Faith-Based Nonprofits and the Delivery of Public Services: An Experimental Study of Sector-Bias

Public services are delivered not just by government agencies but also by nonprofit and for profit organizations often in collaborative networks (Milward & Provan, 2000; O’Toole, 1997). The rise of the New Public Management reforms placed renewed emphasis on using third party organizations to deliver services either via contract, vouchers, or collaborative agreements in a wide variety of countries (Dunleavy & Hood, 1994; Pollitt & Bouckaert, 2017). As examples, the provision of long term elder care and the delivery of hospital services in the US might be delivered at the local level by government agencies, nonprofits, or for profit organizations (Amirkhanyan, 2008; Cheon et al., 2021). Such mixes of delivery systems are not usual, and such combinations can be recognized by service recipients in different countries (Meier, Dhillon, et al., 2022; but see Kissane, 2008; Van Slyke & Roch, 2004).

While a small but growing body of research in behavioral public administration has been concerned with how public perceptions of government programs change when implemented by government agencies or the private^[1] sector (both for profit and nonprofit, see Amirkhanyan et al., 2024; Hvidman, 2019; Hvidman & Andersen, 2016; Marvel, 2015; Meier et al., 2019), the literature has not addressed such questions when the nongovernment organizations are faith based nonprofits. Since public perceptions of public programs respond to a wide variety of different aspects such as objective performance criteria, subjective assessments, secondary outcomes, and who is delivering the services (Song & Meier, 2018), logic suggests that the involvement of a faith based nonprofit might also influence such perceptions. While faith based nonprofits operate in a wide variety of policy areas depending on the country involved (see Amirkhanyan et al., 2009; Graddy, 2006; Riccucci & Meyers, 2008; Watkinson, 2015), they are especially prominent in humanitarian relief efforts that occur as a result of natural disasters or human made disasters (Kim et al., 2010; Mathias et al., 2022; Simo & Bies, 2007). Faith based organizations (FBOs) are major vehicles for the delivery of foreign aid to developing countries when local government capacity might be limited (Austin et al., 2022; Heist & Cnaan, 2016; Lindenberg, 1999; Nunnenkamp & Öhler, 2012).

Within the public administration literature, one related question has been whether faith based organizations are more effective than their secular counterparts with regard to providing social services (Amirkhanyan et al., 2009; Bielefeld & Cleveland, 2013; Feiock & Andrew, 2006; Watkinson, 2015). This question has increased in importance owing to the dynamic nature of the nonprofit arena. Since the passage of the Charity, Aid, and Recovery Act of 2002 in the US under George W. Bush, ever larger amounts of public monies have been supplied to FBOs to deliver public services (Luksetich, 2008). Faith based nonprofits are not new to the social services environment; governments in the US have been providing funding to these organizations since the 1967 amendment to the Social Security Act (Amirkhanyan et al., 2009). The primary difference now is the magnitude of funds. In addition to whether actual FBO performance exceeds that of secular peers, a similar line of inquiry is whether public opinion varies between the typology of sector status: religious or secular (Kissane, 2008). If public support for programs varies by which organizations deliver public services, there could be real world ramifications for how different policies attain their goals. Mackenzie-Liu, Schwegman & Lopoo (2022), for example, find that faith based foster care agencies are more likely to discriminate against same sex couples with the result being fewer available foster parents. Such behaviors that reflect perceptions of bias by faith based organizations might result in individuals less likely to contribute to FBOs, less likely to volunteer in them, and perhaps less likely to accept services from them (see Davey et al., 2021; Gibelman & Gelman, 2002; Mackenzie-Liu et al., 2022). Owing to the interrelationship between delivering services and gaining support, financial or otherwise, to underwrite these services, any biases the public might have in terms of the sector or organization that delivers services and how that might affect public support for such programs deserves additional consideration in the wider literature on public administration and public policy.

Preferences in terms of which organizations deliver services has been termed “sector bias,” and the literature in that area has been dominated by a conversation on the relative differences between the private and public sectors (Hvidman, 2019; Hvidman & Andersen, 2016; Marvel, 2015; Meier et al., 2019; Meier & An, 2020). Evidence in different countries has been found for a variety of claims but nothing has proven conclusive (Baniamin & Jamil, 2023; Berg & Johansson, 2020; Hameduddin & Vivona, 2023). Recent work has divided the nongovernmental sector into for profit and nonprofit organizations (Amirkhanyan et al., 2024; Meier et al., 2019). With respect to the nonprofit sector, there has been some evidence from Europe that nonprofits are perceived as warmer and slightly more competent than their private sector counterparts (Drevs et al., 2014; Xu, 2020); however, nothing definitive has arisen in the US (Amirkhanyan et al., 2024; Meier et al., 2019).

With respect to nonprofits, sector bias findings are important because they influence policy implementation both in terms of domestic policies and international efforts linked to development and humanitarian aid. Nonprofits operating abroad carry the weight of administering funds and enacting policies that reflect the preferences of host governments or their foreign counterparts. How nonprofits interact with different governments and individuals will shape how they are funded by all potential donors. Moreover, how nonprofits are stereotyped will help define how unaffiliated persons engage with them and aid them in their respective missions. Nonprofits clearly vary a great deal in function, orientation, and capacity, and one clear distinction is whether the nonprofit is faith based or secular. Ties to various religions brings the possibility that attitudes about religion in general or specific denominations in particular might bias individuals’ assessments of the policy implementation process. This brings us to our research question: Are nonprofits stereotyped as being more effective, depending on whether they are faith based or secular?

Sector bias in regard to nonprofits has only received a cursory level of consideration in the past (Amirkhanyan et al., 2024; Meier, Song, et al., 2022; Meier & An, 2020), and the application of it to foreign aid provision has not been investigated. A key distinguishing characteristic among nonprofits is whether they are religious or secular. Religious nonprofits can be further sub-divided into generic faith based and church (or denomination) affiliated nonprofits.^[2] We distinguish among these categories of nonprofits in our survey experiment to analyze for potential of bias both against faith based organizations and those that are directly linked to a specific denomination. Our research is important because nonprofits are commonly enlisted by governments to implement policy on the ground (Bielefeld & Cleveland, 2013; Ebaugh et al., 2005). Although the perception might be that faith based nonprofits and humanitarian aid is a developing nations’ phenomena, they frequently participate in developed countries as seen with the operation of numerous nonprofits in the United States after Hurricane Katrina (Eikenberry et al., 2007) and their role in delivering a wide variety of other public services (Amirkhanyan et al., 2009; Feiock & Andrew, 2006). Internationally, these nonprofits frequently provide services that host governments are either unwilling or unable to offer.

Our experimental scenario involves foreign aid delivery in the country of Somalia. Somalia hosts a variety of nonprofits from across the globe, and these organizations are sub-contracted by governmental organizations, such as USAID, to implement developmental policies on the ground (Steenland, 2011; USAID, 2018). The US, only one of many countries involved in aid to Somalia, provided $430 million dollars in aid in 2018 (CRS, 2019). The situation in Somalia is replicated in a wide range of other countries and thus can be informative for policymakers.

Our research also dovetails with related research on non-government organizations (NGOs) and foreign aid programs. Some of this research has examined whether restrictive policies by host governments can reduce funding to foreign NGOs (Bromley et al., 2020; Oelberger & Shachter, 2020). The literature has also examined whether foreign aid can then improve social capital, Gross Domestic Product, and even voting (Das & Sethi, 2020; Dupuy & Prakash, 2020; Karanda & Toledano, 2018; Mallik, 2008). Finally, the literature has examined whether foreign aid dollars may become dominated by the wealthy and thereby reflect the interests of the powerful (Saunders-Hastings, 2018). Our research should inform those operating in these adjacent literatures and create new synergies in the process.

Literature Review

All organizations develop reputations, and the literature on the reputations of public organizations (Carpenter & Krause, 2012) can serve as a model for examining nonprofit organizations. Some public organizations, such as National Aeronautics and Space Administration (NASA), have strong reputations because of their technical capacity; other organizations are sometimes regarded as functionally inept, such as the US Immigration and Customs Enforcement (i.e. ICE) (Pew, 2020). Research on individual organizations is important, but stereotypes of organizational categories may cast shadows over entire sectors of the economy. In other words, macro-level biases may be more significant than micro-level data.

Research in relation to public sector stereotyping has been rising (Hvidman, 2019; Hvidman & Andersen, 2016; Marvel, 2015; Meier et al., 2019; Meier & An, 2020), and public services are often delivered by nonprofit and for profit organizations rather than directly by government. Thus far there has been only modest nonprofit research on stereotyping in the public administration literature (Amirkhanyan et al., 2024; Meier, Song, et al., 2022), but some research has focused on perceptions of benevolence and warmth among different categories of hospitals, especially with nonprofits. Drevs et al. (2014) found in a survey experiment conducted in Germany that nonprofit hospitals were perceived as being more trustworthy and displaying greater warmth, although for-profit hospitals were perceived as being more competent. In a similar series of experiments on day care centers, recycling organizations, and nursing homes, Xu (2020) found that for profits were not perceived as more competent, and he attributed this finding to the profit seeking motives of the latter sector.

Some researchers regard nonprofit organizations as being more effective at the provision of certain services than government agencies (Van Slyke & Roch, 2004). Like government agencies, nonprofits have also been experiencing increased pressures, beginning in the 1990s, to adopt for profit managerial practices (Salamon, 1995). These pressures reflect one view held by some in academia that nonprofits vary so much that this sector is merely a tax status rather than a fundamentally unique form of organization (Meier & An, 2020). At the same time, nonprofits have been found to be more likely to engage in technological investments than for-profits (Freedman & Lin, 2018). From these conflicting views, one article in the literature on nonprofits has found that stereotypes about bureaucracy can transcend into the 3rd sector. Through public opinion surveys, Van Slyke and Roch (2004) found that nonprofits that contract with the federal government to provide services are thought to be part of the government when performing poorly. In short, people project their stereotypes of inefficient government agencies onto nonprofits when they perform poorly, but they may recognize them as being a different organization under more positive circumstances. However, this is arguably not an example of explicit stereotyping but one of misidentification that hints at stereotyping.

There has also been one research article that examines the sub-categorization of nonprofits in depth. Seemann et al. (2015) focus on the perceived differences between secular and religious nonprofits with an emphasis on the healthcare market. The authors find through a survey of German citizens that religious affiliation with hospitals increases perceptions of trustworthiness but not competence. The authors conclude that religious nonprofits have a branding advantage over secular nonprofits and that they should seek to emphasize this to potential clients. Seemann et al. (2015) recommended that this study be replicated in other markets because of issues with external validity. Germany is a country with low church attendance but some overlap in institutional functions between government and churches (e.g. revenue generation). This contrasts with countries such as the US that have relatively high church attendance but a clear formal separation between church and state. In our study we explore the religious vs. secular distinction too, but we do it in the United States and emphasize organizational effectiveness, the general emphasis of the performance management literature (James et al., 2020), which differs from warmth, trustworthiness, and competence.

While our article focuses on perceptions and public opinion, there has been some research on nonprofits and actual outcomes. This line of research focuses on the two central goals of nonprofits: (1) providing services of high quality that are (2) then made accessible to those in great need (Robbins, 1987). In short, the goals are quality and accessibility. Starting with this organizational mission, researchers have explored differences between separate categories of nonprofits. With respect to quality, faith based nonprofits have been found to have fewer program deficiencies than other types of nonprofits (Ragan, 2004). This finding is in line with Jacobs and Polito’s (2012) findings that upper management of faith based oriented nonprofits are efficiency oriented. Other evidence, however, has shown that there is no difference in quality between secular and faith based nonprofits (Reingold et al., 2007). In contrast, Kennedy and Bielefeld (2006) found evidence that secular nonprofits were the better performers. These findings suggest that the relative advantages of secular and faith based nonprofits vary, and that context and industry matters in determining relative performance of nonprofits. With respect to accessibility, nonprofits that are faith based organizations have been found to rarely turn away those in need (Eisinger, 2002), but there is also clear evidence that they do in some circumstances (Mackenzie-Liu et al., 2022). Moreover, Reingold et al. (2007) found that the most underprivileged members of society are likely to receive the most support at such nonprofits. Wuthnow et al. (2004), however, found mixed evidence that suggests that past findings are not absolute; and Amirkhanyan et al. (2009) found evidence against it. In sum, there is a noteworthy amount of conflicting evidence in relation to the performance of different types of nonprofits suggesting the need for more research.

Theory

A variety of possible organizations are involved in foreign aid implementation. Different organizations bring different goals and motives, depending on their skills, history and ideology. Faith based nonprofits in particular may be perceived by the public as having a religious agenda in addition to their service goals (see Mackenzie-Liu et al., 2022). As George W. Bush aptly stated, they have “purpose-driven activities” (Amirkhanyan et al., 2009, p. 492). This agenda may be described as exhibiting common religious features, such as benevolence, mercy, and tolerance, but may also contain an absence of tolerance for some individuals (Davey et al., 2021; Mackenzie-Liu et al., 2022). One principal concern that may arise is that different religions might exhibit these values in different levels or even not at all. Examining a variety of dissimilar religions, therefore, would be necessary when trying to examine any perceived biases.

In our study, we are primarily concerned with one performance dimension: effectiveness. We view effectiveness as being a function of two things: competence and persistence (Semeijn et al., 2014; Wilson, 1989). Because most faith based nonprofits rely heavily on volunteers, they may be perceived as lacking expertise and professionalism. Even though this may not be true, the perception may hold up and have a negative impact on perceived competence. Persistence is likely to be a positive attribute of religious nonprofits that act at a different intensity than their secular counterparts (Amirkhanyan et al., 2009). Moreover, religious nonprofits may provide more individualized care, more direct care, and provide long-term commitments to service recipients (Amirkhanyan et al., 2009; Graddy & Ye, 2006). This leads to cross-cutting features for assessing whether faith based nonprofits are viewed as being more effective than their secular counterparts in the nonprofit sector. In sum, they may be perceived as less competent but more persistent.

The perceptions of effectiveness might also differ among faith based organizations that have affiliation with a specific religious group and those that are faith based without a tie to a specific religion. Ties to a specific religious group bring the advantage of stronger normative ties to the organization as institutionalized by the church structure (persistence) but at the same time might limit access to expertise depending on the composition of the church membership.

Hypothesis 1: Faith based organizations will be viewed as more effective than secular organizations.

Hypothesis 2: Denomination based organizations will be viewed as less effective than secular organizations.

Religions vary in doctrine, evangelism, actions, and behaviors that can readily be observed by the general public. These differences may form stereotypes in popular culture, but they may also differ based on their institutional characteristics. One academic distinction classifies religions on the orthodoxic-orthopraxic dichotomy (King, 2003; McKim, 1996; McKnight, 2007). Orthopraxy emphasizes correct action, whereas orthodoxy emphasizes correct belief. Religions can exist at one extreme, but most fall somewhere in between on this continuum. As an example, most Protestant denominations do not have much ritualistic behavior; however, these same churches may have very strict beliefs. As a result, these religious organizations would be very orthodoxic. Judaism, on the other hand, has some very ritualistic behaviors, pushing it toward the opposite end of the continuum towards increasing levels of orthopraxy.

Similarly, orthopraxic religions may have better networking opportunities, increasing access to better levels of expertise. As a result of these two factors, orthopraxic religions will be hypothesized as more effective at what they do. While these academic distinctions are unlikely to be foremost in the public’s mind, their correlations with behaviors, efforts to assist others, resistance to policy issues, or association with unpopular causes are likely to be viewed by the public and form the basis of judgements about the religion or organizations linked to that religion. In short, the behavioral manifestations of these abstract orientations provide a wealth of information that an individual could use to form perceptions about a faith based nonprofit. Given the greater orthopraxic orientation of Islam and Catholicism than Protestantism in general, we offer the following hypotheses as a first step in determining if denomination based nonprofits vary in stereotypes perceptions:

Hypothesis 3: Catholic organizations will be viewed as more effective than those affiliated with Protestant denominations.

Hypothesis 4: Islamic organizations will be viewed as more effective than Christian organizations.

Empirical Analysis

We examine public reactions through an experiment that is characterized by variation in nonprofit typology and organizational success. Our survey experiment has two 3x3 between-groups factorial designs to assess how the American public conceptualizes nonprofit organizations and rates their performance (N = 617 and 732). The survey participants were drawn from an online convenience sample via Amazon’s Mechanical Turk (hereafter “MTurk”). There have been some concerns about MTurk stressed since its inception as a tool for experimental research; however, the reliability of MTurk has been substantiated through the successful replication of numerous major American surveys (Berinsky et al., 2012). The potency of these findings also suggests a noteworthy degree of generalizability for associated findings (Mullinix et al., 2015). Moreover, empiricists can augment these findings with practical moves in implementation to exclude foreign respondents, VPN users, and access by bots. We implemented these exclusions in our experiments. In our vetted group of participants, individuals were randomly assigned to one of 9 experimental conditions for each experiment, and the findings for both experiments remained robust when controlling for a swathe of demographic covariates. Our findings center upon two separate but related experiments. The first experiment compares secular and religious organizations, and the second experiment compares different religious denominations. Experiments have the advantage of mitigating several of the common methodological concerns in research. The random assignment of treatment conditions guarantees that the independent variables (the treatments) are exogenous to respondents’ perceptions (Dague & Lahey, 2019). Because the treatments are random with respect to each other, collinearity issues are not a concern (Mutz, 2011), and the independent variables are established externally and thus cannot be affected by common source bias (Meier and O’Toole 2015). Both experiments were approved by the Institutional Review Board of Texas A&M University. We will examine each of these experiments separately below.

A). Experiment 1: Secular vs. Religious Comparison

For our first experimental test has two treatments. Our first treatment is on organizational status to probe sector bias and faith based organizations. Respondents are randomly divided into the three groups: (1) a control group where the organization has no religious affiliation (that is, secular), (2) the organization is a “faith based” organization but does not have any specific denominational affiliation, 3) and the organization is directly affiliated with a denomination. In our case, we chose the Methodist church as our religiously affiliated link because of its widespread reputation to openness that should inspire the least opposition by the broader public.

Our second treatment added organizational performance randomly assigned in three groups. The negative performance information group was told that health conditions had stagnated on the ground in Somalia and that a US Federal Agency rated the respective nonprofit 2 out of 5 stars in their performance assessment. The control group was given no performance information. The positive performance information group was told that health outcomes had improved, and the same Federal Agency assigned a 4 out of 5 stars rating to the relevant nonprofit in the vignette. Both positive and negative assessments were included given the frequent finding linked to prospect theory (Kahneman & Tversky (2013) that negative information is considered more salient than positive information (Hong, 2019; Olsen, 2015).

The core vignette for our experiment is below. The purpose of this vignette was to stress the operational context of the organization and its basic structure.

Hope International, (insert sector cue here), has 235 full-time and part-time employees operating a major project in Somalia. The organization’s goal is to provide medical services, taking into account the special needs of the individual people on the ground. This nonprofit organization is organized into three divisions. The organization’s top management division is performed by a management team consisting of an operations director, a chief accounting officer, and chief medical officer. The organization’s central administrative division is responsible for documenting that the organization meets management’s demands for safe and efficient services on the ground. This task involves a comprehensive system of policies and standards in all areas of foreign aid provision.

Negative Performance Cue: United States Agency for International Development (USAID), a federal government agency, has observed that infant mortality rates have stagnated since the first phase of Hope International’s operations were completed in Somalia. Moreover, the need for food aid in Somalia has remained the same. Based on qualitative metrics, the USAID awarded Hope International 2 stars out of 5 for the organization’s current performance.

Positive Performance Cue: United States Agency for International Development (USAID), a federal government agency, has observed that infant mortality rates have improved by 20 percent since the first phase of Hope International’s operations were completed in Somalia. Furthermore, the need for food aid in Somalia has dropped considerably as general health services have greatly increased. Based on qualitative metrics, the USAID awarded Hope International 4 stars out of 5 for the organization’s excellent performance.

In response to the previously mentioned vignette and associated treatments, the survey participants were then asked to rate the organization on multiple dimensions of performance. The questions are listed in Table 1 and drawn from the experimental literature on public program performance (Hvidman & Andersen, 2016; Meier & An, 2020). Participants were asked to rate the organization on a scale of 1 through 7 with 1 indicating “strongly disagree,” 4 “neutral,” and 7 “strongly agree.”

The dimensions of performance were then used in conjunction with principal components analysis to arrive at a latent variable. See Table 2. All questions loaded successfully on one factor with an eigenvalue of 3.3 and a Cronbach’s alpha of 0.9052. The resulting factor scores were then used as our principal dependent variable in this analysis.

A set of demographic questions and manipulation checks were included in our study to ensure robustness of our empirical findings. Finally, in order to test for balance across the treatments, we calculated the F-test of the difference of means across groups and found only one modest problem area (ideology) although it is important to note that recent statistical work suggests that if randomization was used that balance is not a major factor (Mutz et al., 2019). Moreover, robustness checks that included all covariate and interactions found this issue to be an anomalous finding with no impact on any results. The balance table is listed in Table 3.

We use Ordinary Least Square (OLS) regression modeling to analyze our empirical data and our models are presented in Table 4. Model 1 in the table includes dichotomous variables for whether the observation is Methodist affiliated or whether it is a faith based organization. A secular organization (i.e. non-religious) is the excluded base category in Model 1. Model 2 in the table includes dichotomous variables for whether the observation received a positive performance information cue or a negative performance information cue. The no information cue is the excluded base for Model 2. Model 3 combines the variables from Models 1 and 2 into a single model. Findings are consistent across these first 3 models. The variable for the Methodist church, although negatively signed, is not statistically significant at the 0.05 level, but the variable for faith based organizations is statistically significant at the 0.05 level. A noteworthy takeaway here is that the coefficient is negative for faith based organizations, signaling that the public perceives them to be less effective than secular nonprofits that are performing at the same level (about two tenths of a standard deviation less). Furthermore, the models give us intuitive findings that a negative performance information cue will harm perceptions of effectiveness, and a positive performance information cue will boost perceptions of effectiveness. Consistent with the logic of prospect theory and the resulting negativity bias, the absolute value of the negative cue is almost twice as large as the absolute value of the positive cue. The difference in absolute magnitude of the two performance coefficients is statistically significant (t = 2.32).

Model 4 is a robustness check that incorporates an array of demographic covariates in addition to those variables used in Model 3. These covariates are: age, political ideology, sex (female), race (white), and religious attendance (measured on a four point scale from 0 never to 3 weekly). Political ideology and religious attendance are both statistically significant at conventional levels, but they do not affect the core findings of this article. The primary takeaway from Model 4 is that the previous findings are unaffected by these potentially confounding variables.

In Model 5, we incorporate a series of multiplicative interaction terms to test whether there is an indirect effect in terms of bias against faith based organizations. Essentially these tests indicate whether faith based organizations might be getting less benefit from positive assessments (or greater penalties for negative assessments) than secular organizations (see Marvel, 2015 on this form of bias). Between our two organizational cues and two performance cues, we have a total of four interactions. These interactions will determine if individuals are more likely to discount either positive or negative information depending on the sector of the organization. Most importantly, we find zero evidence for there being any statistically significant interactions in Model 5. Substantively, this means that information on performance had the same impact on the overall evaluation of the organization regardless of whether the organization was secular, faith based, or Methodist affiliated. No type of organization got more or less credit for performance than another type of organization.

In sum, we find strong evidence in two areas. The first area is organizational typology. Faith based organizations suffer when it comes to effectiveness in the eyes of survey participants (however this affects only generic faith based organizations and not those affiliated with the Methodist church). The second area involves performance information cues. Positive performance information has a positive impact on effectiveness. On the opposite end of the spectrum, negative performance information cues have a negative impact on perceived effectiveness. These latter findings are both interesting and highly intuitive; although it is important to note that positive information often leads to null findings in the literature, a finding that is attributable to negativity bias (Hvidman & Andersen, 2016; Meier et al., 2019; Olsen, 2015).

B). Experiment 2: Religious Denomination Comparison Experiment

The second experiment seeks more detail on the evaluation of religious affiliated nonprofits using the same general vignette. Similar to the first experiment, there are two treatments that influence the outcomes of this next experiment. Our first treatment is on organizational denomination and individuals were told that the organization was 1) Catholic, 2) Methodist, or 3) Muslim. Our second treatment, organizational performance, was identical to the first experiment with individuals given 1) no information on performance (the control group), 2) the negative performance cue (stagnation and a 2 star rating), or 3) the positive performance cue (improvement and a 4 star rating).

In response to the previously mentioned vignette and associated treatments, the survey participants were then asked to rate the organization on multiple dimensions of performance. The questions are listed in Table 1. Participants were asked to rate the organization on a scale of 1 through 7 with 1 indicating “strongly disagree,” 4 “neutral,” and 7 “strongly agree.”

These dimensions of performance were then used in conjunction with factor analysis to arrive at a latent variable. See Table 6. All questions loaded successfully on one measure with an eigenvalue of 2.98 and a Cronbach’s alpha of 0. 8837. The resulting factor scores were then used as the dependent variable in the second portion of this analysis.

An assorted set of demographic questions and manipulation checks were included in our study to ensure robustness. Finally, in order to test for balance in this second experiment, we calculated the F-test of the difference of means across group - and found no problem areas. The balance table is listed in Table 7.

The regression results are presented in Table 8. Model 1 in the table includes dichotomous variables for whether the observation is Catholic church affiliated and whether it is Muslim affiliated. The Methodist church is the excluded base category in Model 1. Model 2 in the table includes dichotomous variables for whether the observation received a positive performance information cue or a negative performance information cue. The no information cue is the excluded base category for Model 2. Model 3 combines the variables from Models 1 and 2 into a single model. Findings are consistent across these first 3 models. Most importantly, none of the religious variables are statistically significant at the 0.05 level. While this indicates that none of these three denominational factors appears to affect public perceptions, they should also be interpreted in light of the previous experiment which indicates a more positive view of secular nonprofits.^[3] The variable for positive information, although positively signed, is not statistically significant at the 0.05 level, but the variable for negative information is. As previously noted, this is a finding that is consistent with the literature on negativity bias.

The addition of potentially confounding covariates in Model 4 produces no impact on our findings. Only the negative performance information cue remains statistically significant at the 0.05 level. Covariates were statistically significant similar to that of the first experiment. Moreover, the incorporation of interaction terms in Model 5 led to no significant findings once again. Each of the interaction terms were not statistically significant at conventional levels, thus we can conclude that there are only direct effects. And more specifically, only negative performance information matters here.

Conclusions and Discussion

With one exception, we fail to support our hypotheses in regard to the religious status of nonprofits operating in the area of foreign aid. Although the experimental participants appear to view faith based organizations as inferior service providers (at least with respect to organizational effectiveness), they do not appear to distinguish among the various church organizations that might sponsor nonprofits (Catholic, Methodist, Islamic). The actual performance of the organization (particularly if the performance was negative) played a larger role in the evaluation of effectiveness than did the religious affiliation. Nor did religious affiliation influence the credibility of the performance information as assessed by the interaction terms.

Theoretically, inferior human resources may have outweighed any gains from organizational persistence in the eyes of the public and resulted in the lower evaluations of generic faith based organizations. On the other hand, if so then one should expect all religious nonprofits to be sanctioned by experimental participants on the same grounds if these theoretical dimensions are used to assess the organizations. Naturally, one would expect denomination based nonprofits and faith based nonprofits to benefit and suffer from the same traits. An equally feasible explanation is that respondents have no predispositions in regard to the effectiveness of denominational based nonprofits and place any biases in regard to an individual sect aside when dealing with a nonreligious issue such as humanitarian relief. That could especially be the case since it is unlikely that many of the respondents have first-hand experience with any nonprofits operating in Somalia. Another possibility is that three denominations were generally very common and might not be associated with specific traits that a less orthodox sect would be.

Any study with predominantly null findings should consider factors that might have reduced the salience of the treatment (in this case whether faith based or denomination based nonprofits were delivering public services). The experimental context was Somalia and involved humanitarian aid. Humanitarian aid might well be perceived as a generally a positive thing regardless of who delivers the aid (and would be very consistent with the basic philosophy of most religions and thus shared by the respondents). It does not raise political or policy issues like using faith based organizations where religious values might conflict with policy values (e.g., family planning, foster care, transgender rights, etc.). In humanitarian aid the objective might simply overwhelm any existing stereotypes. In this particular case, the experiment indicates a clear policy implication; humanitarian aid is evaluated more on whether those in need get aid than who is delivering the aid.

This study is not without limitations. Although we examined both generic faith based organizations and those associated with specific religious groups, we did so with respondents from one country (US) with an example from only a single country (Somalia) and for only three major religious groups (Catholic, Methodist, Islamic) for a single policy area (humanitarian based foreign assistance). We encourage replication of this paper in different contexts. In terms of the locus of respondents, our findings certainly might differ across countries where religious cleavages are more salient, so external validity of any experiments should be of concern. Reasons for these differences center on these findings being based on stereotypes. Stereotyping should differ based on the degree of religiosity across different countries. Also, institutional differences in countries may also shape stereotypes and perceptions of performance of public programs (Meier et al., 2017). Examples of this may be seen with official religions in northern Europe and Germany’s taxation system being used to collect tithes directly from religious residents. Similarly, the public might have stronger preferences about service delivery within their own country than in terms of distributing humanitarian aid to another country.

Similarly, the plight of Somalia is well known, and this salience might have resulted in more positive responses to any type of organization that is providing relief aid. Responses might vary both by which foreign countries are targets of aid and whether the aid focuses on international or domestic recipients. The provision of medical and food aid is also relatively noncontroversial and may overcome any negative stereotypes held by the public. More controversial activities such as in family planning or those linked to proselytizing could generate stronger responses. Finally, the concept of faith based organizations covers a wide range of possibilities, and the individual religions used also vary significantly internally. It is possible that less mainstream sects might engender a more negative response. In short, replications are needed in a variety of different contexts using different groups of subjects, targets, and religious organizations to provide a fuller picture of stereotyping of faith based nonprofits. The external validity will be highly dependent on context and determining the boundary conditions for studies such as this one.

Conflicts of interest

The author(s) declare that there are no conflicts of interest that relate to the research, authorship, or publication of this article.