa, adding to the existing body of quantitative research that examines charitable behaviour. This kind of behaviour can be quite multidimensional, involving both humanitarian aid as well as advocacy-based actions. First, the study examined generalised recent participation in this kind of help and then investigated three specific types of behaviour: (i) material donations; (ii) volunteer activities; and (iii) information sharing. The study estimated an empirical model that identified those factors that most influenced recent participation in pro-refugee actions. Data from a largescale quantitative survey fielded by Ipsos was used for this study. Four different hypotheses were tested in this article, contributing to our understanding of what determines helping behaviours in the country.

Refugees in South Africa

For much of the twentieth century, the South African government applied its immigration policies without reference to the 1951 United Nations’ Refugee Convention or the United Nations High Commissioner for Refugees (UNHCR). In the case of refugee inflows that the state found acceptable, ad hoc procedures allowed refugees to be admitted without major policy change (Handmaker, 2011). Following the end of white settler rule, the new government signed an agreement with the UNHCR in 1993 and began drafting new refugee legislation. The Refugees Act (No. 130), finally signed in 1998, set out the provisions and protections afforded to refugees and asylum-seekers in the country. To be granted legal recognition as a ‘refugee’, an individual must prove that he/she meets the definition as outlined in the Act. Consequently, a prospective refugee must first apply for asylum-seeker status and then undergo a screening process.

The number of people applying for refugee status in the country grew substantially between 1990 and 2020. The United Nations Development Programme (UNDP, 2021) estimates that about 277,000 people are currently seeking official refugee status in South Africa. According to UNDP data, the majority (68%) of refugees remain classified by the government as ‘asylum-seekers’ in 2020. The government requires refugees to self-settle and most live in the country’s towns and cities. Refugees come mostly from Somali, Zimbabwe and the Democratic Republic of Congo (also see Dinbabo et al., 2021). The official statistics utilised by the UNDP and other agencies, however, probably underestimate the true number of people seeking refuge in South Africa.

According to a report from Amnesty International (2019), South Africa’s asylum management system is inadequate, inaccessible and corrupt. Making a successful asylum claim is often a difficult process, fraught with corruption and malpractice, and there is currently a largescale backlog of pending claims (also see Handmaker, 2011 for a historical perspective on this problem). In addition to these documentation issues, the integration of refugees into South African society is undermined by xenophobia, i.e., the presence of widespread public antipathy towards foreigners (Gordon, 2016). Over the last two decades refugees have regularly found themselves the victims of anti-immigrant riots and violence (also see Gordon, 2020). The COVID-19 pandemic has, at the time of writing, only seemed to worsen the conditions of refugees in South Africa (for a discussion of this issue, see UNDP, 2021).

Pro-refugee philanthropy can take many forms in South Africa and is often quite complex and diverse. Religious organisations (such as Cape Town's Central Methodist Church) play a prominent role here as do international and domestic non-profit organisations (NGOs). At the forefront of these efforts have been NGOs founded by migrants and refugees themselves (Dinbabo et al., 2021). Although there are numerous formal organisations that provide support to refugees, most pro-refugee philanthropy appears to be informal and non-institutionalized. This trend would be consistent with other types of charitable behaviour in the country (Seabe & Burger, 2021). Given that a significant degree of pro-refugee philanthropic behaviour is informal and non-registered, it is difficult to objectively track. Not much is known, therefore, about the scale and character of this kind of philanthropy.

Theoretical Framework and Hypotheses

The article drew on the relevant empirical literature on volunteering and non-profit organisations to construct different hypotheses to be tested. In doing so, it is important to be cognisant of the limitations of existing scholarship. Much of the current work on charitable behaviour tends to operationalize philanthropy as a generalised behaviour and is not concerned with the object of that help. But as existing research by Evers and Gesthuizen (2011) shows, the micro-level factors that determine participation in different types of charitable organisations vary. In addition, there is empirical evidence from Wiepking (2010) demonstrating that the drivers of philanthropic behaviour will differ according to the focus of the charity (also see Neumayr & Handy, 2019). Although scholarship has differentiated giving to certain causes (e.g., cultural, international or health), less is known about giving to refugees.

Empathy is often considered an important driver of charitable behaviour in the large interdisciplinary literature on philanthropy. For example, Bekkers (2010), in his study of intentions to engage in different kinds of philanthropic activities, found that empathy was positively associated with a willingness to participate in charitable giving (also see Bekkers & Wiepking, 2011; Mitani, 2014; Neumayr & Handy, 2019). People routinely fail to empathize with those from different social or cultural groups, and research demonstrates that this can affect philanthropic behaviour. Bruneau et al. (2017), for example, found that altering American participants’ perception of the Arab population influenced their donations to Arab charities (also see Stürmer et al., 2006). This body of research suggests that attitudes towards refugees will have an impact on prosocial behaviour towards this group.

Negative outgroup perceptions are more likely to occur when an outgroup is viewed as threatening (Cikara et al., 2014). Intergroup threats are often characterised by a representation of the outgroup relationship in competitive terms or in terms of negative interdependence. There is significant evidence that perceived intergroup threat is associated with discriminatory behaviour. For example, using laboratory experiments, Shepherd et al. (2018) demonstrated that viewing immigrants as a threat is associated with a greater willingness to participate in collective anti-immigrant action. Gordon (2020), employing nationally representative public opinion data, found that perceived threat was correlated with a willingness to participate in anti-immigrant activities (e.g., boycotts or violence) in South Africa. It would appear that perceived intergroup threat is a major factor that predicts how people will behave towards an outgroup. Given the existing scholarship on perceived threat and intergroup behaviour, the following hypothesis was put forward:

H#1

Viewing refugees as a threat will reduce the likelihood that an individual will participate in pro-refugee philanthropic behaviour.

Social networks can be an important predictor of charitable behaviour. For example, Janoski (2010), documenting the association between recruitment and social ties, found that volunteers tended to be recruited by friends, acquaintances, and neighbours. This work suggests that people are less likely to refuse or disregard volunteer requests from those they know well, and close ties can exercise peer pressure on persons to accede to such requests. An individual’s social network composition should, therefore, influence the likelihood of becoming involved in philanthropic activities. Many scholars see social networks as a structural ‘pull’ factor, encouraging the individual to make good on existing altruistic values (also see Bekkers & Wiepking, 2011; Musick & Wilson, 2008; Wilson, 2012). It may be that contact with friends who volunteer increases the likelihood of learning about volunteering opportunities or being asked to volunteer. Consequently, the following hypothesis was developed to test the effects of social networks on helping behaviour:

H#2

Positive contact with people who do voluntary work for refugees will be positively correlated with charitable behaviour towards refugees.

When a person makes the decision to volunteer or donate, it is often unclear to them how effective this kind of help will be. It is doubtful that people have enough relevant knowledge to make a judgement about the utility of engaging in a specific act of charitable giving. Because charitable giving involves trusting others, people need high levels of trust to partake in that kind of action (Bekkers, 2003). This may be especially true in a context (like South Africa) where information about volunteering and non-profits is quite difficult to access. The need for trust may also be particularly strong in the case of donating to activist organizations (e.g., humanitarian and environmental groups) where the beneficiaries are usually unknown others (also see Wiepking, 2010). Given the importance of trust as a determinant of participation in charitable activities, many scholars have investigated the role played by social trust in determining philanthropic behaviour (for a review of this work, see Chapman et al., 2021).

When in doubt, individuals may utilise their trust in key institutions (e.g., national or local governments) to reach a decision about participation in a certain philanthropic behaviour. The existing literature (e.g., Bekkers & Wiepking, 2011; Evers & Gesthuizen, 2011; Taniguchi & Marshall, 2014) considers institutional trust (via the mechanism of perceived efficacy) to have a particularly powerful effect on giving and volunteering behaviour. Confidence in legal–political institutions can function as a proxy for sector-specific confidence in philanthropic organizations. In fact, because this form of trust pertains to organisations (and not to an abstract notion such as ‘trust in mankind in general’), it is often considered to have an even more pronounced impact on charitable behaviour than generalised trust (also see Chapman et al., 2021). Based on this research, the following hypothesis on institutional trust as a driver of behaviour was constructed:

H#3

Perceptions of political institutions as trustworthy will be positively associated with helping refugees.

Scholars in the social sciences have long linked religious capital with prosocial behaviour. Using a large multi-country dataset, for instance, Bennett and Einolf (2017) found that religiosity was positively correlated with helping outsiders. The authors contend that religiosity encourages values (e.g., gratitude, caring and charity) that promote prosocial behaviour. This thesis is consistent with the empirical scholarship on charitable behaviour in Africa by Compion (2017) which discovered that religiosity was linked to greater involvement in both secular and religious volunteer groups. The empirical scholarship on philanthropic behaviour (e.g., Brooks, 2007; Mitani, 2014; Musick & Wilson, 2008; Wiepking et al., 2014) tends to show that having a religious affiliation increases the chances of participating in charity. All religious traditions, of course, do not endorse the same values and there are important distinctions between them. Different religious affiliations should, therefore, have differing effects on whether an individual participates in charitable behaviour. With this important caveat in mind, the following hypothesis was developed:

H#4

Religious affiliation will improve the likelihood of individual participation in pro-refugee philanthropic behaviour.

Data

Ipsos conducted a survey of the adult population in four (i.e., Limpopo, KwaZulu-Natal, Western Cape and Gauteng) of South Africa’s nine provinces. These provinces (hereafter the LKWG cluster) were selected because they constitute the primary areas where most of the country’s refugees and asylum-seekers are thought to live. In addition, from a socioeconomic and ethnolinguistic perspective, they represent a good mix of the country's different economic, racial and linguistic subgroups. The survey was collected at the household level, respondent selection was randomised, and fieldworkers conducted interviews face-to-face. The questionnaire was translated into multiple languages and (where appropriate) the interview was administered in the home language of the respondent. The sample was restricted to citizens of the country who were adults (18 years or older) living in private households and was design to be representative of all such adults in the four provinces. Fieldwork was conducted during October and November 2019. After cleaning, the Ipsos sample contained 2004 respondents. Benchmark weights were applied to the sample to ensure that the data was representative at the provincial level.

Measurements

Dependent Variables

Ipsos respondents were read a list of different helping behaviours and then asked if they had engaged in any of these during the twelve months prior to the interview. A significant portion (18%) of the LKWG citizenry had engaged in some kind of help in the year prior to the interview (Fig. 1). This finding was consistent with previous research (e.g., Everatt et al., 2005) on more generalised charitable behaviours in the country and shows that South African society is characterised by a diverse and wide spectrum of giving behaviours. Of those who claimed to have helped refugees, the most common form of help was making a material donation (e.g., money or other goods). Interestingly, about a tenth (12%) of the populace was unsure of how to answer this question. This may be due to the diverse nature of the behaviours listed, or it could indicate a degree of social desirability bias on this item.1

Fig. 1

Public volunteering to support and assist refugees in the twelve months prior to the Ipsos interview (multiple response)

Sharing positive information about refugees online or talking with friends and family about the refugee issue were relatively common behaviours listed in Fig. 1. Although not commonly studied as a philanthropic action, this kind of advocacy is especially important in the South African context. One of the primary challenges facing refugees in the country are exclusionary anti-immigrant attitudes and prejudice (Gordon, 2020). Ordinary people are often the best communication agents and studies show that spreading information through their own personal networks (either digital or otherwise) has the potential to change hearts and minds (Kalla & Broockman, 2020). Anti-prejudice communication, especially at the interpersonal level, can be an effective strategy for attenuating intergroup conflict and increasing public support for policies that protect refugees and help them integrate.

In order to analyse determinants of generalised pro-refugee help in the recent past, the response categories under discussion were converted into a variable containing three categories: (i) helped; (ii) did not help; and (iii) undeclared. In addition to an analysis of generalised help, the study also examined specific types of philanthropic behaviours. Three distinct variables were created for this purpose. The first was material donations and the second was volunteering time. A respondent was recognised as having engaged in this behaviour if they had either: (i) ‘volunteered some of their time'; or (ii) ‘personally welcomed one or more refugees into their home’. The third is information sharing and is defined as having either: (i) ‘talked to friends and family about a positive story regarding refugees'; or (ii) ‘shared positive stories about refugees online’. Each of these three variables is coded as follows: (i) helped; (ii) did not help; and (iii) undeclared.

Independent Variables

A number of standard background dummy variables were constructed to capture pertinent demographic characteristics (e.g., gender, age, geographic type, and race group). An educational attainment variable was derived from the following question: ‘[w]hat is the highest level of education you personally have achieved?’ Responses were categorised into three dummy variables: (i) no complete secondary school; (ii) completed secondary school; and (iii) post-secondary school qualification. Respondents were read a list of different work situations and then asked: ‘[w]hich of these statements best describes your working life?’ Using responses to this question, four categorical dummy variables were produced: (i) employed; (ii) student; (iii) unemployed; and (iv) outside labour market. In addition, different independent variables were created to test the study’s four hypotheses, these are described below.

Refugee Threat Index

Respondents were asked if they agreed or disagreed that refugees: (i) are different to me; (ii) are violent; (iii) are good people; (iv) are dishonest and (v) take our jobs. Responses were coded onto a five-point Likert Scale (1 strongly disagree, 5 strongly agree). To produce the index, the direction on the ‘good people’ item was reversed. The five items were combined and then transformed onto a 0–10 composite index.2 The higher the value on the index, the more negatively an individual viewed refugees in South Africa.

Contact with Volunteers

Survey participants were asked if they had positive contact (i.e., friendship) with someone who does voluntary work for refugees. This dichotomous variable was coded as 1 ‘volunteer friend’ and 0 ‘no volunteer friend’.

Institutional Trust

Respondents were required to indicate whether they trusted or distrusted their government at the: (i) national; (ii) provincial; and (iii) local level. Responses to this question were recorded on a 1 (strongly trust) to 5 (strongly distrust) Likert Scale. The three items were combined together into a single index,3 this new metric was converted into a 0 to 10 scale with the higher value indicating the higher level of trust. This variable was used as a proxy for institutional trust in South Africa.

Religious Affiliation

To gauge individual membership of a specific a religious group, survey participants were queried on whether they considered themselves to be religious. Then, if the respondent said yes, they were questioned on which religion they belong to. Responses to these questions were used to construct four dummy affiliation variables: (i) unaffiliated; (ii) non-Christian; (iii) Christian; and (iv) uncertain.

Multinominal Regression of Generalized Pro-Refugee Behaviour

The four hypotheses outlined were tested using a multivariate regression approach. Due to the nature of the categorical generalised pro-refugee philanthropy variable, a multinomial approach was employed to model individual participation in the twelve months prior to the interview. A probit regression model was selected as this specification does not make the erroneous independence of irrelevant alternatives assumption. The model estimates the log odds of the probability of choosing one outcome category versus the base. The base outcome of the model was ‘did not help’. Two models were produced, the first included only the background control variables outlined in Table 1 and the second introduced the four independent variables described in the previous section. The results of the multinomial probit regression models are displayed in Table 2.

Table 1

Summary descriptive of all independent variables

	Mean	Std. Dev	Min	Max
Refugee Threat Index	6.03	2.42	0	10
Contact with volunteers	0.12	0.32	0	1
Institutional Trust	4.56	2.90	0	10
Religion Affiliation (ref. none)
Non-Christian	0.07	0.26	0	1
Christian	0.65	0.48	0	1
Undeclared	0.07	0.26	0	1
Background controls
Gender (ref. male)	0.50	0.50	0	1
Age	35.67	13.72	1	98
Race Group (ref. white)
Black African	0.75	0.44	0	1
Indian	0.03	0.16	0	1
Coloured	0.14	0.35	0	1
Geographic type (ref. urban metro)
Urban non-metro	0.05	0.22	0	1
Rural	0.25	0.43	0	1
Employment Status (ref. employed)
Student	0.13	0.33	0	1
Unemployment	0.26	0.44	0	1
Outside Labour Market	0.13	0.34	0	1
Educational Attainment (ref. incomplete secondary schooling)
Completed Secondary	0.46	0.50	0	1
Post-Secondary	0.16	0.36	0	1
Province of Residence (ref. Western Cape)
KwaZulu-Natal	0.25	0.43	0	1
Gauteng	0.30	0.46	0	1
Limpopo	0.20	0.40	0	1

Unweighted data used

Table 2

Multinomial probit regression on participating in helping refugees in the twelve months prior to the Ipsos interview

	Model I				Model II
	Helped		Undeclared		Helped		Undeclared
Gender (ref. male)	0.123	(0.106)	0.236	(0.123)	0.030	(0.114)	0.209	(0.127)
Age	− 0.015	(0.005)**	− 0.010	(0.006)	− 0.017**	(0.006)	− 0.009	(0.006)
Race Group (ref. Black African)
White	− 0.117	(0.209)	− 0.236	(0.263)	− 0.216	(0.216)	− 0.370	(0.165)
Coloured	− 0.259	(0.179)	0.130	(0.232)	− 0.265	(0.196)	0.115	(0.242)
Indian	0.117	(0.321)	0.233	(0.300)	− 0.013	(0.383)	0.090	(0.319)
Geographic type (ref. urban metro)
Urban non − metro	− 0.309	(0.195)	− 0.528	(0.218)*	− 0.284	(0.207)	− 0.619	(0.227) **
Rural	− 0.095	(0.211)	0.028	(0.208)	0.074	(0.232)	0.046	(0.215)
Employment Status (ref. employed)
Student	− 0.434	(0.197)*	− 0.442	(0.230)	− 0.470	(0.212)*	− 0.412	(0.234)
Unemployment	− 0.654	(0.142)***	− 0.501	(0.162)**	− 0.557	(0.149) ***	− 0.489	(0.171) **
Outside Labour Market	0.030	(0.172)	− 0.372	(0.230)	0.108	(0.185)	− 0.444	(0.248)
Educational Attainment (ref. incomplete secondary schooling)
Completed Secondary	− 0.060	(0.117)	− 0.084	(0.144)	− 0.107	(0.124)	− 0.101	(0.148)
Tertiary Schooling	− 0.042	(0.171)	− 0.000	(0.209)	− 0.003	(0.175)	0.005	(0.215)
Refugee Threat Index					− 0.162	(0.022) ***	− 0.113	(0.025) ***
Contact with volunteers					1.057	(0.164) ***	− 0.079	(0.252)
Institutional Trust					− 0.007	(0.339)	− 0.044	(0.024)
Religion Affiliation (ref. none)
Non − Christian					0.847	(0.237) ***	0.728	(0.269)**
Christian					0.498	(0.146) **	0.324	(0.175)
Undeclared					0.534	(0.317) *	0.636	(0.255)*
Number of obs		1,983					1,939
Wald chi2		175 (30)					284	(42)
Prob > chi2		0.000					0.000
Log pseudolikelihood		− 20,967,331					− 18,935,934

*** p < 0.001, **p < 0.01, * p < 0.05

Linearized standard error in parenthesis; and 2. Models control for province of residence

Age was a robust (and negative) predictor of participation in philanthropy (r = − 0.015; SE = 0.005; p = 0.003) in Model I. This outcome was unanticipated as existing research (e.g., Bekkers, 2010; Brooks, 2007; Compion, 2017) tends to show that older persons are generally more apt to participate in charitable behaviour. The observed age effect cannot be attributed to greater exposure to student politics. Relative to employment, being a student was a negative correlate (r = − 0.434; SE = 0.197; p = 0.028) of helping in the first model. Unemployment was a robust driver of behaviour in that model, reducing the probability of engaging in helpful action. This outcome was unsurprising, as unemployment reduces social interaction and community integration, factors associated with charitable behaviour (Musick & Wilson, 2008). The employed, by contrast, are thought to have extensive and diverse networks which increases the likelihood of receiving requests to donate or volunteer (also see Wilson, 2012).

The Refugee Threat Index had a statistically significant (and negative) correlation with the dependent variable in Model II (H#1). Even when controlling for a range of socio-economic and demographic characteristics, the index was found to reduce the log odds (r = − 0.162; SE = 0.022; p = 0.000) of helping refugees (versus the base outcome). If we accept that people feel less empathy for groups that they see as threatening (as work by Cikara et al., 2014 would suggest), this finding was consistent with past research on charitable behaviour (e.g., Bekkers, 2010; Bekkers & Wiepking, 2011; Mitani, 2014). In addition, anti-refugee sentiment also had a negative (and statistically significant) effect (r = − 0.113; SE = 0.025; p = 0.000) on the dependent variable in the model’s second pairing. This finding suggests that antipathy towards refugees reduces the likelihood of reporting ‘don’t know’ when asked about helping. Having contact with someone who does voluntary work for refugees was associated with recent past participation in pro-refugee philanthropy (H#2). Positive contact with a volunteer improved the log odds of helping refugees by 1.057 (SE = 0.164; p = 0.000) in the LKWG cluster. No statistically significant effect, on the other hand, was noted between volunteer contact and the dependent variable in the second pairing.

Trusting political institutions decreased the log odds of participation by 0.007 (SE = 0.339; p = 0.712) but not at a statistically significant level (H#3). This null finding held true even if the model was adjusted to account for political affiliation.4 These results suggest that trust in legal–political institutions may not be the best proxy for public confidence in pro-refugee charitable organisations. This outcome was in line with the work of Hager & Hedberg (2016) who argued that institutional trust and confidence in charitable organisations should be differentiated in studies of charitable giving. Using data from Arizona, the authors found that confidence in charitable organisations directly influences philanthropic behaviour while institutional trust does not (also see Chapman et al., 2021).

Rejecting a religious orientation made an individual less likely to participate in pro-refugee philanthropy (H#4). Even an uncertain religious orientation was found to significantly increase the likelihood that an individual would help refugees. A robust level of differentiation in the effect of religious affiliation on behaviour was noted. Having a non-Christian affiliation increased the log odds of past participation by 0.847 (SE = 0.237; p = 0.000) in Model II. This can be unfavourably compared to 0.498 (SE = 0.146; p = 0.001) for a Christian affiliation. In other words, belonging to a religious minority (e.g., Islam, Judaism or Hinduism) in the LKWG cluster increased the propensity of individuals to have recently helped refugees. Given the rational choice and social networks approaches used in most studies of charitable behaviour (e.g., Bekkers & Wiepking, 2011; Neumayr & Handy, 2019; Wiepking, 2010), this finding was quite unexpected.

Self-reported racial group status5 had no influence on whether an individual engaged in pro-refugee philanthropy, implying that racial identity was not a determining factor here. This finding was at odds with other research on volunteering and charitable giving in South Africa (e.g., Everatt et al., 2005; Seabe & Burger, 2021). But perhaps linguistic identity was a more powerful driver of charitable behaviour in the South African case. As a robustness check, Model II was modified to substitute race group for language group.6 Using Afrikaans as the reference, speaking English significantly increased the log odds (r = 1.218; SE = 0.210; p = 0.000) of having recently participated in pro-refugee philanthropic behaviour. Similar (albeit weaker) correlations were recorded for Setswana (0.933; SE = 0.299; p = 0.002), isiZulu (r = 0.699; SE = 0.241; p = 0.004) and isiXhosa (r = 0.538; SE = 0.246; p = 0.029). Speaking one of the other languages under review (e.g., Sepedi or Tshivenda) was not correlated with past philanthropic participation. Given that most studies in South Africa have not examined linguistic identity as a predictor of charitable behaviour, this was an intriguing finding.

Multinominal Regression of Specific Pro-Refugee Behaviours

Moving beyond generalised behaviour, three multinomial regression models were produced to identify the determinants of the three specific pro-refugee philanthropic behavioural variables previously outlined. Each model contains all the background control variables produced for this study as well as the four independent variables created to test the different hypotheses. Table 3 depicts model outcomes for donations (Model I), volunteer activities (Model II) and information sharing (Model III). The Refugee Threat Index was found to be a robust predictor of behaviour in all three models (H#1) with the strongest effect observed in Model III (r = − 0.124; SE = 0.025; p = 0.000). Having positive contact with someone who does voluntary work for refugees was associated with all three philanthropic behaviours in Table 3 (H#2). Friendship had the greatest effect on behaviour in Model II (r = 1.007; SE = 0.182; p = 0.000). This outcome was consistent with scholarship that views social networks as a structural ‘pull’ factor, promoting volunteer behaviour (Bekkers & Wiepking, 2011).

Table 3

Multinomial probit regression on participating in different types of pf pro-refugee help in the twelve months prior to the Ipsos interview

	Donations		Volunteer Activities		Information Sharing
	Helped	Undeclared	Helped	Undeclared	Helped	Undeclared
Gender (ref. male)	0.108	0.227	0.280	0.248	− 0.033	0.213
	(0.137)	(0.128)	(0.147)	(0.128)	(0.130)	(0.128)
Age	− 0.004	− 0.005	− 0.013*	− 0.007	− 0.021**	− 0.007
	(0.006)	(0.007)	(0.007)	(0.007)	(0.007)	(0.006)
Race Group (ref. Black African)
White	− 0.174	0.267	0.078	0.337	0.225	0.356
	(0.270)	(0.266)	(0.291)	(0.273)	(0.251)	(0.268)
Coloured	− 0.232	0.449	− 0.265	0.483	− 0.301	0.462
	(0.304)	(0.324)	(0.314)	(0.331)	(0.286)	(0.327)
Indian	0.179	0.408	0.064	0.452	− 0.492	0.372
	(0.506)	(0.402)	(0.465)	(0.408)	(0.522)	(0.406)
Geographic type (ref. urban metro)
Urban non − metro	− 0.148	− 0.586*	0.117	− 0.548*	− 0.425	− 0.621**
	(0.225)	(0.229)	(0.252)	(0.229)	(0.225)	(0.229)
Rural	0.464	0.120	0.037	0.072	− 0.468	0.014
	(0.281)	(0.214)	(0.282)	(0.216)	(0.250)	(0.215)
Employment Status (ref. employed)
Student	0.075	− 0.296	− 0.588*	− 0.386	− 0.681**	− 0.400
	(0.237)	(0.237)	(0.273)	(0.237)	(0.260)	(0.236)
Unemployment	− 0.201	− 0.395*	− 0.598**	− 0.447*	− 0.596***	− 0.456**
	(0.181)	(0.173)	(0.202)	(0.173)	(0.162)	(0.173)
Outside Labour Market	0.110	− 0.477	0.187	− 0.446	− 0.307	− 0.533*
	(0.191)	(0.249)	(0.231)	(0.249)	(0.230)	(0.249)
Educational Attainment (ref. incomplete secondary schooling)
Completed Secondary	− 0.224	− 0.099	0.160	− 0.049	− 0.105	− 0.078
	(0.146)	(0.149)	(0.153)	(0.149)	(0.203)	(0.148)
Tertiary Schooling	0.146	0.025	− 0.122	0.012	− 0.203	− 0.014
	(0.215)	(0.216)	(0.233)	(0.219)	(0.308)	(0.216)
Refugee Threat Index	− 0.112***	− 0.089***	− 0.090**	− 0.085**	− 0.124***	− 0.096***
	(0.029)	(0.024)	(0.027)	(0.025)	(0.025)	(0.248)
Contact with volunteers	0.756***	− 0.346	1.007***	− 0.269	0.540**	− 0.394
	(0.182)	(0.247)	(0.182)	(0.247)	(0.180)	(0.248)
Institutional Trust	− 0.064*	− 0.052*	− 0.068**	− 0.051*	− 0.001	− 0.043
	(0.027)	(0.024)	(0.026)	(0.024)	(0.023)	(0.024)
Religion Affiliation (ref. none)
Non − Christian	0.718*	0.626*	0.539	0.603*	0.771**	0.638*
	(0.455)	(0.270)	(0.290)	(0.271)	(0.261)	(0.271)
Christian	0.400*	0.281	0.261	0.267	0.327*	0.269
	(0.302)	(0.179)	(0.183)	(0.179)	(0.166)	(0.178)
Undeclared	*	0.626*	0.016	0.535*	0.342	0.578*
	(0.304)	(0.257)	(0.310)	(0.260)	(0.274)	(0.257)
Number of obs	1,939		1,939		1,939
Wald chi2	200 (42)		218 (42)		225 (42)
Prob > chi2	0.000		0.000		0.000
Log pseudolikelihood	− 14,696,075		− 14,443,922		− 15,632,483

*** p < 0.001, **p < 0.01, * p < 0.05

Linearized standard error in parenthesis; 2. Models control for province of residence

Institutional trust was found to be a statistically significant correlate in the first pairing of Model II (r = − 0.068; SE = 0.026; p = 0.009) and Model III (r = − 0.064; SE = 0.027; p = 0.016). The direction of the observed correlation contradicted prior expectations (H#3), suggesting that people who distrust political institutions were more liable to make donations and partake in volunteer activities. Having a religious affiliation was positively associated with individual participation in certain types of charitable behaviour (H#4). No effect was noted in Model II but rejecting a religious affiliation decreased the log odds of participation in Model I and Model III. Interestingly, having an uncertain affiliation had a particularly positive (and statistically significant) correlation with the dependent variable in the first pairing of Model I (r = 0.718; SE = 0.304; p = 0.018). If a person was uncertain about their religious affiliation, then they were more likely to participate in donations to help refugees. Regression outcomes show that dissimilar religious affiliations have differing effects on the different types of philanthropic behaviour under review. A non-Christian affiliation had much more robust effect on donating and information sharing that a Christian affiliation. The noted differentiation is comparable with what was observed in the previous section.

Labour market status had a significant relationship with past participation in two of the three multivariate models produced for Table 3. Being unemployed or a student had a statistically significant (and negative) relationship with past participation in volunteer activities and information sharing. Age was also a statistically significant correlate of behaviour in these two models. A one-year increase in age reduced the likelihood that an individual would participate in information sharing (r = − 0.013; SE = 0.007; p = 0.047) or volunteer activities (r = − 0.021; SE = 0.007; p = 0.004) that helped refugees. Other variables (e.g., geographic type, race group or formal education) that were used to capture socio-demographic status were not robust determinants of pro-refugee behaviour in the table. This outcome is consistent with findings presented in Table 2, suggesting that these are not good predictors of recent pro-refugee philanthropic behaviour in the LKWG cluster. Given past research in South Africa on charitable and volunteering behaviour (Seabe & Burger, 2021), this was unanticipated.

Discussion

The outcomes of the quantitative analysis presented in the previous section provided meaningful insights into the drivers of recent pro-refugee philanthropic behaviour in South Africa. Building on the existing empirical research on charitable behaviour, the study identified social and cultural determinants of recent pro-refugee behaviour. The analysis first focused on generalised help and then considered three specific kinds of behaviour. The results showed that different variables exhibited varying effects depending on the behaviour under review. Empirical evidence, for instance, found that H#3 and H#4 were rejected in the case of material donations but were accepted for volunteer activities and information sharing (for a full overview, see Table 4). A discussion of these results is presented in the remainder of this section with areas of future research highlighted.

Table 4

Overview of determinants of public participation in supporting refugees in the twelve months prior to the Ipsos interview

	Generalised Help	Donations	Volunteer activities	Information sharing
Hypotheses
H1: Intergroup threat	−	−	−	−
H2: Contact with volunteers	+	+	+	+
H3: Institutional trust	n.s	n.s	−	−
H4: Religious affiliation	+	n.s	+	+
Background variables
Gender	n.s	n.s	n.s	n.s
Age	−	n.s	−	−
Geographic type	n.s	n.s	n.s	n.s
Race Group	n.s	n.s	n.s	n.s
Education	n.s	n.s	n.s	n.s
Employment	+	n.s	+	+

‘−’ negative effect on participation at least p < 0.05, ‘ + ’ positive effect on participation at least p < 0.05,, ‘n.s.’ effect not significant effect at p < 0.05

Positive contact with someone who volunteers to help refugees was found to improve the likelihood of participating in pro-refugee philanthropy. This finding was in keeping with existing research (e.g., Janoski, 2010; Musick & Wilson, 2008; Wilson, 2012) on how volunteering opportunities were biased by social network composition. It is possible, on the other hand, that engaging in charitable behaviour improves the chances of making like-minded friends. In other words, it is conceivable that charitable behaviour is influencing network composition here rather than the reverse. The data utilised for this analysis was cross-sectional and, unfortunately, the directionality of the observed relationship cannot be discerned.

An association between religious affiliation and prosocial behaviour towards refugees was a positive determinant of information sharing and volunteerism. This may reflect the fact that the affiliated tend to be more socially connected and have values that motivate such behaviour. Furthermore, this group may have greater exposure to requests for aid as well as more opportunities to help. On the other hand, religious norms about altruistic behaviour, as Galen (2012) contends, may lead some affiliates to present themselves as more charitable than they actually are. This kind of social desirability bias cannot be detected using the self-reported data available for this study, and this data limitation can only be acknowledged. Future research could design lab (or even field) experiments to test whether religiosity correlates with behaviours that control for this type of bias. Another issue lies with the limited measures of religiosity available in the Ipsos dataset. Future researchers should investigate the impact of different kinds of religiosity more comprehensively than has been attempted in this study.

One of the more unexpected findings that emerged from the multivariate analysis concerned the robust correlation between non-Christian affiliation and helping refugees. In four of the five models produced, a non-Christian affiliation was a more powerful correlate of behaviour than a Christian affiliation. Although unanticipated, this finding was comparable with the work of Bennett and Einolf (2017) who found that religious minority members were more likely to engage in prosocial behaviours. One possible explanation, put forward by Wiepking et al. (2014), is that faith is more salient to members of religious minorities, so values about charity and altruism are more likely to influence their members. Another possibility is that minority religious organisations must work harder to preserve the devotion of their members when compared to majority organisations. This results in a greater concentration on prosociality (including stronger messages on altruism) within those organisations.

Many studies, as the review by Chapman et al. (2021) shows, have regarded institutional trust as a key determinant of philanthropic behaviour. But the data presented in this paper has demonstrated that trust in political institutions was not a positive correlate of recent participation in donation behaviour. Empirical results indicate, moreover, that institutional trust was a negative correlate of other helping behaviours (i.e., donations and volunteering). One could argue that it is disappointment with 'normal politics' which motivates people to become more engaged in philanthropy. This would be in line with Brooks (2007) who argues that cynicism about the role of government (especially as it applies to state assistance for vulnerable groups) increases the likelihood of charitable giving. If a person was contented with existing legal–political institutions, then they may trust the state to look after vulnerable groups (such as refugees).

A number of different socio-demographic variables were included in the quantitative analysis. Some of the observed effects of these variables were consistent with prior expectations. Labour market status, for example, was a key determinant in the models produced for the study, and the results conformed to previous findings noted in the literature (e.g., Bekkers, 2010; Brooks, 2007; Musick & Wilson, 2008). Other findings were not in line with past quantitative work on the micro-level determinants of philanthropic participation. Formal schooling was, for instance, not a predictor of pro-refugee behaviour in any of the study’s models, a finding that challenges the expectations of resource theory (e.g., Bekkers & Wiepking, 2011; Mitani, 2014; Taniguchi & Marshall, 2014). This may be due to the fact, already aforementioned, that most of this work tends to investigate volunteering or giving as a generalised behaviour. Those studies that examine philanthropic behaviour for specific causes (e.g., Evers & Gesthuizen, 2011; Neumayr & Handy, 2019; Wiepking, 2010) find that the drivers of participation differ according to the focus of the philanthropy.

In contrast to prior expectations, race group identity was not a determinant of past participation in any of the models produced for this study. Linguistic identity, on the other hand, had a much robust correlation with helping refugees. Speaking certain languages (i.e., English, Setswana, isiXhosa or isiZulu) significantly improved the odds that an individual had engaged in pro-refugee philanthropy. Linguistic identity has not been the subject of study in past scholarship on philanthropic behaviour in South Africa. The reasons for these observed differences are, therefore, not clear. The standard theorems used to explain ethnolinguistic differences in charitable behaviour (such as dominant status theory or structural resources theory) do not seem to be applicable here. The result may be biased due to omitted variables, and future research will need to explore this possibility. Linguistic identity could, for example, be acting as a proxy for certain cultural or political values which influence charitable behaviour. Another possibility is that linguistic identity denotes access to certain information networks in South Africa, this may affect exposure to pro-refugee philanthropy.

Conclusion

There is limited empirical research on pro-refugee philanthropic behaviour in the South African context. Part of the reason for this knowledge gap has been the lack of dependable, quantifiable evidence. To overcome this problem, data from a largescale quantitative survey fielded by Ipsos on refugees and cross border migrants was used in this study. The data outlined a diverse and multicultural spectrum of recent pro-refugee behaviours amongst the LKWG population that ranged from donations to anti-prejudice communication. The main determinants of these behaviours were identified, and the study made a number of important academic contributions to the limited empirical research on pro-refugee philanthropy. Scholars should build on these findings, helping us understand why certain types of individuals help refugees and others do not. Beyond the academic importance of this work, it is hoped that the study findings can be useful for practitioners in the non-profit sector and those interested in promoting pro-refugee philanthropy. The data presented here can be utilised to design interventions that can increase public participation in actions that help refugees.