Under What Conditions Does Caseworker-Caregiver Racial/Ethnic Similarity Matter for Housing Service Provision? An Application of Representative Bureaucracy Theory

Abstract

In this article, we examine child welfare caseworkers’ housing-related service strategies when they serve culturally similar versus culturally dissimilar clients. Testing hypotheses drawn from representative bureaucracy theory and using data from the second cohort of the National Survey of Child and Adolescent Well-Being, we find that when non-Caucasian caseworkers share the same racial/ethnic background as caregivers, caseworkers use more active strategies to connect caregivers to needed housing services. The relationship between racial/ethnic matching and frontline workers’ repertoire of service strategies is most pronounced when the need for housing has been registered formally via referrals and case plans and thus legitimated institutionally. These results reinforce basic tenets of representative bureaucracy theory and provide evidence of the benefits of racial and ethnic diversity in the human service workforce. Our findings also highlight the need for research identifying institutional and frontline organizational factors that enhance the quality of service provision.

Child welfare caseworkers commonly aim to ensure that the families they serve have safe, stable, and affordable housing. Housing assistance can relieve families’ poverty-related material needs, help stabilize families in crisis, and enhance families’ engagement with therapeutic services (Kelly, Blacksin, and Mason 2001; Ashley, Marsden, and Brady 2003). In some cases, housing assistance may also be a prerequisite for permanency planning, particularly for the reunification of children and youth in foster care with their biological parents (Farrell et al. 2010). The provision of housing assistance may, therefore, play an important preventative and restorative role in helping primary caregivers care for their children.

Over the past two decades, national and community-based studies have documented high levels of unmet need for housing assistance among child welfare-involved caregivers. Recent analyses based on the first cohort of families from the National Survey of Child and Adolescent Well-Being suggest that 24 percent of child welfare-involved families have significant housing needs, often relating to extreme poverty (Barth, Wildfire, and Green 2006). Yet, the National Study of Protective, Preventive, and Reunification Services Delivered to Children and their Families (US Department of Health and Human Services, Children’s Bureau 1997), which interviewed caseworkers serving a random sample of 2,109 foster children, found that only 5 percent of primary caregivers received any housing assistance and that 67 percent of primary caregivers who needed housing services did not receive them. Community-based studies of child welfare-involved families have also documented the need for housing supports (Cohen-Schlanger et al. 1995; Culhane et al. 2003; Courtney, McMurtry, and Zinn 2004), with some estimates of unmet need as high as 90 percent (Smith and Marsh 2002).

Unmet need for housing services among primary caregivers may be attributed in part to the difficulty child welfare agencies have in securing access to high-demand, community-based services (Freisthler 2013). Child welfare agencies are regularly required to link children and families with services across a diverse set of interrelated domains, including parenting, health and behavioral health, transportation, and housing. Because a single agency rarely provides such a wide range of services, and because the overall availability of public services such as housing assistance tends to be limited, it is common for child welfare agencies to develop formal and informal arrangements with a large number of community service providers. The responsibility for tracking these arrangements, referring families to providers with available space, and ensuring that families receive services in a timely manner is traditionally left to the individual caseworker (Bass, Shields, and Behrman 2004).

The central role that frontline workers play in procuring services for families raises questions about how caseworkers link primary caregivers with high-demand community services. Their role as the principal conduit for service delivery to child welfare-involved primary caregivers may require them to assess needs, develop case plans that identify services, arrange for the timely and coordinated provision of services, and evaluate progress toward predetermined case goals (Stiffman et al. 2004; Kohl et al. 2005). Caseworkers may refer caregivers to services by providing information about different community service providers, scheduling appointments for service, completing eligibility paperwork, accompanying caregivers to service appointments, and following up with service providers and caregivers. In sum, caseworker efforts can range from providing less direct, informational assistance to actively shepherding caregivers toward services (Bunger, Chuang, and McBeath 2012).

In the case of housing assistance, caseworkers may employ different referral strategies depending on factors particular to each case, such as specific allegations of child maltreatment and the presence of housing-related needs identified during the investigation process, the availability of and caregivers’ eligibility for housing services, and the extent to which housing assistance is required by court officials as a precondition for case resolution or noted as a factor requiring immediate redress (Shdaimah 2009a, 2009b). Caseworkers’ service linkage efforts may also differ by caseworker characteristics and, in particular, by whether caseworkers and primary caregivers share key cultural attributes. Numerous human service studies suggest that caseworkers’ cultural knowledge may facilitate service engagement with non-Caucasian clients by increasing frontline workers’ awareness of cultural differences and service preferences, enhancing the quality of assessments in response to clients’ presenting needs, and improving the cultural syntonicity of services (for reviews of this literature, see Kemp et al. 2009; Briggs and McBeath 2010). Over the past three decades, child welfare agencies have sought to recruit and train a diverse workforce with the goal of improving services to children and families from different cultural backgrounds. Examples of this trend include growing attention to child welfare workforce diversity and cultural competency training (Ortega and Faller 2012) and the infusion of cultural diversity content in Title IV-E educational programs (Barbee et al. 2012).

These human service policy and practice developments have mirrored efforts in the health and behavioral health sector to reduce racial/ethnic disparities in service access, delivery, and outcomes via culturally informed organizational interventions (e.g., culturally centered policies, programs, and partnerships) and frontline training (Smedley, Stith, and Nelson 2003; Chin et al. 2007; Fischer et al. 2007). Research on the effects of frontline worker-client racial and ethnic matching suggests that clients generally prefer to be served by practitioners who share their cultural background and that they experience the process of service delivery more positively when a racial/ethnic match is present; however, their health and behavioral health treatment outcomes are often no better than those of clients not served by culturally concordant practitioners (Chinman, Rosenheck, and Lam 2000; Karlsson 2005; Cabral and Smith 2011). Despite these mixed findings, cultural matching of frontline service providers with clients is a common approach to enhancing the quality of health and behavioral health services (Sue et al. 2009; Alegría, Vallas, and Pumariega 2010).

To a large extent, culturally informed service improvement initiatives in health care and the human services are premised on representative bureaucracy theory, which suggests that frontline practitioners who share key cultural attributes with clients will serve them more effectively and advocate on their behalf more readily than they would for culturally dissimilar clients (Meier and Bohte 2001; Sowa and Selden 2003). However, no study to date has examined the relationship between cultural similarity between child welfare caseworkers and clients and how caseworkers connect caregivers with needed services. Nor has representative bureaucracy theory been applied specifically to the child welfare service context.

Our article provides the first description of the relationship between caseworker-caregiver cultural matching and the strategies caseworkers use to link primary caregivers with housing services. We analyze caseworker data from the second cohort of families from the National Survey of Child and Adolescent Well-Being in pursuit of two aims. Our first goal is to describe the range of strategies used by caseworkers to connect primary caregivers with needed housing services. Our second goal is to empirically assess the application of representative bureaucracy theory to child welfare by investigating the strategies Caucasian and non-Caucasian caseworkers use to connect racially or ethnically similar caregivers to housing providers.

Active Representation in Frontline Human Service Work

Representative bureaucracy theory provides a suitable framework for understanding the cultural determinants of caseworkers’ use of different referral strategies to link primary caregivers with needed housing services. The theory was developed to understand how and why public bureaucrats (e.g., policymakers, managers, and frontline workers) represent the interests and respond to the problems of some clients and constituencies more fully than others. In its most basic form, the theory proposes that the individual bureaucrats in an organizational setting should proportionally mirror the demographics of the individuals in their client population (passive representation), under the premise that bureaucrats with similar characteristics will advance the preferences and well-being of clients from their group more readily than other bureaucrats (active representation) (Meier and Bohte 2001; Sowa and Selden 2003). Representative bureaucracy may thus be understood as both a theory and a tool for enhancing organizational effectiveness: agency directors may seek to increase the diversity of their staff and deploy bureaucrats with different group memberships strategically to improve the performance of programs serving diverse client populations.

Prior studies demonstrate a strong association between passive and active representation for minority client groups in numerous sectors, including enhanced child support collection efforts by female workers (Wilkins 2007), student educational achievement in the presence of non-Caucasian educators (Meier, O’Toole, and Nicholson-Crotty 2004; Grissom, Nicholson-Crotty, and Nicholson-Crotty 2009), reduction in police departments’ racial profiling practices in the presence of an African American police force (Wilkins and Williams 2008), and the provision of rural home loans to African American, Hispanic, and Asian applicants (Selden, Brudney, and Kellough 1998). Research has also identified relationships between worker demographic characteristics, worker-client interactions, and program outcomes (Sowa and Selden 2003; Wilkins and Keiser 2006), suggesting that cultural similarities between frontline workers and non-Caucasian clients may be associated with greater frontline efforts around eligibility determination for public programs and thus enhanced public access overall. These studies highlight the prominent role of minority frontline workers in implementing public policies and programs to the benefit of racially/ethnically similar client groups.

Existing literature describes several culturally informed mechanisms for translating passive representation to active representation, including socialization processes that generate concordant preferences, culturally tailored interpersonal processes, and bureaucrats’ active adoption of representational roles. At a frontline level, racially/ethnically diverse practitioners may benefit from socialization processes that afford them greater cultural knowledge and more positive social constructions of minority clients, which may enhance their ability to combat racial stereotyping and prejudicial service practices (Smedley, Stith, and Nelson 2003; Roch, Pitts, and Navarro 2010). Research finds that minority bureaucrats sharing the race/ethnicity of their clients are more likely to hold similar preferences for public programs and policies than culturally different bureaucrats and clients, implying that minority bureaucrats may more accurately translate minority client preferences into action (Bradbury and Kellough 2008). This knowledge-to-action process may benefit from practitioners’ use of culturally informed language and knowledge of cultural symbols, the development of a culturally centered therapeutic alliance, and the promotion of treatment goals and service plans reflecting client cultural preferences (Bernal and Sáez-Santiago 2006; Sawrikar 2013). Finally, passive representation may be translated to action specifically when minority bureaucrats adopt a representational role, and connect their practice efforts to desired policy outcomes by actively representing minority group interests (Sowa and Selden 2003; Bradbury and Kellough 2008).

These mechanisms imply that active representation is most common and meaningful among minority as opposed to majority groups in bureaucratic settings. Although Caucasian bureaucrats may advocate on behalf of Caucasian clients, particularly when their individual and group interests are poorly represented by organizational policies, pressures to adopt an active representational role may be stronger among non-Caucasian than Caucasian bureaucrats due to poorer outcomes for racially/ethnically diverse groups and historical patterns of racially/ethnically influenced discriminatory policies and programming in public human service settings. Additionally, Caucasian workers may have less need to extend themselves actively on behalf of clients if the clients’ majority group status confers important and generalized privileges and opportunities that may influence frontline services (e.g., enhanced access to administrators and social networks)and that are less available to non-Caucasian bureaucrats (Jones 2000; Abrams and Moio 2009). These dynamics may limit Caucasian workers’ understanding of whiteness as a potential factor in client advocacy and frontline decision-making. Thus, pragmatic considerations relating to the daily work of case management may influence how Caucasian frontline workers serve their Caucasian clients more strongly than active representation based on a shared racial identity. Thus, passive representation based on whiteness may have less influence over the shape of bureaucratic encounters than other, pragmatic caseworker and client considerations relating to more essential for decision-making (Bradbury and Kellough 2011).

In sum, cultural concordance between bureaucrats and clients does not always result in active representation, and cultural mismatch does not necessarily lead to poorer frontline service outcomes for minority clients, particularly when workers have relevant cultural knowledge and experience. However, representative bureaucracy theory illuminates the possibility that cultural similarity in the caseworker-caregiver relationship activates minority caseworkers as stewards of minority client interests. At the organizational level, the theory suggests that a culturally diverse workforce can act affirmatively to improve racial/ethnic equity in public policy development and implementation (Harris and Hackett 2008). With respect to caseworkers’ use of different strategies to link primary caregivers with housing services, the theory proposes that racial/ethnic similarity between non-Caucasian, but not Caucasian, caseworkers and caregivers may lead caseworkers to respond to caregivers’ housing needs with more intense service efforts. These culturally informed frontline activities are predicated on the recognition that active representation may be needed to remedy racial/ethnic discrimination and disparities in human service programming.

Under What Conditions Should Caseworker-Caregiver Cultural Similarity Be Expected to Enhance Housing Referral Strategies?

Passive representation may be activated when caseworkers have sufficient discretion and autonomy to translate their cultural values and preferences into enhanced referral efforts (Sowa and Selden 2003; Sosin 2009). However, in practice, frontline workers’ decision-making ability may be enabled or constrained by mandates imposed by policymakers and public officials (Meier and Bohte 2001; Andrews et al. 2005). For example, juvenile or family court officials, who may require families to participate in court-ordered services, can influence caseworkers’ span of control in arranging services for primary caregivers. Supervisors may also direct caseworkers to connect families to specific services identified in formal case plans. In these situations, caseworkers’ service linkage roles maybe activated only after families’ needs are officially and clearly registered through formal, institutional channels (e.g., a written case plan). These institutional signals may serve as formal and normative markers of emphasis when caseworkers are determining how to serve different primary caregivers, and may therefore legitimize minority caseworkers directing different levels of service-related resources and effort to select cases (Hasenfeld 2000; Garrow and Grusky 2013).

It is important to explicate the institutionally bounded nature of representative bureaucracy. Such research tempers expectations concerning the consistency and strength of the relationship between passive and active representation and helps identify the organizational conditions under which primary caregivers may benefit from having culturally similar non-Caucasian caseworkers (Sandfort 2000; Watkins-Hayes 2011; Brodkin 2012). For example, no consistent understanding of the relationship between caseworker characteristics and child welfare outcomes emerges from the literature, with some studies finding no effects of caseworker demographic factors on permanency and service provision (Ryan et al. 2006; McBeath and Meezan 2008) and others finding effects on child maltreatment assessments (Font, Berger, and Slack 2012). Yet, research to date has not been able to disentangle the service efforts of workers from the amount of discretion contextualizing their case-based practices. Questions remain concerning the institutional settings in which minority frontline workers work to improve services for culturally similar children and families. Although some evidence suggests that child welfare agencies may deliberately match African American and Hispanic caseworkers and clients (Perry and Limb 2004), it is unclear whether such matching changes the manner in which frontline casework is carried out, particularly when caseworkers may not be required or able to coordinate service delivery effectively.

In summary, according to representative bureaucracy theory, non-Caucasian caseworkers’ efforts on behalf of culturally similar clients should be conditioned by the degree of caseworker discretion present. The official registration of primary caregivers’ needs for housing services, such as a referral for housing services or the formalization of the need for services via incorporation in a case plan, may enable minority caseworkers to move beyond passive service strategies to actively represent racially/ethnically similar caregivers in the process of accessing housing services. This approach to linking caregivers with needed services reflects the central tenets of representative bureaucracy theory understood within an institutionalized frontline child welfare context.

Research Questions and Core Hypothesis

Our study is organized around two questions that correspond with the two study aims: What referral strategies do child welfare caseworkers employ to link permanent, primary caregivers with needed housing services? Controlling for other caregiver and caseworker factors, to what extent is racial/ethnic matching between caseworkers and caregivers related to the types of housing referral strategies caseworkers use on behalf of caregivers? We expect to find that non-Caucasian caseworkers use more active housing referral strategies when working with culturally similar primary caregivers than when working with culturally dissimilar caregivers. We also expect that this representative bureaucracy effect among non-Caucasian caseworkers and clients is enhanced when the need for housing assistance is recognized formally via referrals and case plans. Under these conditions, non-Caucasian caseworkers will be legitimated to act on behalf of clients, and caseworkers will have discretion to use more active referral strategies to the benefit of culturally similar primary caregivers. In contrast, in situations where clients’ housing needs are not registered formally, minority caseworkers will not be afforded the institutional space to expend additional effort on behalf of clients and their service linkage activities will be limited, regardless of their cultural similarity with primary caregivers. Finally, even where institutionalized discretion is present, we do not expect that matching between Caucasian caseworkers and caregivers is associated with active housing referral strategies due to the reduced salience of whiteness-based client advocacy as a factor in frontline decision-making.

Methods

Data

We report on a set of analyses of data from the second cohort of families from the National Survey of Child and Adolescent Well-Being (NSCAW II). NSCAW is the only national, longitudinal study of families who were subjects of child abuse or neglect investigations or assessments conducted by US Child Protective Services (CPS) agencies. NSCAW was funded by the Administration for Children and Families within the US Department of Health and Human Services, with field data collection carried out by Research Triangle Institute (RTI) International (Dowd et al. 2010). The researchers used a two-stage, stratified sample design. In the first stage, they divided the United States into nine sampling strata. Eight of these strata correspond to the eight states with the largest child welfare caseloads, and the ninth consists of the remaining states and the District of Columbia. They excluded states requiring CPS agency first contact (i.e., that first contact of potential study participants be made by CPS agency staff rather than NSCAW field representatives) from the sampling frame because the researchers had difficulty obtaining approval from child welfare agencies located in states requiring child welfare agency approval prior to contact of families, and due to concerns about protecting the confidentiality of respondents, given the nature of questions about their experience with the child welfare system. They then formed and selected primary sampling units (PSUs) from these nine strata with a probability proportional to the size of the county child welfare population. They defined each PSU as the geographic area served by a single CPS agency. Within these PSUs, they sampled a total of 5,873 children ranging in age from birth to 17.5 years from all child welfare investigations or assessments completed between February 2008 and April 2009. The NSCAW II sample is nationally representative of all children investigated for maltreatment by CPS agencies during that time period.

The researchers collected detailed, retrospective assessments of family context and well-being through face-to-face interviews with current primary caregivers (either permanent or foster) and their investigative caseworkers. They asked investigative caseworkers to provide information on their own background and demographics, their assessment of families’ service needs, and actions taken on behalf of families. Researchers conducted baseline (Wave 1) interviews with investigative caseworkers between March 2008 and September 2009, on average approximately four months after the close of investigation or assessment. The caseworkers’ responses were based on families’ confidential case records as well as their knowledge of the case (Dowd et al. 2010). Each caseworker was only interviewed about one case for NSCAW.

Sample

Given our focus on factors contributing to caseworker housing referral strategies, we restricted the operational study sample to include only permanent, primary caregivers (not foster caregivers) who were identified by investigative caseworkers as needing housing services and who were not already receiving them (Sample 1). Application of these inclusion criteria reduced the study sample to 873 permanent, primary caregivers.

To disentangle the effect of institutionalized discretion from other caseworker-centered determinants of frontline casework processes, we generated two additional, conceptually distinct samples of primary caregivers. First, we restricted the sample to include only permanent, primary caregivers who needed housing services and were referred for such services by investigative caseworkers (Sample 2). A total of 509 caregivers met these criteria. Next, we further restricted the sample to include only permanent, primary caregivers who needed housing services, were referred for such services by investigative caseworkers, and for whom this referral was specified in the case plan (Sample 3). Only 286 permanent primary caregivers met these criteria. To summarize, Sample 1 includes all caregivers with housing service needs, regardless of whether they received a referral, or had their need noted in their case plan. Sample 2 includes only those caregivers who received a referral for housing services from the investigative caseworker. Sample 3 includes only caregivers for whom a referral for housing services was registered in the case plan. Samples 2 and 3 thus constitute organizational settings in which institutional signals direct caseworkers to exert agency in linking caregivers with needed services.

Sampling weights within NSCAW account for differential selection probabilities as well as potential bias resulting from survey nonresponse (Pfefferman et al. 1998; Dowd et al. 2010). However, these weights do not account for item non-response. Levels of item non-response were generally very low in our sample, and did not exceed 10 percent for any variable. Weighted t-tests also indicate that caregivers in each of the three analytic samples did not differ significantly from those excluded due to listwise deletion on observed variables. To reduce the possibility of non-response bias, we conducted multiple imputation using the multivariate normal imputation method within the Stata 12.0 MI module (StataCorp 2011). We used a total of 20 imputations to reduce sampling error. Weighted t-tests following imputation did not reveal any statistically significant differences between imputed and unimputed variables in the three study samples; we employ the imputed data in our analyses.

Measures

Need for Housing Services (Sample Restriction)

Permanent, primary caregivers were identified as needing housing services if investigative caseworkers responded yes to the following question: “In the last 12 months, did the permanent, primary caregiver need help finding a place to live?” Given our interest in testing the relationship between caseworker characteristics and service effort in support of caregivers, we excluded caregivers already receiving public housing at the time of investigation or assessment because it could not be determined whether service receipt for these individuals reflected efforts of caseworkers, other professional staff, or caregivers. This methodology follows the approach used previously to assess need for health and substance abuse treatment services in NSCAW (Wells et al. 2009; Chuang et al. 2013).

Caseworker Referral Strategies (Outcome Variable)

We organized caseworker referral strategies into three categories reflecting the intensity of effort reported by the caseworker in facilitating caregiver access to needed housing services: caseworker did not take any action (lowest degree of effort), caseworker provided caregiver with information (medium degree of effort), and caseworker provided caregiver with both information and active assistance (highest degree of effort). Caseworkers were considered to have provided information (or put forth a medium degree of effort) if they suggested that caregivers receive services, provided caregivers with names and numbers of service providers, and/or followed up with caregivers to determine whether services were provided. Caseworkers were considered to have provided active assistance (or put forth a high degree of effort) if they directly assisted caregivers with completing or filing service applications, made an appointment for the caregiver, and/or accompanied caregivers to appointments. These categories provided a basic weighting of caseworker referral strategies in alignment with prior research (Bunger, Chuang, and McBeath 2012).

Caseworker-Caregiver Racial/Ethnic Match (Key Independent Variable)

Caseworker-caregiver racial/ethnic match was operationalized as a categorical variable comprised of three categories: absence of a racial/ethnic match between the caseworker and caregiver (referent), racial/ethnic match between a Caucasian, non-Hispanic caseworker and the caregiver, and racial/ethnic match between a non-Caucasian caseworker and the caregiver. A non-Caucasian racial/ethnic match was registered if both the caseworker and caregiver self-identified as being African American, both were Hispanic, both Asian/Pacific Islander, or both American Indian/Alaska Native. Non-Caucasian non-matches, which involved minority caseworkers serving caregivers of a different race or ethnicity, were placed in the referent group. These three categories allowed for a parsimonious test of racial/ethnic matching among Caucasian versus non-Caucasian caseworkers and caregivers.

Other Covariates (Control Variables)

We also include as controls a number of caseworker, case-level, and community factors known to influence families’ interaction with the child welfare system. Caseworker attributes include the caseworker’s primary job role, tenure, and caseload. The caseworker’s primary job role is a categorical variable comprised of the following categories: investigation/assessment of child abuse, screening, and/or intake services (referent); ongoing services for in-home and/or out-of-home cases; and other job role such as placement. Caseworker tenure is a continuous variable representing the number of years the investigative caseworker had worked in child welfare. Caseload is the average number of new investigations assigned to the caseworker each month. These factors may be associated with the caseworkers’ ability to engage with clients and respond to the needs of child welfare-involved children and families. Specifically, we expect investigative caseworkers to be less actively involved than ongoing services workers in coordinating and delivering services; and as caseloads increase, caseworkers are expected to have less availability to engage with clients and thus to use less active referral strategies (Hasenfeld 2000; McBeath and Meezan 2008; Foldy and Buckley 2010; Chuang et al. 2013).

We include five case-level variables, either as proxies for caregiver or household need for housing services or because prior research has demonstrated associations with intensity and type of frontline efforts expended by child welfare caseworkers. Factors pertaining to primary caregivers and children include: the most serious type of maltreatment experienced by the child (Barnett, Manly, and Cicchetti 1993); whether the child was placed out-of-home after the close of investigation/assessment; a three-point global measure of cumulative family risk, with values ranging from 1 = low family risk to 3 = high family risk, based on investigative caseworkers’ assessment of the presence of 21 risk factors such as prior child welfare service history, high family stress, and history of domestic violence (Barth et al. 2008b; Mersky et al. 2009); and child age in years. A dichotomous indicator of whether the permanent, primary caregiver was older than 35 years captures differences in the home setting of caregivers based on prior NSCAW-based research suggesting that caregivers less than 35 years old are more likely to be providing in-home care whereas older caregivers are more likely to be kinship, foster, or another type of caregiver (Berkhoff, Leslie, and Stahmer 2006). We expect family cumulative risk, in particular, to capture case-level demand for housing supports and thus to covary positively with caseworker referral strategies (Wells et al. 2009; Bunger, Chuang, and McBeath 2012).

Finally, we include two contextual variables to control for the local supply of housing services as well as community demand for such services, which might influence the ability of caseworkers to locate housing supports and link caregivers to them (Freisthler 2013). Factors associated with urban/rural differences in the location of child welfare systems are captured by a dichotomous measure indicating whether the child welfare agency was located in a non-metropolitan area. A second, agency-level measure pertaining to the percentage of the total local population living below the federal poverty level serves as a proxy for overall community need for housing assistance.

Analyses

NSCAW data have a two-level hierarchical structure, with families and their caseworkers nested within child welfare agencies. However, fully unconditional random effects models indicated low variation between agencies in caregivers’ receipt of housing services, with intraclass correlation coefficients of approximately 1 percent. Therefore, we analyzed all models as single-level multivariate regression equations using the Stata 12.0 -svy- module (StataCorp 2011). This module permits analyses that account for the complex survey design of these data, accommodating probability weights and stratification as well as a post-hoc adjustment to standard errors to account for potential clustering of families and caseworkers within child welfare agencies. The post-hoc adjustment to standard errors in the -svy- module is similar to that used by the robust standard error procedure, differing only by a constant multiplier. We used ordinal logistic regression to examine associations between caseworker-caregiver racial/ethnic match and the type of referral strategies caseworkers employed, controlling for other child, caregiver, caseworker, and contextual variables.

Correlations between independent variables as well as variance inflation factors (VIF) did not indicate problematic collinearity. The proportional odds assumption of ordinal logistic regression, assessed by the Brant test, was satisfied for models run on Samples 1 and 2, but failed to run in Sample 3. The relatively small size of this sample (N=286)and uneven distribution of responses in the dependent variable (only 6 percent of caseworkers reported taking no action to facilitate caregiver access to needed housing services) raised some concerns about power. To test the robustness of key study findings, all models were re-run using Poisson regression; the dependent variable in these models was a count (0–6) of the number of referral strategies the caseworker reported engaging in. Significant and similar effects of racial/ethnic matching for Caucasian as well as non-Caucasian caseworkers and caregivers were maintained in Samples 2 and 3 at p<0.05 (not shown, but available from authors upon request), increasing confidence in study findings. The Institutional Review Board at the second author’s home institution approved this secondary data analysis. The Institutional Review Board at RTI International approved the original NSCAW data collection.

Results

Caseworker Referral Strategies

The first research question concerns the referral strategies caseworkers used to connect primary caregivers to needed housing services. Table 1 describes the overall characteristics of the three study samples. Table 2 presents more detailed information about the extent to which caseworkers reported engaging in different referral strategies on behalf of caregivers with identified need for housing services during the CPS investigatory process.

Table 1.

Weighted Descriptive Statistics Reflecting Characteristics of Each Sample

	Sample 1	Sample 2	Sample 3
	CG needs housing % or Mean ± S.E.	CW made housing referral % or Mean ± S.E.	Referral in case plan % or Mean ± S.E.
Caseworker referral strategies
No action taken	47%	9%	6%
Provided caregiver with information	38%	65%	63%
Actively assisted caregiver	15%	26%	31%
Caseworker-caregiver racial/ethnic match
No racial/ethnic match	38%	37%	26%
Caucasian caseworker-caregiver	43%	46%	52%
Non-Caucasian caseworker-caregiver	19%	17%	22%
Control variables: Caseworker attributes
Caseworker job role: Investigation (referent)	81%	75%	68%
Caseworker job role: Services	12%	16%	23%
Caseworker job role: Other (not investigation)	7%	9%	9%
Caseworker years in child welfare	6.74 ± 0.64	6.93 ± 0.74	6.51 ± 0.71
Caseworker case load	13.20 ± 1.36	13.08 ± 1.87	12.40 ± 3.11
Control variables: Case characteristics
Maltreatment type: Neglect (referent)	36%	39%	35%
Maltreatment type: Physical abuse	12%	13%	5%
Maltreatment type: Sexual abuse	3%	2%	3%
Maltreatment type: Substance abuse	19%	16%	22%
Maltreatment type: Other (not neglect)	30%	30%	35%
Child placed out-of-home	24%	24%	36%
Family cumulative risk	2.15 ± 0.07	2.23 ± 0.08	2.28 ± 0.87
Child age in years	5.09 ± 0.29	4.94 ± 0.44	4.81 ± 0.57
Caregiver >35 years	36%	34%	41%
Control variables: Contextual factors
Agency located in non-metropolitan area	30%	32%	33%
% local population below federal poverty line	14.46 ± 0.56	14.74 ± 0.70	14.89 ± 1.14

Note: CG = caregiver; CW = caseworker; Sample 1 N=873; Sample 2 N=509; Sample 3 N=286.

Table 2.

Extent to Which Caseworkers Reported Engaging In Different Referral Strategies

	Sample 1	Sample 2	Sample 3
	CG needs housing	CW made housing referral	Referral in case plan
Providing caregiver with information
Suggested caregiver should obtain housing services	27%	53%	61%
Provided caregiver with names and numbers of housing service providers	32%	64%	69%
Followed up with caregiver to see if housing services were provided	14%	28%	46%
Actively assisted caregiver
Accompanied caregiver to appointment to apply for housing services	4%	7%	10%
Made appointment for caregiver to apply for housing services	4%	8%	12%
Assist caregiver with completing or filing application for housing services	9%	18%	22%

Note: CG = caregiver; Sample 1 N=873; Sample 2 N=509; Sample 3 N=286; strategies are not mutually exclusive.

Across all three study samples and as shown in Table 2, caseworkers were more likely to provide caregivers with information than active assistance. In the general sample of caregivers with housing service needs (Sample 1), the most commonly reported referral efforts included providing caregivers with names and phone numbers of housing providers in the community (32 percent) and generally recommending that caregivers receive housing assistance (27 percent). The least commonly reported referral efforts involved making appointments with housing providers and accompanying caregivers to these appointments (4 percent each).

These referral strategies were used more frequently when a referral for housing services had been made by the investigative caseworker and when such a referral was noted in the caregiver’s case plan. Specifically, the prevalence of each referral strategy was greater among primary caregivers in Samples 2 and 3 as compared to those in Sample 1. For example, caseworkers followed up with nearly half of caregivers when a referral had been noted in the case plan (46 percent, Sample 3), but only about a quarter of caseworkers followed up when a referral for housing assistance had been made but was not noted in the case plan (28 percent, Sample 2) and roughly a seventh of caseworkers followed up when no referral was registered formally (14 percent, Sample 1). The overall frequency with which caseworkers reported both providing information and actively assisting caregivers (e.g., accompanying caregivers to appointments, making appointments on their behalf, or assisting caregivers with completing or filing housing applications) also increased across the three samples, from 15 percent in Sample 1 to 31 percent in Sample 3.

Representative Bureaucracy in Different Institutional Contexts

The second research question concerns whether non-Caucasian caseworkers utilize more active referral strategies to connect primary caregivers to needed housing services when they share the same racial/ethnic background and when the need for housing assistance is institutionally registered. We first explored the extent to which racial/ethnic matching occurs within our general sample of caregivers. As shown in Table 1, only 38 percent of caregivers needing housing (Sample 1) did not share the same racial/ethnic background as their caseworker. In contrast, 43 percent of caregivers were Caucasians who received services from a Caucasian caseworker while 19 percent were non-Caucasian caregivers who received services from a caseworker of a similar race/ethnicity. The proportion of caregivers with a similar racial/ethnic background as their caseworker was highest when a referral was noted in the case plan (74 percent in Sample 3).

Table 3 presents results from ordinal logistic regression models estimating the relationship between caregiver-caseworker match and type of caseworker referral strategy. As hypothesized, non-Caucasian caseworkers were more likely to use active referral strategies on behalf of same-race/ethnicity primary caregivers when need for housing assistance was recognized formally via referrals and/or in case plans. Specifically, the odds of caseworkers expending a high level of effort (by providing active assistance), as compared to the odds of providing a modest level of effort (through the provision of information only or by taking no action at all), were 6.49 and 5.22 times higher in these respective situations (p<0.05), controlling for other caseworker attributes, case characteristics, and contextual factors. In contrast, when need for housing assistance was not institutionally registered, this relationship was positive but non-significant (OR=1.79, p>0.05). Nor was the relationship between caseworker-caregiver racial/ethnic matching and referral strategies significant among Caucasian dyads, as expected

Table 3.

Ordinal Logistic Regression Estimates of Caseworker Referral Strategies

	Sample 1		Sample 2		Sample 3
	CG needs housing		CW made housing referral		Referral in case plan
	O.R. (S.E.)	95% CI	O.R. (S.E.)	95% CI	O.R. (S.E.)	95% CI
Caseworker-caregiver racial/ethnic match
Caucasian caseworker-caregiver	1.27 (0.50)	0.58–2.80	1.08 (0.59)	0.36–3.20	2.09 (1.64)	0.42–10.30
Non-Caucasian caseworker-caregiver	1.79 (0.88)	0.67–4.80	6.49** (4.06)	1.85–22.71	5.22* (4.26)	1.00–27.24
Control variables: Caseworker attributes
Caseworker job role: Ongoing services	3.23* (1.77)	1.08–9.62	2.41 (1.71)	0.59–9.95	1.94 (1.41)	0.46–8.25
Caseworker job role: Other (not investigation)	3.28** (1.40)	1.40–7.70	3.34 (2.24)	0.88–12.69	1.15 (0.97)	0.21–6.23
Caseworker years in child welfare	1.01 (0.03)	0.96–1.06	0.99 (0.03)	0.94–1.06	0.99 (0.05)	0.90–1.09
Caseworker caseload	1.01 (0.01)	0.99–1.03	1.00 (0.01)	0.98–1.03	0.99 (0.01)	0.97–1.02
Control variables: Case characteristics
Maltreatment type: Physical abuse	1.33 (0.52)	0.61–2.88	2.55* (1.12)	1.06–6.13	2.53 (1.96)	0.54–11.88
Maltreatment type: Sexual abuse	0.72 (0.45)	0.21–2.48	1.85 (1.07)	0.59–5.86	1.13 (0.87)	0.25–5.22
Maltreatment type: Substance abuse	0.65 (0.29)	0.26–1.60	1.11 (0.53)	0.43–2.90	0.24* (0.15)	0.07–0.82
Maltreatment type: Other (not neglect)	0.73 (0.29)	0.33–1.61	0.53 (0.33)	0.15–1.84	0.25 (0.19)	0.06–1.10
Child placed out-of-home	0.70 (0.29)	0.30–1.62	0.77 (0.52)	0.20–2.97	0.57 (0.44)	0.12–2.69
Family cumulative risk	1.85** (0.42)	1.17–2.93	2.32* (0.74)	1.22–4.40	2.41* (0.89)	1.15–5.04
Child age in years	1.01 (0.04)	0.94–1.08	1.03 (0.04)	0.96–1.10	1.04 (0.07)	0.90–1.19
Caregiver >35 years	0.98 (0.37)	0.46–2.09	1.30 (0.76)	0.40–4.20	1.63 (1.39)	0.30–8.97
Control variables: Contextual factors
Agency located in non-metropolitan area	1.51 (0.58)	0.71–3.23	1.50 (0.75)	0.56–4.05	1.70 (0.86)	0.62–4.67
% local population below federal poverty line	1.00 (0.04)	0.93–1.08	0.92* (0.04)	0.84–1.00	0.91* (0.04)	0.84–0.98

Note: CG = caregiver; CW = caseworker; Sample 1 N=873; Sample 2 N=509; Sample 3 N=286;

^*p<0.05;

^**p<0.01. Categories of caseworker referral strategies (outcome variable) include: no action taken by caseworker; caseworker provided caregiver with informational assistance only; and caseworker provided caregiver with informational assistance and active assistance.

Several other control variables were significantly associated with the type of referral strategies caseworkers used to link primary caregivers with needed housing services. Of these, the only variable consistently predictive across all models was cumulative family risk, as assessed by investigative caseworkers at the time of investigation. Odds of caseworkers expending more effort to link caregivers to housing services were between 1.85–2.41 higher when family cumulative risk was high, depending on the sample (p<0.05).

Discussion

Our study examines the referral strategies child welfare caseworkers use to connect primary caregivers to needed housing services, and estimates whether non-Caucasian caseworkers with similar racial/ethnic backgrounds as the caregivers they serve use more intense referral strategies compared to Caucasian caseworker-caregiver dyads or caseworkers with dissimilar racial/ethnic backgrounds. We also examine whether the relationship between cultural concordance between caseworkers and caregivers and housing referral strategies differs in the presence of institutional signals authorizing service activity.

We sought to improve understanding of racial/ethnic differences in caseworker housing referral strategies through the application of representative bureaucracy theory to a particular child welfare service area. We find that caseworkers generally used few active strategies to help primary caregivers secure housing assistance. However, consistent with our hypothesis, when need for housing assistance was formally registered via specific referrals and in case plans, the odds of non-Caucasian caseworkers reporting active, high-effort housing referral strategies on behalf of culturally similar caregivers was higher than for Caucasian caseworkers serving Caucasian caregivers or for culturally dissimilar clients. We also find that the referral strategies used by caseworkers were consistently related to the level of cumulative risk present in the family case, but were unrelated to most caseworker attributes, child and family characteristics, and contextual factors.

These findings contribute to our understanding of the conditions under which caseworkers seek to link child welfare-involved primary caregivers with housing assistance, a topic that is understudied despite evidence of racial/ethnic disparities in housing supports to child welfare-involved families and the importance of housing assistance in facilitating resolution of child welfare cases (Courtney, McMurtry, and Zinn 2004; Farrell et al. 2010; Henwood, Stanhope, and Padgett 2011). In particular, our findings suggest that caseworkers’ use of passive versus active housing referral strategies is only somewhat related to child and family covariates that are traditionally viewed as proxies of need for different child welfare services (i.e., maltreatment type, out-of-home placement, cumulative family risk, and child and caregiver age) and that have been found in prior research to covary with the intensity and type of frontline efforts expended by child welfare caseworkers (Wells et al. 2009; Bunger, Chuang, and McBeath 2012).

Our finding that caregiver or household need for services is not strongly related to caseworkers’ referral strategies, as well as results from aforementioned studies documenting persistent needs for housing assistance among child welfare-involved families, implies that housing supports may be apportioned to caregivers in a manner that is relatively untailored to their case conditions. Moreover, the infrequent use of active referral strategies, which remained low in absolute terms although the prevalence of these strategies increased to 26 percent when need for housing assistance was institutionally registered through formal service referrals and 31 percent when embedded in case plans, suggests that frontline efforts to link caregivers with needed housing services are principally informational. Although individual demand for housing supports is apparent across studies, the level of organizational commitment and resources public child welfare systems have to secure housing supports is called into question by such non-tailored and impersonal caseworker referral strategies (Shdaimah 2009a, 2009b).

In following this line of reasoning, our study joins the small set of empirical studies seeking to understand the institutional and organizational context of child welfare casework. Despite the proliferation of new evidence-based practice models, casework remains the principal organizational technology used to deliver frontline services and achieve desired client outcomes in child welfare and other human service fields (Smith and Donovan 2003; Faller, Grabarek, and Vandervort 2009; Foldy and Buckley 2010). Theoretically informed investigation of human service casework has generally involved consideration of institutionalism and structuration in relation to policy and program implementation (Lipsky 1980; Hasenfeld 2000; Sandfort 2000; Brodkin 2011, 2012; Garrow and Grusky 2013). With its conjoint attention to minority group advocacy, the organization of bureaucrat-client exchanges, and frontline worker discretion and decision-making, representative bureaucracy theory is a useful lens for explicating how the cultural determinants of frontline practice are folded into institutionalized organizational settings.

Our data suggest that cultural matching of frontline workers with clients is common, with roughly two-thirds to three-quarters of the three samples of caseworkers reflecting the race/ethnicity of caregivers. These figures align with other NSCAW-based studies of the demography of public child welfare workers. For example, Sarah A. Font, Lawrence M. Berger, and Kristin S. Slack (2012) find that 54 percent of African American CPS-involved families were served by African American caseworkers; similarly, the finding that most racial/ethnic caseworker-caregiver matches involved Caucasian dyads is unsurprising given the continuing under representation of racially/ethnically diverse workers in child welfare (Barth et al. 2008a). In reference to these studies, it is important to contextualize the finding that Caucasian caseworkers are no more active in securing needed housing support for Caucasian caregivers than for non-Caucasian caregivers. While this finding comports with our expectations, it is unclear why active representation does not appear to be present among Caucasian caseworker-caregiver dyads. Whether this is case-based, or whether other institutional factors are prompting or diluting levels of service effort (e.g., organizational rationing of caseworker referral effort regardless of client race, the presence of other community resources linked to housing), is unclear in these findings. It is also not obvious whether Caucasian and non-Caucasian caregivers served by Caucasian caseworkers receive equivalent levels of actual housing resources. However, active representation does appear to be present among non-Caucasian caseworker-caregiver dyads. Our findings suggest that non-Caucasian caseworkers are more likely to engage in active referral strategies to connect same-race/ethnicity caregivers to needed housing services when need for housing has been officially sanctioned. This may suggest recognition on the part of minority caseworkers of race/ethnicity as a factor in housing discrimination and of the importance of frontline bureaucratic exchange in redressing service disparities. Albeit associational in nature, these results reinforce theoretical expectations concerning the value of racial/ethnic similarity for enhancing frontline service processes for non-Caucasian clients. Other research has highlighted the direct role active referral strategies play in securing needed mental health services (Wells et al. 2009; Villagrana 2010; Bunger, Chuang, and McBeath 2012) and housing assistance (Farrell et al. 2010) for child welfare-involved families. By extrapolation, this study suggests that cultural matching of non-Caucasian caseworkers and clients may help reduce the racial/ethnic disparities in health, mental health, and concrete service utilization found in other NSCAW studies (Wells et al. 2009; Cheng and Lo 2012; Martinez, Gudiño, and Lau 2013) and provides support for policies and programming supporting the recruitment and retention of culturally diverse child welfare practitioners.

Importantly, our findings highlight the possibility that the relationship between non-Caucasian caseworker-caregiver racial/ethnic matching and use of different housing referral strategies may vary depending on whether need for housing services is formally registered. We find indirect evidence of different institutional obligations and opportunities for caseworkers to expend effort in linking primary caregivers with needed housing services. Specifically, in situations where non-Caucasian caseworkers have a requirement to connect caregivers with needed assistance, their culturally based agency may be activated and racial/ethnic matching may promote caregiver service access. In a sense, culturally based service advocacy may exist in safe spaces created by policy and practice requirements that allow extra effort to be directed towards certain clients; itemization of need as expressed through referrals and case plans may serve as a tag, activator, and beacon for this extra effort in compliance with administrative rules and regulations. This interpretation of empirical findings reflects a structural understanding of the frontline human service context, in which caseworker agency is enabled or constrained by how workers negotiate their formal and informal roles in relation to other institutional entities (e.g., court officials, child welfare administrators) (Soss, Fording, and Schram 2011; Maynard-Moody and Musheno 2012).

Thus, possibly in contrast to the perspective that service plans and case plans are bureaucratic constraints on child welfare caseworker agency, we propose that red tape may create opportunities for minority frontline workers to express agency and engage selectively with clients, an argument that is supported by Watkins-Hayes (2011) and Brodkin (2012). In the face of considerable case demands and limited time, non-Caucasian caseworkers may choose to exert extra effort on behalf of children and families who are culturally similar to themselves, such that racial/ethnic matching serves as a heuristic or shortcut in organizing service efforts in complex frontline practice environments (Lipsky 1980). Institutional signals may therefore be used strategically for non-Caucasian frontline actors to express culturally based preferences. That these institutional tags (e.g., registering the need for certain services in a case plan) are putatively race-neutral adds to the potentially non-obvious nature of active representation of racial/ethnic preferences in frontline child welfare practice.

Limitations

Our findings and their interpretation should be considered in relation to a number of study limitations. First, these data represent caseworkers’ reports on their referral actions and may be subject to social desirability or recall biases. Second, to understand the nature of housing referral strategies, the sample was restricted to families with identified need for housing assistance; thus, the study does not reflect the full continuum of child welfare-involved caregivers. Sampling only primary caregivers needing housing supports may have led to selection biases, which may not have been fully accounted for in our model specification and may have resulted in more positive multivariate estimates than if a more universal sample of caregivers had been used. Third, because NSCAW data were gathered via a cross-sectional survey design, it was not possible to test for a causal relationship between caseworker-caregiver racial/ethnic similarity and different caseworker referral strategies; nor was it possible to test for potential mediation or moderation of the influence of passive representation by past versus contemporaneous caseworker or case-based factors. The omission of variables that may influence caseworker referral strategies but that were unavailable in NSCAW (e.g., ties between child welfare agencies and housing providers, caseworkers’ knowledge of housing service providers) also limited internal validity. Fourth, measures pertaining to the specific manner in which caseworkers carried out each referral strategy and/or of the precise degree of caseworker discretion regarding each case were also unavailable, limiting our ability to more definitively identify the frontline mechanisms involved in securing housing supports. Fifth, while based on a national sample, our findings are only generalizable to families subject to child maltreatment investigations or assessments conducted between 2008 and 2009 and living in states not requiring CPS agency first contact of sample members. Therefore, our findings may not be applicable to caregivers whose children were receiving long-term foster care case management services. Finally, statistical power considerations prevented testing for differences in referral strategies between and among individual racial/ethnic groups. As a result, we were unable to estimate the relative contributions of different ethnic identities among Caucasian caseworkers and caregivers as well as specific racial identities for frontline housing referral efforts.

Research and Practice Implications

Our findings emphasize the importance of uncovering the mechanisms and effects of culture-based frontline processes for child welfare-involved families. Research is needed to understand the ways in which caseworkers’ cultural knowledge, willingness, and ability to actively represent diverse client groups and/or use different client engagement practices shapes frontline practice and in turn, service access and quality. Evidence from qualitative studies suggests that non-Caucasian caseworkers serving same-race families may fear breaches of confidentiality and may face challenges in reconciling their cultural identity with their formal organizational roles (Watkins-Hayes 2009; Sawrikar 2013), implying that the benefits of frontline racial/ethnic matching should be considered in relation to potential costs for caseworkers, caregivers, and agencies. Research reflecting these concerns might explore how minority and majority caseworkers balance their often-intersecting social identities (i.e., race, ethnicity, and gender) with their formal organizational roles (Watkins-Hayes 2009; 2011). Yet because worker efforts towards the active representation of culturally similar families may be confounded by institutional leveling factors beyond caseworker control, studies capturing variation in frontline and organizational settings may help determine how caseworker characteristics, active representation, and institutional factors act as levers for enhancing child and family outcomes. Whether quantitative or qualitative, longitudinal studies across different institutional and frontline practice contexts are needed to examine how the caseworker-client relationship develops and the specific manner in which racial/ethnic matching may serve as a resource for non-Caucasian and Caucasian caseworkers and their clients. Such multi-level (by unit of analysis and time) studies may clarify the universal versus culture-specific drivers of client engagement and advocacy involving different groupings of caseworkers and clients and thus may help identify the sources of racial/ethnic bias in frontline caseworker effort and service delivery.

Our findings also have implications for policy implementation and program development by suggesting that the use of culturally concordant staffing and service delivery strategies could propel non-Caucasian caseworkers to take additional actions on behalf of non-Caucasian child welfare-involved families. The results reinforce the importance of hiring a workforce that reflects the diversity in the child welfare client population, matching non-Caucasian clients with culturally similar caseworkers so as to reduce stigmatization and promote engagement, and ensuring that frontline services are delivered in a manner that minimizes racial/ethnic biases (Harris and Hackett 2008; Ortega and Faller 2012). Yet, given that the majority of caseworkers were not serving racially/ethnically matched caregivers, our research also underscores the importance of training and supervising all caseworkers to serve racially/ethnically diverse families sensitively and effectively. Finally, attention should be placed on examining the long-term consequences of passive versus active referral efforts of minority caseworkers who, based on current study findings, may be expected to expend additional effort to link culturally similar caregivers with needed services. Representative bureaucracy theory posits that minority bureaucrats may hold potentially competing roles of service gatekeeper, service provider, and client advocate. For these roles to be active across institutional settings and sustainable over time, administrators might consider ways to create organizational cultures, policies, and practices that support culturally informed frontline practice (Alegría et al. 2010; Briggs and McBeath 2010).

Conclusion

This study, which provides an initial application of representative bureaucracy theory to child welfare service processes, highlights how institutional and cultural processes shape the service strategies implemented by racially/ethnically diverse frontline workers. When non-Caucasian caseworkers receive clear institutional signals to invest effort in meeting the needs of primary caregivers, racial/ethnic similarity between the caseworker and caregiver increases the odds that caseworkers will actively pursue housing services. Based on our findings, institutional signaling may help child welfare agencies and frontline workers realize the advantages of racial/ethnic matching of non-Caucasian clients and caseworkers. Thus, our findings support the development and implementation of institutional and organizational processes and policies that formally sanction caseworkers’ efforts to actively represent the needs of minority caregivers in the child welfare system, with the goal of reducing racial/ethnic disparities in service utilization and improving outcomes for children and families.