Discrimination and Administrative Burden in Public Service Markets: Does a Public–Private Difference Exist?

Discrimination and Administrative Burden in Public Service Markets: Does a Public–Private... Abstract Public services are often provided in markets where both public and private providers operate. Irrespective of ownership status, public services ought to be accessible regardless of clients’ race, gender, ethnicity, or age. However, as theories of statistical discrimination and cream skimming suggest, market-based incentives may lead service providers to focus on nonminority clients because they perceive them as easier-to-serve and therefore less costly. This may lead to discrimination and hence jeopardizes equal access. In this study, we ask whether private, for-profit providers are more likely to discriminate on ethnic grounds compared to publicly owned providers. We implement a field experiment within the Flemish elderly care market by sending out email requests with either a Flemish or a Maghrebian name to all public and privately owned nursing homes. For overall response rates, no statistically significant differences between senders were found. However, we do find that privately owned facilities are about 20 percentage points less likely to provide information on how to enroll when the request is send from a Maghrebian name alias, and blinded coders perceive the information sent to the Maghrebian alias as less comprehensive. In publicly owned facilities, no such differences exist. We conclude that a public–private difference does exist, but that the mechanism of discrimination is subtler than expected. Rather than directly refusing to respond, nursing homes increase administrative burdens and learning costs for minority applicants. Introduction Despite comprehensive anti-discrimination laws, citizen minorities repeatedly report difficulties not only when encountering bureaucrats but also when trying to access important public services such as education, health care, or social welfare. For example, a recent Gallup poll in the United States suggests that a vast number of African Americans perceive having fewer opportunities to access crucial public services (Gallup 2014). Stark inequities in service provision exist within European countries as well, as a recent literature review indicates (Dan and Andrews 2015). Moreover, Eurobarometer data shows that disadvantaged groups are increasingly dissatisfied with the accessibility of public services (Clifton et al. 2012) and reports about ethnic and racial disparities in both access to and use of public services are made repeatedly (e.g., Harris et al. 2006; Kugelmass 2016; Weinick, Zuvekas, and Cohen 2000). Public administration scholarship has a long-lasting tradition of studying the sources and consequences of social inequity in public service delivery (for overviews, see Frederickson 2010; Gooden 2015). Inequality in service access has been often attributed to clients’ personal characteristics (e.g., their social capital, cognitive abilities and more general vulnerabilities in economic exchange relationships) (Clifton et al. 2011; Jilke 2015; Schneider, Teske, and Marschall 2002). Additionally, institutional structures, routines and even public officials’ (implicit) attitudes and behaviors factor in as well (Dabney et al. 2006; Epp, Maynard-Moody, and Haider-Markel 2014). A recent addition to the study of inequality in public services looks at how administrative burdens work as a mechanism of exclusion. The literature on administrative burden distinguishes three types of costs that may increase the difficulty citizens have in accessing public services, namely learning costs (citizens must learn about a service and how to apply for it), psychological costs (citizens may face stigma or stress associated with accessing a certain service) and compliance costs (citizens need to comply with a service’s formal rules and requirements) (Moynihan, Herd, and Harvey 2015; Moynihan et al., 2013). These costs are not stable and can be altered by policy makers or frontline workers. Frontline workers play a particularly important role. Their work discretion allows a considerable margin for increasing or decreasing administrative burden (Brodkin and Lipsky 1983; Keiser and Soss 1998; Lipsky 1984; Soss, Fording, and Schram 2011). Frontline work therefore affects the difficulties clients may have in accessing services to which they are entitled. The result is what Lipsky (1984) refers to as “bureaucratic disentitlement.” Some clients may already face an increased administrative burden in their first encounter with service providers—namely when asking for guidance and information concerning how to enroll in public programs. In this case, learning costs could be easily increased by street-level organizations to dissuade potentially costly clients by not replying to requests, providing incomplete information or simply being rude and discouraging. However, how do street-level bureaucrats know which clients are costly? In line with theories of statistical discrimination (Arrow 1973; Phelps 1972), we argue that in the absence of concrete information about individual clients’ future costliness, frontline workers use imperfect signals such as race, ethnicity, and age to make inferences about a client’s future costliness. As a consequence, employees of street-level organizations may increase the administrative burden for these potentially harder-to-serve clients. Few studies have examined whether (and to what extent) frontline workers increase administrative burden for members of societal minorities (for a notable exception, see Heinrich 2016). This is an important consideration given that discretionary bias in the allocation of opportunities and resources can result in unequal treatment of minority groups—especially in the presence of limited resources (Brodkin 1997; Lipsky 1980; Scott 1997). While work discretion provides frontline workers with the opportunity to discriminate, incentives in public service markets may provide them with a motive. Today most public services are not delivered by state monopolists, but by service providers that compete for clients, thereby often facing performance incentives to succeed economically. Such market-based incentives within (quasi-)markets constitute one of the key theoretical mechanisms related to why minority groups may face an increased administrative burden. It has been argued that frontline providers would be more prone to either avoid costly, labor-intense clients (i.e., engage in dumping) or prioritize clients who they deem most likely to succeed in terms of bureaucratic success criteria (i.e., undertake cream-skimming) (Ellis 1998; Koning and Heinrich 2013; Lacinero-Paquet et al. 2002; Le Grand and Bartlett 1993; Lipsky 1980; Tummers et al. 2015). In other words, street-level organizations not only prioritize motivated clients who are most likely to succeed, but also avoid clients who seem likely to either fail or be costly. Arguably, quasi-markets (e.g., education, health care or employment services) have an inherent incentive structure that may lead both public and private service providers to a focus on easier-to-serve clients (Heinrich and Marschke 2010; Koning and Heinrich 2013; Soss, Fording, and Schram 2011). Yet, there are good reasons to believe that incentives to cream-skim or dump clients are affected by a service provider’s ownership status.1 Since private facilities can retain profits for their shareholders, there is a direct financial gain to be made by profiling applicants with stronger socioeconomic profiles. In public facilities, no such profit-maxim exists. Moreover, a stronger public sector ethos can be expected to mitigate the propensity for dumping or cream skimming. In this study, we look for evidence of discriminatory practices against ethnic minorities who are accessing elderly care services (which are typically provided by public, private, or nonprofit providers within a national care market; see also Amirkhanyan, Kim, and Lambright 2008). We implement an audit study design field experiment to assess the revealed discrimination against ethnic minorities among street-level organizations across all public and private elderly care facilities in Flanders (i.e., the Dutch-speaking part of Belgium). We find that elderly care homes do not discriminate against ethnic minorities in responding to information requests. However, we find public–private differences when it comes to increasing administrative burdens through learning costs. Here, private elderly care homes are approximately 20 percentage point less likely to provide information on how to enroll to a Maghrebian sender than to a sender with a Belgian alias; however, publicly owned homes do not make any differentiation. On this basis, we discuss the theoretical and practical implications of public–private differences in ethnic discrimination in access to elderly care—and public services more broadly—and propose a future research agenda for experimentally studying racial and ethnic discrimination within public service delivery. Discrimination within Public Service Delivery: Theory and Evidence Discrimination within access to public services implies that applicants receive different treatment based on belonging to a certain group, such as a racial or ethnic minority. This unequal treatment, which disadvantages a particular group and its members, is often based on other inadequately justified factors beyond race or ethnicity as well (National Research Council 2004). In this study, we broadly distinguish between three types of discrimination: implicit, unconscious discrimination; explicit, overt discrimination; and rational-strategic discrimination.2 Theories of implicit discrimination argue that discriminatory practices are the product of implicit attitudes or stereotypes against the outgroup: that is, people who are perceived not to belong to the own group (e.g., Chugh 2004; Greenwald and Krieger 2006). As such the act of discrimination is “unintentional and outside the discriminator’s awareness” (Bertrand, Chugh, and Mullainathan 2005, 94). In the field of social psychology, these implicit attitudes are measured using the Implicit Association Test (Greenwald, McGhee, and Schwartz 1998). Many scholars have asserted that implicit attitudes (e.g., toward African Americans) predict discriminatory behavior. However, an important meta-analysis (Oswald et al. 2013) has suggested only a weak association between implicit attitudes and racial and ethnic discrimination. A more recent network meta-analysis of how changes in implicit biases predict explicit bias or discriminatory behavior has similarly found only very limited evidence of such a relationship (Forscher et al. 2017). Hence, while the prevalence of implicit attitudes and stereotypes toward outgroups is widely acknowledged in the literature, the link of implicit attitudes to discriminatory practices seems to be not as straightforward as often suggested. Secondly, theories of explicit, overt discrimination argue that discrimination stems from individuals’ distaste toward outgroup members. In his theory of taste-based discrimination, Becker (1971) claims that discrimination is mainly based on an individual’s animus against members of certain outgroups. This taste for discrimination leads individuals to avoid interacting with those against whom they harbor animus—even if doing so comes at some cost. A disamenity value thus exists, for instance in terms of accepting members of certain outgroups into a public program. Thirdly, the literature on statistical discrimination argues that discrimination is not an emotional response to differences but a product of rational-strategic considerations based on aggregate group characteristics. Those aggregate characteristics are used to stereotype representatives of these groups (Altonji and Pierret 2001; Arrow 1973; Phelps 1972). A classic example is an employer who tries to assess the motivation and future productivity of a job applicant. As this type of information is not directly available in the early stages of the application process, employers use so-called imperfect signals of productivity from the applicant’s curriculum vitae (CV). The less direct information an employer obtains from these sources, the more weight he or she assigns to the average productivity of other workers who belong to the same group as the applicant (based for instance on gender, race, ethnicity, age, or place of residence). The employer subsequently uses this aggregate-level group information as a stereotype against individual applicants. The main difference between the explicit and rational-strategic models of discrimination is that taste-based discrimination stems from an agent’s distaste for interacting with members of a particular group, even if avoiding interaction with them bears some costs. Theories of statistical discrimination in turn argue that in the absence of direct information, agents use stereotypes based on population-specific means to avoid individuals who they (or their organizations) would benefit from the least (Guryan and Charles 2013). In the case of access to public services, this also means that having limited information leads service providers to try to avoid clients deemed costly based on aggregate-level group information. As such they use clients’ group membership as an imperfect signal of their future costliness. Statistical discrimination in access to public services can be seen as a form of cream-skimming—which is defined as the intentional selection or avoidance of certain groups for a public program (Ellis 1998; Koning and Heinrich 2013; Lacinero-Paquet et al. 2002; Le Grand and Bartlett 1993; Lipsky 1980). Cream-skimming (or its functional equivalent of dumping) is founded on rational strategic considerations related to reaching program goals or working in a cost-efficient manner. Nonetheless, like statistical discrimination it involves disadvantaging members of a certain group based merely on their group characteristics. Theories of statistical discrimination provide a framework for explaining the mechanism at play when imperfect signals (such as race or ethnicity) are used in the process of cream-skimming; for example, nursing homes may try to avoid clients they deem labor intense or costly. The absence of full information about the future costliness of individual applicants leads care providers to avoid members of certain groups that on average create higher workloads and increased costs. The Maghreb community in Flanders (with migrants from Morocco, Tunisia, and Algeria) is a good example. Members of this group practice a different religion from the majority of the region’s residents (i.e., Islam instead of Roman Catholicism) and therefore have different nutritional needs (e.g., they require halal meat) and religious ceremonies. In addition, they have a higher likelihood of not speaking the local language as fluent as the part of the Flemish population whose mother language is Flemish. Using this population-specific information as a stereotype against individual Maghrebian applicants may lead service providers to stereotype them as being harder to serve (i.e., costlier) than similar Belgian applicants—and hence try to avoid them. Much of the empirical evidence concerning discriminatory practices draws on theories of taste-based and statistical discrimination (Guryan and Charles 2013). The related literature typically examines discrimination in employment decisions based on applicants’ race, gender, or ethnicity. Many empirical studies have employed an audit-study methodology in which fictitious applications are sent in response to public job openings (e.g., Bertrand and Mullainathan 2004; Kaas and Manger 2011; Pager, Western, and Bonikowski 2009). The application materials (including the CV and cover letter) are usually identical, apart from one important characteristic, such as the applicants’ race, gender, or ethnicity (depending on what is being studied). Bertrand and Mullainathan’s (2004) famous audit study looked at racial discrimination in the Boston and Chicago labor markets. The study entailed submitting similar resumes to more than 1,300 help-wanted advertisements in local newspapers, using names manipulated to sound either white or African American (such as Greg or Jamal). The findings reveal that African Americans face considerable discrimination when searching for jobs; members of this group were approximately 50% less likely to receive a call-back than equally qualified “white” applicants. These results have been largely corroborated by numerous other studies (for a recent meta-analysis, see Quillian et al. 2017). In addition to job applications, audit studies also exist in relation to issues such as housing (Ahmed and Hammarstedt 2008; Turner and Ross 2003; Turner et al. 2002; Yinger 1995), car sales (Ayres and Siegelman 1995) and insurance markets (Galster, Wissoker, and Zimmermann 1998). Activities such as hailing taxis (Ridley, Bayton, and Quttz 1989), being recommended for medical care (Schulman et al. 1999), making an appointment with a psychiatrist (Kugelmass 2016) and even contacting elected officials (Broockman 2013; Butler and Broockman 2011) have been explored as well. The related studies have revealed consistent patterns of racial and ethnic discrimination across a wide range of contexts. Recently, scholars of bureaucracy have started to examine discrimination in public services by using audit studies to explicitly explore individuals’ learning costs in relation to accessing services (see also Costa 2017 for a meta-analysis of audit studies on public officials). There is evidence of racial discrimination in terms of requesting information about how to register to vote (White, Nathan, and Faller 2015), but also in answering information requests send to traditional public service providers like school districts, libraries, police forces, job centers and county clerks (Giulietti, Tonin, and Vlassoupoulos 2017). Einstein and Glick (2016), in contrast, find no evidence for racial discrimination in access to public housing. Response rates to Black, Hispanic, or White senders are not significantly different. However, they report that responses sent to Hispanic senders were less likely to include a salutation. Other studies that have examined ethnic discrimination in information requests that were sent directly to local governments found no systematic differences in responses (Adman and Jansson 2016; Grohs, Adams, and Knill 2015). A recent study on ethnic discrimination in information requests of how to enroll into receiving unemployment benefits also found no evidence of ethnic discrimination in response rates, but in the quality of information provided, with ethnic minorities receiving replies of lower quality (Hemker and Rink 2017). These experimental studies provide some reasons to expect ethnic discrimination in access to elderly care services. However, the public services chosen in prior studies operate mostly within monopolistic service structures (e.g., local government, policing, public housing, or welfare provision), where we would expect only low-powered incentives for cream-skimming/dumping (unlike services that are provided within competitive environments). Indeed, public services delivered within noncompetitive environments typically do not face similar performance pressures as services within (quasi-)markets (Le Grand and Bartlett 1993; Savas 2000). As a result, audit studies on bureaucratic discrimination in monopolistic settings that have examined the nature of discriminatory practices suggest that acts of discrimination are the product of taste-based animosity and not rational-strategic considerations (i.e., statistical discrimination) (Costa 2017; Giulietti, Tonin, and Vlassoupoulos 2017; White, Nathan, and Haller 2015). Our study focuses on a service provided within a (quasi-)market, where we would expect performance incentives that encourage discriminatory practices based on rational-strategic considerations. In addition, our research context allows us to examine differences between public and private (for-profit) providers, and thereby to acknowledge varying degrees of discrimination across institutional structures. Public–Private Differences in Discrimination Good reasons exist for expecting that private, for-profit providers are more likely to engage in cream-skimming (or dumping) than publicly owned providers. Different institutional logics in public and private organizations mold the incentive structure that shapes work discretion (Heinrich and Marschke 2010; Soss, Fording, and Schram 2011). Private providers distribute their profits to shareholders and managers. Public providers cannot do so. Weisbrod (1988) refers to the nondistribution constraint which prevents public and not-for-profit organizations to distribute profits to shareholders. Thereby the nondistribution constraint reduces the incentives for public agencies (and nonprofits) to misrepresent or withhold information about how to access services, reduce quality or prioritize certain types of service users at the “front door” of service delivery. The impact of the incentive structure for organizations on the frontline of service delivery is determined through processes of professional norm-setting (Abbott 1988). Incentives for profit-seeking providers tend to emphasize short-term goals, often at the expense of long-term goals related to client outcomes (Dias and Maynard-Moody 2007). In contrast, the workers at publicly owned street-level organizations lean more heavily on distinct professional standards to help clients to the best of their ability. The literature on public service motivation provides insights into what these standards might entail. Public sector employees assign greater value to public service; work that is beneficial to both others and society as a whole; involvement with important public policies; and self-sacrifice, responsibility, and integrity (Rainey and Bozeman 2000, 460). Perry, Hondeghem, and Wise (2010) refer to an other-regarding orientation, which is represented by notions of self-sacrifice, altruism, and prosocial behavior. The focus on the other, in our case the prospective client (i.e., elderly individuals seeking care), can be expected to constrain cream-skimming and dumping practices. The professional care standards can be an antidote against exclusion. Public sector organizations can be expected to be better capable of buffering external competitive pressures and safeguarding professional standards (Meier and O’Toole 2011). In private sector facilities, economic incentives may crowd out the professional standards and public sector motivations (Kjeldsen and Jacobsen 2013). Lacireno-Paquet et al. (2002) found evidence for differences in dumping (i.e., avoiding needy populations) in market-oriented charter schools; however, they found less evidence of differences among public and nonmarket-oriented private schools. Studies of performance target regimes that introduce a for-profit logic into public services have also found evidence of crowding out effects. Several researchers have documented cases in which performance targets induce frontline workers to focus exclusively on measured tasks (Hood 2006; Kerpershoek, Groenleer, and de Bruijn 2016; Pollitt 2013). The goals as measured by performance indicators (as opposed to the real goals) provide direction to work within street-level organizations. For example, schools may focus on obtaining better test scores rather than on providing better education (teaching for the test) and hospitals may focus on shortening waiting times instead of promoting better health (Bevan and Hood 2006; Radin 2006). Bohte and Meier (2000) discuss this phenomenon as an instance of goal displacement. When performance is evaluated in terms of numerical outputs, organizations have an incentive to maximize their outputs—regardless of whether output maximization is the preferred strategy for achieving desired social outcomes (Bohte and Meier 2000, 173). Similar effects have been found in health care (Bevan and Hood 2006), education (Bohte and Meier 2000), job training (Heinrich 1999), and policing (Epp, Maynard-Moody, and Haider-Markel 2014). We expect that private for-profit providers have an incentive to select applicants who are more profitable and avoid those they regard as costly. Ethnic minorities risk being perceived as harder to serve due to their different cultural backgrounds and demands (in relation to issues such as dietary requirements, language, and cultural practices). We base this expectation on theories concerning the work of frontline workers. Incentives for profit maximization in private facilities may limit the scope for workers to uphold professional standards in their daily work. Hypotheses The study tests two hypotheses related to discrimination against ethnic groups in accessing elderly care services. Our first hypothesis is derived by combining literature on statistical discrimination and cream-skimming. In line with a sizeable number of studies, we assume that street-level organizations respond to competitive pressure by intentionally avoiding groups they deem harder to serve and thereby costly. Statistical discrimination provides a theoretical framework for explaining ethnic discrimination in the context of cream-skimming (including dumping). We assert that for employees of street-level organizations, ethnicity serves as an indication about the future costliness of individual clients. This imperfect signal influences street-level decision-making in the process of cream-skimming (or dumping) when no direct signal of a client’s future costliness exists. Against this background, we argue that street-level organizations within competitive environments (such as the Flemish elderly care market) use applicant clients’ ethnicity to determine if they are harder to serve and therefore costly. Within this study we focus on Maghrebians, who currently constitute the largest immigrant group in Flanders to date (Van den Broucke et al. 2016, 96). Members of this particular group who are at a retirement age typically do not speak the Flemish language well and have certain religion-based nutritional needs. Street-level organizations use this objective, population-specific information to stereotype individual applicants, who they perceive as being harder to serve and therefore costlier than similar Flemish applicants. As a result, elderly care providers may try to avoid individual applicants who belong to this particular group to escape future costs. This leads us to expect that Maghrebians will experience more discrimination than similar, otherwise Flemish applicants when they try to access information concerning how to apply for elderly care places in the Flemish elderly care market. Hypothesis 1: Maghrebian aliases experience greater discrimination than Flemish names in relation to accessing information about applying for elderly care services. We expect discriminatory practices to be more pronounced in the private (for-profit) elderly care service sector than in the public sector. In private organizations, the profit motive creates an incentive structure that reduces the scope for maintaining professional standards of care. Moreover, profit maximization incentives may lead workers to substitute commercial standards acquired through job socialization for altruistic public service standards learned through professional training (Kjeldsen and Jacobsen 2013). In contrast, employees in the public sector are shielded by the nondistribution constraint, which buffers them from competitive pressures and bolsters their maintenance of professional standards. This leads us to our second hypothesis: Hypothesis 2: The extent of discrimination against Maghrebian aliases is stronger in privately owned (for-profit) elderly care homes than in publicly owned homes. Research Context: The Flemish Elderly Care Market In Flanders, elderly care is provided by 783 homes. The field of providers is diverse and features different ownership structures. The largest group has a nonprofit legal status. Facilities in this category are typically run under the auspices of religious (mainly Catholic) organizations. The second category is elderly care homes with public ownership. Homes in this group are administered by municipalities—but like the nonprofits, the public facilities are mostly regulated and financed by the central government. The third category includes homes with private (for-profit) ownership, which again fall under the same regulatory framework as the public and nonprofits. The increase in private, for-profit ownership in the Flemish elderly care market is a trend of the last decade. In this study, we focus on the last two categories, namely public and private facilities. The overall goal of the elderly care policy in Flanders is to provide universal access to affordable and high quality care services (Van den Bosch et al. 2011; Willemé et al. 2012). Residents pay their accommodation costs, while the compulsory national health insurance scheme bears their medical and nursing costs. Means-tested compensation exists for individuals who cannot pay for their share. The monthly accommodation costs are comparable for nonprofit and profit organizations but slightly lower for public institutions (Socialistische Mutualiteiten 2016). About 40% of the income of the elderly care facilities comes from health insurance. Elderly care homes typically provide beds for residents with lower and higher care needs; the beds for individuals with greater requirements (such as those with dementia) are called RVT beds. Higher care needs are discounted for with higher payments from the national health insurance. Another 40% of facilities’ income stems from the daily allowances that residents have to pay. The remaining 20% comes from miscellaneous subsidies and cost recuperation. The nonprofit sector is the most profitable sector, with a profit of EUR 4.60 a day/bed; followed by the private sector, at EUR 0.90 a day/bed.3 The public sector is losing approximately EUR 13 a day/bed (Pacolet and De Coninck 2015). The deficit is covered by the municipality in which the elderly care home is located. The central government regulates the elderly care sector. All facilities need to be certified by an inspectorate before they can claim subsidies. This inspectorate monitors compliance with regulations as well as the quality of care being provided. Inspection reports are publicly available. In addition, the central government determines how many beds are required within each region, and elderly care homes can apply for additional beds based on the forecasted need. Based on residents’ care needs, the central government also determines the daily amount that the health insurance scheme has to pay facilities. Nonetheless, elderly care homes have a significant level of autonomy within this regulatory framework. They are separate legal entities that can independently make hiring decisions, borrow from financial institutions, and accumulate assets. They also have significant freedom in determining what they charge residents and can manage their waiting lists independently. In summary, the elderly care sector in Flanders can be described as a regulated public service market. Residents are free to choose between facilities. In addition, facilities are autonomous decision making units that independently manage their admission processes and waiting lists. With an occupancy rate of 96% in 2014 (Socialistische Mutualiteiten 2016), we can safely assume that creaming is a realistic option—which makes it an ideal sector to test our theoretical predictions. Experimental Design In this study, we examine the discriminatory behavior of elderly care facilities in Flanders by studying their responsiveness to information requests by prospective clients. Withholding or providing incomplete information when employees respond to queries concerning how to apply for an elderly care place is an important administrative burden for potential applicants. To explore potential discriminatory practices related to the provision of information about how to enroll in elderly care, we build on the audit study tradition of randomized field experiments (for an overview, see Pager 2007). The basic experimental design of the audit study methodology relies on sending identical information requests that differ by one attribute, namely the race, ethnicity, or gender of the putative sender. The discriminatory behavior of the audited agents is assessed by comparing response rates and the information provided across randomly assigned requester names. Treatment We tested the effectiveness of randomized names on discrimination in access to nursing homes for the elderly in Flanders. Identical information requests were sent to Flemish elderly care homes. To test public–private differences, we focus on public and private, for-profit nursing homes as they most clearly have the incentive structures we are theoretically interested in.4 The treatment consisted of two randomly allocated, hypothetical names. We used one typical Maghrebian name (Mohammed El Markini), with the goal of representing the predominant ethnic minority in Flanders, and one that is typically Flemish (Kenny Maes). These names were selected from a pretest. In Flanders, as in other countries, names are social identifiers not only of ethnicity, but also of socioeconomic status (SES) (Elchardus and Siongers 2011). For our study, we wanted to choose names that signal different ethnicities but a similar SES. Elchardus and Siongers (2011) found that parental educational levels are considerably higher for some names than for others, such as Thomas versus Kenny. We used this study as the basis for selecting Flemish names for our pretest. Therefore, we identified the most popular Maghrebian first names in Belgium from the Belgian national census. Given the underprivileged position of the Maghreb community in Belgium, we generally expected that Maghrebian names signal lower SES (Van den Broucke et al. 2016). The Flemish and Maghrebian last names for the pretest were selected based on the most recent Belgian census (Statistics Belgium). In the pretest, we used an existing online panel of the general population from the political science department of a Belgian university. Out of the 5,837 people who were asked to participate in the survey, 2,114 took part—which reflects a response rate of 36.2%. Each respondent was asked to assess three names based on four factors: education level, age, ethnicity, and income level.5 To avoid answer contamination, we randomized the order of the four items as well as the combination of the first and last names. Respondents randomly received three Flemish names, three Maghrebian names, or a combination of the two. In total, 10 Flemish and 10 Maghrebian names were assessed (see table A1 in the Appendix). As expected, all of the Maghrebian names triggered lower SES assessments than the Flemish names. Of the Flemish names, “Kenny” accounted for the lowest SES ratings. Based on these results, we opted to use Mohamed El Makrini and Kenny Maes, as these names are perceived as being very similar in terms of SES but not vis-à-vis ethnicity (see table A1 in the Appendix). The treatment was implemented in an email sent to all Flemish elderly care homes that asked whether the facilities currently had a place available, whether they had a waiting list and how to enroll. Two email accounts were created (kenny.maes100@gmail.com and mohamed.elmakrini100@gmail.com) and used to send the requests. The email itself was fairly short to decrease the burden for employees; box 1 in Appendix provides the full text.6 Each elderly care facility received only one email. Population and Unit of Analysis The unit of analysis for our field experiment was individual elderly care facilities. While we contacted individuals, we assumed that they would act in their professional capacity as a representative of the organization for which they work. The primary contact information for the facilities was collected from publicly available records (Agentschap Zorg and Gezondheid 2017; in English: Agency Care & Health). Public records also display the total number of elderly care homes in each municipality and indicate which facilities operate under public, private, and nonprofit ownership (according to their legal status; see table A2 in the Appendix). When elderly care facilities shared the same contact email address as a result of operating within the same umbrella organization, we randomly chose one address (a total of 18 facilities fell into this category: 7 public and 11 private). Allocation to the experimental conditions was performed randomly in strata of public and private organizations, thereby employing a randomized block design. As such we conducted an independent randomization procedure for each of the two facility categories (i.e., public and private). We sent 214 emails to public elderly care facilities and 118 emails to private elderly care facilities. Of the 332 emails that were sent in total, 21 (15 to public facilities and 6 to private facilities) could not be successfully delivered due to invalid addresses. For our analysis, we limited the sample to emails that were successfully delivered, which left us with an analytical sample of 311 observations (including 199 public and 112 private elderly care homes).7 Measurement We used two outcome measures to assess the discriminatory behavior of Flemish elderly care homes: response rates and the information provided. We chose these outcome variables because they represent core aspects of elderly care employees’ discretion in granting clients access to public services more widely. From those nursing homes that responded, only 5 had no waiting list and 49 did not provide any information on this account. Therefore, we did not include this question as an outcome measure. The response rate was measured by looking at whether service providers responded to our information requests. We counted systematic differences between ethnic names as evidence of ethnic discrimination.8 Second, we looked at whether those who responded provided information about how to apply for a place at their facility. Indeed, withholding information or providing incomplete information may result in a disadvantaged position for clients trying to access these services. This is therefore an important measure of whether elderly care facilities systematically increase administrative burden (i.e., learning costs) for certain types of clients. We coded responses that included the requested information as “1” and those that did not respond to the inquiry as “0.” Responses were considered valid if they were received within 2 weeks of the request email being sent (after 2 weeks no additional replies were received). Several elderly care homes provided us with information in attachments; for instance, one facility sent us an attached informational brochure that contained details concerning its enrollment procedure. These emails were coded as “1” for our second outcome measure (i.e., providing information about how to enroll). In addition, we report the results of two secondary measures to complement the results of our primary measures of interest: information comprehensiveness and reply friendliness. These were created through ratings of three blinded coders (research assistants). In particular, we asked coders to rate all received replies (which we anonymized with regard to its sender, but also with respect to the experimental condition) on a 7-point scale, ranging from “no information was provided,” “written in an unfriendly manner” to “all asked information was provided,” “written in a friendly manner,” respectively. In this sense, the rating of information comprehensiveness is strongly correlated with our primary measure of interest on information provision. Coders were given the initial request that we sent out. They were also provided examples for each scale’s middle category and respective endpoints to make sure equivalent rating standards are applied across coders (Krippendorff 2013). We have to note, however, that while the intercoder reliability of the information scale was somewhat acceptable, the friendliness ratings’ reliability was rather poor.9 Empirical Results Response Rates Of the 332 emails we sent to elderly care facilities, 223 received responses (table 1). If the 21 emails that bounced back are not taken into account, this yields an overall response rate of 71.7%. The overall response rate for the public facilities were approximately 16 percentage points higher than for the private facilities (77.4% versus 61.5%). Flemish elderly care homes can thus be seen as moderately responsive overall, which exemplifies that not replying to email requests is indeed a realistic option for facilities. Arguably, not responding to emails could be an easy way for nursing homes to “select at the door.” Table 1. Summary statistics for the response rates All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails sent 166 166 107 107 59 59 Incorrect address/ bounce 11 10 8 7 3 3 Response rate 112 (72.3%) 111 (71.2%) 75 (75.8%) 79 (79%) 37 (66.1%) 32 (57.1%) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails sent 166 166 107 107 59 59 Incorrect address/ bounce 11 10 8 7 3 3 Response rate 112 (72.3%) 111 (71.2%) 75 (75.8%) 79 (79%) 37 (66.1%) 32 (57.1%) View Large Table 1. Summary statistics for the response rates All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails sent 166 166 107 107 59 59 Incorrect address/ bounce 11 10 8 7 3 3 Response rate 112 (72.3%) 111 (71.2%) 75 (75.8%) 79 (79%) 37 (66.1%) 32 (57.1%) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails sent 166 166 107 107 59 59 Incorrect address/ bounce 11 10 8 7 3 3 Response rate 112 (72.3%) 111 (71.2%) 75 (75.8%) 79 (79%) 37 (66.1%) 32 (57.1%) View Large Table 1 presents the response rates by treatment condition (i.e., type of name) within public and private elderly care homes. We find no significant differences between emails signed with a Flemish name versus those signed with a Maghrebian name (the difference is 1.1 percentage point; χ2(1) = 0.047, p = .829). When a logistic regression model is estimated instead, using standard errors clustered by municipality to account for potential clustering effects across facilities located in the same area, the results are identical (table 2). In public elderly care homes, signing an email request with a Maghrebian alias led to a 3.2 percentage point greater likelihood of receiving a response than a similar request signed with a Flemish name; however, these differences are statistically insignificant (χ2(1) = 0.944, p = .331). Identical results were obtained for public and private organizations when potential clustering effects within municipalities were accounted for. In relation to private facilities, emails signed with a Maghrebian name were approximately 9 percentage points less likely to receive a reply than those that used a Flemish name. While this effect is nontrivial in size, it is not statistically significant (χ2(1) = 0.299, p = .585). Table 2 summarizes the average treatment effect (ATE) of Maghrebian versus Flemish name aliases for public, private, and both facilities combined. Table 2. Response Rates (Logistic Regression Models) (1) (2) (3) All Public Private Maghrebian name −0.011 0.032 −0.089 (0.050) (0.059) (0.091) Observations 311 199 112 (1) (2) (3) All Public Private Maghrebian name −0.011 0.032 −0.089 (0.050) (0.059) (0.091) Observations 311 199 112 Note: Standard errors in parentheses (clustered by municipality). Coefficients are marginal effects. ***p < .01. **p < .05. +p < .1. View Large Table 2. Response Rates (Logistic Regression Models) (1) (2) (3) All Public Private Maghrebian name −0.011 0.032 −0.089 (0.050) (0.059) (0.091) Observations 311 199 112 (1) (2) (3) All Public Private Maghrebian name −0.011 0.032 −0.089 (0.050) (0.059) (0.091) Observations 311 199 112 Note: Standard errors in parentheses (clustered by municipality). Coefficients are marginal effects. ***p < .01. **p < .05. +p < .1. View Large Information Provision Of the 332 emails sent to elderly care facilities, 196 (59.0%) generated responses that included information about how to enroll in an elderly care facility. While this figure represents a reasonable share of the elderly care facilities, it also demonstrates that a significant number of facilities did not share such information with requesters—and thereby increased their learning costs vis-à-vis access to elderly services. Below we assess whether this increase is systematic, which includes examining the number of responses that contained this information relative to the total number of emails we sent. We do so because focusing on only the subset of responses could break the experimental randomization,10 which could in turn potentially introduce bias into the identification of a causal effect of a Maghrebian versus a Flemish name in increasing the administrative burden. In table 3, the number and percentage of responses that include information about how to apply for a place at a Flemish elderly care home are provided by experimental condition. Of all of the valid requests sent, Maghrebian senders were approximately 8.1 percentage points less likely to receive information than identical requests from Flemish requesters. While this difference is of substantive magnitude, it is not statistically significant (χ2(1) = 2.201, p = .198). When the ATE within a logistic regression framework is estimated with standard errors clustered by municipality, the results are identical. In relation to public facilities, neither substantive nor statistically significant differences exist between the experimental conditions (the difference is only 1.7 percentage points; χ2(1) = 0.067, p = .796). However, examining the differences between name aliases in relation to private elderly care homes reveals substantive differences; here Maghrebian name aliases were approximately 19.6 percentage points less likely to receive enrollment information than identical requests from Flemish aliases. This finding is not only of a substantive magnitude, it is also statistically significant (χ2(1) = 4.334, p = .037). When the ATE of Flemish versus Maghrebian names is estimated in the context of a logistic regression model with clustered standard errors, the results are identical (table 4). Table 3. Summary Statistics for the Information Provided in Responses All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails with information on how to apply 104 (67.1%) 92 (59.0%) 69 (69.7%) 68 (68%) 35 (62.5%) 24 (42.9%) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails with information on how to apply 104 (67.1%) 92 (59.0%) 69 (69.7%) 68 (68%) 35 (62.5%) 24 (42.9%) View Large Table 3. Summary Statistics for the Information Provided in Responses All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails with information on how to apply 104 (67.1%) 92 (59.0%) 69 (69.7%) 68 (68%) 35 (62.5%) 24 (42.9%) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails with information on how to apply 104 (67.1%) 92 (59.0%) 69 (69.7%) 68 (68%) 35 (62.5%) 24 (42.9%) View Large Table 4. Information Provided (Logistic Regression Models) (1) (2) (3) All Public Private Maghrebian name −0.081 −0.017 −0.196** (0.053) (0.065) (0.086) Observations 311 199 112 (1) (2) (3) All Public Private Maghrebian name −0.081 −0.017 −0.196** (0.053) (0.065) (0.086) Observations 311 199 112 Note: Standard errors in parentheses (clustered by municipality). Coefficients are marginal effects. ***p < .01. **p < .05. +p < .1. View Large Table 4. Information Provided (Logistic Regression Models) (1) (2) (3) All Public Private Maghrebian name −0.081 −0.017 −0.196** (0.053) (0.065) (0.086) Observations 311 199 112 (1) (2) (3) All Public Private Maghrebian name −0.081 −0.017 −0.196** (0.053) (0.065) (0.086) Observations 311 199 112 Note: Standard errors in parentheses (clustered by municipality). Coefficients are marginal effects. ***p < .01. **p < .05. +p < .1. View Large Coder Ratings Next we assess whether blinded coders rated responses as more or less comprehensive in terms of the provided information. We also examine the perceived friendliness of responses (i.e., ratings using a 7-point scale), where nonresponses were treated as the lowest endpoints of the scale in order to not break the experimental randomization. Re-estimating these models for the subset of received responses yield no substantively different results. In table 5, mean differences in coder ratings between experimental conditions are summarized. In terms of the information completeness ratings we find very little differences between experimental conditions (mean difference of 0.11). These differences are not only trivial but also not statistically significant. Within public facilities, Maghrebian senders receive slightly more comprehensive replies on average (mean difference of 0.20) and this small difference does not reach acceptable levels of statistical significance. Only in private facilities the difference is stronger, with Maghrebian senders receiving replies of lower quality (mean difference of 0.66). This difference is statically significant at the 10% level, and hence corresponds well with our findings from table 4. Table 5. Mean Differences for Coder Ratings (Standard Errors in Parenthesis) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Information 3.42 (0.15) 3.31 (0.16) 3.49 (0.19) 3.68 (0.20) 3.29 (0.26) 2.64 (0.26) Friendliness 3.57 (0.14) 3.53 (0.15) 3.71 (0.18) 3.86 (0.17) 3.32 (0.25) 2.93 (0.25) Mean difference information 0.11 (0.22) −0.20 (0.27) 0.66 (0.37)+ Mean difference friendliness 0.04 (0.20) −0.15 (0.25) 0.38 (0.35) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Information 3.42 (0.15) 3.31 (0.16) 3.49 (0.19) 3.68 (0.20) 3.29 (0.26) 2.64 (0.26) Friendliness 3.57 (0.14) 3.53 (0.15) 3.71 (0.18) 3.86 (0.17) 3.32 (0.25) 2.93 (0.25) Mean difference information 0.11 (0.22) −0.20 (0.27) 0.66 (0.37)+ Mean difference friendliness 0.04 (0.20) −0.15 (0.25) 0.38 (0.35) ***p < .01. **p < .05. +p < .1. View Large Table 5. Mean Differences for Coder Ratings (Standard Errors in Parenthesis) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Information 3.42 (0.15) 3.31 (0.16) 3.49 (0.19) 3.68 (0.20) 3.29 (0.26) 2.64 (0.26) Friendliness 3.57 (0.14) 3.53 (0.15) 3.71 (0.18) 3.86 (0.17) 3.32 (0.25) 2.93 (0.25) Mean difference information 0.11 (0.22) −0.20 (0.27) 0.66 (0.37)+ Mean difference friendliness 0.04 (0.20) −0.15 (0.25) 0.38 (0.35) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Information 3.42 (0.15) 3.31 (0.16) 3.49 (0.19) 3.68 (0.20) 3.29 (0.26) 2.64 (0.26) Friendliness 3.57 (0.14) 3.53 (0.15) 3.71 (0.18) 3.86 (0.17) 3.32 (0.25) 2.93 (0.25) Mean difference information 0.11 (0.22) −0.20 (0.27) 0.66 (0.37)+ Mean difference friendliness 0.04 (0.20) −0.15 (0.25) 0.38 (0.35) ***p < .01. **p < .05. +p < .1. View Large With regard to friendliness ratings, overall no substantive differences between experimental conditions could be detected (mean difference of 0.04). In public facilities small differences exist, with Maghrebian senders receiving slightly friendlier replies (0.15 mean difference). But again, this small difference is not statistically significant. Within private facilities, small evidence for ethnic discrimination exist (mean difference of 0.38). However, the difference is too small to reach conventional levels of statistical significance. Because the data were left-censored, we estimated tobit models for coder ratings (accounting for potential clustering effects at the municipality level). Results do not change substantively (see table 6). With regard to information completeness ratings, we find that responses sent to the Maghrebian name were 2.6% less likely to be to be rated as providing all asked information. This difference is not only of trivial nature but also statistically insignificant. When considering public facilities alone, the opposite holds true in terms of the effect direction (i.e., a 3.4% increase for the Maghrebian name), but, again, this findings did not reach conventional levels of statistical significance and cannot be regarded as substantive. Within private facilities, however, systematic differences exist. Maghrebian senders received a response that provided less complete information on average (a 15.7% difference). This difference is both, of nontrivial nature, but also statistically significant (at the 10% level), and thus corresponds well with previous findings from table 4 (information provided on how to enroll). Table 6. Coder Ratings (Tobit Regression Models) Information Friendliness (1) (2) (3) (4) (5) (6) All Public Private All Public Private Maghrebian name −0.183 0.239 −1.103+ −0.0787 0.190 −0.669 (0.309) (0.351) (0.651) (0.275) (0.317) (0.548) Observations 311 199 112 311 199 112 Information Friendliness (1) (2) (3) (4) (5) (6) All Public Private All Public Private Maghrebian name −0.183 0.239 −1.103+ −0.0787 0.190 −0.669 (0.309) (0.351) (0.651) (0.275) (0.317) (0.548) Observations 311 199 112 311 199 112 Note: Robust standard errors in parentheses (clustered by municipality). Coefficients are unstandardized. ***p < .01. **p < .05. +p < .1. View Large Table 6. Coder Ratings (Tobit Regression Models) Information Friendliness (1) (2) (3) (4) (5) (6) All Public Private All Public Private Maghrebian name −0.183 0.239 −1.103+ −0.0787 0.190 −0.669 (0.309) (0.351) (0.651) (0.275) (0.317) (0.548) Observations 311 199 112 311 199 112 Information Friendliness (1) (2) (3) (4) (5) (6) All Public Private All Public Private Maghrebian name −0.183 0.239 −1.103+ −0.0787 0.190 −0.669 (0.309) (0.351) (0.651) (0.275) (0.317) (0.548) Observations 311 199 112 311 199 112 Note: Robust standard errors in parentheses (clustered by municipality). Coefficients are unstandardized. ***p < .01. **p < .05. +p < .1. View Large With regard to the friendliness ratings, no statistically significant difference could be detected in any of the estimated regression models. In terms of effect directions, we found very minor differences in terms of ethnic discrimination for all facilities combines (1.2%), so as for the public facilities (2.7%). Differences in the private group were larger (i.e., 9.6%), but here two factors have to be kept in mind, (1) our relatively small sample size of private nursing homes, and (2) the poor intercoder reliability for this measure.11 Discussion This study examined public–private differences in ethnic discrimination. Using a field experiment, we investigated access to elderly care for applicants with a Maghrebian and Flemish background. We implemented an audit study design by sending requests for information about enrolling in an elderly care home. Facilities could easily put up barriers for applicants by either not responding to requests, or not providing sufficient information. Indeed, both nonresponses as well as responses with poor information increase the learning costs of the applicant (Moynihan, Herd, and Harvey 2015). We hypothesized that the response rate and the quality of information provided would be higher for public providers than for private, for-profit homes, because the profit motive in the private sector was expected to create an adverse incentive structure to cream-skim easier to serve applicants. Ethnic minorities risk being perceived as harder to serve due to their different cultural backgrounds and demands. In the public sector, we expected that greater scope in the application of professional care standards would limit cream-skimming. Discrimination Is Lower Than Expected When assessing the overall effect for public and private facilities combined, our study found no evidence for ethnic discrimination in response rates. This is surprising given the evidence of racial and ethnic discrimination in other fields. For instance, studies of the Belgian labor market have discovered discriminatory practices toward ethnic minorities (Arrijn, Feld, and Nayer 1998; Baert et al. 2015). Baert et al. (2015) report that job seekers with Turkish names had to send out twice as many applications as candidates with Flemish-sounding names to be invited to a job interview. Moreover, a field study undertaken at the University of Leuven found strong evidence of discrimination in the Belgian private housing market (Van den Broeck, Heylen, and Winters 2014); 42% of real estate agents were prepared to discriminate against foreigners, which is a criminal offence under Belgian anti-discrimination laws. Given these local studies and prior audit study evidence, it is remarkable that elderly care homes in Flanders overall do not discriminate in their response rates—although the effort it would take to do so is particularly small (i.e., simply not responding to an email). An explanation for the lower-than-expected levels of ethnic discrimination in responding to information requests may be found in the names that were selected for this study. Names serve as social identifiers (Elchardus and Siongers 2011). Seeing as we wanted to study only ethnic discrimination, we pretested the used names connotations of SES or age. Indeed, some names are associated with higher levels of education and income, whereas others are associated with a working-class background. We selected a Flemish name with an SES association that is as close as possible to the low SES association of Maghrebian names. It may therefore be that past studies have attributed to ethnic discrimination what in reality was SES discrimination. One suggestion for future research is to further tease these potential SES effects out in a multi-factorial design, where SES is experimentally manipulated alongside applicants’ ethnicity. Administrative barriers and discriminatory practices can also be introduced downstream in the application process. The next step after an initial contact email and its response is finding information on how to enroll. In our study, we therefore also examined whether enrollment information was provided in the response. The provision of enrollment information lowers the learning costs for applicants. We found that the likelihood that this information would be provided to the Maghrebian applicant was substantially lower—approximately 8 percentage points—than for the Flemish applicant; however, this difference is not statistically significant. In subsequent stages of the admission process (which we did not study), other opportunities for cream skimming occur. Most elderly care homes suggest that applicants make a personal visit to the facility. While very good reasons for establishing personal contact with prospective residents do exist, site visits create the risk that frontline workers will dissuade applicants from minority groups and/or with a lower SES from further pursuing their application. In summary, although the responses we received to the email query is an encouraging finding, we cannot extrapolate it to the entire admission process. But Public–Private Differences Do Exist While we did not find evidence of discrimination for public and private facilities combined, the analysis of public–private differences suggests that ethnic discrimination does exist in private, for-profit nursing homes. There is suggestive evidence for differences (albeit statistically nonsignificant) in response rates. The extent of the difference, however, is smaller than expected. The relatively small population of private facilities may have provided insufficient statistical power to detect an effect, and hence sample size considerations need to be taken into account when interpreting this finding. However, when public–private differences in providing enrollment information were examined, our study revealed that privately owned facilities are less likely to provide information to an ethnic minority applicant; in this context, Maghrebian name aliases were approximately 20 percentage points less likely to receive enrollment information than Flemish aliases. This finding is not only of a substantive magnitude, but also statistically significant. In contrast, Maghrebian discrimination was almost zero within public facilities.11 In addition, when examining blinded coders’ assessments of the information completeness of received responses, a similar picture emerges. This suggests that public–private differences in ethnic discrimination are a realistic concern. The provision of information concerning how to apply for a place at an elderly care facility is not trivial, because it increases learning costs that lead to increased administrative burden for individuals who need to access care. Overall, our findings demonstrate that the institutional context does seem to have an impact on discriminatory behavior within street-level organizations. Private, for-profit organizations are less likely to provide information (and less comprehensive information) to Maghrebian names, whereas public organizations barely differentiate between names. Our findings also suggest that the causal mechanism for ethnic discrimination is probably more complex than initially assumed in the literature. Discrimination exists in many forms. However, in our study the low-hanging fruit (i.e., a simple nonresponse) is left untouched, whereas the more subtle discrimination form (namely increasing learning costs through the response to the email) is applied. Maynard-Moody and Musheno (2003) provide an explanation for why the most accessible form of discrimination, not responding, is the least strong. We can expect that in our research context the frontline imperative is particularly strong. Welfare workers—such as employees of elderly care homes—have direct and intense contact with the people they serve. These personal contacts inform their socialization processes and reinforce an “other-regarding” professional identity that the decision to not discriminate based on ethnicity conforms well to. Nonetheless, adverse institutional incentives seem to not be entirely neutralized. Private, for-profit facilities are significantly less likely to provide enrollment information when responding to Maghrebian aliases, but no such differences exist within public facilities. This indicates that profit incentives may indeed be crowding professional care standards out to some extent. While frontline professionalism seems to prevent the manifest act of ignoring a minority request, more subtle mechanisms of increasing learning costs are at play. Future research should attempt to disentangle this bundle of institutional influences and professional resistance. Appendix Box 1. Email Sent to Flemish Elderly Care Facilities From: [Randomized Treatment Name] To: [Email Address of Primary Contact at Elderly Care Facility] Subject: Admission information Hello, I am contacting you because I am looking for a place in a rest home for my father. We are interested in your facility. Do you have a place available at this moment? And how can I subscribe my father for this? I also heard there is a waiting list. Do you have one and how long is it? Thanks, [Randomized Treatment Name] Box 2. Email Sent to Elderly Care Facilities in Flemish From: [Randomized Treatment Name] To: [Email Address of Primary Contact at Elderly Care Facility] Subject: Informatie opname Hallo, Ik contacteer jullie omdat ik op zoek ben naar een rusthuis voor mijn vader. Wij zijn geïnteresseerd in jullie verblijf. Hebben jullie op dit moment een plaats beschikbaar? En hoe kan ik mijn vader daarvoor inschrijven? Ik heb ook gehoord dat er een wachtlijst is. Hebben jullie er één, en hoe lang is die? Bedankt, [Randomized Treatment Name] Table A1. Means of Pretest (1) Education Level (2) Age (3) Ethnicitya (4) Income Level Sven Janssens 2.526 3.007 98.3 3.137 (0.563) (0.815) (0.3; 1.4) (0.547) Joris Janssens 2.587 3.259 96.7 3.206 (0.525) (0.959) (0.9; 2.4) (0.464) Kenny Janssens 2.079 2.303 98.6 2.742 (0.664) (0.754) (0.3; 1) (0.672) Marc Janssens 2.483 4.157 98.4 3.294 (0.595) (0.895) (0.3; 1.3) (0.610) Thomas Janssens 2.682 2.794 98.6 3.244 (0.497) (0.855) (0.3; 1) (0.527) Sven Maes 2.610 2.927 99.1 3.136 (0.533) (0.750) (0.0; 0.9) (0.554) Joris Maes 2.602 3.272 98.1 3.237 (0.536) (1.084) (0.4; 1.5) (0.532) Kenny Maes 2.230 2.345 97.7 2.723 (0.583) (0.783) (0.7; 1.6) (0.612) Marc Maes 2.478 4.083 98.7 3.184 (0.577) (0.863) (0.0; 1.3) (0.568) Thomas Maes 2.673 2.768 97.6 3.280 (0.541) (0.832) (0.3; 2.1) (0.543) Mohamed El Makrini 2.136 2.803 74.5 2.863 (0.657) (0.500) (16.3; 9.1) (0.568) Ahmed El Makrini 2.167 2.903 71.6 2.902 (0.626) (0.917) (16.2; 11.6) (0.696) Rachid El Makrini 2.266 2.658 73.2 2.875 (0.645) (0.834) (14.9; 12.0) 1(0.534) Hassan El Makrini 2.272 2.804 71.1 2.919 (0.658) (0.924) (15.8; 13.2) (0.553) Youssef El Makrini 2.199 2.678 67.3 2.963 (0.627) (0.899) (21.8; 10.9) (0.635) Mohamed Bouzian 2.194 2.952 70.2 2.941 (0.670) (0.949) (17.8; 12.0) (0.637) Ahmed Bouzian 2.174 2.896 57.6 2.885 (0.634) (0.918) (19.8; 22.7) (0.564) Rachid Bouzian 2.292 2.729 60.2 2.952 (0.634) (0.876) (21.8; 18.0) (0.611) Hassan Bouzian 2.138 2.818 62.0 2.876 (0.630) (0.890) (16.3; 21.6) (0.554) Youssef Bouzian 2.161 2.709 65.4 2.858 (0.665) (0.880) (12.5; 22.1) (0.522) (1) Education Level (2) Age (3) Ethnicitya (4) Income Level Sven Janssens 2.526 3.007 98.3 3.137 (0.563) (0.815) (0.3; 1.4) (0.547) Joris Janssens 2.587 3.259 96.7 3.206 (0.525) (0.959) (0.9; 2.4) (0.464) Kenny Janssens 2.079 2.303 98.6 2.742 (0.664) (0.754) (0.3; 1) (0.672) Marc Janssens 2.483 4.157 98.4 3.294 (0.595) (0.895) (0.3; 1.3) (0.610) Thomas Janssens 2.682 2.794 98.6 3.244 (0.497) (0.855) (0.3; 1) (0.527) Sven Maes 2.610 2.927 99.1 3.136 (0.533) (0.750) (0.0; 0.9) (0.554) Joris Maes 2.602 3.272 98.1 3.237 (0.536) (1.084) (0.4; 1.5) (0.532) Kenny Maes 2.230 2.345 97.7 2.723 (0.583) (0.783) (0.7; 1.6) (0.612) Marc Maes 2.478 4.083 98.7 3.184 (0.577) (0.863) (0.0; 1.3) (0.568) Thomas Maes 2.673 2.768 97.6 3.280 (0.541) (0.832) (0.3; 2.1) (0.543) Mohamed El Makrini 2.136 2.803 74.5 2.863 (0.657) (0.500) (16.3; 9.1) (0.568) Ahmed El Makrini 2.167 2.903 71.6 2.902 (0.626) (0.917) (16.2; 11.6) (0.696) Rachid El Makrini 2.266 2.658 73.2 2.875 (0.645) (0.834) (14.9; 12.0) 1(0.534) Hassan El Makrini 2.272 2.804 71.1 2.919 (0.658) (0.924) (15.8; 13.2) (0.553) Youssef El Makrini 2.199 2.678 67.3 2.963 (0.627) (0.899) (21.8; 10.9) (0.635) Mohamed Bouzian 2.194 2.952 70.2 2.941 (0.670) (0.949) (17.8; 12.0) (0.637) Ahmed Bouzian 2.174 2.896 57.6 2.885 (0.634) (0.918) (19.8; 22.7) (0.564) Rachid Bouzian 2.292 2.729 60.2 2.952 (0.634) (0.876) (21.8; 18.0) (0.611) Hassan Bouzian 2.138 2.818 62.0 2.876 (0.630) (0.890) (16.3; 21.6) (0.554) Youssef Bouzian 2.161 2.709 65.4 2.858 (0.665) (0.880) (12.5; 22.1) (0.522) Note: Standard deviations in parentheses. aPercentage that predicted the intended ethnicity (Belgian/Moroccan). In parentheses, the percentages of the other nationality guess (i.e., Belgian/Moroccan, or Other). Missing percentage points represent missing values. View Large Table A1. Means of Pretest (1) Education Level (2) Age (3) Ethnicitya (4) Income Level Sven Janssens 2.526 3.007 98.3 3.137 (0.563) (0.815) (0.3; 1.4) (0.547) Joris Janssens 2.587 3.259 96.7 3.206 (0.525) (0.959) (0.9; 2.4) (0.464) Kenny Janssens 2.079 2.303 98.6 2.742 (0.664) (0.754) (0.3; 1) (0.672) Marc Janssens 2.483 4.157 98.4 3.294 (0.595) (0.895) (0.3; 1.3) (0.610) Thomas Janssens 2.682 2.794 98.6 3.244 (0.497) (0.855) (0.3; 1) (0.527) Sven Maes 2.610 2.927 99.1 3.136 (0.533) (0.750) (0.0; 0.9) (0.554) Joris Maes 2.602 3.272 98.1 3.237 (0.536) (1.084) (0.4; 1.5) (0.532) Kenny Maes 2.230 2.345 97.7 2.723 (0.583) (0.783) (0.7; 1.6) (0.612) Marc Maes 2.478 4.083 98.7 3.184 (0.577) (0.863) (0.0; 1.3) (0.568) Thomas Maes 2.673 2.768 97.6 3.280 (0.541) (0.832) (0.3; 2.1) (0.543) Mohamed El Makrini 2.136 2.803 74.5 2.863 (0.657) (0.500) (16.3; 9.1) (0.568) Ahmed El Makrini 2.167 2.903 71.6 2.902 (0.626) (0.917) (16.2; 11.6) (0.696) Rachid El Makrini 2.266 2.658 73.2 2.875 (0.645) (0.834) (14.9; 12.0) 1(0.534) Hassan El Makrini 2.272 2.804 71.1 2.919 (0.658) (0.924) (15.8; 13.2) (0.553) Youssef El Makrini 2.199 2.678 67.3 2.963 (0.627) (0.899) (21.8; 10.9) (0.635) Mohamed Bouzian 2.194 2.952 70.2 2.941 (0.670) (0.949) (17.8; 12.0) (0.637) Ahmed Bouzian 2.174 2.896 57.6 2.885 (0.634) (0.918) (19.8; 22.7) (0.564) Rachid Bouzian 2.292 2.729 60.2 2.952 (0.634) (0.876) (21.8; 18.0) (0.611) Hassan Bouzian 2.138 2.818 62.0 2.876 (0.630) (0.890) (16.3; 21.6) (0.554) Youssef Bouzian 2.161 2.709 65.4 2.858 (0.665) (0.880) (12.5; 22.1) (0.522) (1) Education Level (2) Age (3) Ethnicitya (4) Income Level Sven Janssens 2.526 3.007 98.3 3.137 (0.563) (0.815) (0.3; 1.4) (0.547) Joris Janssens 2.587 3.259 96.7 3.206 (0.525) (0.959) (0.9; 2.4) (0.464) Kenny Janssens 2.079 2.303 98.6 2.742 (0.664) (0.754) (0.3; 1) (0.672) Marc Janssens 2.483 4.157 98.4 3.294 (0.595) (0.895) (0.3; 1.3) (0.610) Thomas Janssens 2.682 2.794 98.6 3.244 (0.497) (0.855) (0.3; 1) (0.527) Sven Maes 2.610 2.927 99.1 3.136 (0.533) (0.750) (0.0; 0.9) (0.554) Joris Maes 2.602 3.272 98.1 3.237 (0.536) (1.084) (0.4; 1.5) (0.532) Kenny Maes 2.230 2.345 97.7 2.723 (0.583) (0.783) (0.7; 1.6) (0.612) Marc Maes 2.478 4.083 98.7 3.184 (0.577) (0.863) (0.0; 1.3) (0.568) Thomas Maes 2.673 2.768 97.6 3.280 (0.541) (0.832) (0.3; 2.1) (0.543) Mohamed El Makrini 2.136 2.803 74.5 2.863 (0.657) (0.500) (16.3; 9.1) (0.568) Ahmed El Makrini 2.167 2.903 71.6 2.902 (0.626) (0.917) (16.2; 11.6) (0.696) Rachid El Makrini 2.266 2.658 73.2 2.875 (0.645) (0.834) (14.9; 12.0) 1(0.534) Hassan El Makrini 2.272 2.804 71.1 2.919 (0.658) (0.924) (15.8; 13.2) (0.553) Youssef El Makrini 2.199 2.678 67.3 2.963 (0.627) (0.899) (21.8; 10.9) (0.635) Mohamed Bouzian 2.194 2.952 70.2 2.941 (0.670) (0.949) (17.8; 12.0) (0.637) Ahmed Bouzian 2.174 2.896 57.6 2.885 (0.634) (0.918) (19.8; 22.7) (0.564) Rachid Bouzian 2.292 2.729 60.2 2.952 (0.634) (0.876) (21.8; 18.0) (0.611) Hassan Bouzian 2.138 2.818 62.0 2.876 (0.630) (0.890) (16.3; 21.6) (0.554) Youssef Bouzian 2.161 2.709 65.4 2.858 (0.665) (0.880) (12.5; 22.1) (0.522) Note: Standard deviations in parentheses. aPercentage that predicted the intended ethnicity (Belgian/Moroccan). In parentheses, the percentages of the other nationality guess (i.e., Belgian/Moroccan, or Other). Missing percentage points represent missing values. View Large Table A2. Categorization of Addressed Facilities According to their Legal Status (1) (2) (3) (4) (5) Abbreviation of Belgian Legal Status Belgian Legal Status Translation of Belgian Legal Status Ownership Amount of Facilities APB Autonoom Provinciaal Bedrijf Autonomous Provincial Company Public 1 AV Autonome Vereniging Autonomous Association Public 4 EVA Extern Verzelfstandigd Agentschap External Independent Agency Public 4 OCMW Openbaar Centrum voor Maatschappelijk Welzijn Public Welfare Agency Public 205 BV Besloten Vennootschap Private Company Private 1 BVBA Besloten Vennootschap met Beperkte Aansprakelijkheid Private Company with Limited Liability Private 45 CV Commanditaire Vennootschap Limited Partnership Private 1 CVBA CV met Beperkte Aansprakelijkheid Limited Partnership with Limited Liability Private 2 GCV Gewone Commanditaire Vennootschap Normal Limited Partnership Private 3 NP Natuurlijke Persoon Natural Person Private 5 NV Naamloze Vennootschap Limited Liability Company Private 60 Stichting Stichting Foundation Private 1 VZW Vereniging Zonder Winstoogmerk Association without Lucrative Purpose Nonprofit 387 (1) (2) (3) (4) (5) Abbreviation of Belgian Legal Status Belgian Legal Status Translation of Belgian Legal Status Ownership Amount of Facilities APB Autonoom Provinciaal Bedrijf Autonomous Provincial Company Public 1 AV Autonome Vereniging Autonomous Association Public 4 EVA Extern Verzelfstandigd Agentschap External Independent Agency Public 4 OCMW Openbaar Centrum voor Maatschappelijk Welzijn Public Welfare Agency Public 205 BV Besloten Vennootschap Private Company Private 1 BVBA Besloten Vennootschap met Beperkte Aansprakelijkheid Private Company with Limited Liability Private 45 CV Commanditaire Vennootschap Limited Partnership Private 1 CVBA CV met Beperkte Aansprakelijkheid Limited Partnership with Limited Liability Private 2 GCV Gewone Commanditaire Vennootschap Normal Limited Partnership Private 3 NP Natuurlijke Persoon Natural Person Private 5 NV Naamloze Vennootschap Limited Liability Company Private 60 Stichting Stichting Foundation Private 1 VZW Vereniging Zonder Winstoogmerk Association without Lucrative Purpose Nonprofit 387 View Large Table A2. Categorization of Addressed Facilities According to their Legal Status (1) (2) (3) (4) (5) Abbreviation of Belgian Legal Status Belgian Legal Status Translation of Belgian Legal Status Ownership Amount of Facilities APB Autonoom Provinciaal Bedrijf Autonomous Provincial Company Public 1 AV Autonome Vereniging Autonomous Association Public 4 EVA Extern Verzelfstandigd Agentschap External Independent Agency Public 4 OCMW Openbaar Centrum voor Maatschappelijk Welzijn Public Welfare Agency Public 205 BV Besloten Vennootschap Private Company Private 1 BVBA Besloten Vennootschap met Beperkte Aansprakelijkheid Private Company with Limited Liability Private 45 CV Commanditaire Vennootschap Limited Partnership Private 1 CVBA CV met Beperkte Aansprakelijkheid Limited Partnership with Limited Liability Private 2 GCV Gewone Commanditaire Vennootschap Normal Limited Partnership Private 3 NP Natuurlijke Persoon Natural Person Private 5 NV Naamloze Vennootschap Limited Liability Company Private 60 Stichting Stichting Foundation Private 1 VZW Vereniging Zonder Winstoogmerk Association without Lucrative Purpose Nonprofit 387 (1) (2) (3) (4) (5) Abbreviation of Belgian Legal Status Belgian Legal Status Translation of Belgian Legal Status Ownership Amount of Facilities APB Autonoom Provinciaal Bedrijf Autonomous Provincial Company Public 1 AV Autonome Vereniging Autonomous Association Public 4 EVA Extern Verzelfstandigd Agentschap External Independent Agency Public 4 OCMW Openbaar Centrum voor Maatschappelijk Welzijn Public Welfare Agency Public 205 BV Besloten Vennootschap Private Company Private 1 BVBA Besloten Vennootschap met Beperkte Aansprakelijkheid Private Company with Limited Liability Private 45 CV Commanditaire Vennootschap Limited Partnership Private 1 CVBA CV met Beperkte Aansprakelijkheid Limited Partnership with Limited Liability Private 2 GCV Gewone Commanditaire Vennootschap Normal Limited Partnership Private 3 NP Natuurlijke Persoon Natural Person Private 5 NV Naamloze Vennootschap Limited Liability Company Private 60 Stichting Stichting Foundation Private 1 VZW Vereniging Zonder Winstoogmerk Association without Lucrative Purpose Nonprofit 387 View Large References Abbott , Andrew . 1988 . The system of professions: An essay on the division of expert labor . Chicago : University of Chicago Press . Adman , Per , and Hanna Jansson . 2016 . A field experiment on ethnic discrimination among local Swedish public officials . Local Government Studies 43 : 44 – 63 . Google Scholar CrossRef Search ADS Agentschap Zorg and Gezondheid . 2017 . Adressen woonzorgcentra en rust -en verzorgingstehuizen . https://www.zorg-en-gezondheid.be/adressen-van-woonzorgcentra-en-rust-en-verzorgingstehuizen (accessed March 30, 2017 ). Ahmed , Ali M. , and Mats Hammarstedt . 2008 . Discrimination in the rental housing market. A field experiment on the internet . Journal of Urban Economics 64 : 362 – 72 . Google Scholar CrossRef Search ADS Altonji , Joseph G. , and Charles R. Pierret . 2001 . Employer learning and statistical discrimination . The Quarterly Journal of Economics 116 : 313 – 50 . Google Scholar CrossRef Search ADS Amirkhanyan , Anna , Hyun Joon Kim , and Kristina Lambright . 2008 . Does the public sector outperform the nonprofit and for-profit sectors? Evidence from a national panel study on nursing home quality and access . Journal of Policy Analysis and Management 27 : 326 – 53 . Google Scholar CrossRef Search ADS PubMed Arrijn , Peter , Serge Feld , and André Nayer . 1998 . Discrimination in access to employment on grounds of foreign origin: The case of Belgium . Geneva : International Labour Office, Conditions of Work Branch . Arrow , Kenneth J . 1973 . The theory of discrimination . In Orley Ashenfelter, Albert Rees, eds. Discrimination labor markets , 3 – 33 . Princeton : Princeton Univ. Press . Google Scholar CrossRef Search ADS Ayres , Ian , and Peter Siegelman . 1995 . Race and gender in bargaining for a new car . American Economic Review 85 : 304 – 21 . Baert , Stijn , Bart Cockx , Niels Gheyle , and Cora Vandamme . 2015 . Is there less discrimination in occupations where recruitment is difficult ? ILR Review 68 : 467 – 500 . Google Scholar CrossRef Search ADS Becker , Gary S . 1971 . The economics of discrimination . Chicago : The University of Chicago Press . Google Scholar CrossRef Search ADS Bertrand , Marianne , Dolly Chugh , and Sendhil Mullainathan . 2005 . Implicit Discrimination. American Economic Review 95 : 94 – 8 . CrossRef Search ADS Bertrand , Marianne , and Sendhil Mullainathan . 2004 . Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination . American Economic Review 94 : 991 – 1013 . Google Scholar CrossRef Search ADS Bevan , Gwyn , and Christopher Hood . 2006 . What’s measured is what matters: Targets and gaming in the British health care sector . Public Administration 84 : 517 – 38 . Google Scholar CrossRef Search ADS Bobo , Lawrence . 1983 . Whites’ opposition to busing: Symbolic racism or realistic group conflict ? Journal of Personality and Social Psychology 45 : 1196 – 210 . Google Scholar CrossRef Search ADS Bohte , John , and Kenneth J. Meier . 2000 . Goal displacement: Assessing the motivation for organisational cheating . Public Administration Review 60 : 173 – 82 . Google Scholar CrossRef Search ADS Bozeman , Barry , and Stuart Bretschneider . 1994 . The “publicness puzzle” in organization theory: A test of alternative explanations of differences between public and private organizations . Journal of Public Administration Research and Theory 4 : 197 – 23 . Broockman , David E . 2013 . Black politicians are more intrinsically motivated to advance blacks’ interests: A field experiment manipulating political incentives . American Journal of Political Science 57 : 521 – 36 . Google Scholar CrossRef Search ADS Brodkin , Evelyn . 1997 . Inside the welfare contract: Discretion and accountability in state welfare administration . Social Service Review 71 : 1 – 33 . Google Scholar CrossRef Search ADS Brodkin , Evelyn , and Michael Lipsky . 1983 . Quality control in AFDC as an administrative strategy . Social Service Review 57 : 1 – 34 . Google Scholar CrossRef Search ADS Butler , Daniel M. , and David E. Broockman . 2011 . Do politicians racially discriminate against constituents? A field experiment on state legislators . American Journal of Political Science 55 : 463 – 77 . Google Scholar CrossRef Search ADS Chugh , Dolly . 2004 . Societal and managerial implications of implicit social cognition: Why milliseconds matter . Social Justice Research 17 : 203 – 22 . Google Scholar CrossRef Search ADS Clifton , Judith , Daniel Díaz-Fuentes , Marcos Fernández-Gutiérrez , and Julio Revuelta . 2011 . Is market-oriented reform producing a ‘two-track’ Europe? Evidence from electricity and telecommunications . Annals of Public and Cooperative Economics 82 : 495 – 13 . Google Scholar CrossRef Search ADS Clifton , Judith , Daniel Diaz Fuentes , Marcos Fernandez-Guttierrez , Oliver James , Sebastian Jilke , and Steven Van de Walle . ( 2012 ): Regulatory reform for services of general interest and trends in citizen satisfaction . COCOPS Project Report. Rotterdam : Erasmus Univ. Rotterdam . Costa , Mia . 2017 . How responsive are political elites? A meta-analysis of experiments on public officials . Journal of Experimental Political Science 4 : 241 – 54 . Google Scholar CrossRef Search ADS Dabney , Dean A. , Laura Dugan , Volkan Topalli , and Richard C. Hollinger . 2006 . The impact of implicit stereotyping on offender profiling. Unexpected results from an observational study of shoplifting . Criminal Justice and Behavior 33 : 646 – 74 . Google Scholar CrossRef Search ADS Dan , Sorin , and Rhys Andrews . 2015 . Market-type mechanisms and public service equity: A review of experiences in European public services . Public Organization Review 34 : 543 – 59 . Dias , Janice J. , and Steven Maynard-Moody . 2007 . For-profit welfare: Contracts, conflicts, and the performance paradox . Journal of Public Administration Research and Theory 17 : 189 – 11 . Google Scholar CrossRef Search ADS Einstein , Katherine L. , and David M. Glick . 2016 . Does race affect access to government services? An experiment exploring street-level bureaucrats and access to public housing . American Journal of Political Science 61 : 100 – 16 . Google Scholar CrossRef Search ADS Elchardus , Mark , and Jessy Siongers . 2011 . First names as collective identifiers: An empirical analysis of the social meanings of first names . Cultural Sociology 5 : 403 – 22 . Google Scholar CrossRef Search ADS Ellis , Randall P . 1998 . Creaming, skimping and dumping: Provider competition on the intensive and extensive margins . Journal of Health Economics 17 : 537 – 55 . Google Scholar CrossRef Search ADS PubMed Emker , Johannes , and Anselm Rink . 2017 . Multiple dimensions of bureaucratic discrimination: Evidence from German welfare officers . American Journal of Political Science 61 : 786 – 803 . Google Scholar CrossRef Search ADS Epp , Charles R. , Steven Maynard-Moody , and Donald P. Haider-Markel . 2014 . How police stops define race and citizenships . Chicago : The Univ. of Chicago Press . Google Scholar CrossRef Search ADS Forscher , Patrick S. , Calvin K. Lai , Jordan R. Axt , Charles R. Ebersole , Michelle Herman , Patricia G. Devine , and Brian A. Nosek . 2017 . A meta-analysis of change in implicit bias . Retrieved from Psyarxiv.com/dv8tu . Frederickson , George H . 2010 . Social equity and public administration: Origins, developments, and applications . New York : M.E. Sharpe . Gallup . 2014 . Gallup review: Black and white differences in views on race . http://www.gallup.com/poll/180107/gallup-review-black-white-differences-views-race.aspx (accessed March 30, 2017 ). Giulietti , Corrado , Mirco Tonin , and Michael Vlassopoulos . 2017 . Racial discrimination in local public services: A field experiment in the US . Journal of the European Economic Association, online first. doi: 10.1093/jeea/jvx045. Gooden , Susan T . 2015 . PAR’s social equity footprint . Public Administration Review 75 : 372 – 81 . Google Scholar CrossRef Search ADS Greenwald , Anthony G. , and Linda Hamilton Krieger . 2006 . Implicit bias: Scientific foundations . California Law Review 94 : 945 – 67 . Google Scholar CrossRef Search ADS Greenwald , Anthony G. , Debbie E. McGhee , and Jordan L. K. Schwartz . 1998 . Measuring individual differences in implicit cognition: The implicit association test . Journal of Personality and Social Psychology 74 : 1464 – 80 . Google Scholar CrossRef Search ADS PubMed Grohs , Stephan , Christin Adam , and Christoph Knill . 2015 . Are some citizens more equal than others? Evidence from a field experiment . Public Administration Review 76 : 155 – 64 . Google Scholar CrossRef Search ADS Guryan , Jonathan , and Kerwin Kofi Charles . 2013 . Taste-based or statistical discrimination: The economics of discrimination returns to its roots . The Economic Journal 123 : F417 – 32 . Google Scholar CrossRef Search ADS Harris , Ricci , Martin Tobias , Mona Jeffreys , Kiri Waldegrave , Saffron Karlsen , and James Nazroo . 2006 . Racism and health: The relationship between experience of racial discrimination and health in New Zealand . Social Science & Medicine 63 : 1428 – 41 . Google Scholar CrossRef Search ADS Heinrich , Carolyn J . 1999 . Do government bureaucrats make effective use of performance management information ? Journal of Public Administration Research and Theory 9 : 363 – 93 . Google Scholar CrossRef Search ADS ———. 2016 . The bite of administrative burden: A theoretical and empirical investigation . Journal of Public Administration Research and Theory 3 : 403 – 20 . Heinrich , Carolyn J. , and Gerald Marschke . 2010 . Incentives and their dynamics in public sector performance management systems . Journal of Policy Analysis and Management 29 : 183 – 208 . Google Scholar CrossRef Search ADS Galster , George , Douglas Wissoker , and Wendy Zimmermann . 1998 . Testing for discrimination in home insurance . Washington, DC : Urban Institute Press . Hood , Christopher . 2006 . Gaming in target world: The targets approach to managing British public services . Public Administration Review 66 : 515 – 21 . Google Scholar CrossRef Search ADS Jackson , Jay W . 1993 . Realistic group conflict theory: A review and evaluation of the theoretical and empirical literature . The Psychological Record 43 : 395 – 13 . Jilke , Sebastian . 2015 . Essays on the microfoundations of competition and choice in public service delivery . PhD diss. Rotterdam: Erasmus University Rotterdam. Kaas , Leo , and Christian Manger . 2011 . Ethnic discrimination in Germany’s labour market: A field experiment . German Economic Review 13 : 1 – 20 . Google Scholar CrossRef Search ADS Keiser , Lael R. , and Joe Soss . 1998 . With good cause: Bureaucratic discretion and the politics of child support enforcement . American Journal of Political Science 42 : 1133 – 56 . Google Scholar CrossRef Search ADS Kerpershoek , Emiel , Martijn Groenleer , and Hans de Bruijn . 2016 . Unintended responses to performance management in Dutch hospital care: Bringing together the managerial and professional perspectives . Public Management Review 18 : 417 – 36 . Google Scholar CrossRef Search ADS Kjeldsen , Anne Mette , and Christian Botcher Jacobsen . 2013 . Public service motivation and employment sector: Attraction or socialization ? Journal of Public Administration Research and Theory 23 : 899 – 26 . Google Scholar CrossRef Search ADS Koning , Pierre , and Carolyn J. Heinrich . 2013 . Cream-skimming, parking and other intended and unintended effects of high-powered, performance-based contracts . Journal of Policy Analysis and Management 32 : 461 – 83 . Google Scholar CrossRef Search ADS Krippendorff , Klaus . 2013 . Content analysis: An introduction to its methodology . Thousand Oaks : SAGE . Kugelmass , Heather . 2016 . “Sorry, I’m not accepting new patients”: An audit study of access to mental health care . Journal of Health and Social Behavior 57 : 168 – 83 . Google Scholar CrossRef Search ADS PubMed Lacireno-Paquet , Natalie , Thomas T. Holyoke , Michele Moser , and Jeffrey R. Henig . 2002 . Creaming versus cropping: Charter school enrollment practices in response to market incentives . Educational Evaluation and Policy Analysis 24 : 145 – 58 . Google Scholar CrossRef Search ADS Le Grand , Julian , and Will Bartlett . 1993 . Quasi-markets and social policy . London : Macmillan . Google Scholar CrossRef Search ADS Lipsky , Michael . 1980 . Street-level bureaucracy: Dilemmas of the individual in public services . New York : Russell Sage Foundation . ———. 1984 . Bureaucratic disentitlement in social welfare programs . Social Service Review 58 : 3 – 27 . CrossRef Search ADS Maynard-Moody , Steven , and Michael Musheno . 2000 . State agent or citizen agent: Two narratives of discretion . Journal of Public Administration Research & Theory 10 : 329 . Google Scholar CrossRef Search ADS ———. 2003 . Cops, Teachers, Counselors: Stories from the front lines of public service . Michigan : Univ. of Michigan Press . Meier , Kenneth J. , and Laurence J. O’Toole . 2011 . Comparing public and private management: Theoretical expectations . Journal of Public Administration Research and Theory 21 ( Suppl 3 ): i283 – 99 . Google Scholar CrossRef Search ADS Moynihan , Donald P. , Pamela Herd , and Hope Harvey . 2015 . Administrative burden: Learning, psychological, and compliance costs in citizen-state interactions . Journal of Public Administration Research and Theory 25 : 43 – 69 . Google Scholar CrossRef Search ADS Moynihan , Donald P. , Pamela Herd , and Elizabeth Ribgy . 2013 . Policymaking by other means. Do states use administrative barriers to limit access to Medicaid ? Administration & Society 48 : 497 – 24 . Google Scholar CrossRef Search ADS National Research Council . 2004 . Measuring racial discrimination . Washington, DC : The National Academies Press . Oswald , Frederick L. , Gregory Mitchell , Hart Blanton , and Philip E. Tetlock . 2013 . Predicting ethnic and racial discrimination: A meta-analysis of IAT criterion studies . Journal of Personality and Social Psychology 105 : 171 – 92 . Google Scholar CrossRef Search ADS PubMed Pacolet , Jozef , and Annelies De Coninck . 2015 . Financiering van de residentiële ouderenzorg: Het perspectief van de voorzieningen . KU Leuven met steun van de Vlaamse Overheid . Pager , Devah . 2007 . The use of field experiments for studies of employment discrimination: Contributions, critiques, and directions for the future . Annals of the American Academy of Political and Social Sciences 609 : 104 – 33 . Google Scholar CrossRef Search ADS Pager , Devah , Bruce Western , and Bart Bonikowski . 2009 . Discrimination in a low-wage labor market: A field experiment . American Sociological Review 74 : 777 – 99 . Google Scholar CrossRef Search ADS PubMed Perry , James L. , Annie Hondeghem , and Lois Recascino Wise . 2010 . Revisiting the motivational bases of public service: Twenty years of research and an agenda for the future . Public Administration Review 70 : 681 – 90 . Google Scholar CrossRef Search ADS Phelps , Edmund S . 1972 . The statistical theory of racism and sexism . American Economic Review 62 : 659 – 61 . Pollitt , Christopher . 2013 . The logics of performance management . Evaluation 19 : 346 – 63 . Google Scholar CrossRef Search ADS Quillian , Lincoln , Devah Pager , Ole Hexel , and Arnfinn Midboen . 2017 . Meta-analysis of field experiments shows no change in racial discrimination in hiring over time . Proceedings of the National Academy of Sciences of the United States of America 114 : 10870 – 75 . Google Scholar CrossRef Search ADS PubMed Radin , Beryl A . 2006 . Challenging the performance movement: Accountability, complexity, and democratic values . Washington, DC : Georgetown Univ. Press . Rainey , Hal G. , and Barry Bozeman . 2000 . Comparing public and private organizations: Empirical research and the power of the a priori . Journal of Public Administration Research and Theory 10 : 447 – 70 . Google Scholar CrossRef Search ADS Ridley , Stanley E. , James A. Bayton , and Janice H. Quttz . 1989 . Taxi services in the District of Columbia: Is it influenced by patron’s race and destination . Washington, DC : Washington Lawyers’ Comm. Civil Rights Law . Savas , Emanuel S . 2000 . Privatization and public–private partnerships . New York : Seven Bridges Press . Schneider , Mark , Paul Teske , and Melissa Marschall . 2002 . Choosing schools: Consumer choice and the quality of American schools . Princeton : Princeton Univ. Press . Schulman , Kevin A. , A. Berlin Jesse , William Harless , Jon F. Kerner , Shyrl Sistrunk , Bernard J. Gersh , Ross Dubé , et al. 1999 . The effect of race and sex on physicians’ recommendations for cardiac catheterization . The New England Journal of Medicine 340 : 618 – 26 . Google Scholar CrossRef Search ADS PubMed Scott , Patrick G . 1997 . Assessing determinants of bureaucratic discretion: An experiment in street-level decision making . Journal of Public Administration Research and Theory 35 : 35 – 57 . Google Scholar CrossRef Search ADS Socialistische Mutualiteiten. 2016. Rusthuisbarometer: Analyse bewonersfacturen in woonzorgcentra. Brussels: Socialistische Mutualiteiten. Soss , Joe , Richard C. Fording and Sanford F. Schram . 2011 . The organization of discipline: From performance management to perversity and punishment . Journal of Public Administration Research and Theory 21 ( Suppl 2 ): i203 – 32 . Google Scholar CrossRef Search ADS Statistics Belgium . 2017 . Meest voorkomende familienamen in 2016—België en Gewesten (Top 10) . http://statbel.fgov.be/nl/statistieken/cijfers/bevolking/namen/fam/ (accessed March 30, 2017 ). Tummers , Lars , Victor Bekkers , Evelien Vink , and Michael Musheno . 2015 . Coping during public service delivery: A conceptualization and systematic review of the literature . Journal of Public Administration Review and Theory 25 : 1099 – 126 . Google Scholar CrossRef Search ADS Turner , Margery A. , and Stephen L. Ross . 2003 . Discrimination in metropolitan housing markets phase II: Asians and Pacific islanders . Washington, DC : The Urban Institute . Turner , Margery A. , Stephen L. Ross , George C. Galster , and John Yinger . 2002 . Discrimination in metropolitan housing markets: National results from phase I HDS 2000 . Washington, DC : The Urban Institute . Van den Bosch , Karel , Peter Willemé , Joanna Geerts , Jef Breda , Stéphanie Peeters , Stefaan Van De Sande , France Vrijens , Carine Van de Voorde , and Sabine Stordeur . ( 2011 ). Toekomstige behoefte aan residentiële ouderenzorg in België: Projecties 2011–2025 . Federaal Kenniscentrum voor de Gezondheidszorg . https://kce.fgov.be/sites/default/files/page_documents/KCE_167A_residentiele_zorg_in_Belgie.pdf (accessed March 30, 2017 ). Van den Broeck , Katleen , Kristof Heylen , and Sien Winters . 2014 . Discriminatie en Selectie op de Private Huurmarkt in België . Brussels: Hoger Instituut voor de Arbeid . Van den Broucke Sarah , Jo Noppe , Karen Stuyck , Philippe Buysschaert , Gerlinde Doyen , and Johan Wets . 2016 . Vlaamse migratie en integratiemonitor 2015 . Brussel : Vlaamse Overheid . Weinick , Robin M. , Samuel H. Zuvekas , and Joel W. Cohen . 2000 . Racial and ethnic differences in access to and use of health care services . Medical Care Research and Review 57 ( Suppl 1 ): 36 – 54 . Google Scholar CrossRef Search ADS PubMed Weisbrod , Burton A . 1988 . The nonprofit economy . Cambridge : Harvard Univ. Press . White , Ariel L. , Noah L. Nathan , and Julie K. Faller . 2015 . What do I need to vote? Bureaucratic discretion and discrimination of local election officials . American Political Science Review 109 : 129 – 40 . Google Scholar CrossRef Search ADS Willemé , Peter , Joanna Geerts , Bea Cantillon , and Ninke Mussche . 2012 . Long-term care financing in Belgium . In Financing long-term care in Europe , 300 – 21 . UK : Palgrave Macmillan . Google Scholar CrossRef Search ADS Yinger , John . 1995 . Closed doors, opportunities lost: The continuing costs of housing discrimination . New York : Russell Sage Foundation . Footnotes 1 The determination of the level of publicness of a service provider is not straightforward. In an important contribution to the field, Bozeman and Bretschneider (1994) distinguished between a core approach based on legal ownership and a dimensional approach based on criteria related to public funding, public value, and public communication networks. Bozeman and Bretschneider (1994) argue that the core and dimensional approaches are both useful, depending on the research design. We look only at ownership in our study given that the nondistribution constraint of public facilities creates an incentive structure that is fundamentally different in comparison to private facilities. 2 A fourth type of discrimination comes from realistic group conflict theory (Bobo 1983; Jackson 1993). It suggests that the discrimination of outgroup members stems mainly from competition over limited resources. Here, it is assumed that Belgian frontline workers (as ingroup members) would try to disproportionally distribute elderly care places to fellow Belgians and not to ethnic minorities (i.e., outgroup members), given the perceived competition with minorities over a limited number of places. However, this mechanism seems rather unrealistic given the low societal valence of elderly care places, which are in principle available to everybody. 3 Nonprofits are more profitable because they started to operate elderly care facilities decades before the private facilities. As a result, many buildings are written off in their accounts. Consequently, lower capital duties need to be paid. 4 The nonprofit sector is not included in this study, because the incentive structure of nonprofits is less straightforward than in the public and private sectors. The legal person of nonprofits prohibits personal gain. Profits cannot be distributed to shareholders. However, profits can be retained within organizations for future investment. Moreover, private equity providers increasingly operate a nonprofit as a shell company for accounting reasons. Therefore, the incentive structure in the nonprofit sector is more diffuse. The public and private sector are clearer: potential profits end up in the general municipal budget or in the hand of owners/shareholders. 5 Education level: “Which degree does this person have?” (1) no degree/elementary school, (2) secondary school, (3) higher studies/university. Age: “Which age group does this person belong to?”: (1) 20 years or younger, (2) 21–30 years, (3) 31–40 years, (4) 41–50 years, (5) 51–60 years, (6) 61 years or older. Ethnicity: “Which ethnicity does this person have?”: (1) Belgian, (2) Moroccan, (3) other. Income “How wealthy is this person?”: (1) poor, (2) below average, (3) average, (4) above average, (5) wealthy. 6 The email was originally written in Flemish (see box 2 in the Appendix). 7 Some of the responses that we received were untraceable back to the original request or corresponding elderly care home (including seven that were sent to Kenny Maes and seven that were sent to Mohamed El Makrini); these responses were coded as “n/a.” In four instances, a facility sent an email to both Mohamed and Kenny, even though we did not send requests from both senders to any elderly care home. This was probably due to behind-the-scenes organizational and managerial structures at different homes, as a few of the facilities belong to a “mother” organization (which may have been forwarded the requests from individual facilities). Nonetheless, judging from the contents of the responses, we have no reason to believe that the senders were suspicious. Moreover, excluding these responses did not alter the results in any meaningful way. 8 Autoreplies and forwarded messages were coded as no replies. Nevertheless, when a direct message was later received, it was coded as a response. 9 We estimated a Krippendorf’s alpha (Krippendorff 2013) of 0.56 for information comprehensiveness, and 0.28 for friendliness ratings. 10 We also estimated all of the models presented in this section only for the subset of emails that generated responses. The results are identical to those presented in the text, except that the ATE of the combined public and private facilities is statistically significant. We focus on the more conservative approach of taking all experimental units into account. 11 One alternative interpretation for this finding would be that the ethnic composition of public and private nursing homes differs substantially. A more diverse nursing home, in turn, would have lower associated costs with accepting an additional Maghrebian client. While official figures for client composition on Flemish elderly care homes are not available to test this proposition, we used the ethnic composition within municipalities where nursing homes are situated as a crude proxy, finding no systematic differences between public and private facilities’ localities. © The Author(s) 2018. Published by Oxford University Press on behalf of the Public Management Research Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Public Administration Research and Theory Oxford University Press

Discrimination and Administrative Burden in Public Service Markets: Does a Public–Private Difference Exist?

Loading next page...
 
/lp/ou_press/discrimination-and-administrative-burden-in-public-service-markets-US0Ipk0O93
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Public Management Research Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
1053-1858
eISSN
1477-9803
D.O.I.
10.1093/jopart/muy009
Publisher site
See Article on Publisher Site

Abstract

Abstract Public services are often provided in markets where both public and private providers operate. Irrespective of ownership status, public services ought to be accessible regardless of clients’ race, gender, ethnicity, or age. However, as theories of statistical discrimination and cream skimming suggest, market-based incentives may lead service providers to focus on nonminority clients because they perceive them as easier-to-serve and therefore less costly. This may lead to discrimination and hence jeopardizes equal access. In this study, we ask whether private, for-profit providers are more likely to discriminate on ethnic grounds compared to publicly owned providers. We implement a field experiment within the Flemish elderly care market by sending out email requests with either a Flemish or a Maghrebian name to all public and privately owned nursing homes. For overall response rates, no statistically significant differences between senders were found. However, we do find that privately owned facilities are about 20 percentage points less likely to provide information on how to enroll when the request is send from a Maghrebian name alias, and blinded coders perceive the information sent to the Maghrebian alias as less comprehensive. In publicly owned facilities, no such differences exist. We conclude that a public–private difference does exist, but that the mechanism of discrimination is subtler than expected. Rather than directly refusing to respond, nursing homes increase administrative burdens and learning costs for minority applicants. Introduction Despite comprehensive anti-discrimination laws, citizen minorities repeatedly report difficulties not only when encountering bureaucrats but also when trying to access important public services such as education, health care, or social welfare. For example, a recent Gallup poll in the United States suggests that a vast number of African Americans perceive having fewer opportunities to access crucial public services (Gallup 2014). Stark inequities in service provision exist within European countries as well, as a recent literature review indicates (Dan and Andrews 2015). Moreover, Eurobarometer data shows that disadvantaged groups are increasingly dissatisfied with the accessibility of public services (Clifton et al. 2012) and reports about ethnic and racial disparities in both access to and use of public services are made repeatedly (e.g., Harris et al. 2006; Kugelmass 2016; Weinick, Zuvekas, and Cohen 2000). Public administration scholarship has a long-lasting tradition of studying the sources and consequences of social inequity in public service delivery (for overviews, see Frederickson 2010; Gooden 2015). Inequality in service access has been often attributed to clients’ personal characteristics (e.g., their social capital, cognitive abilities and more general vulnerabilities in economic exchange relationships) (Clifton et al. 2011; Jilke 2015; Schneider, Teske, and Marschall 2002). Additionally, institutional structures, routines and even public officials’ (implicit) attitudes and behaviors factor in as well (Dabney et al. 2006; Epp, Maynard-Moody, and Haider-Markel 2014). A recent addition to the study of inequality in public services looks at how administrative burdens work as a mechanism of exclusion. The literature on administrative burden distinguishes three types of costs that may increase the difficulty citizens have in accessing public services, namely learning costs (citizens must learn about a service and how to apply for it), psychological costs (citizens may face stigma or stress associated with accessing a certain service) and compliance costs (citizens need to comply with a service’s formal rules and requirements) (Moynihan, Herd, and Harvey 2015; Moynihan et al., 2013). These costs are not stable and can be altered by policy makers or frontline workers. Frontline workers play a particularly important role. Their work discretion allows a considerable margin for increasing or decreasing administrative burden (Brodkin and Lipsky 1983; Keiser and Soss 1998; Lipsky 1984; Soss, Fording, and Schram 2011). Frontline work therefore affects the difficulties clients may have in accessing services to which they are entitled. The result is what Lipsky (1984) refers to as “bureaucratic disentitlement.” Some clients may already face an increased administrative burden in their first encounter with service providers—namely when asking for guidance and information concerning how to enroll in public programs. In this case, learning costs could be easily increased by street-level organizations to dissuade potentially costly clients by not replying to requests, providing incomplete information or simply being rude and discouraging. However, how do street-level bureaucrats know which clients are costly? In line with theories of statistical discrimination (Arrow 1973; Phelps 1972), we argue that in the absence of concrete information about individual clients’ future costliness, frontline workers use imperfect signals such as race, ethnicity, and age to make inferences about a client’s future costliness. As a consequence, employees of street-level organizations may increase the administrative burden for these potentially harder-to-serve clients. Few studies have examined whether (and to what extent) frontline workers increase administrative burden for members of societal minorities (for a notable exception, see Heinrich 2016). This is an important consideration given that discretionary bias in the allocation of opportunities and resources can result in unequal treatment of minority groups—especially in the presence of limited resources (Brodkin 1997; Lipsky 1980; Scott 1997). While work discretion provides frontline workers with the opportunity to discriminate, incentives in public service markets may provide them with a motive. Today most public services are not delivered by state monopolists, but by service providers that compete for clients, thereby often facing performance incentives to succeed economically. Such market-based incentives within (quasi-)markets constitute one of the key theoretical mechanisms related to why minority groups may face an increased administrative burden. It has been argued that frontline providers would be more prone to either avoid costly, labor-intense clients (i.e., engage in dumping) or prioritize clients who they deem most likely to succeed in terms of bureaucratic success criteria (i.e., undertake cream-skimming) (Ellis 1998; Koning and Heinrich 2013; Lacinero-Paquet et al. 2002; Le Grand and Bartlett 1993; Lipsky 1980; Tummers et al. 2015). In other words, street-level organizations not only prioritize motivated clients who are most likely to succeed, but also avoid clients who seem likely to either fail or be costly. Arguably, quasi-markets (e.g., education, health care or employment services) have an inherent incentive structure that may lead both public and private service providers to a focus on easier-to-serve clients (Heinrich and Marschke 2010; Koning and Heinrich 2013; Soss, Fording, and Schram 2011). Yet, there are good reasons to believe that incentives to cream-skim or dump clients are affected by a service provider’s ownership status.1 Since private facilities can retain profits for their shareholders, there is a direct financial gain to be made by profiling applicants with stronger socioeconomic profiles. In public facilities, no such profit-maxim exists. Moreover, a stronger public sector ethos can be expected to mitigate the propensity for dumping or cream skimming. In this study, we look for evidence of discriminatory practices against ethnic minorities who are accessing elderly care services (which are typically provided by public, private, or nonprofit providers within a national care market; see also Amirkhanyan, Kim, and Lambright 2008). We implement an audit study design field experiment to assess the revealed discrimination against ethnic minorities among street-level organizations across all public and private elderly care facilities in Flanders (i.e., the Dutch-speaking part of Belgium). We find that elderly care homes do not discriminate against ethnic minorities in responding to information requests. However, we find public–private differences when it comes to increasing administrative burdens through learning costs. Here, private elderly care homes are approximately 20 percentage point less likely to provide information on how to enroll to a Maghrebian sender than to a sender with a Belgian alias; however, publicly owned homes do not make any differentiation. On this basis, we discuss the theoretical and practical implications of public–private differences in ethnic discrimination in access to elderly care—and public services more broadly—and propose a future research agenda for experimentally studying racial and ethnic discrimination within public service delivery. Discrimination within Public Service Delivery: Theory and Evidence Discrimination within access to public services implies that applicants receive different treatment based on belonging to a certain group, such as a racial or ethnic minority. This unequal treatment, which disadvantages a particular group and its members, is often based on other inadequately justified factors beyond race or ethnicity as well (National Research Council 2004). In this study, we broadly distinguish between three types of discrimination: implicit, unconscious discrimination; explicit, overt discrimination; and rational-strategic discrimination.2 Theories of implicit discrimination argue that discriminatory practices are the product of implicit attitudes or stereotypes against the outgroup: that is, people who are perceived not to belong to the own group (e.g., Chugh 2004; Greenwald and Krieger 2006). As such the act of discrimination is “unintentional and outside the discriminator’s awareness” (Bertrand, Chugh, and Mullainathan 2005, 94). In the field of social psychology, these implicit attitudes are measured using the Implicit Association Test (Greenwald, McGhee, and Schwartz 1998). Many scholars have asserted that implicit attitudes (e.g., toward African Americans) predict discriminatory behavior. However, an important meta-analysis (Oswald et al. 2013) has suggested only a weak association between implicit attitudes and racial and ethnic discrimination. A more recent network meta-analysis of how changes in implicit biases predict explicit bias or discriminatory behavior has similarly found only very limited evidence of such a relationship (Forscher et al. 2017). Hence, while the prevalence of implicit attitudes and stereotypes toward outgroups is widely acknowledged in the literature, the link of implicit attitudes to discriminatory practices seems to be not as straightforward as often suggested. Secondly, theories of explicit, overt discrimination argue that discrimination stems from individuals’ distaste toward outgroup members. In his theory of taste-based discrimination, Becker (1971) claims that discrimination is mainly based on an individual’s animus against members of certain outgroups. This taste for discrimination leads individuals to avoid interacting with those against whom they harbor animus—even if doing so comes at some cost. A disamenity value thus exists, for instance in terms of accepting members of certain outgroups into a public program. Thirdly, the literature on statistical discrimination argues that discrimination is not an emotional response to differences but a product of rational-strategic considerations based on aggregate group characteristics. Those aggregate characteristics are used to stereotype representatives of these groups (Altonji and Pierret 2001; Arrow 1973; Phelps 1972). A classic example is an employer who tries to assess the motivation and future productivity of a job applicant. As this type of information is not directly available in the early stages of the application process, employers use so-called imperfect signals of productivity from the applicant’s curriculum vitae (CV). The less direct information an employer obtains from these sources, the more weight he or she assigns to the average productivity of other workers who belong to the same group as the applicant (based for instance on gender, race, ethnicity, age, or place of residence). The employer subsequently uses this aggregate-level group information as a stereotype against individual applicants. The main difference between the explicit and rational-strategic models of discrimination is that taste-based discrimination stems from an agent’s distaste for interacting with members of a particular group, even if avoiding interaction with them bears some costs. Theories of statistical discrimination in turn argue that in the absence of direct information, agents use stereotypes based on population-specific means to avoid individuals who they (or their organizations) would benefit from the least (Guryan and Charles 2013). In the case of access to public services, this also means that having limited information leads service providers to try to avoid clients deemed costly based on aggregate-level group information. As such they use clients’ group membership as an imperfect signal of their future costliness. Statistical discrimination in access to public services can be seen as a form of cream-skimming—which is defined as the intentional selection or avoidance of certain groups for a public program (Ellis 1998; Koning and Heinrich 2013; Lacinero-Paquet et al. 2002; Le Grand and Bartlett 1993; Lipsky 1980). Cream-skimming (or its functional equivalent of dumping) is founded on rational strategic considerations related to reaching program goals or working in a cost-efficient manner. Nonetheless, like statistical discrimination it involves disadvantaging members of a certain group based merely on their group characteristics. Theories of statistical discrimination provide a framework for explaining the mechanism at play when imperfect signals (such as race or ethnicity) are used in the process of cream-skimming; for example, nursing homes may try to avoid clients they deem labor intense or costly. The absence of full information about the future costliness of individual applicants leads care providers to avoid members of certain groups that on average create higher workloads and increased costs. The Maghreb community in Flanders (with migrants from Morocco, Tunisia, and Algeria) is a good example. Members of this group practice a different religion from the majority of the region’s residents (i.e., Islam instead of Roman Catholicism) and therefore have different nutritional needs (e.g., they require halal meat) and religious ceremonies. In addition, they have a higher likelihood of not speaking the local language as fluent as the part of the Flemish population whose mother language is Flemish. Using this population-specific information as a stereotype against individual Maghrebian applicants may lead service providers to stereotype them as being harder to serve (i.e., costlier) than similar Belgian applicants—and hence try to avoid them. Much of the empirical evidence concerning discriminatory practices draws on theories of taste-based and statistical discrimination (Guryan and Charles 2013). The related literature typically examines discrimination in employment decisions based on applicants’ race, gender, or ethnicity. Many empirical studies have employed an audit-study methodology in which fictitious applications are sent in response to public job openings (e.g., Bertrand and Mullainathan 2004; Kaas and Manger 2011; Pager, Western, and Bonikowski 2009). The application materials (including the CV and cover letter) are usually identical, apart from one important characteristic, such as the applicants’ race, gender, or ethnicity (depending on what is being studied). Bertrand and Mullainathan’s (2004) famous audit study looked at racial discrimination in the Boston and Chicago labor markets. The study entailed submitting similar resumes to more than 1,300 help-wanted advertisements in local newspapers, using names manipulated to sound either white or African American (such as Greg or Jamal). The findings reveal that African Americans face considerable discrimination when searching for jobs; members of this group were approximately 50% less likely to receive a call-back than equally qualified “white” applicants. These results have been largely corroborated by numerous other studies (for a recent meta-analysis, see Quillian et al. 2017). In addition to job applications, audit studies also exist in relation to issues such as housing (Ahmed and Hammarstedt 2008; Turner and Ross 2003; Turner et al. 2002; Yinger 1995), car sales (Ayres and Siegelman 1995) and insurance markets (Galster, Wissoker, and Zimmermann 1998). Activities such as hailing taxis (Ridley, Bayton, and Quttz 1989), being recommended for medical care (Schulman et al. 1999), making an appointment with a psychiatrist (Kugelmass 2016) and even contacting elected officials (Broockman 2013; Butler and Broockman 2011) have been explored as well. The related studies have revealed consistent patterns of racial and ethnic discrimination across a wide range of contexts. Recently, scholars of bureaucracy have started to examine discrimination in public services by using audit studies to explicitly explore individuals’ learning costs in relation to accessing services (see also Costa 2017 for a meta-analysis of audit studies on public officials). There is evidence of racial discrimination in terms of requesting information about how to register to vote (White, Nathan, and Faller 2015), but also in answering information requests send to traditional public service providers like school districts, libraries, police forces, job centers and county clerks (Giulietti, Tonin, and Vlassoupoulos 2017). Einstein and Glick (2016), in contrast, find no evidence for racial discrimination in access to public housing. Response rates to Black, Hispanic, or White senders are not significantly different. However, they report that responses sent to Hispanic senders were less likely to include a salutation. Other studies that have examined ethnic discrimination in information requests that were sent directly to local governments found no systematic differences in responses (Adman and Jansson 2016; Grohs, Adams, and Knill 2015). A recent study on ethnic discrimination in information requests of how to enroll into receiving unemployment benefits also found no evidence of ethnic discrimination in response rates, but in the quality of information provided, with ethnic minorities receiving replies of lower quality (Hemker and Rink 2017). These experimental studies provide some reasons to expect ethnic discrimination in access to elderly care services. However, the public services chosen in prior studies operate mostly within monopolistic service structures (e.g., local government, policing, public housing, or welfare provision), where we would expect only low-powered incentives for cream-skimming/dumping (unlike services that are provided within competitive environments). Indeed, public services delivered within noncompetitive environments typically do not face similar performance pressures as services within (quasi-)markets (Le Grand and Bartlett 1993; Savas 2000). As a result, audit studies on bureaucratic discrimination in monopolistic settings that have examined the nature of discriminatory practices suggest that acts of discrimination are the product of taste-based animosity and not rational-strategic considerations (i.e., statistical discrimination) (Costa 2017; Giulietti, Tonin, and Vlassoupoulos 2017; White, Nathan, and Haller 2015). Our study focuses on a service provided within a (quasi-)market, where we would expect performance incentives that encourage discriminatory practices based on rational-strategic considerations. In addition, our research context allows us to examine differences between public and private (for-profit) providers, and thereby to acknowledge varying degrees of discrimination across institutional structures. Public–Private Differences in Discrimination Good reasons exist for expecting that private, for-profit providers are more likely to engage in cream-skimming (or dumping) than publicly owned providers. Different institutional logics in public and private organizations mold the incentive structure that shapes work discretion (Heinrich and Marschke 2010; Soss, Fording, and Schram 2011). Private providers distribute their profits to shareholders and managers. Public providers cannot do so. Weisbrod (1988) refers to the nondistribution constraint which prevents public and not-for-profit organizations to distribute profits to shareholders. Thereby the nondistribution constraint reduces the incentives for public agencies (and nonprofits) to misrepresent or withhold information about how to access services, reduce quality or prioritize certain types of service users at the “front door” of service delivery. The impact of the incentive structure for organizations on the frontline of service delivery is determined through processes of professional norm-setting (Abbott 1988). Incentives for profit-seeking providers tend to emphasize short-term goals, often at the expense of long-term goals related to client outcomes (Dias and Maynard-Moody 2007). In contrast, the workers at publicly owned street-level organizations lean more heavily on distinct professional standards to help clients to the best of their ability. The literature on public service motivation provides insights into what these standards might entail. Public sector employees assign greater value to public service; work that is beneficial to both others and society as a whole; involvement with important public policies; and self-sacrifice, responsibility, and integrity (Rainey and Bozeman 2000, 460). Perry, Hondeghem, and Wise (2010) refer to an other-regarding orientation, which is represented by notions of self-sacrifice, altruism, and prosocial behavior. The focus on the other, in our case the prospective client (i.e., elderly individuals seeking care), can be expected to constrain cream-skimming and dumping practices. The professional care standards can be an antidote against exclusion. Public sector organizations can be expected to be better capable of buffering external competitive pressures and safeguarding professional standards (Meier and O’Toole 2011). In private sector facilities, economic incentives may crowd out the professional standards and public sector motivations (Kjeldsen and Jacobsen 2013). Lacireno-Paquet et al. (2002) found evidence for differences in dumping (i.e., avoiding needy populations) in market-oriented charter schools; however, they found less evidence of differences among public and nonmarket-oriented private schools. Studies of performance target regimes that introduce a for-profit logic into public services have also found evidence of crowding out effects. Several researchers have documented cases in which performance targets induce frontline workers to focus exclusively on measured tasks (Hood 2006; Kerpershoek, Groenleer, and de Bruijn 2016; Pollitt 2013). The goals as measured by performance indicators (as opposed to the real goals) provide direction to work within street-level organizations. For example, schools may focus on obtaining better test scores rather than on providing better education (teaching for the test) and hospitals may focus on shortening waiting times instead of promoting better health (Bevan and Hood 2006; Radin 2006). Bohte and Meier (2000) discuss this phenomenon as an instance of goal displacement. When performance is evaluated in terms of numerical outputs, organizations have an incentive to maximize their outputs—regardless of whether output maximization is the preferred strategy for achieving desired social outcomes (Bohte and Meier 2000, 173). Similar effects have been found in health care (Bevan and Hood 2006), education (Bohte and Meier 2000), job training (Heinrich 1999), and policing (Epp, Maynard-Moody, and Haider-Markel 2014). We expect that private for-profit providers have an incentive to select applicants who are more profitable and avoid those they regard as costly. Ethnic minorities risk being perceived as harder to serve due to their different cultural backgrounds and demands (in relation to issues such as dietary requirements, language, and cultural practices). We base this expectation on theories concerning the work of frontline workers. Incentives for profit maximization in private facilities may limit the scope for workers to uphold professional standards in their daily work. Hypotheses The study tests two hypotheses related to discrimination against ethnic groups in accessing elderly care services. Our first hypothesis is derived by combining literature on statistical discrimination and cream-skimming. In line with a sizeable number of studies, we assume that street-level organizations respond to competitive pressure by intentionally avoiding groups they deem harder to serve and thereby costly. Statistical discrimination provides a theoretical framework for explaining ethnic discrimination in the context of cream-skimming (including dumping). We assert that for employees of street-level organizations, ethnicity serves as an indication about the future costliness of individual clients. This imperfect signal influences street-level decision-making in the process of cream-skimming (or dumping) when no direct signal of a client’s future costliness exists. Against this background, we argue that street-level organizations within competitive environments (such as the Flemish elderly care market) use applicant clients’ ethnicity to determine if they are harder to serve and therefore costly. Within this study we focus on Maghrebians, who currently constitute the largest immigrant group in Flanders to date (Van den Broucke et al. 2016, 96). Members of this particular group who are at a retirement age typically do not speak the Flemish language well and have certain religion-based nutritional needs. Street-level organizations use this objective, population-specific information to stereotype individual applicants, who they perceive as being harder to serve and therefore costlier than similar Flemish applicants. As a result, elderly care providers may try to avoid individual applicants who belong to this particular group to escape future costs. This leads us to expect that Maghrebians will experience more discrimination than similar, otherwise Flemish applicants when they try to access information concerning how to apply for elderly care places in the Flemish elderly care market. Hypothesis 1: Maghrebian aliases experience greater discrimination than Flemish names in relation to accessing information about applying for elderly care services. We expect discriminatory practices to be more pronounced in the private (for-profit) elderly care service sector than in the public sector. In private organizations, the profit motive creates an incentive structure that reduces the scope for maintaining professional standards of care. Moreover, profit maximization incentives may lead workers to substitute commercial standards acquired through job socialization for altruistic public service standards learned through professional training (Kjeldsen and Jacobsen 2013). In contrast, employees in the public sector are shielded by the nondistribution constraint, which buffers them from competitive pressures and bolsters their maintenance of professional standards. This leads us to our second hypothesis: Hypothesis 2: The extent of discrimination against Maghrebian aliases is stronger in privately owned (for-profit) elderly care homes than in publicly owned homes. Research Context: The Flemish Elderly Care Market In Flanders, elderly care is provided by 783 homes. The field of providers is diverse and features different ownership structures. The largest group has a nonprofit legal status. Facilities in this category are typically run under the auspices of religious (mainly Catholic) organizations. The second category is elderly care homes with public ownership. Homes in this group are administered by municipalities—but like the nonprofits, the public facilities are mostly regulated and financed by the central government. The third category includes homes with private (for-profit) ownership, which again fall under the same regulatory framework as the public and nonprofits. The increase in private, for-profit ownership in the Flemish elderly care market is a trend of the last decade. In this study, we focus on the last two categories, namely public and private facilities. The overall goal of the elderly care policy in Flanders is to provide universal access to affordable and high quality care services (Van den Bosch et al. 2011; Willemé et al. 2012). Residents pay their accommodation costs, while the compulsory national health insurance scheme bears their medical and nursing costs. Means-tested compensation exists for individuals who cannot pay for their share. The monthly accommodation costs are comparable for nonprofit and profit organizations but slightly lower for public institutions (Socialistische Mutualiteiten 2016). About 40% of the income of the elderly care facilities comes from health insurance. Elderly care homes typically provide beds for residents with lower and higher care needs; the beds for individuals with greater requirements (such as those with dementia) are called RVT beds. Higher care needs are discounted for with higher payments from the national health insurance. Another 40% of facilities’ income stems from the daily allowances that residents have to pay. The remaining 20% comes from miscellaneous subsidies and cost recuperation. The nonprofit sector is the most profitable sector, with a profit of EUR 4.60 a day/bed; followed by the private sector, at EUR 0.90 a day/bed.3 The public sector is losing approximately EUR 13 a day/bed (Pacolet and De Coninck 2015). The deficit is covered by the municipality in which the elderly care home is located. The central government regulates the elderly care sector. All facilities need to be certified by an inspectorate before they can claim subsidies. This inspectorate monitors compliance with regulations as well as the quality of care being provided. Inspection reports are publicly available. In addition, the central government determines how many beds are required within each region, and elderly care homes can apply for additional beds based on the forecasted need. Based on residents’ care needs, the central government also determines the daily amount that the health insurance scheme has to pay facilities. Nonetheless, elderly care homes have a significant level of autonomy within this regulatory framework. They are separate legal entities that can independently make hiring decisions, borrow from financial institutions, and accumulate assets. They also have significant freedom in determining what they charge residents and can manage their waiting lists independently. In summary, the elderly care sector in Flanders can be described as a regulated public service market. Residents are free to choose between facilities. In addition, facilities are autonomous decision making units that independently manage their admission processes and waiting lists. With an occupancy rate of 96% in 2014 (Socialistische Mutualiteiten 2016), we can safely assume that creaming is a realistic option—which makes it an ideal sector to test our theoretical predictions. Experimental Design In this study, we examine the discriminatory behavior of elderly care facilities in Flanders by studying their responsiveness to information requests by prospective clients. Withholding or providing incomplete information when employees respond to queries concerning how to apply for an elderly care place is an important administrative burden for potential applicants. To explore potential discriminatory practices related to the provision of information about how to enroll in elderly care, we build on the audit study tradition of randomized field experiments (for an overview, see Pager 2007). The basic experimental design of the audit study methodology relies on sending identical information requests that differ by one attribute, namely the race, ethnicity, or gender of the putative sender. The discriminatory behavior of the audited agents is assessed by comparing response rates and the information provided across randomly assigned requester names. Treatment We tested the effectiveness of randomized names on discrimination in access to nursing homes for the elderly in Flanders. Identical information requests were sent to Flemish elderly care homes. To test public–private differences, we focus on public and private, for-profit nursing homes as they most clearly have the incentive structures we are theoretically interested in.4 The treatment consisted of two randomly allocated, hypothetical names. We used one typical Maghrebian name (Mohammed El Markini), with the goal of representing the predominant ethnic minority in Flanders, and one that is typically Flemish (Kenny Maes). These names were selected from a pretest. In Flanders, as in other countries, names are social identifiers not only of ethnicity, but also of socioeconomic status (SES) (Elchardus and Siongers 2011). For our study, we wanted to choose names that signal different ethnicities but a similar SES. Elchardus and Siongers (2011) found that parental educational levels are considerably higher for some names than for others, such as Thomas versus Kenny. We used this study as the basis for selecting Flemish names for our pretest. Therefore, we identified the most popular Maghrebian first names in Belgium from the Belgian national census. Given the underprivileged position of the Maghreb community in Belgium, we generally expected that Maghrebian names signal lower SES (Van den Broucke et al. 2016). The Flemish and Maghrebian last names for the pretest were selected based on the most recent Belgian census (Statistics Belgium). In the pretest, we used an existing online panel of the general population from the political science department of a Belgian university. Out of the 5,837 people who were asked to participate in the survey, 2,114 took part—which reflects a response rate of 36.2%. Each respondent was asked to assess three names based on four factors: education level, age, ethnicity, and income level.5 To avoid answer contamination, we randomized the order of the four items as well as the combination of the first and last names. Respondents randomly received three Flemish names, three Maghrebian names, or a combination of the two. In total, 10 Flemish and 10 Maghrebian names were assessed (see table A1 in the Appendix). As expected, all of the Maghrebian names triggered lower SES assessments than the Flemish names. Of the Flemish names, “Kenny” accounted for the lowest SES ratings. Based on these results, we opted to use Mohamed El Makrini and Kenny Maes, as these names are perceived as being very similar in terms of SES but not vis-à-vis ethnicity (see table A1 in the Appendix). The treatment was implemented in an email sent to all Flemish elderly care homes that asked whether the facilities currently had a place available, whether they had a waiting list and how to enroll. Two email accounts were created (kenny.maes100@gmail.com and mohamed.elmakrini100@gmail.com) and used to send the requests. The email itself was fairly short to decrease the burden for employees; box 1 in Appendix provides the full text.6 Each elderly care facility received only one email. Population and Unit of Analysis The unit of analysis for our field experiment was individual elderly care facilities. While we contacted individuals, we assumed that they would act in their professional capacity as a representative of the organization for which they work. The primary contact information for the facilities was collected from publicly available records (Agentschap Zorg and Gezondheid 2017; in English: Agency Care & Health). Public records also display the total number of elderly care homes in each municipality and indicate which facilities operate under public, private, and nonprofit ownership (according to their legal status; see table A2 in the Appendix). When elderly care facilities shared the same contact email address as a result of operating within the same umbrella organization, we randomly chose one address (a total of 18 facilities fell into this category: 7 public and 11 private). Allocation to the experimental conditions was performed randomly in strata of public and private organizations, thereby employing a randomized block design. As such we conducted an independent randomization procedure for each of the two facility categories (i.e., public and private). We sent 214 emails to public elderly care facilities and 118 emails to private elderly care facilities. Of the 332 emails that were sent in total, 21 (15 to public facilities and 6 to private facilities) could not be successfully delivered due to invalid addresses. For our analysis, we limited the sample to emails that were successfully delivered, which left us with an analytical sample of 311 observations (including 199 public and 112 private elderly care homes).7 Measurement We used two outcome measures to assess the discriminatory behavior of Flemish elderly care homes: response rates and the information provided. We chose these outcome variables because they represent core aspects of elderly care employees’ discretion in granting clients access to public services more widely. From those nursing homes that responded, only 5 had no waiting list and 49 did not provide any information on this account. Therefore, we did not include this question as an outcome measure. The response rate was measured by looking at whether service providers responded to our information requests. We counted systematic differences between ethnic names as evidence of ethnic discrimination.8 Second, we looked at whether those who responded provided information about how to apply for a place at their facility. Indeed, withholding information or providing incomplete information may result in a disadvantaged position for clients trying to access these services. This is therefore an important measure of whether elderly care facilities systematically increase administrative burden (i.e., learning costs) for certain types of clients. We coded responses that included the requested information as “1” and those that did not respond to the inquiry as “0.” Responses were considered valid if they were received within 2 weeks of the request email being sent (after 2 weeks no additional replies were received). Several elderly care homes provided us with information in attachments; for instance, one facility sent us an attached informational brochure that contained details concerning its enrollment procedure. These emails were coded as “1” for our second outcome measure (i.e., providing information about how to enroll). In addition, we report the results of two secondary measures to complement the results of our primary measures of interest: information comprehensiveness and reply friendliness. These were created through ratings of three blinded coders (research assistants). In particular, we asked coders to rate all received replies (which we anonymized with regard to its sender, but also with respect to the experimental condition) on a 7-point scale, ranging from “no information was provided,” “written in an unfriendly manner” to “all asked information was provided,” “written in a friendly manner,” respectively. In this sense, the rating of information comprehensiveness is strongly correlated with our primary measure of interest on information provision. Coders were given the initial request that we sent out. They were also provided examples for each scale’s middle category and respective endpoints to make sure equivalent rating standards are applied across coders (Krippendorff 2013). We have to note, however, that while the intercoder reliability of the information scale was somewhat acceptable, the friendliness ratings’ reliability was rather poor.9 Empirical Results Response Rates Of the 332 emails we sent to elderly care facilities, 223 received responses (table 1). If the 21 emails that bounced back are not taken into account, this yields an overall response rate of 71.7%. The overall response rate for the public facilities were approximately 16 percentage points higher than for the private facilities (77.4% versus 61.5%). Flemish elderly care homes can thus be seen as moderately responsive overall, which exemplifies that not replying to email requests is indeed a realistic option for facilities. Arguably, not responding to emails could be an easy way for nursing homes to “select at the door.” Table 1. Summary statistics for the response rates All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails sent 166 166 107 107 59 59 Incorrect address/ bounce 11 10 8 7 3 3 Response rate 112 (72.3%) 111 (71.2%) 75 (75.8%) 79 (79%) 37 (66.1%) 32 (57.1%) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails sent 166 166 107 107 59 59 Incorrect address/ bounce 11 10 8 7 3 3 Response rate 112 (72.3%) 111 (71.2%) 75 (75.8%) 79 (79%) 37 (66.1%) 32 (57.1%) View Large Table 1. Summary statistics for the response rates All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails sent 166 166 107 107 59 59 Incorrect address/ bounce 11 10 8 7 3 3 Response rate 112 (72.3%) 111 (71.2%) 75 (75.8%) 79 (79%) 37 (66.1%) 32 (57.1%) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails sent 166 166 107 107 59 59 Incorrect address/ bounce 11 10 8 7 3 3 Response rate 112 (72.3%) 111 (71.2%) 75 (75.8%) 79 (79%) 37 (66.1%) 32 (57.1%) View Large Table 1 presents the response rates by treatment condition (i.e., type of name) within public and private elderly care homes. We find no significant differences between emails signed with a Flemish name versus those signed with a Maghrebian name (the difference is 1.1 percentage point; χ2(1) = 0.047, p = .829). When a logistic regression model is estimated instead, using standard errors clustered by municipality to account for potential clustering effects across facilities located in the same area, the results are identical (table 2). In public elderly care homes, signing an email request with a Maghrebian alias led to a 3.2 percentage point greater likelihood of receiving a response than a similar request signed with a Flemish name; however, these differences are statistically insignificant (χ2(1) = 0.944, p = .331). Identical results were obtained for public and private organizations when potential clustering effects within municipalities were accounted for. In relation to private facilities, emails signed with a Maghrebian name were approximately 9 percentage points less likely to receive a reply than those that used a Flemish name. While this effect is nontrivial in size, it is not statistically significant (χ2(1) = 0.299, p = .585). Table 2 summarizes the average treatment effect (ATE) of Maghrebian versus Flemish name aliases for public, private, and both facilities combined. Table 2. Response Rates (Logistic Regression Models) (1) (2) (3) All Public Private Maghrebian name −0.011 0.032 −0.089 (0.050) (0.059) (0.091) Observations 311 199 112 (1) (2) (3) All Public Private Maghrebian name −0.011 0.032 −0.089 (0.050) (0.059) (0.091) Observations 311 199 112 Note: Standard errors in parentheses (clustered by municipality). Coefficients are marginal effects. ***p < .01. **p < .05. +p < .1. View Large Table 2. Response Rates (Logistic Regression Models) (1) (2) (3) All Public Private Maghrebian name −0.011 0.032 −0.089 (0.050) (0.059) (0.091) Observations 311 199 112 (1) (2) (3) All Public Private Maghrebian name −0.011 0.032 −0.089 (0.050) (0.059) (0.091) Observations 311 199 112 Note: Standard errors in parentheses (clustered by municipality). Coefficients are marginal effects. ***p < .01. **p < .05. +p < .1. View Large Information Provision Of the 332 emails sent to elderly care facilities, 196 (59.0%) generated responses that included information about how to enroll in an elderly care facility. While this figure represents a reasonable share of the elderly care facilities, it also demonstrates that a significant number of facilities did not share such information with requesters—and thereby increased their learning costs vis-à-vis access to elderly services. Below we assess whether this increase is systematic, which includes examining the number of responses that contained this information relative to the total number of emails we sent. We do so because focusing on only the subset of responses could break the experimental randomization,10 which could in turn potentially introduce bias into the identification of a causal effect of a Maghrebian versus a Flemish name in increasing the administrative burden. In table 3, the number and percentage of responses that include information about how to apply for a place at a Flemish elderly care home are provided by experimental condition. Of all of the valid requests sent, Maghrebian senders were approximately 8.1 percentage points less likely to receive information than identical requests from Flemish requesters. While this difference is of substantive magnitude, it is not statistically significant (χ2(1) = 2.201, p = .198). When the ATE within a logistic regression framework is estimated with standard errors clustered by municipality, the results are identical. In relation to public facilities, neither substantive nor statistically significant differences exist between the experimental conditions (the difference is only 1.7 percentage points; χ2(1) = 0.067, p = .796). However, examining the differences between name aliases in relation to private elderly care homes reveals substantive differences; here Maghrebian name aliases were approximately 19.6 percentage points less likely to receive enrollment information than identical requests from Flemish aliases. This finding is not only of a substantive magnitude, it is also statistically significant (χ2(1) = 4.334, p = .037). When the ATE of Flemish versus Maghrebian names is estimated in the context of a logistic regression model with clustered standard errors, the results are identical (table 4). Table 3. Summary Statistics for the Information Provided in Responses All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails with information on how to apply 104 (67.1%) 92 (59.0%) 69 (69.7%) 68 (68%) 35 (62.5%) 24 (42.9%) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails with information on how to apply 104 (67.1%) 92 (59.0%) 69 (69.7%) 68 (68%) 35 (62.5%) 24 (42.9%) View Large Table 3. Summary Statistics for the Information Provided in Responses All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails with information on how to apply 104 (67.1%) 92 (59.0%) 69 (69.7%) 68 (68%) 35 (62.5%) 24 (42.9%) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Emails with information on how to apply 104 (67.1%) 92 (59.0%) 69 (69.7%) 68 (68%) 35 (62.5%) 24 (42.9%) View Large Table 4. Information Provided (Logistic Regression Models) (1) (2) (3) All Public Private Maghrebian name −0.081 −0.017 −0.196** (0.053) (0.065) (0.086) Observations 311 199 112 (1) (2) (3) All Public Private Maghrebian name −0.081 −0.017 −0.196** (0.053) (0.065) (0.086) Observations 311 199 112 Note: Standard errors in parentheses (clustered by municipality). Coefficients are marginal effects. ***p < .01. **p < .05. +p < .1. View Large Table 4. Information Provided (Logistic Regression Models) (1) (2) (3) All Public Private Maghrebian name −0.081 −0.017 −0.196** (0.053) (0.065) (0.086) Observations 311 199 112 (1) (2) (3) All Public Private Maghrebian name −0.081 −0.017 −0.196** (0.053) (0.065) (0.086) Observations 311 199 112 Note: Standard errors in parentheses (clustered by municipality). Coefficients are marginal effects. ***p < .01. **p < .05. +p < .1. View Large Coder Ratings Next we assess whether blinded coders rated responses as more or less comprehensive in terms of the provided information. We also examine the perceived friendliness of responses (i.e., ratings using a 7-point scale), where nonresponses were treated as the lowest endpoints of the scale in order to not break the experimental randomization. Re-estimating these models for the subset of received responses yield no substantively different results. In table 5, mean differences in coder ratings between experimental conditions are summarized. In terms of the information completeness ratings we find very little differences between experimental conditions (mean difference of 0.11). These differences are not only trivial but also not statistically significant. Within public facilities, Maghrebian senders receive slightly more comprehensive replies on average (mean difference of 0.20) and this small difference does not reach acceptable levels of statistical significance. Only in private facilities the difference is stronger, with Maghrebian senders receiving replies of lower quality (mean difference of 0.66). This difference is statically significant at the 10% level, and hence corresponds well with our findings from table 4. Table 5. Mean Differences for Coder Ratings (Standard Errors in Parenthesis) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Information 3.42 (0.15) 3.31 (0.16) 3.49 (0.19) 3.68 (0.20) 3.29 (0.26) 2.64 (0.26) Friendliness 3.57 (0.14) 3.53 (0.15) 3.71 (0.18) 3.86 (0.17) 3.32 (0.25) 2.93 (0.25) Mean difference information 0.11 (0.22) −0.20 (0.27) 0.66 (0.37)+ Mean difference friendliness 0.04 (0.20) −0.15 (0.25) 0.38 (0.35) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Information 3.42 (0.15) 3.31 (0.16) 3.49 (0.19) 3.68 (0.20) 3.29 (0.26) 2.64 (0.26) Friendliness 3.57 (0.14) 3.53 (0.15) 3.71 (0.18) 3.86 (0.17) 3.32 (0.25) 2.93 (0.25) Mean difference information 0.11 (0.22) −0.20 (0.27) 0.66 (0.37)+ Mean difference friendliness 0.04 (0.20) −0.15 (0.25) 0.38 (0.35) ***p < .01. **p < .05. +p < .1. View Large Table 5. Mean Differences for Coder Ratings (Standard Errors in Parenthesis) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Information 3.42 (0.15) 3.31 (0.16) 3.49 (0.19) 3.68 (0.20) 3.29 (0.26) 2.64 (0.26) Friendliness 3.57 (0.14) 3.53 (0.15) 3.71 (0.18) 3.86 (0.17) 3.32 (0.25) 2.93 (0.25) Mean difference information 0.11 (0.22) −0.20 (0.27) 0.66 (0.37)+ Mean difference friendliness 0.04 (0.20) −0.15 (0.25) 0.38 (0.35) All Facilities Public Facilities Private Facilities Flemish Maghrebian Flemish Maghrebian Flemish Maghrebian Information 3.42 (0.15) 3.31 (0.16) 3.49 (0.19) 3.68 (0.20) 3.29 (0.26) 2.64 (0.26) Friendliness 3.57 (0.14) 3.53 (0.15) 3.71 (0.18) 3.86 (0.17) 3.32 (0.25) 2.93 (0.25) Mean difference information 0.11 (0.22) −0.20 (0.27) 0.66 (0.37)+ Mean difference friendliness 0.04 (0.20) −0.15 (0.25) 0.38 (0.35) ***p < .01. **p < .05. +p < .1. View Large With regard to friendliness ratings, overall no substantive differences between experimental conditions could be detected (mean difference of 0.04). In public facilities small differences exist, with Maghrebian senders receiving slightly friendlier replies (0.15 mean difference). But again, this small difference is not statistically significant. Within private facilities, small evidence for ethnic discrimination exist (mean difference of 0.38). However, the difference is too small to reach conventional levels of statistical significance. Because the data were left-censored, we estimated tobit models for coder ratings (accounting for potential clustering effects at the municipality level). Results do not change substantively (see table 6). With regard to information completeness ratings, we find that responses sent to the Maghrebian name were 2.6% less likely to be to be rated as providing all asked information. This difference is not only of trivial nature but also statistically insignificant. When considering public facilities alone, the opposite holds true in terms of the effect direction (i.e., a 3.4% increase for the Maghrebian name), but, again, this findings did not reach conventional levels of statistical significance and cannot be regarded as substantive. Within private facilities, however, systematic differences exist. Maghrebian senders received a response that provided less complete information on average (a 15.7% difference). This difference is both, of nontrivial nature, but also statistically significant (at the 10% level), and thus corresponds well with previous findings from table 4 (information provided on how to enroll). Table 6. Coder Ratings (Tobit Regression Models) Information Friendliness (1) (2) (3) (4) (5) (6) All Public Private All Public Private Maghrebian name −0.183 0.239 −1.103+ −0.0787 0.190 −0.669 (0.309) (0.351) (0.651) (0.275) (0.317) (0.548) Observations 311 199 112 311 199 112 Information Friendliness (1) (2) (3) (4) (5) (6) All Public Private All Public Private Maghrebian name −0.183 0.239 −1.103+ −0.0787 0.190 −0.669 (0.309) (0.351) (0.651) (0.275) (0.317) (0.548) Observations 311 199 112 311 199 112 Note: Robust standard errors in parentheses (clustered by municipality). Coefficients are unstandardized. ***p < .01. **p < .05. +p < .1. View Large Table 6. Coder Ratings (Tobit Regression Models) Information Friendliness (1) (2) (3) (4) (5) (6) All Public Private All Public Private Maghrebian name −0.183 0.239 −1.103+ −0.0787 0.190 −0.669 (0.309) (0.351) (0.651) (0.275) (0.317) (0.548) Observations 311 199 112 311 199 112 Information Friendliness (1) (2) (3) (4) (5) (6) All Public Private All Public Private Maghrebian name −0.183 0.239 −1.103+ −0.0787 0.190 −0.669 (0.309) (0.351) (0.651) (0.275) (0.317) (0.548) Observations 311 199 112 311 199 112 Note: Robust standard errors in parentheses (clustered by municipality). Coefficients are unstandardized. ***p < .01. **p < .05. +p < .1. View Large With regard to the friendliness ratings, no statistically significant difference could be detected in any of the estimated regression models. In terms of effect directions, we found very minor differences in terms of ethnic discrimination for all facilities combines (1.2%), so as for the public facilities (2.7%). Differences in the private group were larger (i.e., 9.6%), but here two factors have to be kept in mind, (1) our relatively small sample size of private nursing homes, and (2) the poor intercoder reliability for this measure.11 Discussion This study examined public–private differences in ethnic discrimination. Using a field experiment, we investigated access to elderly care for applicants with a Maghrebian and Flemish background. We implemented an audit study design by sending requests for information about enrolling in an elderly care home. Facilities could easily put up barriers for applicants by either not responding to requests, or not providing sufficient information. Indeed, both nonresponses as well as responses with poor information increase the learning costs of the applicant (Moynihan, Herd, and Harvey 2015). We hypothesized that the response rate and the quality of information provided would be higher for public providers than for private, for-profit homes, because the profit motive in the private sector was expected to create an adverse incentive structure to cream-skim easier to serve applicants. Ethnic minorities risk being perceived as harder to serve due to their different cultural backgrounds and demands. In the public sector, we expected that greater scope in the application of professional care standards would limit cream-skimming. Discrimination Is Lower Than Expected When assessing the overall effect for public and private facilities combined, our study found no evidence for ethnic discrimination in response rates. This is surprising given the evidence of racial and ethnic discrimination in other fields. For instance, studies of the Belgian labor market have discovered discriminatory practices toward ethnic minorities (Arrijn, Feld, and Nayer 1998; Baert et al. 2015). Baert et al. (2015) report that job seekers with Turkish names had to send out twice as many applications as candidates with Flemish-sounding names to be invited to a job interview. Moreover, a field study undertaken at the University of Leuven found strong evidence of discrimination in the Belgian private housing market (Van den Broeck, Heylen, and Winters 2014); 42% of real estate agents were prepared to discriminate against foreigners, which is a criminal offence under Belgian anti-discrimination laws. Given these local studies and prior audit study evidence, it is remarkable that elderly care homes in Flanders overall do not discriminate in their response rates—although the effort it would take to do so is particularly small (i.e., simply not responding to an email). An explanation for the lower-than-expected levels of ethnic discrimination in responding to information requests may be found in the names that were selected for this study. Names serve as social identifiers (Elchardus and Siongers 2011). Seeing as we wanted to study only ethnic discrimination, we pretested the used names connotations of SES or age. Indeed, some names are associated with higher levels of education and income, whereas others are associated with a working-class background. We selected a Flemish name with an SES association that is as close as possible to the low SES association of Maghrebian names. It may therefore be that past studies have attributed to ethnic discrimination what in reality was SES discrimination. One suggestion for future research is to further tease these potential SES effects out in a multi-factorial design, where SES is experimentally manipulated alongside applicants’ ethnicity. Administrative barriers and discriminatory practices can also be introduced downstream in the application process. The next step after an initial contact email and its response is finding information on how to enroll. In our study, we therefore also examined whether enrollment information was provided in the response. The provision of enrollment information lowers the learning costs for applicants. We found that the likelihood that this information would be provided to the Maghrebian applicant was substantially lower—approximately 8 percentage points—than for the Flemish applicant; however, this difference is not statistically significant. In subsequent stages of the admission process (which we did not study), other opportunities for cream skimming occur. Most elderly care homes suggest that applicants make a personal visit to the facility. While very good reasons for establishing personal contact with prospective residents do exist, site visits create the risk that frontline workers will dissuade applicants from minority groups and/or with a lower SES from further pursuing their application. In summary, although the responses we received to the email query is an encouraging finding, we cannot extrapolate it to the entire admission process. But Public–Private Differences Do Exist While we did not find evidence of discrimination for public and private facilities combined, the analysis of public–private differences suggests that ethnic discrimination does exist in private, for-profit nursing homes. There is suggestive evidence for differences (albeit statistically nonsignificant) in response rates. The extent of the difference, however, is smaller than expected. The relatively small population of private facilities may have provided insufficient statistical power to detect an effect, and hence sample size considerations need to be taken into account when interpreting this finding. However, when public–private differences in providing enrollment information were examined, our study revealed that privately owned facilities are less likely to provide information to an ethnic minority applicant; in this context, Maghrebian name aliases were approximately 20 percentage points less likely to receive enrollment information than Flemish aliases. This finding is not only of a substantive magnitude, but also statistically significant. In contrast, Maghrebian discrimination was almost zero within public facilities.11 In addition, when examining blinded coders’ assessments of the information completeness of received responses, a similar picture emerges. This suggests that public–private differences in ethnic discrimination are a realistic concern. The provision of information concerning how to apply for a place at an elderly care facility is not trivial, because it increases learning costs that lead to increased administrative burden for individuals who need to access care. Overall, our findings demonstrate that the institutional context does seem to have an impact on discriminatory behavior within street-level organizations. Private, for-profit organizations are less likely to provide information (and less comprehensive information) to Maghrebian names, whereas public organizations barely differentiate between names. Our findings also suggest that the causal mechanism for ethnic discrimination is probably more complex than initially assumed in the literature. Discrimination exists in many forms. However, in our study the low-hanging fruit (i.e., a simple nonresponse) is left untouched, whereas the more subtle discrimination form (namely increasing learning costs through the response to the email) is applied. Maynard-Moody and Musheno (2003) provide an explanation for why the most accessible form of discrimination, not responding, is the least strong. We can expect that in our research context the frontline imperative is particularly strong. Welfare workers—such as employees of elderly care homes—have direct and intense contact with the people they serve. These personal contacts inform their socialization processes and reinforce an “other-regarding” professional identity that the decision to not discriminate based on ethnicity conforms well to. Nonetheless, adverse institutional incentives seem to not be entirely neutralized. Private, for-profit facilities are significantly less likely to provide enrollment information when responding to Maghrebian aliases, but no such differences exist within public facilities. This indicates that profit incentives may indeed be crowding professional care standards out to some extent. While frontline professionalism seems to prevent the manifest act of ignoring a minority request, more subtle mechanisms of increasing learning costs are at play. Future research should attempt to disentangle this bundle of institutional influences and professional resistance. Appendix Box 1. Email Sent to Flemish Elderly Care Facilities From: [Randomized Treatment Name] To: [Email Address of Primary Contact at Elderly Care Facility] Subject: Admission information Hello, I am contacting you because I am looking for a place in a rest home for my father. We are interested in your facility. Do you have a place available at this moment? And how can I subscribe my father for this? I also heard there is a waiting list. Do you have one and how long is it? Thanks, [Randomized Treatment Name] Box 2. Email Sent to Elderly Care Facilities in Flemish From: [Randomized Treatment Name] To: [Email Address of Primary Contact at Elderly Care Facility] Subject: Informatie opname Hallo, Ik contacteer jullie omdat ik op zoek ben naar een rusthuis voor mijn vader. Wij zijn geïnteresseerd in jullie verblijf. Hebben jullie op dit moment een plaats beschikbaar? En hoe kan ik mijn vader daarvoor inschrijven? Ik heb ook gehoord dat er een wachtlijst is. Hebben jullie er één, en hoe lang is die? Bedankt, [Randomized Treatment Name] Table A1. Means of Pretest (1) Education Level (2) Age (3) Ethnicitya (4) Income Level Sven Janssens 2.526 3.007 98.3 3.137 (0.563) (0.815) (0.3; 1.4) (0.547) Joris Janssens 2.587 3.259 96.7 3.206 (0.525) (0.959) (0.9; 2.4) (0.464) Kenny Janssens 2.079 2.303 98.6 2.742 (0.664) (0.754) (0.3; 1) (0.672) Marc Janssens 2.483 4.157 98.4 3.294 (0.595) (0.895) (0.3; 1.3) (0.610) Thomas Janssens 2.682 2.794 98.6 3.244 (0.497) (0.855) (0.3; 1) (0.527) Sven Maes 2.610 2.927 99.1 3.136 (0.533) (0.750) (0.0; 0.9) (0.554) Joris Maes 2.602 3.272 98.1 3.237 (0.536) (1.084) (0.4; 1.5) (0.532) Kenny Maes 2.230 2.345 97.7 2.723 (0.583) (0.783) (0.7; 1.6) (0.612) Marc Maes 2.478 4.083 98.7 3.184 (0.577) (0.863) (0.0; 1.3) (0.568) Thomas Maes 2.673 2.768 97.6 3.280 (0.541) (0.832) (0.3; 2.1) (0.543) Mohamed El Makrini 2.136 2.803 74.5 2.863 (0.657) (0.500) (16.3; 9.1) (0.568) Ahmed El Makrini 2.167 2.903 71.6 2.902 (0.626) (0.917) (16.2; 11.6) (0.696) Rachid El Makrini 2.266 2.658 73.2 2.875 (0.645) (0.834) (14.9; 12.0) 1(0.534) Hassan El Makrini 2.272 2.804 71.1 2.919 (0.658) (0.924) (15.8; 13.2) (0.553) Youssef El Makrini 2.199 2.678 67.3 2.963 (0.627) (0.899) (21.8; 10.9) (0.635) Mohamed Bouzian 2.194 2.952 70.2 2.941 (0.670) (0.949) (17.8; 12.0) (0.637) Ahmed Bouzian 2.174 2.896 57.6 2.885 (0.634) (0.918) (19.8; 22.7) (0.564) Rachid Bouzian 2.292 2.729 60.2 2.952 (0.634) (0.876) (21.8; 18.0) (0.611) Hassan Bouzian 2.138 2.818 62.0 2.876 (0.630) (0.890) (16.3; 21.6) (0.554) Youssef Bouzian 2.161 2.709 65.4 2.858 (0.665) (0.880) (12.5; 22.1) (0.522) (1) Education Level (2) Age (3) Ethnicitya (4) Income Level Sven Janssens 2.526 3.007 98.3 3.137 (0.563) (0.815) (0.3; 1.4) (0.547) Joris Janssens 2.587 3.259 96.7 3.206 (0.525) (0.959) (0.9; 2.4) (0.464) Kenny Janssens 2.079 2.303 98.6 2.742 (0.664) (0.754) (0.3; 1) (0.672) Marc Janssens 2.483 4.157 98.4 3.294 (0.595) (0.895) (0.3; 1.3) (0.610) Thomas Janssens 2.682 2.794 98.6 3.244 (0.497) (0.855) (0.3; 1) (0.527) Sven Maes 2.610 2.927 99.1 3.136 (0.533) (0.750) (0.0; 0.9) (0.554) Joris Maes 2.602 3.272 98.1 3.237 (0.536) (1.084) (0.4; 1.5) (0.532) Kenny Maes 2.230 2.345 97.7 2.723 (0.583) (0.783) (0.7; 1.6) (0.612) Marc Maes 2.478 4.083 98.7 3.184 (0.577) (0.863) (0.0; 1.3) (0.568) Thomas Maes 2.673 2.768 97.6 3.280 (0.541) (0.832) (0.3; 2.1) (0.543) Mohamed El Makrini 2.136 2.803 74.5 2.863 (0.657) (0.500) (16.3; 9.1) (0.568) Ahmed El Makrini 2.167 2.903 71.6 2.902 (0.626) (0.917) (16.2; 11.6) (0.696) Rachid El Makrini 2.266 2.658 73.2 2.875 (0.645) (0.834) (14.9; 12.0) 1(0.534) Hassan El Makrini 2.272 2.804 71.1 2.919 (0.658) (0.924) (15.8; 13.2) (0.553) Youssef El Makrini 2.199 2.678 67.3 2.963 (0.627) (0.899) (21.8; 10.9) (0.635) Mohamed Bouzian 2.194 2.952 70.2 2.941 (0.670) (0.949) (17.8; 12.0) (0.637) Ahmed Bouzian 2.174 2.896 57.6 2.885 (0.634) (0.918) (19.8; 22.7) (0.564) Rachid Bouzian 2.292 2.729 60.2 2.952 (0.634) (0.876) (21.8; 18.0) (0.611) Hassan Bouzian 2.138 2.818 62.0 2.876 (0.630) (0.890) (16.3; 21.6) (0.554) Youssef Bouzian 2.161 2.709 65.4 2.858 (0.665) (0.880) (12.5; 22.1) (0.522) Note: Standard deviations in parentheses. aPercentage that predicted the intended ethnicity (Belgian/Moroccan). In parentheses, the percentages of the other nationality guess (i.e., Belgian/Moroccan, or Other). Missing percentage points represent missing values. View Large Table A1. Means of Pretest (1) Education Level (2) Age (3) Ethnicitya (4) Income Level Sven Janssens 2.526 3.007 98.3 3.137 (0.563) (0.815) (0.3; 1.4) (0.547) Joris Janssens 2.587 3.259 96.7 3.206 (0.525) (0.959) (0.9; 2.4) (0.464) Kenny Janssens 2.079 2.303 98.6 2.742 (0.664) (0.754) (0.3; 1) (0.672) Marc Janssens 2.483 4.157 98.4 3.294 (0.595) (0.895) (0.3; 1.3) (0.610) Thomas Janssens 2.682 2.794 98.6 3.244 (0.497) (0.855) (0.3; 1) (0.527) Sven Maes 2.610 2.927 99.1 3.136 (0.533) (0.750) (0.0; 0.9) (0.554) Joris Maes 2.602 3.272 98.1 3.237 (0.536) (1.084) (0.4; 1.5) (0.532) Kenny Maes 2.230 2.345 97.7 2.723 (0.583) (0.783) (0.7; 1.6) (0.612) Marc Maes 2.478 4.083 98.7 3.184 (0.577) (0.863) (0.0; 1.3) (0.568) Thomas Maes 2.673 2.768 97.6 3.280 (0.541) (0.832) (0.3; 2.1) (0.543) Mohamed El Makrini 2.136 2.803 74.5 2.863 (0.657) (0.500) (16.3; 9.1) (0.568) Ahmed El Makrini 2.167 2.903 71.6 2.902 (0.626) (0.917) (16.2; 11.6) (0.696) Rachid El Makrini 2.266 2.658 73.2 2.875 (0.645) (0.834) (14.9; 12.0) 1(0.534) Hassan El Makrini 2.272 2.804 71.1 2.919 (0.658) (0.924) (15.8; 13.2) (0.553) Youssef El Makrini 2.199 2.678 67.3 2.963 (0.627) (0.899) (21.8; 10.9) (0.635) Mohamed Bouzian 2.194 2.952 70.2 2.941 (0.670) (0.949) (17.8; 12.0) (0.637) Ahmed Bouzian 2.174 2.896 57.6 2.885 (0.634) (0.918) (19.8; 22.7) (0.564) Rachid Bouzian 2.292 2.729 60.2 2.952 (0.634) (0.876) (21.8; 18.0) (0.611) Hassan Bouzian 2.138 2.818 62.0 2.876 (0.630) (0.890) (16.3; 21.6) (0.554) Youssef Bouzian 2.161 2.709 65.4 2.858 (0.665) (0.880) (12.5; 22.1) (0.522) (1) Education Level (2) Age (3) Ethnicitya (4) Income Level Sven Janssens 2.526 3.007 98.3 3.137 (0.563) (0.815) (0.3; 1.4) (0.547) Joris Janssens 2.587 3.259 96.7 3.206 (0.525) (0.959) (0.9; 2.4) (0.464) Kenny Janssens 2.079 2.303 98.6 2.742 (0.664) (0.754) (0.3; 1) (0.672) Marc Janssens 2.483 4.157 98.4 3.294 (0.595) (0.895) (0.3; 1.3) (0.610) Thomas Janssens 2.682 2.794 98.6 3.244 (0.497) (0.855) (0.3; 1) (0.527) Sven Maes 2.610 2.927 99.1 3.136 (0.533) (0.750) (0.0; 0.9) (0.554) Joris Maes 2.602 3.272 98.1 3.237 (0.536) (1.084) (0.4; 1.5) (0.532) Kenny Maes 2.230 2.345 97.7 2.723 (0.583) (0.783) (0.7; 1.6) (0.612) Marc Maes 2.478 4.083 98.7 3.184 (0.577) (0.863) (0.0; 1.3) (0.568) Thomas Maes 2.673 2.768 97.6 3.280 (0.541) (0.832) (0.3; 2.1) (0.543) Mohamed El Makrini 2.136 2.803 74.5 2.863 (0.657) (0.500) (16.3; 9.1) (0.568) Ahmed El Makrini 2.167 2.903 71.6 2.902 (0.626) (0.917) (16.2; 11.6) (0.696) Rachid El Makrini 2.266 2.658 73.2 2.875 (0.645) (0.834) (14.9; 12.0) 1(0.534) Hassan El Makrini 2.272 2.804 71.1 2.919 (0.658) (0.924) (15.8; 13.2) (0.553) Youssef El Makrini 2.199 2.678 67.3 2.963 (0.627) (0.899) (21.8; 10.9) (0.635) Mohamed Bouzian 2.194 2.952 70.2 2.941 (0.670) (0.949) (17.8; 12.0) (0.637) Ahmed Bouzian 2.174 2.896 57.6 2.885 (0.634) (0.918) (19.8; 22.7) (0.564) Rachid Bouzian 2.292 2.729 60.2 2.952 (0.634) (0.876) (21.8; 18.0) (0.611) Hassan Bouzian 2.138 2.818 62.0 2.876 (0.630) (0.890) (16.3; 21.6) (0.554) Youssef Bouzian 2.161 2.709 65.4 2.858 (0.665) (0.880) (12.5; 22.1) (0.522) Note: Standard deviations in parentheses. aPercentage that predicted the intended ethnicity (Belgian/Moroccan). In parentheses, the percentages of the other nationality guess (i.e., Belgian/Moroccan, or Other). Missing percentage points represent missing values. View Large Table A2. Categorization of Addressed Facilities According to their Legal Status (1) (2) (3) (4) (5) Abbreviation of Belgian Legal Status Belgian Legal Status Translation of Belgian Legal Status Ownership Amount of Facilities APB Autonoom Provinciaal Bedrijf Autonomous Provincial Company Public 1 AV Autonome Vereniging Autonomous Association Public 4 EVA Extern Verzelfstandigd Agentschap External Independent Agency Public 4 OCMW Openbaar Centrum voor Maatschappelijk Welzijn Public Welfare Agency Public 205 BV Besloten Vennootschap Private Company Private 1 BVBA Besloten Vennootschap met Beperkte Aansprakelijkheid Private Company with Limited Liability Private 45 CV Commanditaire Vennootschap Limited Partnership Private 1 CVBA CV met Beperkte Aansprakelijkheid Limited Partnership with Limited Liability Private 2 GCV Gewone Commanditaire Vennootschap Normal Limited Partnership Private 3 NP Natuurlijke Persoon Natural Person Private 5 NV Naamloze Vennootschap Limited Liability Company Private 60 Stichting Stichting Foundation Private 1 VZW Vereniging Zonder Winstoogmerk Association without Lucrative Purpose Nonprofit 387 (1) (2) (3) (4) (5) Abbreviation of Belgian Legal Status Belgian Legal Status Translation of Belgian Legal Status Ownership Amount of Facilities APB Autonoom Provinciaal Bedrijf Autonomous Provincial Company Public 1 AV Autonome Vereniging Autonomous Association Public 4 EVA Extern Verzelfstandigd Agentschap External Independent Agency Public 4 OCMW Openbaar Centrum voor Maatschappelijk Welzijn Public Welfare Agency Public 205 BV Besloten Vennootschap Private Company Private 1 BVBA Besloten Vennootschap met Beperkte Aansprakelijkheid Private Company with Limited Liability Private 45 CV Commanditaire Vennootschap Limited Partnership Private 1 CVBA CV met Beperkte Aansprakelijkheid Limited Partnership with Limited Liability Private 2 GCV Gewone Commanditaire Vennootschap Normal Limited Partnership Private 3 NP Natuurlijke Persoon Natural Person Private 5 NV Naamloze Vennootschap Limited Liability Company Private 60 Stichting Stichting Foundation Private 1 VZW Vereniging Zonder Winstoogmerk Association without Lucrative Purpose Nonprofit 387 View Large Table A2. Categorization of Addressed Facilities According to their Legal Status (1) (2) (3) (4) (5) Abbreviation of Belgian Legal Status Belgian Legal Status Translation of Belgian Legal Status Ownership Amount of Facilities APB Autonoom Provinciaal Bedrijf Autonomous Provincial Company Public 1 AV Autonome Vereniging Autonomous Association Public 4 EVA Extern Verzelfstandigd Agentschap External Independent Agency Public 4 OCMW Openbaar Centrum voor Maatschappelijk Welzijn Public Welfare Agency Public 205 BV Besloten Vennootschap Private Company Private 1 BVBA Besloten Vennootschap met Beperkte Aansprakelijkheid Private Company with Limited Liability Private 45 CV Commanditaire Vennootschap Limited Partnership Private 1 CVBA CV met Beperkte Aansprakelijkheid Limited Partnership with Limited Liability Private 2 GCV Gewone Commanditaire Vennootschap Normal Limited Partnership Private 3 NP Natuurlijke Persoon Natural Person Private 5 NV Naamloze Vennootschap Limited Liability Company Private 60 Stichting Stichting Foundation Private 1 VZW Vereniging Zonder Winstoogmerk Association without Lucrative Purpose Nonprofit 387 (1) (2) (3) (4) (5) Abbreviation of Belgian Legal Status Belgian Legal Status Translation of Belgian Legal Status Ownership Amount of Facilities APB Autonoom Provinciaal Bedrijf Autonomous Provincial Company Public 1 AV Autonome Vereniging Autonomous Association Public 4 EVA Extern Verzelfstandigd Agentschap External Independent Agency Public 4 OCMW Openbaar Centrum voor Maatschappelijk Welzijn Public Welfare Agency Public 205 BV Besloten Vennootschap Private Company Private 1 BVBA Besloten Vennootschap met Beperkte Aansprakelijkheid Private Company with Limited Liability Private 45 CV Commanditaire Vennootschap Limited Partnership Private 1 CVBA CV met Beperkte Aansprakelijkheid Limited Partnership with Limited Liability Private 2 GCV Gewone Commanditaire Vennootschap Normal Limited Partnership Private 3 NP Natuurlijke Persoon Natural Person Private 5 NV Naamloze Vennootschap Limited Liability Company Private 60 Stichting Stichting Foundation Private 1 VZW Vereniging Zonder Winstoogmerk Association without Lucrative Purpose Nonprofit 387 View Large References Abbott , Andrew . 1988 . The system of professions: An essay on the division of expert labor . Chicago : University of Chicago Press . Adman , Per , and Hanna Jansson . 2016 . A field experiment on ethnic discrimination among local Swedish public officials . Local Government Studies 43 : 44 – 63 . Google Scholar CrossRef Search ADS Agentschap Zorg and Gezondheid . 2017 . Adressen woonzorgcentra en rust -en verzorgingstehuizen . https://www.zorg-en-gezondheid.be/adressen-van-woonzorgcentra-en-rust-en-verzorgingstehuizen (accessed March 30, 2017 ). Ahmed , Ali M. , and Mats Hammarstedt . 2008 . Discrimination in the rental housing market. A field experiment on the internet . Journal of Urban Economics 64 : 362 – 72 . Google Scholar CrossRef Search ADS Altonji , Joseph G. , and Charles R. Pierret . 2001 . Employer learning and statistical discrimination . The Quarterly Journal of Economics 116 : 313 – 50 . Google Scholar CrossRef Search ADS Amirkhanyan , Anna , Hyun Joon Kim , and Kristina Lambright . 2008 . Does the public sector outperform the nonprofit and for-profit sectors? Evidence from a national panel study on nursing home quality and access . Journal of Policy Analysis and Management 27 : 326 – 53 . Google Scholar CrossRef Search ADS PubMed Arrijn , Peter , Serge Feld , and André Nayer . 1998 . Discrimination in access to employment on grounds of foreign origin: The case of Belgium . Geneva : International Labour Office, Conditions of Work Branch . Arrow , Kenneth J . 1973 . The theory of discrimination . In Orley Ashenfelter, Albert Rees, eds. Discrimination labor markets , 3 – 33 . Princeton : Princeton Univ. Press . Google Scholar CrossRef Search ADS Ayres , Ian , and Peter Siegelman . 1995 . Race and gender in bargaining for a new car . American Economic Review 85 : 304 – 21 . Baert , Stijn , Bart Cockx , Niels Gheyle , and Cora Vandamme . 2015 . Is there less discrimination in occupations where recruitment is difficult ? ILR Review 68 : 467 – 500 . Google Scholar CrossRef Search ADS Becker , Gary S . 1971 . The economics of discrimination . Chicago : The University of Chicago Press . Google Scholar CrossRef Search ADS Bertrand , Marianne , Dolly Chugh , and Sendhil Mullainathan . 2005 . Implicit Discrimination. American Economic Review 95 : 94 – 8 . CrossRef Search ADS Bertrand , Marianne , and Sendhil Mullainathan . 2004 . Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination . American Economic Review 94 : 991 – 1013 . Google Scholar CrossRef Search ADS Bevan , Gwyn , and Christopher Hood . 2006 . What’s measured is what matters: Targets and gaming in the British health care sector . Public Administration 84 : 517 – 38 . Google Scholar CrossRef Search ADS Bobo , Lawrence . 1983 . Whites’ opposition to busing: Symbolic racism or realistic group conflict ? Journal of Personality and Social Psychology 45 : 1196 – 210 . Google Scholar CrossRef Search ADS Bohte , John , and Kenneth J. Meier . 2000 . Goal displacement: Assessing the motivation for organisational cheating . Public Administration Review 60 : 173 – 82 . Google Scholar CrossRef Search ADS Bozeman , Barry , and Stuart Bretschneider . 1994 . The “publicness puzzle” in organization theory: A test of alternative explanations of differences between public and private organizations . Journal of Public Administration Research and Theory 4 : 197 – 23 . Broockman , David E . 2013 . Black politicians are more intrinsically motivated to advance blacks’ interests: A field experiment manipulating political incentives . American Journal of Political Science 57 : 521 – 36 . Google Scholar CrossRef Search ADS Brodkin , Evelyn . 1997 . Inside the welfare contract: Discretion and accountability in state welfare administration . Social Service Review 71 : 1 – 33 . Google Scholar CrossRef Search ADS Brodkin , Evelyn , and Michael Lipsky . 1983 . Quality control in AFDC as an administrative strategy . Social Service Review 57 : 1 – 34 . Google Scholar CrossRef Search ADS Butler , Daniel M. , and David E. Broockman . 2011 . Do politicians racially discriminate against constituents? A field experiment on state legislators . American Journal of Political Science 55 : 463 – 77 . Google Scholar CrossRef Search ADS Chugh , Dolly . 2004 . Societal and managerial implications of implicit social cognition: Why milliseconds matter . Social Justice Research 17 : 203 – 22 . Google Scholar CrossRef Search ADS Clifton , Judith , Daniel Díaz-Fuentes , Marcos Fernández-Gutiérrez , and Julio Revuelta . 2011 . Is market-oriented reform producing a ‘two-track’ Europe? Evidence from electricity and telecommunications . Annals of Public and Cooperative Economics 82 : 495 – 13 . Google Scholar CrossRef Search ADS Clifton , Judith , Daniel Diaz Fuentes , Marcos Fernandez-Guttierrez , Oliver James , Sebastian Jilke , and Steven Van de Walle . ( 2012 ): Regulatory reform for services of general interest and trends in citizen satisfaction . COCOPS Project Report. Rotterdam : Erasmus Univ. Rotterdam . Costa , Mia . 2017 . How responsive are political elites? A meta-analysis of experiments on public officials . Journal of Experimental Political Science 4 : 241 – 54 . Google Scholar CrossRef Search ADS Dabney , Dean A. , Laura Dugan , Volkan Topalli , and Richard C. Hollinger . 2006 . The impact of implicit stereotyping on offender profiling. Unexpected results from an observational study of shoplifting . Criminal Justice and Behavior 33 : 646 – 74 . Google Scholar CrossRef Search ADS Dan , Sorin , and Rhys Andrews . 2015 . Market-type mechanisms and public service equity: A review of experiences in European public services . Public Organization Review 34 : 543 – 59 . Dias , Janice J. , and Steven Maynard-Moody . 2007 . For-profit welfare: Contracts, conflicts, and the performance paradox . Journal of Public Administration Research and Theory 17 : 189 – 11 . Google Scholar CrossRef Search ADS Einstein , Katherine L. , and David M. Glick . 2016 . Does race affect access to government services? An experiment exploring street-level bureaucrats and access to public housing . American Journal of Political Science 61 : 100 – 16 . Google Scholar CrossRef Search ADS Elchardus , Mark , and Jessy Siongers . 2011 . First names as collective identifiers: An empirical analysis of the social meanings of first names . Cultural Sociology 5 : 403 – 22 . Google Scholar CrossRef Search ADS Ellis , Randall P . 1998 . Creaming, skimping and dumping: Provider competition on the intensive and extensive margins . Journal of Health Economics 17 : 537 – 55 . Google Scholar CrossRef Search ADS PubMed Emker , Johannes , and Anselm Rink . 2017 . Multiple dimensions of bureaucratic discrimination: Evidence from German welfare officers . American Journal of Political Science 61 : 786 – 803 . Google Scholar CrossRef Search ADS Epp , Charles R. , Steven Maynard-Moody , and Donald P. Haider-Markel . 2014 . How police stops define race and citizenships . Chicago : The Univ. of Chicago Press . Google Scholar CrossRef Search ADS Forscher , Patrick S. , Calvin K. Lai , Jordan R. Axt , Charles R. Ebersole , Michelle Herman , Patricia G. Devine , and Brian A. Nosek . 2017 . A meta-analysis of change in implicit bias . Retrieved from Psyarxiv.com/dv8tu . Frederickson , George H . 2010 . Social equity and public administration: Origins, developments, and applications . New York : M.E. Sharpe . Gallup . 2014 . Gallup review: Black and white differences in views on race . http://www.gallup.com/poll/180107/gallup-review-black-white-differences-views-race.aspx (accessed March 30, 2017 ). Giulietti , Corrado , Mirco Tonin , and Michael Vlassopoulos . 2017 . Racial discrimination in local public services: A field experiment in the US . Journal of the European Economic Association, online first. doi: 10.1093/jeea/jvx045. Gooden , Susan T . 2015 . PAR’s social equity footprint . Public Administration Review 75 : 372 – 81 . Google Scholar CrossRef Search ADS Greenwald , Anthony G. , and Linda Hamilton Krieger . 2006 . Implicit bias: Scientific foundations . California Law Review 94 : 945 – 67 . Google Scholar CrossRef Search ADS Greenwald , Anthony G. , Debbie E. McGhee , and Jordan L. K. Schwartz . 1998 . Measuring individual differences in implicit cognition: The implicit association test . Journal of Personality and Social Psychology 74 : 1464 – 80 . Google Scholar CrossRef Search ADS PubMed Grohs , Stephan , Christin Adam , and Christoph Knill . 2015 . Are some citizens more equal than others? Evidence from a field experiment . Public Administration Review 76 : 155 – 64 . Google Scholar CrossRef Search ADS Guryan , Jonathan , and Kerwin Kofi Charles . 2013 . Taste-based or statistical discrimination: The economics of discrimination returns to its roots . The Economic Journal 123 : F417 – 32 . Google Scholar CrossRef Search ADS Harris , Ricci , Martin Tobias , Mona Jeffreys , Kiri Waldegrave , Saffron Karlsen , and James Nazroo . 2006 . Racism and health: The relationship between experience of racial discrimination and health in New Zealand . Social Science & Medicine 63 : 1428 – 41 . Google Scholar CrossRef Search ADS Heinrich , Carolyn J . 1999 . Do government bureaucrats make effective use of performance management information ? Journal of Public Administration Research and Theory 9 : 363 – 93 . Google Scholar CrossRef Search ADS ———. 2016 . The bite of administrative burden: A theoretical and empirical investigation . Journal of Public Administration Research and Theory 3 : 403 – 20 . Heinrich , Carolyn J. , and Gerald Marschke . 2010 . Incentives and their dynamics in public sector performance management systems . Journal of Policy Analysis and Management 29 : 183 – 208 . Google Scholar CrossRef Search ADS Galster , George , Douglas Wissoker , and Wendy Zimmermann . 1998 . Testing for discrimination in home insurance . Washington, DC : Urban Institute Press . Hood , Christopher . 2006 . Gaming in target world: The targets approach to managing British public services . Public Administration Review 66 : 515 – 21 . Google Scholar CrossRef Search ADS Jackson , Jay W . 1993 . Realistic group conflict theory: A review and evaluation of the theoretical and empirical literature . The Psychological Record 43 : 395 – 13 . Jilke , Sebastian . 2015 . Essays on the microfoundations of competition and choice in public service delivery . PhD diss. Rotterdam: Erasmus University Rotterdam. Kaas , Leo , and Christian Manger . 2011 . Ethnic discrimination in Germany’s labour market: A field experiment . German Economic Review 13 : 1 – 20 . Google Scholar CrossRef Search ADS Keiser , Lael R. , and Joe Soss . 1998 . With good cause: Bureaucratic discretion and the politics of child support enforcement . American Journal of Political Science 42 : 1133 – 56 . Google Scholar CrossRef Search ADS Kerpershoek , Emiel , Martijn Groenleer , and Hans de Bruijn . 2016 . Unintended responses to performance management in Dutch hospital care: Bringing together the managerial and professional perspectives . Public Management Review 18 : 417 – 36 . Google Scholar CrossRef Search ADS Kjeldsen , Anne Mette , and Christian Botcher Jacobsen . 2013 . Public service motivation and employment sector: Attraction or socialization ? Journal of Public Administration Research and Theory 23 : 899 – 26 . Google Scholar CrossRef Search ADS Koning , Pierre , and Carolyn J. Heinrich . 2013 . Cream-skimming, parking and other intended and unintended effects of high-powered, performance-based contracts . Journal of Policy Analysis and Management 32 : 461 – 83 . Google Scholar CrossRef Search ADS Krippendorff , Klaus . 2013 . Content analysis: An introduction to its methodology . Thousand Oaks : SAGE . Kugelmass , Heather . 2016 . “Sorry, I’m not accepting new patients”: An audit study of access to mental health care . Journal of Health and Social Behavior 57 : 168 – 83 . Google Scholar CrossRef Search ADS PubMed Lacireno-Paquet , Natalie , Thomas T. Holyoke , Michele Moser , and Jeffrey R. Henig . 2002 . Creaming versus cropping: Charter school enrollment practices in response to market incentives . Educational Evaluation and Policy Analysis 24 : 145 – 58 . Google Scholar CrossRef Search ADS Le Grand , Julian , and Will Bartlett . 1993 . Quasi-markets and social policy . London : Macmillan . Google Scholar CrossRef Search ADS Lipsky , Michael . 1980 . Street-level bureaucracy: Dilemmas of the individual in public services . New York : Russell Sage Foundation . ———. 1984 . Bureaucratic disentitlement in social welfare programs . Social Service Review 58 : 3 – 27 . CrossRef Search ADS Maynard-Moody , Steven , and Michael Musheno . 2000 . State agent or citizen agent: Two narratives of discretion . Journal of Public Administration Research & Theory 10 : 329 . Google Scholar CrossRef Search ADS ———. 2003 . Cops, Teachers, Counselors: Stories from the front lines of public service . Michigan : Univ. of Michigan Press . Meier , Kenneth J. , and Laurence J. O’Toole . 2011 . Comparing public and private management: Theoretical expectations . Journal of Public Administration Research and Theory 21 ( Suppl 3 ): i283 – 99 . Google Scholar CrossRef Search ADS Moynihan , Donald P. , Pamela Herd , and Hope Harvey . 2015 . Administrative burden: Learning, psychological, and compliance costs in citizen-state interactions . Journal of Public Administration Research and Theory 25 : 43 – 69 . Google Scholar CrossRef Search ADS Moynihan , Donald P. , Pamela Herd , and Elizabeth Ribgy . 2013 . Policymaking by other means. Do states use administrative barriers to limit access to Medicaid ? Administration & Society 48 : 497 – 24 . Google Scholar CrossRef Search ADS National Research Council . 2004 . Measuring racial discrimination . Washington, DC : The National Academies Press . Oswald , Frederick L. , Gregory Mitchell , Hart Blanton , and Philip E. Tetlock . 2013 . Predicting ethnic and racial discrimination: A meta-analysis of IAT criterion studies . Journal of Personality and Social Psychology 105 : 171 – 92 . Google Scholar CrossRef Search ADS PubMed Pacolet , Jozef , and Annelies De Coninck . 2015 . Financiering van de residentiële ouderenzorg: Het perspectief van de voorzieningen . KU Leuven met steun van de Vlaamse Overheid . Pager , Devah . 2007 . The use of field experiments for studies of employment discrimination: Contributions, critiques, and directions for the future . Annals of the American Academy of Political and Social Sciences 609 : 104 – 33 . Google Scholar CrossRef Search ADS Pager , Devah , Bruce Western , and Bart Bonikowski . 2009 . Discrimination in a low-wage labor market: A field experiment . American Sociological Review 74 : 777 – 99 . Google Scholar CrossRef Search ADS PubMed Perry , James L. , Annie Hondeghem , and Lois Recascino Wise . 2010 . Revisiting the motivational bases of public service: Twenty years of research and an agenda for the future . Public Administration Review 70 : 681 – 90 . Google Scholar CrossRef Search ADS Phelps , Edmund S . 1972 . The statistical theory of racism and sexism . American Economic Review 62 : 659 – 61 . Pollitt , Christopher . 2013 . The logics of performance management . Evaluation 19 : 346 – 63 . Google Scholar CrossRef Search ADS Quillian , Lincoln , Devah Pager , Ole Hexel , and Arnfinn Midboen . 2017 . Meta-analysis of field experiments shows no change in racial discrimination in hiring over time . Proceedings of the National Academy of Sciences of the United States of America 114 : 10870 – 75 . Google Scholar CrossRef Search ADS PubMed Radin , Beryl A . 2006 . Challenging the performance movement: Accountability, complexity, and democratic values . Washington, DC : Georgetown Univ. Press . Rainey , Hal G. , and Barry Bozeman . 2000 . Comparing public and private organizations: Empirical research and the power of the a priori . Journal of Public Administration Research and Theory 10 : 447 – 70 . Google Scholar CrossRef Search ADS Ridley , Stanley E. , James A. Bayton , and Janice H. Quttz . 1989 . Taxi services in the District of Columbia: Is it influenced by patron’s race and destination . Washington, DC : Washington Lawyers’ Comm. Civil Rights Law . Savas , Emanuel S . 2000 . Privatization and public–private partnerships . New York : Seven Bridges Press . Schneider , Mark , Paul Teske , and Melissa Marschall . 2002 . Choosing schools: Consumer choice and the quality of American schools . Princeton : Princeton Univ. Press . Schulman , Kevin A. , A. Berlin Jesse , William Harless , Jon F. Kerner , Shyrl Sistrunk , Bernard J. Gersh , Ross Dubé , et al. 1999 . The effect of race and sex on physicians’ recommendations for cardiac catheterization . The New England Journal of Medicine 340 : 618 – 26 . Google Scholar CrossRef Search ADS PubMed Scott , Patrick G . 1997 . Assessing determinants of bureaucratic discretion: An experiment in street-level decision making . Journal of Public Administration Research and Theory 35 : 35 – 57 . Google Scholar CrossRef Search ADS Socialistische Mutualiteiten. 2016. Rusthuisbarometer: Analyse bewonersfacturen in woonzorgcentra. Brussels: Socialistische Mutualiteiten. Soss , Joe , Richard C. Fording and Sanford F. Schram . 2011 . The organization of discipline: From performance management to perversity and punishment . Journal of Public Administration Research and Theory 21 ( Suppl 2 ): i203 – 32 . Google Scholar CrossRef Search ADS Statistics Belgium . 2017 . Meest voorkomende familienamen in 2016—België en Gewesten (Top 10) . http://statbel.fgov.be/nl/statistieken/cijfers/bevolking/namen/fam/ (accessed March 30, 2017 ). Tummers , Lars , Victor Bekkers , Evelien Vink , and Michael Musheno . 2015 . Coping during public service delivery: A conceptualization and systematic review of the literature . Journal of Public Administration Review and Theory 25 : 1099 – 126 . Google Scholar CrossRef Search ADS Turner , Margery A. , and Stephen L. Ross . 2003 . Discrimination in metropolitan housing markets phase II: Asians and Pacific islanders . Washington, DC : The Urban Institute . Turner , Margery A. , Stephen L. Ross , George C. Galster , and John Yinger . 2002 . Discrimination in metropolitan housing markets: National results from phase I HDS 2000 . Washington, DC : The Urban Institute . Van den Bosch , Karel , Peter Willemé , Joanna Geerts , Jef Breda , Stéphanie Peeters , Stefaan Van De Sande , France Vrijens , Carine Van de Voorde , and Sabine Stordeur . ( 2011 ). Toekomstige behoefte aan residentiële ouderenzorg in België: Projecties 2011–2025 . Federaal Kenniscentrum voor de Gezondheidszorg . https://kce.fgov.be/sites/default/files/page_documents/KCE_167A_residentiele_zorg_in_Belgie.pdf (accessed March 30, 2017 ). Van den Broeck , Katleen , Kristof Heylen , and Sien Winters . 2014 . Discriminatie en Selectie op de Private Huurmarkt in België . Brussels: Hoger Instituut voor de Arbeid . Van den Broucke Sarah , Jo Noppe , Karen Stuyck , Philippe Buysschaert , Gerlinde Doyen , and Johan Wets . 2016 . Vlaamse migratie en integratiemonitor 2015 . Brussel : Vlaamse Overheid . Weinick , Robin M. , Samuel H. Zuvekas , and Joel W. Cohen . 2000 . Racial and ethnic differences in access to and use of health care services . Medical Care Research and Review 57 ( Suppl 1 ): 36 – 54 . Google Scholar CrossRef Search ADS PubMed Weisbrod , Burton A . 1988 . The nonprofit economy . Cambridge : Harvard Univ. Press . White , Ariel L. , Noah L. Nathan , and Julie K. Faller . 2015 . What do I need to vote? Bureaucratic discretion and discrimination of local election officials . American Political Science Review 109 : 129 – 40 . Google Scholar CrossRef Search ADS Willemé , Peter , Joanna Geerts , Bea Cantillon , and Ninke Mussche . 2012 . Long-term care financing in Belgium . In Financing long-term care in Europe , 300 – 21 . UK : Palgrave Macmillan . Google Scholar CrossRef Search ADS Yinger , John . 1995 . Closed doors, opportunities lost: The continuing costs of housing discrimination . New York : Russell Sage Foundation . Footnotes 1 The determination of the level of publicness of a service provider is not straightforward. In an important contribution to the field, Bozeman and Bretschneider (1994) distinguished between a core approach based on legal ownership and a dimensional approach based on criteria related to public funding, public value, and public communication networks. Bozeman and Bretschneider (1994) argue that the core and dimensional approaches are both useful, depending on the research design. We look only at ownership in our study given that the nondistribution constraint of public facilities creates an incentive structure that is fundamentally different in comparison to private facilities. 2 A fourth type of discrimination comes from realistic group conflict theory (Bobo 1983; Jackson 1993). It suggests that the discrimination of outgroup members stems mainly from competition over limited resources. Here, it is assumed that Belgian frontline workers (as ingroup members) would try to disproportionally distribute elderly care places to fellow Belgians and not to ethnic minorities (i.e., outgroup members), given the perceived competition with minorities over a limited number of places. However, this mechanism seems rather unrealistic given the low societal valence of elderly care places, which are in principle available to everybody. 3 Nonprofits are more profitable because they started to operate elderly care facilities decades before the private facilities. As a result, many buildings are written off in their accounts. Consequently, lower capital duties need to be paid. 4 The nonprofit sector is not included in this study, because the incentive structure of nonprofits is less straightforward than in the public and private sectors. The legal person of nonprofits prohibits personal gain. Profits cannot be distributed to shareholders. However, profits can be retained within organizations for future investment. Moreover, private equity providers increasingly operate a nonprofit as a shell company for accounting reasons. Therefore, the incentive structure in the nonprofit sector is more diffuse. The public and private sector are clearer: potential profits end up in the general municipal budget or in the hand of owners/shareholders. 5 Education level: “Which degree does this person have?” (1) no degree/elementary school, (2) secondary school, (3) higher studies/university. Age: “Which age group does this person belong to?”: (1) 20 years or younger, (2) 21–30 years, (3) 31–40 years, (4) 41–50 years, (5) 51–60 years, (6) 61 years or older. Ethnicity: “Which ethnicity does this person have?”: (1) Belgian, (2) Moroccan, (3) other. Income “How wealthy is this person?”: (1) poor, (2) below average, (3) average, (4) above average, (5) wealthy. 6 The email was originally written in Flemish (see box 2 in the Appendix). 7 Some of the responses that we received were untraceable back to the original request or corresponding elderly care home (including seven that were sent to Kenny Maes and seven that were sent to Mohamed El Makrini); these responses were coded as “n/a.” In four instances, a facility sent an email to both Mohamed and Kenny, even though we did not send requests from both senders to any elderly care home. This was probably due to behind-the-scenes organizational and managerial structures at different homes, as a few of the facilities belong to a “mother” organization (which may have been forwarded the requests from individual facilities). Nonetheless, judging from the contents of the responses, we have no reason to believe that the senders were suspicious. Moreover, excluding these responses did not alter the results in any meaningful way. 8 Autoreplies and forwarded messages were coded as no replies. Nevertheless, when a direct message was later received, it was coded as a response. 9 We estimated a Krippendorf’s alpha (Krippendorff 2013) of 0.56 for information comprehensiveness, and 0.28 for friendliness ratings. 10 We also estimated all of the models presented in this section only for the subset of emails that generated responses. The results are identical to those presented in the text, except that the ATE of the combined public and private facilities is statistically significant. We focus on the more conservative approach of taking all experimental units into account. 11 One alternative interpretation for this finding would be that the ethnic composition of public and private nursing homes differs substantially. A more diverse nursing home, in turn, would have lower associated costs with accepting an additional Maghrebian client. While official figures for client composition on Flemish elderly care homes are not available to test this proposition, we used the ethnic composition within municipalities where nursing homes are situated as a crude proxy, finding no systematic differences between public and private facilities’ localities. © The Author(s) 2018. Published by Oxford University Press on behalf of the Public Management Research Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Journal of Public Administration Research and TheoryOxford University Press

Published: Mar 15, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off