Do Faith-Based Foster Care Agencies Respond Equally to All Clients?

Introduction

The passage of the Charity, Aid, and Recovery Act of 2002 is often seen as the watershed moment when Faith Based Organizations (FBOs) began playing a larger role in the provision of public services in the United States. By some estimates, congregations, religious charities, and other FBOs are now the third largest component of the non-profit sector, only after secular health and educational organizations (Vindal et al., 2001).

FBOs have provided a wide range of social services throughout American history, either directly or by contracting with government. In particular, FBOs have long played a foundational role in providing child welfare services in the United States. Faith based agencies, such as the Child Aid Society, which was founded in 1853, were the first to provide nascent forms of child welfare services prior to the development of the modern social welfare state. Today, almost all states contract with FBOs to provide foster care services. Services can include directly caring for and housing children, training and certifying prospective foster parents, and placing children in foster homes. Many FBOs also provide domestic and international adoption services.

In recent years, FBOs that provide child welfare services, as well as other health and human services, have been at the center of highly contentious (and highly litigious) policy debates regarding the extent to which FBOs should be exempt from providing, or be compelled to provide, services to all individuals regardless of background. Several states have passed targeted religious exemption laws that allow child welfare providers, notably foster care and adoption agencies, to refuse to provide services to individuals based on deeply held religious beliefs (Movement Advancement Project, 2021).^[1] These exemption laws allow for FBOs to potentially exercise discretion regarding which clients or individuals with whom they work.

Given this policy landscape, and the continued importance of FBOs in child welfare, it is critical for policymakers and researchers to evaluate the extent to which FBOs may be more (or less) responsive and accessible to certain clients relative to others. The focus of this paper is to examine the extent to which religiously-affiliated child welfare agencies, specifically religiously-affiliated foster care agencies, respond to same-sex couples differently from heterosexual couples.

In this paper, we use data from Mackenzie-Liu, Schwegman, and Lopoo (2021), an email correspondence study which tested whether foster care agencies responded similarly to same-sex and heterosexual couples. Whereas Mackenzie-Liu et al. (2021) was focused primarily on examining if foster care agencies discriminate against same-sex married couples, this paper examines whether religiously-affiliated foster care agencies treat same-sex couples differently from secular and public foster care agencies.

We find that religiously-affiliated foster care agencies are over 14 percentage points less likely to respond to same-sex male couples compared to heterosexual couples. However, due to a number of statistical limitations which we detail below, this finding is not robust. We find no evidence that religiously-affiliated foster care agencies respond differently to same-sex female couples relative to heterosexual couples.

Despite our limited ability to isolate robust correlations between organizational characteristics and measures of discrimination, this paper contributes to the literature on social equity and public sector discrimination in several ways. First, we examine a context in which there is a high degree of organizational and bureaucratic discretion: child welfare services. Private foster care agencies have broad discretion regarding with whom they work, and religious exemption laws have reinforced (and provided legal protections for) this discretion. Moreover, in June 2021, the U.S. Supreme Court ruled in Fulton v. City of Philadelphia that, under certain circumstances, municipal governments cannot penalize religious non-profits for refusing to certify same-sex couples as foster parents. Despite this broad and expanded discretion afforded to private foster care agencies, little is known about how religiously-affiliated foster care agencies, or other FBOs, treat same-sex couples.

Second, this paper contributes to a rapidly expanding literature on public sector discrimination in both the United States and Europe (Ahmed & Hammarstedt, 2019; Butler & Broockman, 2011; Einstein & Glick, 2017; Giulietti et al., 2019; Jilke et al., 2018; Mackenzie-Liu et al., 2021). To our knowledge it is the first paper that explicitly examines the extent to which religiously-affiliated non-profits make their publicly funded or subsidized services more or less accessible to certain clients based on a client’s personal attributes. In doing so, this paper contributes to a much smaller literature that documents, using data from field experiments, how organizational attributes influence or are correlated with organizational behavior and responses to different types of clients (Jilke et al., 2018).

Third, our study also highlights the limitations of conducting experiments in the public sector. This is particularly true in contexts in which there is a finite number of organizations providing a particular service, when data collection is costly, and where there are ethical concerns about imposing a burden on organizations providing public services. In certain contexts, it is either impossible to collect a sufficient sample size or would arguably impose unethical burdens on audited organizations. Thus audit studies examining heterogeneity in discriminatory behavior across different types of organizations may be infeasible (or less feasible) in certain public sector contexts (Balfe et al., 2021).

Despite these limitations, it is increasingly important for scholars of public administration and public policy to examine intersectional discrimination based not only on group membership (e.g., race, ethnicity, gender), but also based on organizational type (e.g., secular-private, religious-private, and public), funding source for the public service (e.g., own-source revenue v. state aid), demographic composition of the local bureaucracy, etc. While exploring this intersectional discrimination may be limited in certain contexts, understanding how and why organizations and public servants are more or less likely to respond to particular groups is an important first step in designing interventions or crafting policy to reduce this discrimination.

Faith Based Organizations and Child Welfare in the United States

The child welfare system in the United States has developed organically, albeit somewhat haphazardly, over time through a network of organizations, including public organizations, faith-based and secular organizations, and private families. The faith-based provision of child welfare services predates the public sector. Until the 19^th century, child welfare was largely seen as a private matter for which the state should not intervene (Sethi, 2019). As such, social services were almost exclusively provided by faith-based organizations (Cnaan, 1999).

Over the course of the 19^th century, states and local communities constructed orphanages, which were considered to be a more humane alternative to almshouses or alternative forms of child “welfare” programs designed to use children as cheap labor (Rymph, 2017). The vast majority of orphanages were privately constructed and privately funded and thus these institutions could be selective in the children they chose to serve. Many orphanages only served specific religious communities, or they were for whites only (Rymph, 2017). By 1880, there were over six hundred orphanages in the United States serving more than 50,000 children (Hacsi, 1997).

Beginning in the 20^th century, the child welfare system became more formalized as the role of government in the lives of families changed. The child welfare system no longer primarily existed to house orphans, but instead sought to protect children from maltreatment and neglect. States were given the authority to remove children from their parents’ care. State child welfare agencies (and thus the state itself) assumed care and welfare responsibilities for the child. The public child welfare agency would then place the child in an out-of-home environment, preferably with a family.^[2] The social welfare agency oversaw screening, supervising, and licensing foster parents with whom they placed children.

The late 20^th century was a time of increased privatization of public services and the rise of “public choice,” which continues with some segments of the public sector to this day. From 1977 to 1997, the number of public charities more than doubled (Berry & Arons, 2005). Many of these public non-profits were faith-based. Advocates of faith-based sector involvement argued that faith-based organizations were able to uniquely deliver services given the prominence of religious institutions in American’s daily lives and their distinctive ties to the community (Hula et al., 2008). FBOs, it was argued, could act as effective mediating structures, which are defined by Berger and Neuhaus (1977) as “[institutions] that [stand] between the individual and his private life and the large institutions of public life.” While government institutions were viewed as ineffective and impersonal, FBOs could be more effective and accessible to their clientele.

In the early 2000s, the U.S. Department of Health and Human Services described FBOs as “uniquely situated” to provide services to individuals in need (Fischer, 2008). This trend in increasing reliance on FBOs continued after the passage of “Charitable Choice” (Amirkhanyan et al., 2009. In the modern era, foster care is jointly administered at the state level by both the public and non-profit sectors, with both receiving public funding from the Federal and state governments. Many of the private non-profits that work in this space are religiously affiliated.

Critics of FBOs raise concerns over the separation of church and state. Despite also being supported by private contributions, a majority of funding for FBOs comes from taxpayer dollars. The influx of federal funding also generated concerns regarding who these institutions would serve. In the 1960s, federal regulators were met with some resistance when they began to require nonsectarian intake policies. As described by Rosenthal (2000), many religiously-affiliated charities had historically only worked with members of their own religion, for instance Catholic agencies only serving Catholic children. Though many non-profits adopted policies to not discriminate on the basis of race, for example, discrimination on the basis of religion and religious belief was considered different (Rosenthal, 2000). This distinction remains contentious.

The literature has largely focused on differences in organizational structure, perception, and quality between religious and secular non-profits. Graddy and Ye (2006) compare the services offered by secular and religious organizations. They find that faith-based organizations tend to be more concentrated in the services they offer. Graddy and Ye posit that, while faith-based and secular non-profits may use the same service delivery methods, faith-based organizations may be more effective at providing social services because faith provides them with increased internal motivation and because congregations provide a network that can be leveraged for service delivery. Some recent work has examined public perception of religiously-affiliated non-profits compared to other social service providers (see A. Johnson et al., 2021).

The literature also examines differences in the effectiveness of religious and non-religious institutions. Etindi (1999) argues that religiously-affiliated non-profits may be more likely to make longer-term commitments to clients, they may engage in more personalized interactions with clients, and they may provide more personalized care. For these reasons, faith-based non-profits may be more effective than their secular counterparts. However, the results from other studies are mixed. Many scholars find no differences in performance. These conclusions span a diverse group of sectors and metrics, including nursing home accessibility (as measured by ability to pay) (Amirkhanyan et al., 2009), as well as incarceration and recidivism (Brazzell & La Vigne, 2008; B. R. Johnson & Larson, 2003). Kennedy and Bielefeld (2006) suggest secular providers are more effective than religiously-affiliated providers, while Weisbrod and Schlesinger (1986) find the opposite.

Several studies emphasize that FBO and secular organizations excel in different aspects of service provision. For example, Monsma and Soper (2003) find that secular welfare-to-work programs in Los Angeles have higher placement rates while faith-based organizations were more likely to receive better ratings by their clientele. Many of these studies raise concerns over how to best measure service quality and are plagued by issues of selection bias.

There is a notably smaller literature examining if FBOs and secular non-profits tend to serve different types of individuals, but again, the findings are not conclusive. Deb and Jones (2004) find that FBOs serve clientele that are similar in terms of race, gender, and education, as other providers. In contrast, Reingold, Pirog, and Brady (2007) find that individuals who receive services from FBOs in Indiana are more likely to be older, white, and married. Both studies are only able to capture who receives services and not how accessible their services are. Amirkhanyan et al. (2009) measure accessibility as ability to pay. While this proxies for race and ethnicity in certain localities in the United States, they were not explicitly considering the demographic aspect of accessibility.

Lipsky and Smith (1989) theorize that one should expect differences between FBOs and non-FBOs. While both nonprofit and government service providers are subject to bias in client selection, nonprofits are more likely to be mission-oriented and therefore are more likely to self-select clients who are in line with that mission. Religious non-profits, therefore, may focus on serving individuals either from their own specific community or whose beliefs or lives are in line with their religious teaching. To our knowledge, no study has explicitly examined the extent to which religiously-affiliated non-profits are more or less accessible to certain clients measured by a client’s demographic characteristics.

This paper contributes to this gap in the literature by testing the degree to which religiously-affiliated foster care agencies provide differential treatment to potential foster care parents on the basis of sexual orientation. In a correspondence study, the most obvious form of differential treatment is whether inquiries made to an organization are returned at different rates for one group relative to another (e.g., same-sex couples compared to heterosexual couples). However, by examining the content of the responses, one can examine if certain groups create more administrative burden for one group relative to another by selectively increasing search and compliance costs. Thus, in this paper, we also test if religiously-affiliated foster care agencies create more administrative burden for same-sex couples by providing less detailed responses which may discourage them from engaging in the foster care certification process more than heterosexual couples.

This paper posits that religiously-affiliated non-profits are less likely to respond to same-sex couples compared to heterosexual couples for two reasons. First, as described by Lipsky and Smith (1989), non-profits are more likely than public institutions to be mission-based and select their clients accordingly. Despite a recent dramatic increase in acceptance of homosexuality in American society overall, homosexuality is still considered against the tenants of some faiths. The United Methodist church, for example, one of the largest Protestant denominations in the United States, is currently breaking into two denominations over the question of whether to ordain and marry individuals who identify as members of the LGBTQ community. As a consequence, we expect some religious non-profits to choose not to reply to same-sex couples.

Second, as noted above, religious non-profits have received legal protections in a number of states that enable them to legally refuse to work with same-sex couples. Since the early 2000s, several U.S. states have adopted “targeted” religious exemptions that specifically allow state-licensed child welfare agencies to refuse to provide services to individuals and families, notably members of the LGBTQ community, if doing so conflicts with a deeply held religious belief. While North Dakota adopted their exemption in 2003 (North Dakota Century Code §50-12-07.1), most states have adopted their targeted religious exemption since 2010: Virginia (§63.2-1709.3 in 2012), Michigan (HB 4189-4190 in 2015), Mississippi (HB 1523 in 2016), Alabama (Alabama HB 24 in 2017), South Dakota (SB 149 in 2017), Texas (HB 3889 in 2017), Kansas (SB 284 in 2018), South Carolina (HB 4950, §38.29 in 2018), and Tennessee (HB 836 in 2020).^[3]

Three states are particularly interesting, and highlight the contentious, litigious, and confusing nature of these religious exemption laws and how they conflict with agency policies or other anti-discrimination statutes. Michigan has both an agency policy prohibiting discrimination on the basis of sexual orientation, as well as a statewide religious exemption. There is a pending court case that will decide the status and applicability of the religious exemption there. South Dakota has an agency policy that prohibits discrimination against all individuals based on sexual orientation or gender identity, as well as a law (SB 149 passed in 2017) that allows state-licensed child welfare agencies to refuse to provide services to children and families if doing so conflicts with their religious beliefs. Nebraska had an agency policy that restricted placing children in the homes of “persons who identify themselves as homosexuals” (from NE DHHS Administrative Memorandum #1-95, 1995 quoted in MAP, 2021). However, this agency regulation was officially declared unconstitutional in 2017 (MAP, 2021).

Currently, 28 states either have regulations or agency rules or policies that prohibit discrimination against foster parents based on sexual orientation, or child welfare agencies are prohibited from discriminating on the basis of sexual orientation. The remaining states have no explicit protection for same-sex couples in the child welfare sector (MAP, 2021). It is also possible that instead of failing to respond to individuals with whom they do not wish to work, FBOs may respond more positively to individuals with whom they do wish to work. Our data allow us to test this hypothesis as well.

Even in the absence of these protections, religious foster care agencies have been willing to explicitly declare their refusal to work with same-sex couples based on deeply held religious beliefs. For instance, Catholic Social Services of Philadelphia (CSS), which is part of Catholic Charities USA^[4], states that it holds two beliefs: (1) marriage is a sacred bond between a man and a woman, and (2) the certification of prospective foster families is both an implicit and explicit endorsement of their relationship. Due to these beliefs, in 2018, CSS stated that it was not willing to certify same-sex married couples or unmarried couples (regardless of their sexual orientation). In response, the City of Philadelphia informed CSS that this behavior violated their anti-discrimination policies, and it would no longer refer children to the agency or enter into a foster care contract with it unless the organization began certifying same-sex couples.^[5] CSS sued Philadelphia. This lawsuit (Fulton v. City of Philadelphia) reached the Supreme Court in November of 2020. In June of 2021, the U.S. Supreme Court sided with CSS. This case draws attention to a difficult issue that can arise when religiously-affiliated agencies work as agents of the government: they may treat same-sex couples differently than heterosexual couples.

Data

To examine whether faith-based nonprofits respond to couples with different sexual orientations in the same manner, this paper uses data from an email correspondence study in 2019 (see Mackenzie-Liu et al., 2021), a period prior to the COVID-19 pandemic.^[6] To briefly summarize the field experiment: in the summer of 2019, Mackenzie-Liu et al. (2021) sent two emails, one from a heterosexual couple and one from a same-sex couple, to 1,147 foster care agencies around the United States. These foster care agencies had a publicly posted email address. In these emails, Mackenzie-Liu et al. (2021) signaled sexual orientation by having each inquirer state the name of their spouse. Inquirers who mentioned a spouse with the same gendered name as their own thus signaled that they were in a same-sex relationship.^[7] Previous audit studies have used this procedure to examined discrimination on the basis of sexual orientation (Ahmed et al., 2008; Schwegman, 2019). For more information on the study, please see Mackenzie-Liu et al. (2021).

Identifying Religious Affiliation

We gathered new data on the potential religious affiliations for the foster care agency. To determine if the foster care agency was religiously-affiliated, we did a web search on each one in the winter of 2020/21. This paper uses the following decision rules to determine if the foster care agency had a religious affiliation.

1. All agencies that had a religion in the name (for instance “Christian” or “Jewish”) were determined to be religiously-affiliated.

2. For non-profits that did not have religion referenced in their name, we did a web search on either their mission statement or “About Us” section. Agencies that had religious references in either section were categorized as religiously-affiliated.

3. All public agencies, i.e., those run by individual state’s Department of Social Services were categorized as non-religiously-affiliated.

Based on these decision rules, 12% of the total sample is classified as religiously-affiliated.

There are important limitations to this process. First, foster care agencies are only classified as religious if there is some explicit religious content on their current websites (not their websites as of 2019, when the response data was collected). Second, while in some cases, it is fairly easy to identify the religion (i.e., Christian, Muslim, or Jewish) of the organization, it was not always possible to classify the foster care agency by its denomination (e.g., Catholic), conditional on it being a Christian FBO. Given the likelihood of misclassifying an agency, the relatively small size of the sample that is religious, and the disclosure concerns raised from stratifying the sample based on certain religions, we do not report or stratify the sample based on religion or Christian denomination.

Outcomes

We first consider two outcomes. As with Mackenzie-Liu et al. (2021), and most other email correspondence studies, this study’s primary focus is on the initial step, the first contact a prospective parent has with an agency. Our first outcome, therefore, is whether each email inquiry receives a response (the variable is called “Response”). This action is the clearest signal of interest on behalf of the agency. Next, we consider the number of words in each response (called “Word Count”). We use word count as a measure of the effort exerted to recruit a prospective parent.

The process to become a licensed foster parent is a long, multistep process. Interested parties typically must contact an agency, attend information sessions, receive training, and have their home evaluated before a child is placed in their care. As a result, following Mackenzie-Liu et al. (2021), this paper examines twelve fostering process measures. These are references to later steps in the process by the agency in response to an inquiry. We examine whether the organization forwarded the email internally (“Forward”); if the foster care agency provided information about a home study (“Home Study”), licensure (“Licensure”), or information session (“Session” and “Session Plus”^[8]); if the organization asked for contact information (“Solicits Contact Info”) or provided their own (“Provides Contact Info”); and if someone from the organization asked to set up an appointment (“Sets Appointments”) or offered to talk on the phone (“Talk on Phone”). We also look at if the organization attaches an application (“Application”) or any other attachments (“Attachment”). Finally, we look at whether agencies question the location of the inquirer, as agencies only work with potential families who reside in their region.

For each of these outcome variables, we estimate the following regression model:

$$\begin{align} y_{ir} &= \ \beta_{1}{GayMale}_{i} + \beta_{2}{GayFemale}_{i} + \beta_{3}{Religious}_{i} \\ &\quad+ \beta_{4}\biggl( {Religious}_{i}*\ {GayMale}_{i} \biggr) \\ &\quad+ \beta_{5}\biggl( {Religious}_{i}*\ {GayFemale}_{i} \biggr) + \ \theta_{r} + \varepsilon_{ir} \end{align}$$

where $y_{ir}$ is the outcome of interest for foster care agency i in region r. ${GayMale}_{ir}$ is an indicator variable equal to one if the email contained two male names. ${GayFemale}_{ir}$ is an indicator variable that is set equal to one if the email contained two female names. ${Religious}_{i}$ is set equal to 1 if the foster care agency i is classified as religious. We include a Census region fixed effect, $\theta_{r}.$^[9] These fixed effects control for unmeasured factors common to the different regions of the country thereby reducing the bias in any estimates of the differences in outcomes for same-sex couples relative to heterosexual couples by foster care agency religious affiliation. For instance, certain areas of the country may face a higher demand for foster care or have cultural practices that could influence a foster care agency’s actions. These fixed effects should reduce the bias generated by the inability to control for these factors. The coefficients of interest are β₄ and β₅. These two measures tell us if the difference in response outcomes for same-sex couples relative to heterosexual couples differ if the foster care agency has a religious affiliation.

Results

We begin by examining if religious agencies responded differently than non-religious agencies, regardless of sexual preference. Table 1 shows the mean of all outcome measures, as well as providing descriptive information about the analytical sample. As noted above, 12% of the sample is classified as religious (n = 135), 34% of the sample (n = 386) are public agencies, and the remaining 55% (n = 626) of the sample are secular organizations providing fostering services.

On average, these agencies responded to 55% of all emails. We report the average response rate and email length for all FCAs in column 1, religious agencies in column 2, and non-religious agencies in column 3 of Table 1. On average, non-religious foster care agencies (both publics and privates) respond to 55.3% of all inquiries, whereas religious foster care agencies respond at a slightly lower rate (51.9%). As shown in panel 2 of Table 1, while the average reply from the average foster care agency contains 35 words, religious foster care agencies send slightly shorter responses (32.85 words) compared to non-religious agencies (35.05) words. These differences, however, are not statistically significantly different.

However, in both Panel A and Panel B of Table 1, there is clear descriptive evidence that religious agencies are less likely to respond to same-sex male couples than non-religious foster care agencies. While the average response rate for same-sex males is 49.9% for all agencies (row 3 of Panel A), religious foster care agencies only respond to 37.3% of all inquiries sent by same-sex male couples. This is a 12.6 percentage point difference. Compared to non-religious foster care agencies (column 3), who respond to 51.4% of all inquiries from same-sex male couples, religious foster care agencies are 14.1 percentage points less likely to respond.

Religious foster care agencies are also 3.1 percentage points less likely to respond to same-sex female couples (average response rate is 56.6%) compared to non-religious agencies (59.7%). There is no substantive difference in how religious foster care agencies respond to heterosexual couples compared to non-religious foster care agencies.

As shown in Panel C, there is significant geographical variation in where the foster care agencies in the study are located. 12% of the agencies are located in the Northeast, but only 9.6% of the religious foster care sample (n = 13) are located in the Northeast. 33.6% of the agencies are located in the South, but the South accounts for over 62% of the religious foster care sample. Most of the religious foster care agencies in the study are located in the southern United States. 39.1% of the sample is in the Midwest, but the Midwest only accounts for 21.5% of the religious sample. Lastly, 15.3% of the sample is located in the West. The West has the fewest religious foster care agencies in our sample—only 6.7%.

Before proceeding to any regression estimates, based on the field data, we are underpowered to identify our interaction effects based on the underlying covariance between each sexual orientation indicator $({GayMale}_{i}$ and ${GayFemale}_{i}$ in Equation 1), the religious FBO indicator $({Religious}_{i}),$ and the interaction between the two. As noted by Gelman (2018), identifying an interaction effect requires a significantly larger sample size than identifying the main effect. In preliminary results (described in the Appendix “Sample Size Simulations”), we simulate the required sample size required to obtain the requisite 80% to rule out a Type II error. Our results show that we would need a sample that is at least twice as large as our current sample size. Thus, going into this exercise, we understood that even if our results are significant, we have to be concerned about Type II error.^[10]

With those caveats, in Column 1 of Table 2, we test for a difference in response rates between religious and non-religious public and private agencies, and, in Column 2, we test for a difference in response rates between religious and private non-profits (i.e., excluding public agencies).^[11] We find that the differences in response rate are all relatively small in magnitude and not statistically significant. Thus, religious agencies respond, on average, to inquiries about foster care at a similar rate as secular public and private agencies.

We then test for any measurable difference in the average word count of the responses between religious and non-religious agencies (both public and private) in Column 3 and between religious and non-religious private agencies in Column 4. These differences are also small in magnitude and statistically insignificant.^[12] Overall, these results suggest that religiously-affiliated foster care agencies are nearly identical to non-religiously-affiliated organizations with respect to email inquiries (response rate and word count) about foster care.

Table 3 presents the results from Equation 1 with two separate samples. We include all agencies—public organizations, religious agencies, and non-religious agencies—in Columns 1 and 3. However, to examine differences in behavior between private agencies, we restrict our sample to private religious and secular foster care agencies in Columns 2 and 4. In Column 1 of Table 3, we find that religious agencies are 14.3 percentage points (unadjusted p-value = 0.036) less likely to respond to same-sex male couples compared to non-religious agencies. This coefficient remains statistically significant at the 10% level (p-value = 0.10) once we adjust for multiple hypothesis testing using the Westfall and Young (1993) resampling method.^[13] As noted above, these results are not sufficiently powered to rule out the possibility of a Type II error. Nonetheless, this correlation suggests that religiously-affiliated agencies are 26.05% less^[14] likely to respond to same-sex male couples compared to responses from non-religious public and private agencies to heterosexual couples. While this is measurably larger in terms of magnitude than the percentage point differences presented in Table 1, it is consistent with religious foster care agencies responding less to same-sex male couples than heterosexual couples. If we restrict the sample to private foster care agencies (see Column 2), the point estimate suggests that religious foster care agencies are less likely to respond than private secular foster care agencies. However, this difference is not statistically significantly different from zero (unadjusted p-value = 0.174). Regardless of how we restrict the sample, we do not find statistically significant differences between how religious and non-religious agencies respond to same-sex, female couples.

In Columns 3 and 4 of Table 4, we find some suggestive evidence that, compared to heterosexual couples, religious foster care agencies send shorter email responses to same-sex male and female couples than non-religious agencies (Column 3) or private non-profits (Column 4). However, these differences, which range from approximately 3 fewer words per email to 11 fewer words per email, are not statistically significantly different in any of the models.

In Table 4, using the classification from Mackenzie-Liu et al. (2021) and described above in the Outcomes Section, we examine the text of the responses sent by foster care agencies, looking for references to future steps in the fostering process. As we lack sufficient statistical power based on our sample size, we largely fail to reject the null of no difference in response. Religiously-affiliated agencies are only suggestively less likely to mention information sessions (unadjusted p-value = 0.038) and information sessions with specific times/days (i.e., “Sessions Plus,” unadjusted p-value = 0.049) than non-religiously-affiliated agencies in response to gay males relative to heterosexual couples. Given that we are testing two different hypotheses for 12 different outcomes in Table 4 and have concerns about Type I error, we also control for the family-wise error rate using the Westfall and Young (1993) resampling method. With this correction, none of the results reported in Table 4 remain statistically significant.^[15]

In Appendix Table A1, we restrict our sample to private agencies only. This is a pre-treatment characteristic, and thus it should not result in post-treatment bias. Overall, the results are similar. In the United States, foster care agencies appear less likely to provide detailed responses to same-sex male couples. However, there are no systematic differences in the content of responses between religious and non-religious agencies.^[16]

Discussion of Results

Descriptively, our results suggest that religious foster care agencies are less likely to respond to same-sex male couples relative to heterosexual couples, although these results are not robust to multiple hypothesis tests. Further, we have concerns that our sample sizes are too small to rule out Type II error. However, despite our inability to identify correlations between organizational characteristics and our outcomes of interest definitively, the results presented above suggest that organizational characteristics may influence the ways in which discriminatory behavior manifests.

This paper speaks to broader issues regarding equity in public policy and administration. If it is true that one of the objectives of public administration in the United States is to ensure that public administrators, including those individuals and organizations who are contracted to provide services by the state, treat all persons with “fairness, justice, and equality,” then it is important to understand if there is public sector discrimination. This includes understanding which groups are more or less likely to be discriminated against, and how (and ideally why) these organizations are discriminating (ASPA, 2021).

This paper explicitly examines if one type of organization—faith-based / religiously-affiliated foster care agencies— operationalize their discretion differently for same-sex couples relative to heterosexual couples. Such differences have the potential to either exacerbate or reduce social inequities. While this article cannot conclusively demonstrate that FBOs were less likely to respond to same-sex male couples, the fact that these organizations were noticeably less likely to respond to inquiries from these couples suggests future work is warranted. Given that recently enacted religious exemption laws provide legal protection for such behavior, and the Supreme Court has ruled it is within agencies rights to proactively refuse to provide services, such actions may become increasingly common.

While future research is needed to examine how bureaucratic discrimination varies by organizational characteristics or geographical location, this paper also highlights how a key methodology for analyzing this discrimination—the correspondence audit study—may be impractical in certain contexts, such as if the sample size is both too small and finite.

In terms of social equity and maximizing child welfare, if any foster care organization deters individuals from becoming foster parents, they may also inadvertently undermine child welfare. There is a large shortage of foster parents in the U.S. If same-sex male couples (and other members of the LGBTQ community) face additional barriers to become foster parents in the certification process, they may be less likely to participate. In turn, these hurdles could reduce social welfare as both children in need of foster placements, and same-sex couples (and individuals or couples who identify as members of the LGBTQ community) who want to provide a loving home for these children, are harmed by these practices. This is particularly concerning because we know that same-sex couples are more likely to foster children than heterosexual couples (S. K. Goldberg & Conron, 2018). Furthermore, Goldberg and colleagues find that same-sex couples are much more likely to adopt, a frequent outcome of fostering a child, older and disabled children as well as children of color (Brooks & Goldberg, 2001; A. E. Goldberg, 2009; A. E. Goldberg & Smith, 2009; Matthews & Cramer, 2006). Finally, it suggests that many same-sex male couples, a group that has limited options to become parents, may have less opportunity for family formation.

It is important to note that this paper is limited in certain ways beyond the power issues discussed above. First, like most email-based correspondence audit studies, this paper can only examine the initial interaction between foster care agencies and prospective foster parents. As such, it is unclear how responsive foster care agencies are to in-person visits, phone calls, or other modes of communication. It is also unclear the extent to which these organizations actually certify a foster parent, i.e., simply because an agency is responsive and provides a great deal of information in their responses does not mean that they are willing to certify all types of foster parents and foster families.

Moreover, this paper does not include the full universe of foster care agencies in the United States in 2019, or the current universe of foster care agencies. This is for several reasons. First, there is no centralized public database of foster care agencies, and not all foster care agencies put a contact email on their website. Many have only a phone number or an online-message system. Only agencies with a publicly posted email are in this study. It also remains unclear how generalizable our results are today given the Coronavirus-19 pandemic. Many small non-profits, notably those without sufficient financial reserves, may have closed or changed the services they provide, e.g., some non-profits may have ended, or started, programs to certify foster parents. Moreover, foster care agencies may have developed an improved online presence and become more responsive to online inquiries.

Future research should examine how organizational attributes—such as geography, composition of the workforce, religious affiliation, or organizational mission—influence the propensity for these organizations to respond to different types of clients. These studies should occur in contexts in which there is a large number of organizations in order to attain a sufficient sample size to identify effects with a high degree of confidence.

Acknowledgements

We are grateful for the feedback from Ken Meier, as well as the participants of the 2021 Miniconference on Social Equity and Public Administration.

Funding acknowledgement

We acknowledge no funding for this manuscript