Punishment on the Frontlines of Public Service Delivery: Client Ethnicity and Caseworker Sanctioning Decisions in a Scandinavian Welfare State

Punishment on the Frontlines of Public Service Delivery: Client Ethnicity and Caseworker... Abstract Many public welfare programs give public employees discretionary authority to dispense sanctions when clients do not follow or comply with the policies and procedures required for receiving welfare benefits. Yet research also shows that public employees’ use of discretion in decision-making that affects clients can occasionally be marked by racial biases and disparities. Drawing on the Racial Classification Model (RCM) for a theoretical model, this article examines how client ethnicity shapes public employees’ decisions to sanction clients. Using Danish employment agencies as our empirical setting, we present findings from two complementary studies. Study 1 uses nationwide administrative data. Examining sanctioning activity at the employment agency-level, we find that agencies with a larger percentage of clients being non-Western immigrants or their descendants impose a greater overall number of sanctions and dispense them with greater frequency. Study 2 uses survey experimental data to build on this finding. Addressing concerns about internal validity and a need for analyses at the individual employee-level, we present survey experimental evidence that employment agency caseworkers are more likely to recommend sanctions for ethnic minority (Middle-Eastern origin) clients than for ethnic majority (Danish origin) clients. Moreover, we investigate how three caseworker characteristics—ethnicity, gender, and work experience—condition the relationship between client ethnicity and caseworkers’ decisions to sanction clients. Although we find no moderation effects for ethnicity or gender, work experience appears to diminish the influence of client ethnicity on the caseworkers’ sanctioning decisions. Overall, our studies support the likelihood that ethnic minority clients will be punished more often for policy infractions than ethnic majority clients—and that caseworker work experience mitigates part of this bias. Introduction In the wake of neoliberal public policy reforms, public welfare programs worldwide have become more directive in setting behavioral expectations for clients and more punitive in responding to client noncompliance (Hood and Peters 2004; Kettl 2005). Social welfare programs use a range of incentives, monitoring mechanisms, and restrictive regulations and rules aimed at modifying client behaviors and ensuring that client groups practice appropriate behaviors (Danziger and Seefeldt 2003; Pavetti, Derr, and Hesketh 2003; Schram 2006). Many of these programs involve a system for dispensing sanctions—penalties that reduce or terminate welfare benefits in response to client infractions of policy, that is, noncompliance (Pavetti, Derr, and Hesketh 2003; Schram et al. 2009). However, the dispensing of sanctions is not an automated process devoid of personal cognitions or beliefs. Although program rules may allow for and prescribe sanctions under certain conditions, rules are not self-executing (Blau 1955; Merton 1957): They require judgments from frontline workers in their application (Hupe and Hill 2007, p. 281). In many areas of public welfare administration, public employees have substantial discretionary authority to evaluate and decide whether and, if so, when clients should be penalized for failure to comply with policy prescriptions and procedures (Lipsky 1980; Hasenfeld, Ghose, and Larson 2004; Brodkin 2011; Soss, Fording, and Schram 2011a). Public employees have a responsibility to ensure that all citizens, in their dealings with public agencies and services, receive fair and equitable treatment, free from discrimination on the basis of race, ethnicity, or other attributes (Frederickson 1990). In addition to equity as a normative ideal of democratic administration and a principle of international law, empirical work suggests that procedural fairness is important in ensuring the legitimacy of public institutions in the eyes of citizens (Sunshine and Tyler 2003; Tyler 2003). Thus, an understanding of the potential effects of client characteristics such as race and ethnicity on public employees’ decisions to sanction is of broad societal interest and importance. This article examines public employees’ decisions to sanction clients, focusing first on the role of client ethnicity: What are the effects of client ethnicity on public employees’ discretionary decisions to sanction clients? Second, we also consider how the interplay between client ethnicity and various employee characteristics shapes these decisions: How and to what extent do public employee characteristics—ethnicity, gender, and work experience—moderate the potential disparity between sanctioning decisions for ethnic minority clients and for ethnic majority clients? Our analyses help explain the ways that bureaucratic discretion may lead to disparate treatment of comparable citizens of different races or ethnicities in their dealings with government and public administration. Research reveals that citizen or client race may affect public employees’ decision-making and exercise of discretion in the judicial system (Pager 2007; Johnson 2014), law enforcement (Wilkins and Williams 2008; Pickerill, Mosher, and Pratt 2009), incarceration (Olson 2016), education and schooling (Meier and Stewart 1992; Meier 1993b; Grissom, Nicholson-Crotty, and Nicholson-Crotty 2009; Atkins and Wilkins 2014; Okonofua and Eberhardt 2015), and other public agency settings (Hindera 1993; Selden 1997; Selden, Brudney, and Kellough 1998; Bertrand and Mullainathan 2004). This research provides valuable insights into both racial disparity in public services provision and representative bureaucracy (Krislov 1974; Meier 1993a). However, systematic studies of how client race or ethnicity may affect the micro-level decisions of frontline workers are few (see Bradbury and Kellough 2011). Little empirical research—using individual-level data and research designs that can allow for a causal interpretation of the results—has examined racial or ethnic bias in individual-level welfare sanction decisions. One exception is a study by Schram et al. (2009); see also Soss, Fording, and Schram 2011b). In terms of theory, they draw on the Racial Classification Model (RCM) developed by Soss, Fording, and Schram (2008), and examine how implicit racial bias shapes employment caseworkers’ decisions to dispense sanctions for clients. With the United States as their empirical setting, the results indicate differences in the treatment of female clients of different races. Both Latina and black clients are more likely to be recommended for sanctions than white clients. Moreover, Schram et al. (2009) show that two other client characteristics—family size and sanctioning history—may widen these racial disparities. Our study extends Schram et al.’s (2009) work and the RCM with new insights. Specifically, our study expands scholarly understanding of racial or ethnic disparity in public service provision in at least three ways. First, using the Scandinavian welfare state of Denmark as our empirical setting, we provide insights into the applicability and generalizability of both the RCM and Schram et al.’s (2009) findings outside of the US context. Does the RCM predict employment caseworkers’ decisions to sanction clients in a non-US country setting? As we elaborate later on, Denmark is radically different from the United States in terms of public welfare services, racial/ethnic composition, and the historical development of race/ethnicity as a salient policy issue. Second, we consider the role of the employment caseworkers’ ethnicity, gender, and professional work experience as important moderators, by drawing from different streams of behavioral economics (Kahneman and Frederick 2002; Chen, Moskowitz, and Shue 2016), social psychology (Tajfel and Turner 1986; Eagly et al. 2004; Schwartz and Rubel 2005; Rand et al. 2016), and public administration research (Downs 1967; Meier 1993a; Selden 1997; Flyvbjerg 2001; DeHart-Davis, Marlowe, and Pandey 2006; Bradbury and Kellough 2008, 2011). In doing so, we take an important step in unlocking both the micro-foundations of representation on the frontlines of public service and the importance of work experience in minimizing the possible consequences of cognitive biases affecting the distribution of public benefits. Our examination also constitutes a theoretical extension of the RCM. Although Schram et al. (2009) theorize and test how various client markers interact with client race in shaping caseworkers’ sanctioning decisions—and thus extend the RCM with insights into the moderating effects of client characteristics—we examine and expand our knowledge about the moderating effects of caseworker characteristics. By identifying how caseworker characteristics moderate disparity in sanctioning decisions for clients of a different ethnicity, we expand scholarly understanding of when and why welfare programs involving systems for dispensing sanctions may (or may not) produce racial and ethnic disparities. Third, we advance current public administration research by examining disparities in sanctioning activity based on ethnicity using both agency-level administrative data and individual-level experimental data. Study 1 uses nationwide administrative data from the national employment program. We look at three sanctioning indicators at the employment agency-level and test for differences between agencies with different percentages of clients being non-Western immigrants or their descendants. Study 2 uses individual-level experimental data from a national survey among Danish employment caseworkers to build on the findings from Study 1. We present the respondents with a vignette containing a violation of program rules by a client. In the vignette, the client’s ethnicity (majority/minority) is randomly assigned.1 Specifically, client ethnicity was either blinded (unknown to the respondents) or cued as either ethnic Danish or a non-Western ethnic minority (of Middle-Eastern origin). After having read the vignette, the respondents were asked whether they would impose a sanction. The two studies thus complement each other in important ways. Study 1 relates to agency-level sanctioning decisions as they naturally occur in real-life cases and work environments and, therefore, complements Study 2 in terms of ecological validity. Study 2 relates to individual-level sanctioning decisions and allows for causal inferences and, therefore, complements Study 1 in terms of internal validity and findings at the individual employee-level. The article proceeds as follows. First, we describe our application of the RCM in the Danish context, and hypothesize on the nature of the relationship between client ethnicity and a caseworker’s decision to sanction a client. Second, we develop hypotheses on the moderating effects of caseworker characteristics—ethnicity, gender, and work experience–on the decision to sanction. Third, we review the rules and procedures of the Danish unemployment program regarding citizens’ rights and entitlements to unemployment benefits and caseworkers’ decisions to sanction. Fourth, we present the results of our empirical analyses. Finally, we discuss our findings and their implications for both policy and future research. Theory and Hypotheses The Racial Classification Model (RCM) and the Danish Context To understand how client ethnicity may shape employment caseworkers’ decisions to sanction clients, we turn to Soss, Fording, and Schram’s (2008) RCM. In brief, the RCM explains how the race or ethnicity of policy targets (clients) affects the policy implementation decisions of public employees. The RCM represents “a cognitive model of policy decision making built on the necessity of social classification and consequences of group reputation […] specifying how and when racial classifications should affect target group constructions and, hence, officials’ choice regarding policy design and implementation” (Soss, Fording, and Schram 2008, p. 539). Although Soss, Fording, and Schram (2008) discuss the RCM in terms of race, they explicitly state that the RCM may also apply to social classifications defined by race or ethnicity (p. 540). According to Soss, Fording, and Schram (2008), the RCM rests on three premises: (1) To be effective at designing and applying policies to specific target groups, policy actors must rely on salient social classifications; (2) when racial (or ethnic) minorities are salient in a policy context, race or ethnicity is more likely to provide a basis for classifications of targets; and (3) the likelihood of racially or ethnically patterned policy outcomes will be positively associated with the degree of contrast in policy actors’ perceptions of targets. The extent to which policy actors differentiate among target groups and treat them differently is a function of the perceived contrasts among the groups. The RCM has been the basis for empirical inquiry into how the racial characteristics of a population affect social welfare policy choices by legislatures (Soss, Fording, and Schram 2008) as well as disparities faced by African-Americans and Latinas in the United States with respect to sanctions for policy noncompliance (Schram et al. 2009). Relative to Schram et al. (2009), our study covers a different employment policy and a different country, Denmark, with a setting that is radically different from that of the United States in terms of the racial and ethnic majority and minority composition. Ethnic Danes constitute the large majority of the Danish population (88%), whereas non-Western ethnic minorities constitute a much smaller but growing group (1% in 1980, 5% in 2000, and 8% in 2016).2 The four most common countries of origin among non-Western ethnic minorities in Denmark are Turkey, Iraq, Syria, and Lebanon,3 with Islam the most prevalent religion among them (Jacobsen 2012). In seeking to extend the RCM, we apply the model and its underlying assumptions to the Danish context. The Scandinavian countries, including Denmark, are typically viewed as more collectivist and homogenous than the United States, with their preferences for universal and more generous social welfare policies (Esping-Andersen 1990; Caswell, Eskelinen, and Olesen 2011). Scholars root the historical development of the Danish welfare state in a social democratic ideology originating from a strong sense of community and social solidarity among Danes as a homogenous ethnic group (Christiansen et al. 2006), starting in the 1870s and predating the first waves of immigration from the Middle East to Denmark in the 1980s. Although traditionally Denmark has not been marked by the same racial and ethnic cleavages as the United States, the increasing influx of non-Western immigrants since the 1980s has made questions about immigration, national culture, and the societal integration of people of Middle-Eastern origin an important topic in current public and political debates. For example, a recent poll shows that 65 percent of Danes view “refugees, immigration, and integration” as the most important political issue (Redder and Christensen 2016). One particularly salient policy issue concerns the gap in employment rates for non-Western immigrants and their descendants relative to ethnic Danes—and how to minimize this gap (Liebig 2007). In terms of labor market attachment, the employment rates among non-Western ethnic minorities are lower than among ethnic Danes. The employment rates among ethnic Danish males and females in 2014 were 76 and 73 percent, respectively. In comparison, the employment rates among non-Western male and female immigrants were 53 and 46 percent, respectively—and, among non-Western male and female descendants of immigrants, 53 and 52 percent, respectively.4 At the same time, ethnic minorities of non-Western origin are also subject to stereotyping in Denmark about their willingness to work. One preconception is that many non-Western immigrants come to Denmark for the benefits of the Danish welfare system and that more non-Western immigrants than ethnic Danes prefer living off welfare rather than pursuing paid employment. Although academic inquiry into the Danes’ attitudes towards the work ethic of non-Western immigrants is somewhat limited, a 20-year-old survey study shows that 65 percent of Danes agree that “[m]any immigrants come to Denmark to take advantage of our social welfare system” (Gaasholt and Togeby 1995, p. 40)—and scholars suggest that the general attitudes towards immigrants in Denmark have remained fairly stable over time (Togeby 2004; Thomsen 2006). Moreover, comparative research based on General Social Survey data supports the existence of Danish stereotypes of “laziness” and “preference for welfare relative to work” for non-Western immigrants, stereotypes similar to those existing for African-Americans in the United States (Larsen 2012). US researchers have found that this notion of laziness is one of the strongest US racial stereotypes, influencing citizen attitudes towards welfare and deservingness of public assistance (Gilens 1999, see also Hancock 2004; Brown-Iannuzzi et al. 2017).5 The RCM does not argue that the presence or influence of these ethnic stereotypes of a certain group’s willingness to work is a subject about which individual caseworkers would necessarily be cognizant of. Indeed, social psychology suggests that cognitive biases relating to race and ethnicity may be unconscious—and that the individual acting in a biased manner may be unaware of his or her bias (Devine et al. 2002; Amodio and Devine 2006). In the context of a caseworker’s decision to sanction a client who does not comply with the requirements of, say an unemployment program policy, the ethnic stereotypes previously discussed may influence the caseworker’s cognitive classification and perceptions of that client, thereby resulting in systematically different sanctioning behaviors for clients of different ethnicity. These stereotypes may particularly affect how the caseworker perceives and assesses the deservingness of clients of Middle-Eastern origin relative to clients of Danish origin. Stereotypes and group reputations may lead caseworkers to evaluate clients appearing to be non-Western immigrants of Middle-Eastern origin as less deserving of a second chance (or the benefit of the doubt) when they are in policy noncompliance. Because of cognitive biases rooted in ethnicity and ethnic stereotypes, caseworkers may therefore be more likely to recommend sanctions for clients of Middle-Eastern origin than for clients of Danish origin: Hypothesis 1: Caseworkers are more likely to recommend sanctions for ethnic minority clients (here, clients of Middle-Eastern origin) than for ethnic majority clients (here, clients of ethnic Danish origin). In addition to the direct effect of client ethnicity on a caseworker’s decision to sanction, a range of caseworker characteristics could potentially moderate the ways in which client ethnicity affects a caseworker’s sanctioning decisions. Our focus on the caseworker characteristics of ethnicity, gender, and work experience is guided by research that highlights these variables as possible moderators of the effects of client ethnicity on employee decision-making. Both representative bureaucracy research (Bradbury and Kellough 2008; Meier 1993a; Selden 1997) and social psychology theory (Tajfel and Turner 1986) suggest that the caseworkers’ ethnicity can have important implications for the way they understand, identify with, and treat ethnic minority clients. Similarly, insights from social psychology research (Eagly et al. 2004; Schwartz and Rubel 2005; Rand et al. 2016) and research on public administration and public opinion (DeHart-Davis, Marlowe, and Pandey 2006; Crowder-Meyer 2007) suggest that the caseworkers’ gender may moderate the effect of client ethnicity on sanctioning decisions. Finally, insights from behavioral economics research (Kahneman and Frederick 2002; Chen, Moskowitz, and Shue 2016) and management research (Downs 1967; Flyvbjerg 2001) suggest that work experience may mitigate ethnic bias in caseworkers’ decisions to sanction. We believe that our focus on ethnicity, gender, and work experience constitutes a solid foundation for understanding the role of caseworker characteristics in the context of the RCM and the occurrence of racial or ethnic disparity in public services provision. Caseworker Ethnicity Several research perspectives suggest that human beings tend to hold cognitive biases that favor their own social in-group, in turn possibly moderating how client ethnicity shapes caseworkers’ decisions to sanction. As Bradbury and Kellough (2011, p. 160) note, the public administration literature on representative bureaucracy shows that “public organizations with larger proportions of women and/or minorities in decision-making roles are more likely to produce outcomes compatible with the interests of women and/or minorities than similar organizations with fewer women and/or minorities” (e.g., Meier and Stewart 1992; Hindera 1993; Keiser et al. 2002; Wilkins and Keiser 2006). The theory of representative bureaucracy suggests that the demographic characteristics of public employees shape their attitudes, thereby shaping their behavior and decision-making (Meier 1993a). Employees and citizens with similar demographic characteristics (e.g., ethnicity) share common experiences and often values and attitudes (Dolan 2000; Bradbury and Kellough 2008). This shared experience explains why female and/or ethnic minority employees are more likely to make decisions and adopt an advocacy role that favor female and/or ethnic minority clients (Meier 1993a; Selden 1997; Bradbury and Kellough 2008). Social psychological theories elaborate on the similarity-dissimilarity process providing individuals with cognitive biases toward individuals of the same ethnicity as themselves. Similarity/attraction theory (Byrne 1971) and social identity theory (Tajfel and Turner 1986) emphasize how human beings tend to better understand and identify with people exhibiting the same characteristics as themselves. Demographic similarity on a dimension such as ethnicity promotes compatibility, mutual understanding, and identity reinforcement, whereas dissimilarity creates incompatibility and disagreement (Hogg and Abrams 1998). Social identity theory further explains how motives and behavior are affected by “in group”–“out group” stereotyping, group cohesion, and emotional sympathy. The theory posits that individual identities are based on self-categorization into social groups, that is, people classify those who are similar (e.g., in terms of ethnicity) as the in-group, and those they perceive as being different as the out-group (Tajfel and Turner 1986). Individuals who belong to the in-group coalesce and bond based upon shared traits or experiences (Currarini and Mengel 2016). The perceived connection between members of the same in-group (e.g., caseworkers and clients who belong to an ethnic minority group) might inform the basis of an in-group favorability bias or a debiasing of intra-in-group judgments and decision-making. At the minimum, this connection may mitigate the application of ethnic stereotyping. Combining these insights, we theorize that caseworkers have cognitive biases that make them tend to favor clients of their own ethnicity—and that these biases may moderate how client ethnicity shapes their decisions to sanction. Drawing on social identity theory (Tajfel and Turner 1986), we expect that caseworkers tend to engage clients of the same ethnicity as themselves with greater sympathy, understanding, and equity. Because of ethnicity-similarity bias, ethnic stereotyping is less likely to affect decision-making—that is, create a disparity in sanctioning decisions between ethnic minority and ethnic majority clients—when those clients are of the same ethnicity (in-group) as the caseworker. In summary, we expect the propensity to sanction ethnic minority clients more than ethnic majority clients to be smaller for caseworkers who themselves belong to an ethnic minority group: Hypothesis 2a: Caseworkers are less likely to sanction ethnic minority clients more than ethnic majority clients when the caseworker belongs to an ethnic minority group. Caseworker Gender The public administration literature on bureaucratic gender representation focuses mainly on how passive representation leads to active representation for women (Keiser et al. 2002; Wilkins and Keiser 2006; Bradbury and Kellough 2011), that is, the association between the presence of female bureaucrats within public organizations and bureaucratic outcomes consistent with the interests of female citizens. However, other streams of research suggest that caseworkers’ gender may moderate the effect of client ethnicity on caseworkers’ decisions to sanction in other ways. Social role theory suggests that women tend to be more altruistic, benevolent, and compassionate than men. For example, meta-analytical evidence suggests that women, but not men, internalize altruism as their intuitive social response (Rand et al. 2016). Women are more likely to express concern and responsibility for the well-being of others (Beutel and Marini 1995). Similarly, Dehart-Davis, Marlowe, and Pandey (2006) find that women score higher on the “compassion” subdimension of “public service motivation” than their male counterparts. In addition, public opinion research suggests that women are more favorable towards policies promoting equity and provision of social welfare benefits (Shapiro and Mahajan 1986; Crowder-Meyer 2007). Eagly et al. (2004) show that women, more than men, endorse policies that are socially compassionate, traditionally moral, and supportive of equal rights. In a cross-cultural and multi-method study based on 127 samples in 70 countries, Schwartz and Rubel (2005) find that women consistently attribute more importance than men do to both universalism values (understanding, appreciation, tolerance, and protection for the welfare of all people) and benevolence values (preservation and enhancement of the welfare of people with whom one is in personal contact). Combining these insights, we theorize that female caseworkers exhibit social roles and value orientations that make them more likely to engage clients of all ethnicities with greater sympathy, understanding, and equity than male caseworkers. Greater concern for the welfare, well-being, and equitable treatment of others among women may entail that ethnic stereotyping is less likely to shape decision-making—that is, create a disparity in sanctioning decisions between ethnic minority and ethnic majority clients—among female caseworkers. In summary, we expect the propensity to sanction ethnic minority clients more than ethnic majority clients to be smaller for female caseworkers: Hypothesis 2b: Caseworkers are less likely to sanction ethnic minority clients more than ethnic majority clients when the caseworker is female. Work Experience In addition to a caseworker’s ethnicity and gender, we are also interested in the role that work experience plays in caseworkers’ exercise of discretion, i.e., how work experience may moderate the effect of client ethnicity on the decision to sanction. Research suggests that stereotypical beliefs may have stronger effects in situations where the individual finds knowledge about the right course of action (what to do and how to do it) is unclear. In instances where the right action is unclear, Kahneman and colleagues (Kahneman, Slovic, and Tversky 1982; Kahneman 2011) show how bias in an individual’s judgment and decision-making may arise from the application of cognitive heuristics (information-processing shortcuts), such as stereotypes. Kahneman and Frederick (2002) emphasize the role of a process they call attribute substitution, which happens without conscious awareness. According to this theory, stereotypes can be a source of heuristic attributes that people substitute in when judgment of a target attribute is computationally complex. For example, judging whether a client’s behavior in a particular situation warrants sanctioning is more computationally complex than judging their ethnicity. The pre-conscious, intuitive nature of attribute substitution explains how a caseworker’s decisions to sanction may be influenced by client ethnicity, even though the caseworker believes that he or she has made an unbiased evaluation. At the same time, scholars have long known that professional socialization—i.e., the acquisition of not only attitudes and values but also skills and know-how pertaining to a professional subculture—increases over time (Downs 1967). As caseworkers gain work experience, they accumulate increasing knowledge about typical and atypical cases and casework proceedings and practices. They develop and internalize work routines and standard operating procedures that increasingly guide their work activities, discretion, and decision-making. Flyvbjerg (2001) describes such accumulation of work experience as a learning process through which a novice may gradually become a competent performer and even an expert over time. The accumulation of work experience and the development of expertise might lead to professional norms or standards serving as a guide to action, replacing individual caseworker attitudes, motives, or unconscious beliefs, such as stereotypes. Jensen and Vestergaard (2017) find that experience can moderate (suppress) the effect of motives with respect to predicting behavior. They argue that as people accumulate experiences, they rely on those experiences to formulate judgments about the best way to act in a situation. In the process, their personal motives or attitudes become less important. We, therefore, theorize that ethnic stereotypes are more likely to shape caseworkers’ decision-making when the caseworker has less work experience with employment casework. Moreover, Chen, Moskowitz, and Shue (2016) offer robust empirical evidence for the importance of work experience in mitigating biased decision-making. Across three high-stakes field settings (asylum, loan, and officiating decisions), they show how people tend to underestimate the likelihood of sequential streaks occurring by chance. This underestimation leads to negatively auto-correlated decisions that result in error. Due to misperception of random processes, people tend to believe that streaks of good or bad cases are unlikely to occur by chance (Rabin and Vayanos 2010). Therefore, decision makers may approach a new case with the subconscious belief that the case is likely to be positive if they considered the previous case negative, and vice versa. Decisions are affected by prior decisions in an inverse relationship, that is, decisions are negatively auto-correlated. Importantly, Chen, Moskowitz, and Shue (2016) find that the negative autocorrelation (bias) is less strong among more experienced decision makers. Similar to the way that work experience diminishes this particular form of cognitive bias in decision-making, we expect that work experience may mitigate ethnic bias in caseworkers’ sanctioning decisions. In line with Flyvbjerg (2001), we expect that an experienced performer or expert is less likely than a novice to resort to a (simplified) decision-making process based on ethnic cues. In summary, we expect that caseworkers’ work experience moderates the influence of client ethnicity on caseworkers’ decision to sanction in the following way: Hypothesis 2c: Caseworkers are less likely to sanction ethnic minority clients more than ethnic majority clients when the caseworker has more work experience. Empirical Setting We test the hypotheses using Danish employment agencies as our empirical setting. Denmark comprises 98 municipalities, each with its own employment agency. All unemployed citizens of working age and fit to work, but who are not members of an unemployment insurance fund (typically individuals having only a secondary education or lower), are eligible for a monthly means-tested unemployment cash benefit. To receive the cash benefit, unemployed citizens have to register at their local employment agency. Each citizen is then assigned a caseworker, who is responsible for helping that citizen obtain employment. A client’s entitlement to an unemployment benefit is conditioned by compliance with a set of rules and procedures. For example, all clients are expected to regularly consult with their caseworker (every month for the first half year, then every third month), actively seek employment, and be willing to accept employment offers. All clients must accept and attend employability-increasing activities that their local employment agency arranges, such as on-the-job training. The caseworkers have the authority to sanction clients for noncompliance with these rules and procedures. For example, a client may be assigned to an unpaid on-the-job training program in a private firm or public organization. If the client fails to show up without reasonable cause, the caseworker has the discretionary authority to reduce or terminate the client’s unemployment benefits until the client either requalifies for assistance or provides evidence of reasonable cause (see the Online Appendix for a visual presentation of the sanctioning procedure). Study 1: Administrative Data Our administrative data comprises Danish employment agency-level records for the first quarter of 2016. In the data, we have measures capturing the percentage of non-Western immigrants or their descendants who were cash benefit recipients, as well as three agency-level sanctioning indicators: (1) the average number of sanctions per cash benefit recipient; (2) the percentage of cash benefit recipients receiving a sanction; and (3) the average number of sanctions per cash benefit recipient receiving a sanction. We obtained the records from jobindsats.dk, an online platform hosted by the Danish Agency for Labour Market and Recruitment.6 We combine these data with administrative data from 2015 on the characteristics of the municipality where each employment agency is geographically located. All the data were available from Statistic Denmark’s population register and from noegletal.dk, an online platform hosted by the Danish Ministry of Social Affairs and the Interior. The dataset gives us access to the following municipal characteristics: number of residents, number of residents of Danish origin, percentage of residents living in urban agglomeration, percentage of residents aged 25–64 without vocational training, percentage of residents aged 25–64 with a higher education, and the unemployment rate (the average number of full-time unemployed who are fit to work per 100 residents aged 17–64). We include these variables in our analysis as controls. Table 1 gives the descriptive statistics. Table 1. Employment Agency Characteristics Employment Agency Characteristics  Mean  SD  Min  Max  N  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .37  .10  0  .62  98  Average number of sanctions per cash benefit recipient  .11  .10  0  .68  98  Percentage of cash benefit recipients receiving sanction  .06  .03  0  .22  98  Average number of sanctions per cash benefit recipient receiving sanction  1.69  .60  0  4.46  98  Number of municipal residents (1000s)  58.16  69.38  1.82  589.70  98  Number of municipal residents of Danish origin (1000s)  51.05  55.21  1.72  450.35  98  Percentage of municipal residents living in urban agglomeration  83.37  12.72  34.7  100  98  Percentage of municipal residents aged 25–64 without vocational training  20.59  4.95  8.2  31.4  98  Percentage of municipal residents aged 25–64 with a higher education  25.10  8.96  14.1  52  98  Municipal unemployment rate  3.99  1.10  2  8.4  98  Employment Agency Characteristics  Mean  SD  Min  Max  N  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .37  .10  0  .62  98  Average number of sanctions per cash benefit recipient  .11  .10  0  .68  98  Percentage of cash benefit recipients receiving sanction  .06  .03  0  .22  98  Average number of sanctions per cash benefit recipient receiving sanction  1.69  .60  0  4.46  98  Number of municipal residents (1000s)  58.16  69.38  1.82  589.70  98  Number of municipal residents of Danish origin (1000s)  51.05  55.21  1.72  450.35  98  Percentage of municipal residents living in urban agglomeration  83.37  12.72  34.7  100  98  Percentage of municipal residents aged 25–64 without vocational training  20.59  4.95  8.2  31.4  98  Percentage of municipal residents aged 25–64 with a higher education  25.10  8.96  14.1  52  98  Municipal unemployment rate  3.99  1.10  2  8.4  98  View Large Table 1. Employment Agency Characteristics Employment Agency Characteristics  Mean  SD  Min  Max  N  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .37  .10  0  .62  98  Average number of sanctions per cash benefit recipient  .11  .10  0  .68  98  Percentage of cash benefit recipients receiving sanction  .06  .03  0  .22  98  Average number of sanctions per cash benefit recipient receiving sanction  1.69  .60  0  4.46  98  Number of municipal residents (1000s)  58.16  69.38  1.82  589.70  98  Number of municipal residents of Danish origin (1000s)  51.05  55.21  1.72  450.35  98  Percentage of municipal residents living in urban agglomeration  83.37  12.72  34.7  100  98  Percentage of municipal residents aged 25–64 without vocational training  20.59  4.95  8.2  31.4  98  Percentage of municipal residents aged 25–64 with a higher education  25.10  8.96  14.1  52  98  Municipal unemployment rate  3.99  1.10  2  8.4  98  Employment Agency Characteristics  Mean  SD  Min  Max  N  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .37  .10  0  .62  98  Average number of sanctions per cash benefit recipient  .11  .10  0  .68  98  Percentage of cash benefit recipients receiving sanction  .06  .03  0  .22  98  Average number of sanctions per cash benefit recipient receiving sanction  1.69  .60  0  4.46  98  Number of municipal residents (1000s)  58.16  69.38  1.82  589.70  98  Number of municipal residents of Danish origin (1000s)  51.05  55.21  1.72  450.35  98  Percentage of municipal residents living in urban agglomeration  83.37  12.72  34.7  100  98  Percentage of municipal residents aged 25–64 without vocational training  20.59  4.95  8.2  31.4  98  Percentage of municipal residents aged 25–64 with a higher education  25.10  8.96  14.1  52  98  Municipal unemployment rate  3.99  1.10  2  8.4  98  View Large Results We analyze the administrative data using multivariate OLS regression with robust standard errors. We run three models in which we regress the percentage of cash benefit recipients who are non-Western immigrants or their descendants on the average number of sanctions per cash benefit recipient (model 1), the percentage of cash benefit recipients receiving a sanction (model 2), and the average number of sanctions per cash benefit recipient receiving a sanction (model 3). All three models include a set of statistical control variables reflecting the municipal characteristics retrieved from Statistic Denmark’s population register and noegletal.dk. Table 2 shows the results of the analyses. Table 2. OLS Regression, Administrative Data Independent variables  Model 1  Model 2  Model 3  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .255* (.124)  .089† (.047)  1.92* (.948)  Municipal characteristics        Number of residents (1000s)  .001 (.001)  .001 (.001)  −.008 (.008)  Number of residents of Danish origin (1000s)  −.001 (.002)  −.001 (.001)  .011 (.011)  Percentage of residents living in urban agglomeration  −.002 (.001)  −.001 (.001)  −.004 (.004)  Percentage of residents aged 25–64 without vocational training  .017* (.008)  .005* (.002)  .084 (.056)  Percentage of residents aged 25–64 with a higher education  .004 (.003)  .001 (.001)  .023 (.203)  Unemployment rate  −.033* (.014)  −.009* (.004)  −.202** (.072)  R2  .17  .16  .16  N  98  98  98  Independent variables  Model 1  Model 2  Model 3  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .255* (.124)  .089† (.047)  1.92* (.948)  Municipal characteristics        Number of residents (1000s)  .001 (.001)  .001 (.001)  −.008 (.008)  Number of residents of Danish origin (1000s)  −.001 (.002)  −.001 (.001)  .011 (.011)  Percentage of residents living in urban agglomeration  −.002 (.001)  −.001 (.001)  −.004 (.004)  Percentage of residents aged 25–64 without vocational training  .017* (.008)  .005* (.002)  .084 (.056)  Percentage of residents aged 25–64 with a higher education  .004 (.003)  .001 (.001)  .023 (.203)  Unemployment rate  −.033* (.014)  −.009* (.004)  −.202** (.072)  R2  .17  .16  .16  N  98  98  98  Notes: Entries are coefficients with robust standard errors (in parentheses). †p ≤ .1; *p ≤ .05; **p ≤ .01. View Large Table 2. OLS Regression, Administrative Data Independent variables  Model 1  Model 2  Model 3  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .255* (.124)  .089† (.047)  1.92* (.948)  Municipal characteristics        Number of residents (1000s)  .001 (.001)  .001 (.001)  −.008 (.008)  Number of residents of Danish origin (1000s)  −.001 (.002)  −.001 (.001)  .011 (.011)  Percentage of residents living in urban agglomeration  −.002 (.001)  −.001 (.001)  −.004 (.004)  Percentage of residents aged 25–64 without vocational training  .017* (.008)  .005* (.002)  .084 (.056)  Percentage of residents aged 25–64 with a higher education  .004 (.003)  .001 (.001)  .023 (.203)  Unemployment rate  −.033* (.014)  −.009* (.004)  −.202** (.072)  R2  .17  .16  .16  N  98  98  98  Independent variables  Model 1  Model 2  Model 3  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .255* (.124)  .089† (.047)  1.92* (.948)  Municipal characteristics        Number of residents (1000s)  .001 (.001)  .001 (.001)  −.008 (.008)  Number of residents of Danish origin (1000s)  −.001 (.002)  −.001 (.001)  .011 (.011)  Percentage of residents living in urban agglomeration  −.002 (.001)  −.001 (.001)  −.004 (.004)  Percentage of residents aged 25–64 without vocational training  .017* (.008)  .005* (.002)  .084 (.056)  Percentage of residents aged 25–64 with a higher education  .004 (.003)  .001 (.001)  .023 (.203)  Unemployment rate  −.033* (.014)  −.009* (.004)  −.202** (.072)  R2  .17  .16  .16  N  98  98  98  Notes: Entries are coefficients with robust standard errors (in parentheses). †p ≤ .1; *p ≤ .05; **p ≤ .01. View Large The analyses suggest that employment agencies in which a larger percentage of cash benefit recipients are non-Western immigrants or their descendants impose a greater overall number of sanctions and dispense them with greater frequency. In model 1, we find that employment agencies in which a greater percentage of the cash benefit recipients are non-Western immigrants or their descendants are characterized by a greater average number of sanctions per cash benefit recipient (β = .255, p ≤ .05). Similarly, models 2 and 3 show, respectively, that the percentage of non-Western immigrants or their descendants receiving cash benefits is positively associated with a higher percentage of cash benefit recipients receiving a sanction (β = .089, p ≤ .10) and a greater number of sanctions per sanctioned cash benefit recipient (β = 1.92, p ≤ .05). The results are thus consistent with the notion that caseworkers are more likely to recommend sanctions for ethnic minority clients than for ethnic majority clients (Hypothesis 1). As the administrative data reflects the real world of caseworker decision-making, the findings are characterized by high ecological validity. Moreover, the nationwide nature of the data eliminates the risk of bias caused by missing observations or nonresponse. However, the analyses of the administrative data are marked by one key limitation: a modest potential for causal inferences. Although the analyses indicate that ethnic minority clients are sanctioned more than ethnic majority clients, we recognize that other factors than ethnic bias among caseworkers may possibly explain this finding. For example, the agency-level differences in sanctioning activity that we observe could be the result of systematic differences in client behavior (i.e., clients of Middle-Eastern origin exhibiting more sanctionable behavior than clients of Danish origin). To better understand how ethnic bias shapes caseworkers’ decisions to sanction clients, we complement the analyses of the administrative data with a survey experiment that relates to individual-level sanctioning decisions and allows for a causal interpretation of the results. Study 2: Survey Experiment Our experimental data originates from a national e-survey among Danish employment caseworkers. To build our sample frame of employment agency caseworkers, we sent a request to employment agency managers from all 98 Danish municipalities for the contact information of each of their local employment caseworkers and for permission to allow us to invite the caseworkers to participate in a research project involving a survey. Managers from 63 municipalities responded, with the email addresses of a total of 1,335 caseworkers and permission from their supervisors to request their participation in our study. We conducted the survey experiment in April 2016. We received complete survey responses from 497 caseworkers (≈37%). In line with Danish guidelines for ethics in social sciences research, the respondents were instructed that while their responses would be used for research purposes, their individual responses were anonymous and that the researchers would comply with standard confidentiality procedures. Furthermore, we made clear that participation in our study was voluntary and that respondents could choose not to participate or to stop at any time. Table 3 shows a demographic profile of the survey respondents. Table 3. Respondent Characteristics Respondent characteristics  Mean  SD  Min  Max  N  Gender (female)  .81  .39  0  1  497  Age (years)  43.21  11.16  21  69  497  Ethnicity (non-Dane)  .09  .26  0  1  497  Education: Lower secondary  .01  .11  0  1  497      : Secondary  .02  .15  0  1  497      : Youth vocational training  .08  .28  0  1  497      : Higher, short-cycle  .05  .23  0  1  497      : Higher, medium-cycle  .73  .44  0  1  497      : Higher, long-cycle  .09  .29  0  1  497  Primary client group: Cash benefit recipients  .57  .50  0  1  497  Tenure (years)  4.73  4.94  0  35  497  Work experience, caseworker (years)  8.62  7.78  0  38  497  Respondent characteristics  Mean  SD  Min  Max  N  Gender (female)  .81  .39  0  1  497  Age (years)  43.21  11.16  21  69  497  Ethnicity (non-Dane)  .09  .26  0  1  497  Education: Lower secondary  .01  .11  0  1  497      : Secondary  .02  .15  0  1  497      : Youth vocational training  .08  .28  0  1  497      : Higher, short-cycle  .05  .23  0  1  497      : Higher, medium-cycle  .73  .44  0  1  497      : Higher, long-cycle  .09  .29  0  1  497  Primary client group: Cash benefit recipients  .57  .50  0  1  497  Tenure (years)  4.73  4.94  0  35  497  Work experience, caseworker (years)  8.62  7.78  0  38  497  View Large Table 3. Respondent Characteristics Respondent characteristics  Mean  SD  Min  Max  N  Gender (female)  .81  .39  0  1  497  Age (years)  43.21  11.16  21  69  497  Ethnicity (non-Dane)  .09  .26  0  1  497  Education: Lower secondary  .01  .11  0  1  497      : Secondary  .02  .15  0  1  497      : Youth vocational training  .08  .28  0  1  497      : Higher, short-cycle  .05  .23  0  1  497      : Higher, medium-cycle  .73  .44  0  1  497      : Higher, long-cycle  .09  .29  0  1  497  Primary client group: Cash benefit recipients  .57  .50  0  1  497  Tenure (years)  4.73  4.94  0  35  497  Work experience, caseworker (years)  8.62  7.78  0  38  497  Respondent characteristics  Mean  SD  Min  Max  N  Gender (female)  .81  .39  0  1  497  Age (years)  43.21  11.16  21  69  497  Ethnicity (non-Dane)  .09  .26  0  1  497  Education: Lower secondary  .01  .11  0  1  497      : Secondary  .02  .15  0  1  497      : Youth vocational training  .08  .28  0  1  497      : Higher, short-cycle  .05  .23  0  1  497      : Higher, medium-cycle  .73  .44  0  1  497      : Higher, long-cycle  .09  .29  0  1  497  Primary client group: Cash benefit recipients  .57  .50  0  1  497  Tenure (years)  4.73  4.94  0  35  497  Work experience, caseworker (years)  8.62  7.78  0  38  497  View Large While the sampling procedure and response rate suggest some caution in extrapolating from this sample, the respondents appear relatively diverse and suit the needs of our experimental design (see Auginis and Bradley 2014). The sampling procedure—while imperfect—nonetheless provides us with a population of actual caseworkers. Experimental Design The design involves a vignette portraying an unemployed male cash benefit recipient who is assigned to an unpaid on-the-job training program in a public organization. The vignette explains that the client did not show up for the program on some occasions—without notifying the employer—and that the client is unable to document having been ill (his stated reason for being absent). The caseworkers were asked to decide, based on the vignette, whether to impose a sanction on the client or not.7 Each caseworker respondent received one of three versions of the vignette. For capturing the influence of client ethnicity on caseworker responses, we randomly assigned the client described in the vignette (1) a “Middle-Eastern-sounding” name, (2) an “ethnic Danish-sounding” name, or (3) a blinded name. In all other respects, all respondents received identical vignette information, including some biographical information to add to the contextual realism of the case (Auginis and Bradley 2014). Similar “name manipulation” procedures are used in other studies examining the effects of client race in the area of unemployment (Bertrand and Mullainathan 2004; Schram et al. 2009). We use the name “Mohammed” as the ethnic minority of Middle-Eastern origin-sounding name, and “Peter” as the ethnic Danish-sounding name. We selected those names from administrative data from Statistic Denmark (2016), data suggesting that Peter is the most common male first name in Denmark and Mohammed the most common male first name among non-Western immigrants and their descendants in Denmark (Sahin and Dreyer 2015) and among Muslims in Denmark (Meldgaard 2005). Their ethnic connotations notwithstanding, neither name is associated with any known social stigma. For example, the name Brian is also fairly common in Denmark but is associated with the stereotype of an ethnic Danish working-class troublemaker. The version of the vignette involving a blinded name simply showed a black box where the name should be, signaling that we had anonymized the client’s name. The random assignment process resulted in a statistically equal distribution of client names to respondents (t-tests): 169 received the Mohammed treatment (34%), 173 received the Peter treatment (35%), and 155 received a blinded name (31%). The Vignette The vignette was inspired by case descriptions in Beer et al. (2008) and Ejrnæs and Monrad (2012). For contextual realism, two experienced unemployment caseworkers and an expert in immigrants and multicultural relations were involved in developing the vignette. The full text is as follows (translated from Danish to English with the assistance of a professional language editor): “[Peter/Mohammed/”blinded”] is a 31-year-old male. He is fit to work and receives unemployment cash benefits. Peter left Folkeskolen [Danish compulsory education] after 8th grade and has not received any further education. He has about five years of work experience from different minor jobs. He has been dismissed from all of them due to absenteeism. He has received cash benefits or been self-supporting on a rotating basis between these jobs. He has been unemployed and received cash benefits since he was fired from his latest job five months ago. Peter was diagnosed with depression two years ago and received treatment with antidepressants for a period of six months. Peter started on-the-job training in a public organization in January 2016. During the start-up phase, Peter was informed in writing about the consequences for his cash benefits should he fail to attend the training without reasonable cause, and about what he could do to regain his cash benefits entitlement in such a case. Peter’s time sheet from the on-the-job training for February 2016 shows that he has been absent five days without informing the employer. In a consultation, Peter has attributed his absences to illness caused by migraines. However, he did not consult a general practitioner. He has not been sanctioned previously.” The vignette allows us to test whether caseworkers may be more likely to recommend sanctions for clients appearing to be an ethnic minority (of a Middle-Eastern origin) than for clients of Danish ethnic origin. The vignette portrays a situation in which a caseworker would have substantial discretion to decide whether to impose a sanction. With the blinded version of the vignette as an additional baseline, we expect that the Middle-Eastern-sounding name (Mohammed) may cue negative stereotypes related to unemployed clients of Middle-Eastern origin—in turn resulting in a greater sanctioning propensity among recipients of the “Mohammed” version of the vignette than among recipients of the “Peter” or “blinded” versions. Following the vignette was a question asking the respondents to indicate the extent to which they would impose a sanction on this client. Responses were measured on an 11-point scale, anchored at 0 (“definitely not sanction”) and at 10 (“definitely sanction”). All scale points were labeled with a numeral. The mean score was 7.28, with a standard deviation of 2.88.8 As an experimental manipulation check, at a later stage of the survey the respondents were asked to recall the name of the welfare recipient (response categories: “Anonymized,” “Peter,” “Michael,” “Mohammed,” “Ahmed,” “do not remember”). The “correct” name (the name in the vignette they received) was reported by 78 percent of the caseworkers. The number of false or “do not remember” responses was distributed equally across the three experimental conditions. We reran all model specifications with inclusion of a manipulation check dummy variable (“correct” response or not), and we tested all model specifications on a subsample comprising only respondents with “correct” responses. Neither of these robustness checks resulted in any qualitative changes of the findings. Analysis and Results One-way analysis of variance tests show balance across the three experimental groups in the distributions of gender, age, ethnicity, education, primary client group, tenure, and work experience (at p < .1), indicating that the randomization process worked. Moreover, one-way analysis of variance shows a significant difference in sanctioning recommendation across the three experimental conditions (p = .04). Caseworkers assigned to the “Mohammed” vignette reported a greater inclination to sanction the client than caseworkers assigned to the Peter or the “blinded” vignette. The mean sanctioning score was 7.73 for Mohammed, 7.06 for Peter, and 7.02 for the version where the name was blinded. T-tests show that the sanctioning score for Mohammed is significantly higher than the other two sanctioning scores (p = .03 and p = .02). While this result confirms Hypothesis 1, we also performed multivariate OLS regression analysis that minimizes the risk of bias due to potential imbalances in caseworker characteristics across the experimental conditions. The regression model includes two binary variables representing the Mohammed and the Peter treatment (the “blinded” version serves as the reference category). We control for the caseworker characteristics in Table 1 (gender, age, ethnicity, education, primary client group, tenure, and work experience). We estimate robust standard errors that allow for the presence of heteroscedasticity. Figure 1 illustrates the key findings: the coefficients for our experimental conditions (the underlying regression model appear in the Online Appendix, Table A-1). The point estimates (X-axis) show the differences in sanctioning recommendations among the caseworkers assigned to the Peter and Mohammed treatments relative to those assigned to the blinded vignette (Y-axis). The horizontal lines signify 95 percent confidence bands. Figure 1. View largeDownload slide Plot of the Peter and Mohammed Coefficients, Vignette Experiment Figure 1. View largeDownload slide Plot of the Peter and Mohammed Coefficients, Vignette Experiment In support of Hypothesis 1, we see that caseworkers receiving the Mohammed treatment report being more likely to sanction the client than those assigned to the blinded version of the vignette (β = .824, p = .01). The effect translates to an increase of .29 of a standard deviation in sanctioning inclination, which is comparable to an effect size of d ≈ .29, more than a “small” effect as defined by Cohen (1988). A likelihood-ratio test reveals that caseworkers receiving the Mohammed treatment are also more likely to sanction the client than those assigned to the Peter treatment (p = .03) (d ≈ .23). Assignment of the Peter treatment does not result in a statistically different sanctioning recommendation relative to assignment of the blinded version of the vignette. Moderation Analyses We examine the moderating effects of caseworker ethnicity, gender, and work experience using the survey experimental data. In particular, we employ multivariate OLS regression analysis with robust standard errors and test how assignment of experimental condition (Peter, Mohammed, or “blinded”) interact with each of the three caseworker characteristics in predicting caseworker sanctioning responses. We thus perform three regression analyses. Each regression includes interaction terms in which the binary variables representing the Mohammed and the Peter treatments are multiplied, respectively, by one of the three caseworker characteristics (gender, ethnicity, or work experience). The analysis provides a partial confirmation of our hypotheses related to the moderating effects of caseworker characteristics (the full models appear in the Online Appendix, Table A-2). Neither caseworker ethnicity nor gender appears to moderate the effect of assignment to the Mohammed treatment or the Peter treatment relative to the “blinded” version of the vignette (Hypotheses 2a and 2b). However, as anticipated, we find that assignment to the Mohammed treatment relative to the “blinded” version of the vignette results in a lesser propensity to sanction the client among caseworkers with more work experience (Hypothesis 2c). This finding supports our theory-based expectation that work experience suppresses or reduces reliance on ethnic stereotypes, thereby lessening bias in frontline employee decision-making. Figure 2 illustrates this finding. We compute and plot the marginal effects of assignment to the Mohammed treatment relative to the “blinded” version of the vignette (Y-axis) at different levels of caseworker work experience (X-axis). Figure 2. View largeDownload slide Plot of the Effect of the Mohammed Treatment Relative to the Blinded Version of the Vignette at Different Levels of Caseworker Work Experience Figure 2. View largeDownload slide Plot of the Effect of the Mohammed Treatment Relative to the Blinded Version of the Vignette at Different Levels of Caseworker Work Experience As Figure 2 shows, the statistical difference in sanctioning response between recipients of the Mohammed treatment and those of the “blinded” version of the vignettes goes towards zero as the work experience of the caseworker increases. Caseworker assignment to the Mohammed treatment relative to the “blinded” version of the vignette is associated with a higher sanctioning propensity for caseworkers with 12 years or less of work experience as a caseworker (≈75% of the sample respondents). In contrast, assignment to the Mohammed treatment relative to the “blinded” version of the vignette does not appear to influence the sanctioning recommendations of caseworkers with more than 12 years of work experience (≈25% of the sample respondents). Caseworker work experience is naturally correlated with caseworker age. Importantly, however, similar interaction term estimates across models without and with control for age (see the Online Appendix, Table A-2) diminishes the concern that the moderation finding is driven by a generational cohort effect rather than by work experience.9 Discussion and Conclusion In many public welfare programs, public employees have discretionary authority to decide when and, if so, whether welfare recipients should be penalized for failure to comply with policy requirements. In terms of formal rules and prescriptions, such welfare sanctions should be imposed as responses to the behavior of the individual client. Client characteristics, such as race or ethnicity, should not influence a caseworker’s decisions to sanction. This article, however, provides evidence that client characteristics do affect casework sanctioning decisions. Drawing on the RCM (Soss, Schram, and Fording 2008) for understanding race and ethnic biases in employee decision-making, we examine and show how client ethnicity influences sanctioning behavior in the Danish unemployment program. Drawing on the RCM, we theorized that ethnic stereotypes may shape how public employees perceive and assess the deservingness of clients of different ethnicity, and thus affect their propensity for imposing a sanction in response to client noncompliance with unemployment policy requirements. We hypothesized that employment caseworkers in Denmark are more likely to sanction a client of Middle-Eastern descent than one of Danish origin. Our empirical findings are consistent with this theoretical expectation. Examining correlates of sanctioning activity at the agency-level, our administrative data indicate that employment agencies serving a larger population of non-Western immigrants or their descendants impose a larger overall number of sanctions and dispense them to clients with greater frequency. Addressing concerns about internal validity and a need for analyses at the individual employee-level, our survey experimental data show that employment caseworkers are more likely to recommend sanctions for a client named Mohammed than for a comparable client named Peter or an unnamed client. Research suggests a proliferation of ethnic stereotypes in Denmark, implying that immigrants and people of Middle-Eastern origin in Denmark are occasionally typecast as “taking advantage” of the social welfare system (Gaasholt and Togeby 1995; Larsen 2012). In line with the RCM, such ethnic stereotypes may explain the causal mechanism underlying our empirical finding, that is, that caseworkers use discretion in a biased manner that is punitive when the client appears Middle-Eastern, not “Danish.” In addition to the direct effect of the client’s ethnicity on a caseworker’s likelihood of administering a sanction, this article extends the RCM by considering the moderating effects of three caseworker characteristics. Both theoretically and empirically, we considered the role of the employment caseworker’s ethnicity, gender, and professional work experience as important moderator variables. We find no evidence that caseworker ethnicity or gender moderates the effect of a client’s ethnicity on a caseworker’s decision to sanction. Some caution is advised when interpreting these results. The null finding for caseworker ethnicity may be a product of limited empirical variance in caseworkers’ ethnicity (91% are ethnic Danes). Moreover, a full test of the gender moderation hypothesis would require a 2 × 3 factorial design (manipulation of both client gender and ethnicity). As the client in our vignette is a male, we do not reject the possibility that female caseworkers may be less likely than male caseworkers to sanction female Middle-Eastern origin clients more than female Danish origin clients. Nevertheless, Schram et al. (2009) found similar null results in their study of caseworkers in the United States. They found no evidence that white case managers differ from their non-white peers in their sanction decisions (see also Soss, Fording, and Schram 2011b: pp. 251–261). While our null findings are inconsistent with the theoretical rationales of representative bureaucracy (Meier 1993a), social identity theory (Tajfel and Turner 1986), and theories related to group identity and in-group biases and loyalties (Currarini and Mengel 2016), the results nonetheless match the logic and underpinnings of the RCM. As Schram et al. (2009, p. 415) note: “Racial classifications operate in implicit ways—without conscious racism—to generate racial disparities (Quillian 2008). Race matters in more subtle ways than overt hostility or loyalty; race is built into the cognitive processes that provide the foundation for decisions about how target groups should be treated in welfare policy settings.” Thus, caseworker attitudes related to ethnic stereotypes might be more important in driving behaviors, actions, or decisions than shared (or unshared) ethnic characteristics. This insight has important implications for our understanding of representation in public organizations. The effects of representation found in studies at the organizational level may be a function of dynamic, group processes that may not be explained by simple aggregation of individual employee actions based on in-group bias or favoritism. Work experience constituted the only caseworker characteristic that had the hypothesized moderation effect. Theoretically, heuristics are thought most important for caseworkers when a high level of uncertainty exists in a particular case. As people gain professional work experience, they obtain professional knowledge. Over time, they rely more on knowledge gained through experience—and less on stereotypes—in their situational decision-making and navigating uncertainty. The result is greater consistency in how they treat clients and process cases under conditions of uncertainty, resulting in less biased outcomes. By using our vignette, we found that the professional work experience of a caseworker mitigates bias against an individual presumably of Middle-Eastern origin. This finding is consistent with studies suggesting that professional work experience mitigates biased decision-making (Chen, Moskowitz, and Shue 2016) and suppresses the importance of motives in predicting behavior (Jensen and Vestergaard 2017). In summary, we find that welfare sanctions are not imposed exclusively as responses to client behavior. Sanctions are also imposed as responses to client ethnicity—and the caseworkers’ professional work experience may alleviate this relationship. In line with Schram et al. (2009), the results reinforce the conclusion that policy choices not only reflect but also create the elements that underpin racial or ethnic disparity in the provision of public goods and services. Our findings contest the idea of race- and ethnic-neutral policies, instead showing that real-world social welfare systems operate in ways in which the imposing of sanctions may reflect racial and ethnic stereotypes and (re)produce racial and ethnic inequities. In times where sanctioning practices have become an increasingly integral part of most social welfare programs (Pavetti, Derr and Hesketh 2003; Hood and Peters 2004; Kettl 2005), our findings call for the attention of both policy makers and scholars. Policy makers need to know, and factor into their decision-making, that while present welfare arrangements may appear ostensibly race- and ethnic-neutral, the employees carry them out in ways that allow for racial and ethnic stereotyping, thus marking official decision-making by cognitive biases against certain races or ethnic groups. Similarly, scholars should seek to provide further insights into how and why disciplinary welfare practices produce racial and ethnic disadvantages and inequity. More specifically, we believe that our moderation result for work experience warrants future research inquiry. Managerial transformation of employee demographic characteristics, such as gender and ethnicity, is infeasible. However, professional training and development initiatives could target and cultivate the formation of the specific elements of knowledge and know-how that come with professional work experience and mitigate the influence of client ethnicity on decisions to sanction or other types of decisions. This perspective is in line with recent hermeneutic research suggesting that biases and prejudices can be corrected through dialogical inquiry (Metselaar, Meynen, and Widdershoven 2016) and research demonstrating how training can effectively debias decision makers over the long term (Dhami, Schlottmann, and Waldmann 2013; Morewedge et al. 2015). Although identifying bias in public service provision is critical, a next step for future research should be to identify the precise knowledge components that come with increasing work experience and that diminish cognitive bias in decision-making at the frontlines of public service delivery. Supplementary Material Supplementary data are available at Journal of Public Administration Research and Theory online. Acknowledgements We thank the editor and three anonymous reviewers for helpful comments and suggestions. We also thank Isak K. Houe for his assistance in generating the administrative data set. A previous version of this article was presented at the 14th Biennial Public Management Research Conference (PMRC), American University, Washington, D.C., June 8–10, 2017; at the 2nd Conference on Street-level Research in the Employment and Social Policy Area, Aalborg University Copenhagen Campus, Copenhagen, Denmark, June 21–22, 2017; and at the 77th Annual Meeting of the Academy of Management (AOM), The Public and Nonprofit Division (PNP), Atlanta, GA, August 4–5, 2017. We thank conference participants for useful feedback and suggestions. References ALS Research. 2014. Rapport: Myter om minoriteter i københavn . Copenhagen, Denmark: Als research. Amodio, David M., and Patricia G. Devine. 2006. Stereotyping and evaluation in implicit race bias: Evidence for independent constructs and unique effects on behavior. Journal of Personality and Social Psychology  91: 652– 661. Google Scholar CrossRef Search ADS PubMed  Atkins, Danielle N., and Vicky M. Wilkins. 2014. Going beyond reading, writing, and arithmetic: The effects of teacher representation on teen pregnancy rates. Journal of Public Administration Research and Theory  4: 771– 90. Auginis, Herman, and Kyle J. Bradley. 2014. Best practice recommendations for designing and implementing experimental vignette methodology studies. Organizational Research Methods  17: 351– 71. Google Scholar CrossRef Search ADS   Beer, Frederikke, Søren C. Winter, Mette H. Skou, Mads V. Stigaard, Annemette C. Henriksen, and Nina Friisberg. 2008. Statslig og kommunal Beskæftigelsesindsats . SFI Rapport 08:19. Copenhagen: SFI—The Danish National Centre for Social Research. 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   Beutel, Ann M., and Maragret M. Marini. 1995. Gender and values. American Sociological Review  60: 436– 448. Google Scholar CrossRef Search ADS   Blau, Peter M. 1955. The dynamics of bureaucracy: A study of interpersonal relations in two government agencies . Chicago, IL: University of Chicago Press. Bradbury, Mark, and J. Edward Kellough. 2008. Representative bureaucracy: Exploring the potential for active representation in local government. Journal of Public Administration Research and Theory  18: 697– 714. Google Scholar CrossRef Search ADS   Bradbury, Mark, and J. Edward Kellough. 2011. Representative bureaucracy: Assessing the evidence on active representation. The American Review of Public Administration  41: 157– 67. Google Scholar CrossRef Search ADS   Brodkin, Evelyn Z. 2011. Policy work: Street-level organizations under new managerialism. Journal of Public Administration Research and Theory  21( suppl 2): i253– i277. Google Scholar CrossRef Search ADS   Brown-Iannuzzi, Jazmin L., Ron Dotsch, Erin Cooley, and B. Keith Payne. 2017. The relationship between mental representations of welfare recipients and attitudes toward welfare. Psychological Science  (Epub ahead of print). doi: 10.1177/0956797616674999 Bursell, Moa. 2012. Name change and destigmatization among middle eastern immigrants in sweden. Ethnic and Racial Studies  35: 471– 487. Google Scholar CrossRef Search ADS   Byrne, Donn. 1971. The attraction paradigm . New York, NY: Academic Press. Caswell, Dorte, Henrik L. Andersen, Mathilde Høybye-Mortensen, Anna M. Markussen, and Sille L. Thuesen. 2011. Når kassen smækkes i. Analyser af økonomiske sanktioner over for kontanthjælpsmodtagere . Copenhagen: AKF. Caswell, Dorte, Leena Eskelinen, and Søren P. Olesen. 2011. Identity work and client resistance underneath the canopy of employment policy. Qualitative Social Work  12: 8– 23. Google Scholar CrossRef Search ADS   Chen, Daniel L., Tobias J. Moskowitz, and Kelly Shue. 2016. Decision making under the gambler’s fallacy: Evidence from asylum judges, loan officers, and baseball umpires. The Quarterly Journal of Economics  131: 1181– 1242. Google Scholar CrossRef Search ADS   Christiansen, Niels F., Klaus Petersen, Nils Edling, and Per Haave. 2006. The nordic model of welfare: A historical reappraissal . Copenhagen, Denmark: Museum Tusculanum Press. Cohen, Jacob. 1988. Statistical power analysis for the behavioral sciences . Hillsdale, NJ: Lawrence Erlbaum. Crowder-Meyer, Melody. 2007. Gender differences in policy preferences and priorities . Paper presented at the annual meeting of the Midwest Political Science Association, Chicago, IL, April 12, 2007. Currarini, Sergio, and Friederike Mengel. 2016. Identity, homophily and in-group bias. European Economic Review  90: 40– 55. Google Scholar CrossRef Search ADS   Dhami, Mandeep, Anne Schlottmann, and Michael R. Waldmann. 2013. Judgment and decision making as a skill: Learning, development, and evolution . Cambridge, UK: Cambridge Univ. Press. Danziger, Sandra K., and Kristin S. Seefeldt. 2003. Barriers to employment and the ‘hard to serve’: Implications for services, sanctions, and time limits. Social Policy and Society  2: 151– 160. Google Scholar CrossRef Search ADS   DeHart-Davis, Leisha, Justin Marlowe, and Sanjay K. Pandey. 2006. Gender dimensions of public service motivation. Public Administration Review  66: 837– 887. Google Scholar CrossRef Search ADS   Devine, Patricia G., E. Ashby Plant, David M. Amodio, Eddie Harmon-Jones, and Stephanie L. Vance. 2002. The regulation of explicit and implicit race bias: The role of motivations to respond without prejudice. Journal of Personality and Social Psychology  82: 835. Google Scholar CrossRef Search ADS PubMed  Dolan, Julie. 2000. The senior executive service: Gender, attitudes, and representative bureaucracy. Journal of Public Administration Research and Theory  10: 513– 30. Google Scholar CrossRef Search ADS   Downs, Anthony. 1967. Inside bureaucracy . Boston: Little, Brown and Company. Google Scholar CrossRef Search ADS   Eagly, Alice H., Amanda B. Diekman, Mary C. Johannesen-Schmidt, and Anne M. Koenig. 2004. gender gaps in sociopolitical attititudes: A social psychological analysis. Journal of Personality and Social Psychology  87: 796– 816. Google Scholar CrossRef Search ADS PubMed  Ejrnæs, Morten, and Merete Monrad. 2012. Vignetmetoden: Sociologiske metode og redskab til faglig udvikling . Copenhagen: Akademisk Forlag. Esping-Andersen, Gøsta. 1990. The three worlds of welfare capitalism. Princeton, NJ: Princeton University Press. Flyvbjerg, Bent. 2001. Making social science matter: Why social inquiry fails and how it can succeed again . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Frederickson, H. George. 1990. Public administration and social equity. Public Administration Review  50: 228– 237. Google Scholar CrossRef Search ADS   Gaasholt, Øystein, and Lise Togeby. 1995. I syv sind: Danskernes holdninger til flygtninge og indvandrere . Aarhus, Denmark: Forlaget Politica. Gilens, Martin. 1999. Why americans hate welfare: Race, media, and the politics of antipoverty policy . Chicago, IL: Univ. of Chicago Press. Google Scholar CrossRef Search ADS   Grissom, Jason A., Jill Nicholson-Crotty, and Sean Nicholson-Crotty. 2009. Race, region, and representative bureaucracy. Public Administration Review  69: 911– 19. Google Scholar CrossRef Search ADS   Hancock, Ange-Marie. 2004. The politics of disgust: The public identity of the welfare queen . New York: New York Univ. Press. Hasenfeld, Yeheskel, Toorjo Ghose, and Kadyce Larson. 2004. The logic of sanctioning welfare recipients: An empirical assessment. Social Service Review  78: 304– 19. Google Scholar CrossRef Search ADS   Hindera, John J. 1993. Representative bureaucracy: Further evidence of active representation in the EEOC district offices. Journal of Public Administration Research and Theory  3: 415– 29. Hogg, Michael A., and Dominic Abrams. 1998. Social identifications: A social psychology of intergroup relations and group processes . London: Routledge. Hood, Christopher, and Guy Peters. 2004. The middle aging of new public management: Into the age of paradox? Journal of Public Administration Research and Theory  14: 267– 82. Google Scholar CrossRef Search ADS   Hupe, Peter, and Michael Hill. 2007. Street-level bureaucracy and public accountability. Public Administration  85: 279– 99. Google Scholar CrossRef Search ADS   Jacobsen, Brian A. 2012. Muslims in Denmark: A critical evaluation of estimations. In Islam in Denmark: The challenge of diversity , ed. Jørgen S. Nielsen, 31– 56. Plymouth, UK: Lexington Books. Jensen, Ulrich T., and Christian F. Vestergaard. 2017. Public service motivation and public service behaviors: Testing the moderating effect of tenure. Journal of Public Administration Research and Theory  27: 52– 67. Google Scholar CrossRef Search ADS   Johnson, Brian D. 2014. Judges on trial: A reexamination of judicial race and gender. Effects across modes of conviction. Criminal Justice Policy Review  25: 159– 84. Google Scholar CrossRef Search ADS   Kahneman, Daniel. 2011. Thinking, fast and slow . New York: Farrar, Straus and Giroux. Kahneman, Daniel, and Shane Frederick. 2002. Representativeness revisited: Attribute substitution in intuitive judgment. In Heuristics and biases: The psychology of intuitive judgment , eds. Thomas Gilovich, Dale Griffin, and Daniel Kahneman, 49– 81. Cambridge, UK: Cambridge Univ. Press. Google Scholar CrossRef Search ADS   Kahneman, Daniel, Amos Tversky, and Paul Slovic. 1982. Judgment under uncertainty: Heuristics & biases . Cambridge, UK: Cambridge University Press. Google Scholar CrossRef Search ADS   Keiser, Lael R., Vicky M. Wilkins, Kenneth J. Meier, and Catherine A. Holland. 2002. Lipstick and logarithms: Gender, institutional context, and representative bureaucracy. American Political Science Review  96: 553– 563. Google Scholar CrossRef Search ADS   Kettl, Donald F. 2005. The global public management revolution (2e) . Washington, DC: Brookings Institution Press. Krislov, Samuel J. 1974. Representative bureaucracy . Englewood Cliffs, NJ: Prentice Hall. Larsen, Christian A. 2012. Fordomme over for etniske minoriteter i Danmark, Sverige, Storbritannien og USA. Politica  44: 488– 504. Liebig, Thomas. 2007. The labour market integration of immigrants in denmark . OECD Social, Employment, and Migration Working Papers No. 50. Paris: OECD. Lipsky, Michael. 1980. Street-level bureaucracy: Dilemmas of the individual in public services . New York: Russell Sage Foundation Publications. Meier, Kenneth J. 1993a. Representative bureaucracy: A theoretical and empirical exposition. In Research in public administration , ed. James L. Perry,vol. 2, 1– 35. New Greenwich, CT: JAI Press. Meier, Kenneth J. 1993b. Latinos and representative bureaucracy testing the thompson and henderson hypotheses. Journal of Public Administration Research and Theory  3: 393– 414. Meier, Kenneth J., and Joseph Stewart. 1992. The impact of representative bureaucracies: Educational systems and public policies. The American Review of Public Administration  22: 157– 71. Google Scholar CrossRef Search ADS   Meldgaard, Eva V. 2005. Muslimske fornavne i Danmark. http://nfi.ku.dk/publikationer/webpublikationer/muslimske_fornavne (accessed October 7, 2016). Merton, Robert K. 1957. Social theory and social structure . Glencoe, IL: The Free Press. Metselaar, Suzanne, Gerben Meynen, and Guy Widdershoven. 2016. Reconsidering bias: A hermeneutic perspective. The American Journal of Bioethics  16: 33– 35. Google Scholar CrossRef Search ADS PubMed  Morewedge, Carey K., Haewon Yoon, Irene Scopelliti, Carl W. Symborski, James H. Korris, and Karim S. Kassam. 2015. Debiasing decisions: Improved decision making with a single training intervention. Policy Insights form the Behavioral and Brain Sciences  2: 129– 140. Google Scholar CrossRef Search ADS   Okonofua, Jason A., and Jennifer L. Eberhardt. 2015. Two strikes: Race and the disciplining of young students. Psychological Science  26: 617– 24. Google Scholar CrossRef Search ADS PubMed  Olson, Jeremiah C. 2016. Race and punishment in american prisons. Journal of Public Administration Research and Theory  26: 758– 68. Google Scholar CrossRef Search ADS   Pager, Devah. 2007. Marked: Race, crime, and finding work in an era of mass incarceration . Chicago, IL: Univ. of Chicago Press. Google Scholar CrossRef Search ADS   Pavetti, LaDonna, Michelle K. Derr, and Heather Hesketh. 2003. Review of sanction policies and research studies: final literature review . Washington, DC: US Department of Health and Human Services. https://aspe.hhs.gov/sites/default/files/pdf/72646/full-report.pdf (accessed January 3, 2017). Pickerill, J. Mitchell, Clayton Mosher, and Travis Pratt. 2009. Search and seizure, racial profiling, and traffic stops: A disparate impact framework. Law & Policy  31: 1– 30. Google Scholar CrossRef Search ADS   Quillian, Lincoln. 2008. Does unconscious racism exist. Social Psychology Quarterly  71: 6– 11. Google Scholar CrossRef Search ADS   Rabin, Matthew, and Dmitri Vayanos. 2010. The gambler’s and hot-hand fallacies: Theory and applications. Review of Economic Studies  77: 730– 78. Google Scholar CrossRef Search ADS   Rand, David G., Victoria L. Brescoll, Jim A. C. Everett, Valerio Capraro, and Hélène Barcelo. 2016. Social heuristics and social roles: Intuition favors altruism for women but not for men. Journal of Experimental Psychology  145: 389– 96. Google Scholar CrossRef Search ADS PubMed  Redder, Gitte, and Allan Christensen. 2016. Danskerne: Flygtninge og integration er vores allerstørste problem. Ugebrevet A4 (January 4, 2016). http://www.ugebreveta4.dk/danskerne-flygtninge-og-integration-er-vores-allerst_20333.aspx (accessed January 4, 2017). Sahin, Sultan, and Pernille Dreyer. 2015. Hvem er fatma og muhammed – og hvad mener de om Danmark? Berlingske, 8 August 2015. http://www.b.dk/nationalt/hvem-er-fatma-og-muhammed-og-hvad-mener-de-om-danmark (accessed October 7, 2016). Schram, Sanford F. 2006. Welfare discipline: Discourse, governance, and globalization . Philadelphia, PA: Temple Univ. Press. Schram, Sanford F., Joe Soss, Richard C. Fording, and Lina Houser. 2009. Deciding to discipline: race, choice, and punishment at the frontlines of welfare reform. American Sociological Review  74: 398– 422. Google Scholar CrossRef Search ADS   Schwartz, Shalom H., and Tammy Rubel. 2005. Sex differences in value priorities: cross-cultural and multimethod studies. Journal of Personality and Social Psychology  89: 1010– 1028. Google Scholar CrossRef Search ADS PubMed  Selden, Sally C. 1997. Representative bureaucracy: examining the linkage between passive and active representation in the farmers home administration. The American Review of Public Administration  27: 22– 42. Google Scholar CrossRef Search ADS   Selden, Sally C., Jeffrey Brudney, and J. Edward Kellough. 1998. Bureaucracy as a representative institution: toward a reconciliation of bureaucratic government and democratic theory. American Journal of Political Science  42: 719– 744. Shapiro, Robert Y., and Harpreet Mahajan. 1986. Gender differences in policy preferences: a summary of trends from the 1960s to the 1980s. Public Opinion Quarterly  50: 42– 61. Google Scholar CrossRef Search ADS   Soss, Joe, Richard C. Fording, and Sanford F. Schram. 2008. The color of devolution: the politics of local punishment in an era of neo-liberal welfare reform. American Journal of Political Science  52: 536– 53. Google Scholar CrossRef Search ADS   Soss, Joe, Richard C. Fording, and Sanford F. Schram. 2011a. The organization of discipline: from performance management to perversity and punishment. Journal of Public Administration Research and Theory  21( suppl 2): i203– i232. Google Scholar CrossRef Search ADS   Soss, Joe, Richard C. Fording, and Sanford F. Schram. 2011b. Diciplining the poor: neoliberal paternalsm and the persistent power of race . Chicago, IL: Univ. of Chicago Press. Statistics Denmark. 2015. Indvandrere i Danmark 2015 [Immigrants in Denmark 2015] . Copenhagen: Danmarks Statistik. Statistics Denmark. 2016. Navne i hele befolkningen [Names of the Total Population]. http://www.dst.dk/da/Statistik/emner/navne/navne-i-hele-befolkningen (accessed October 7, 2016). Sunshine, Jason, and Tom R. Tyler. 2003. The role of procedural justice and legitimacy in shaping public support for policing. Law & Society Review  37: 513– 548. Google Scholar CrossRef Search ADS   Tajfel, Henri, and John C. Turner. 1986. The social identity theory of intergroup behavior. In Psychology of intergroup relations , 2nd ed., eds. Stephen Worschel and William G. Austin, 7– 24. Chicago, IL: Nelson Hall. Thomsen, Jens Peter F. 2006. Konflikten om de nye danskere: Om danskernes holdninger til etniske minoriteters kultur og rettigheder . Copenhagen, Denmark: Akademisk Forlag. Togeby, Lise. 2004. Man har et standpunkt…: Om stabilitet og forandring i befolkningens holdninger . Magtudredningen. Aarhus, Denmark: Aarhus Universitetsforlag. Tyler, Tom R. 2003. Procedural justice, legitimacy, and the effective rule of law. Crime and Justice  30: 283– 357. Google Scholar CrossRef Search ADS   Wilkins, Vicky M., and Lael R. Keiser. 2006. Linking active and passive representation for gender: The case of child support. Journal of Public Administration Research and Theory  16: 87– 102. Google Scholar CrossRef Search ADS   Wilkins, Vicky M., and Brian N. Williams. 2008. Black or blue: Racial profiling and representative bureaucracy. Public Administration Review  68: 654– 64. Google Scholar CrossRef Search ADS   Footnotes 1 For the purpose of this study, “ethnic minorities” refers to non-Western immigrants having settled in Denmark and their descendants (the “second” or “third” generation). Statistics Denmark classifies an individual as a native Dane, or immigrant (“first-generation immigrant”), or a descendant of immigrants (“second-generation immigrant”) (Statistics Denmark 2015, pp. 11–12). A person is classified as native Danish if at least one of the person’s parents was born in Denmark and holds Danish citizenship. Individuals born outside of Denmark are considered first-generation immigrants, whereas individuals with parents born outside of Denmark are classified as descendants (second-generation immigrants). We also follow Statistic Denmark in terms of classifying non-Western immigrants and their descendants as those individuals not originating in the EU-28, Iceland, Norway, Switzerland, the European micro-states, the United States, Canada, Australia, or New Zealand (Statistics Denmark 2015, 11–12). 2 Figures based on authors’ calculations, using population data (FOLK2) from Statistics Denmark, first quarter of 2016. Available at www.statistikbanken.dk. 3 Ranking based on authors’ calculations, using population data (FOLK1) from Statistics Denmark, first quarter of 2016. Available at www.statistikbanken.dk. 4 Figures based on authors’ calculations, using the latest available unemployment data (RAS1F) from Statistics Denmark. Available at www.statistikbanken.dk. 5 Furthermore, non-Western immigrants are occasionally stereotyped as having a penchant for petty crime and lower levels of educational attainment (ALS Research 2014), as well as being unwilling to accept or live by the democratic values of the countries in which they settle (Bursell 2012)—group reputations that may also affect perceptions of deservingness with respect to public assistance. 6 The administrative data contain information on the share of cash benefit recipients who are non-Western immigrants or their descendants for each employment agency but do not allow us to focus especially on those of Middle-Eastern origin. However, analyses conducted by Statistics Denmark show that a larger percentage of non-Western immigrants than ethnic Danes live of public support (cash or other benefits). The countries of origin most prevalent among those non-Western immigrants are Turkey, Syria, Iraq, and Lebanon (Statistics Denmark 2016). 7 The survey experimental findings have a modest claim to ecological validity. The caseworkers’ responses are based on a description of a hypothetical client in a survey. This situation is clearly different from actual casework. For example, the respondents did not interact with a real person, and they were asked to base their responses on only the information that the vignette provided. In real life, caseworkers can request or seek additional data, and face-to-face client meetings are likely to inform decision-making. However, Study 1 complements Study 2 in terms of ecological validity (similar to the way that Study 2 complements Study 1 in terms of internal validity). Although the administrative data analysis does not offer the same potential for causal inference as the survey experimental analysis, it allows us to motivate the survey experiment with data reflecting sanctioning decisions as they naturally occur in real-life cases and work environments. 8 A skewness-kurtosis test for normality shows that the score distribution is slightly left-skewed (−.93, p < .001) but has a kurtosis similar to that of a standard normal distribution (2.84, p = .50). As a robustness check, because of the left-skewed distribution, we have re-analyzed the survey experimental data using quantile (median) regression instead of OLS. Although the median regression estimates are slightly smaller and accompanied by larger standard errors, the estimates do not change the substantive findings. 9 As a robustness test, we also estimated the work experience interaction model with the inclusion of an index variable capturing the caseworkers’ general perceptions of clients’ efforts and motivations for obtaining employment. In theory, the moderation effect could be driven by more negative perceptions of clients among less experienced caseworkers (rather than by work experience per se). We constructed the index measure using three Likert-scale items from the caseworker survey: “Almost all clients are sincerely very much motivated to obtain employment”; “Many clients are at times unreasonable or troublesome (reversed)”; and “A fair share of clients do not, in reality, wish to work (reversed).” Cronbach’s alpha was .79. As expected, a more positive perception of clients’ efforts and motivations is negatively associated with decisions to sanction (β = −.557; p < .001). More importantly, however, including the index variable in the model does not qualitatively change the work experience interaction term estimate (β = −.083, p = .039)—thus suggesting that our work experience moderation result is not driven (or biased) by general perceptions of the clients’ efforts and motivations for obtaining employment. A second concern relates to the dynamics of policy contexts. Although the use of sanctions has been an integral part of the Danish unemployment regime since the late 1980s, the cash benefit program has been marked by a gradually increasing emphasis on sanctioning clients for noncompliance (Caswell et al. 2011). Given this trajectory, could the moderating effect of work experience be the result of a policy socialization process in which the caseworkers’ sanctioning norms are shaped by the unemployment policy operating at the point in time when they entered the profession? In the attempt to consider this concern, we examined the correlation between work experience and overall sanctioning propensity—for the full sample and by treatment group. Our expectation is that said policy socialization process would be reflected in a lesser overall propensity to sanction among the more experienced caseworkers. However, the analyses do not support this notion. To the contrary, we find weak evidence that the more experienced caseworkers are more likely to recommend sanction than their less experienced peers (p = .09). © 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

Punishment on the Frontlines of Public Service Delivery: Client Ethnicity and Caseworker Sanctioning Decisions in a Scandinavian Welfare State

Loading next page...
 
/lp/ou_press/punishment-on-the-frontlines-of-public-service-delivery-client-cT4j0VdDop
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/muy018
Publisher site
See Article on Publisher Site

Abstract

Abstract Many public welfare programs give public employees discretionary authority to dispense sanctions when clients do not follow or comply with the policies and procedures required for receiving welfare benefits. Yet research also shows that public employees’ use of discretion in decision-making that affects clients can occasionally be marked by racial biases and disparities. Drawing on the Racial Classification Model (RCM) for a theoretical model, this article examines how client ethnicity shapes public employees’ decisions to sanction clients. Using Danish employment agencies as our empirical setting, we present findings from two complementary studies. Study 1 uses nationwide administrative data. Examining sanctioning activity at the employment agency-level, we find that agencies with a larger percentage of clients being non-Western immigrants or their descendants impose a greater overall number of sanctions and dispense them with greater frequency. Study 2 uses survey experimental data to build on this finding. Addressing concerns about internal validity and a need for analyses at the individual employee-level, we present survey experimental evidence that employment agency caseworkers are more likely to recommend sanctions for ethnic minority (Middle-Eastern origin) clients than for ethnic majority (Danish origin) clients. Moreover, we investigate how three caseworker characteristics—ethnicity, gender, and work experience—condition the relationship between client ethnicity and caseworkers’ decisions to sanction clients. Although we find no moderation effects for ethnicity or gender, work experience appears to diminish the influence of client ethnicity on the caseworkers’ sanctioning decisions. Overall, our studies support the likelihood that ethnic minority clients will be punished more often for policy infractions than ethnic majority clients—and that caseworker work experience mitigates part of this bias. Introduction In the wake of neoliberal public policy reforms, public welfare programs worldwide have become more directive in setting behavioral expectations for clients and more punitive in responding to client noncompliance (Hood and Peters 2004; Kettl 2005). Social welfare programs use a range of incentives, monitoring mechanisms, and restrictive regulations and rules aimed at modifying client behaviors and ensuring that client groups practice appropriate behaviors (Danziger and Seefeldt 2003; Pavetti, Derr, and Hesketh 2003; Schram 2006). Many of these programs involve a system for dispensing sanctions—penalties that reduce or terminate welfare benefits in response to client infractions of policy, that is, noncompliance (Pavetti, Derr, and Hesketh 2003; Schram et al. 2009). However, the dispensing of sanctions is not an automated process devoid of personal cognitions or beliefs. Although program rules may allow for and prescribe sanctions under certain conditions, rules are not self-executing (Blau 1955; Merton 1957): They require judgments from frontline workers in their application (Hupe and Hill 2007, p. 281). In many areas of public welfare administration, public employees have substantial discretionary authority to evaluate and decide whether and, if so, when clients should be penalized for failure to comply with policy prescriptions and procedures (Lipsky 1980; Hasenfeld, Ghose, and Larson 2004; Brodkin 2011; Soss, Fording, and Schram 2011a). Public employees have a responsibility to ensure that all citizens, in their dealings with public agencies and services, receive fair and equitable treatment, free from discrimination on the basis of race, ethnicity, or other attributes (Frederickson 1990). In addition to equity as a normative ideal of democratic administration and a principle of international law, empirical work suggests that procedural fairness is important in ensuring the legitimacy of public institutions in the eyes of citizens (Sunshine and Tyler 2003; Tyler 2003). Thus, an understanding of the potential effects of client characteristics such as race and ethnicity on public employees’ decisions to sanction is of broad societal interest and importance. This article examines public employees’ decisions to sanction clients, focusing first on the role of client ethnicity: What are the effects of client ethnicity on public employees’ discretionary decisions to sanction clients? Second, we also consider how the interplay between client ethnicity and various employee characteristics shapes these decisions: How and to what extent do public employee characteristics—ethnicity, gender, and work experience—moderate the potential disparity between sanctioning decisions for ethnic minority clients and for ethnic majority clients? Our analyses help explain the ways that bureaucratic discretion may lead to disparate treatment of comparable citizens of different races or ethnicities in their dealings with government and public administration. Research reveals that citizen or client race may affect public employees’ decision-making and exercise of discretion in the judicial system (Pager 2007; Johnson 2014), law enforcement (Wilkins and Williams 2008; Pickerill, Mosher, and Pratt 2009), incarceration (Olson 2016), education and schooling (Meier and Stewart 1992; Meier 1993b; Grissom, Nicholson-Crotty, and Nicholson-Crotty 2009; Atkins and Wilkins 2014; Okonofua and Eberhardt 2015), and other public agency settings (Hindera 1993; Selden 1997; Selden, Brudney, and Kellough 1998; Bertrand and Mullainathan 2004). This research provides valuable insights into both racial disparity in public services provision and representative bureaucracy (Krislov 1974; Meier 1993a). However, systematic studies of how client race or ethnicity may affect the micro-level decisions of frontline workers are few (see Bradbury and Kellough 2011). Little empirical research—using individual-level data and research designs that can allow for a causal interpretation of the results—has examined racial or ethnic bias in individual-level welfare sanction decisions. One exception is a study by Schram et al. (2009); see also Soss, Fording, and Schram 2011b). In terms of theory, they draw on the Racial Classification Model (RCM) developed by Soss, Fording, and Schram (2008), and examine how implicit racial bias shapes employment caseworkers’ decisions to dispense sanctions for clients. With the United States as their empirical setting, the results indicate differences in the treatment of female clients of different races. Both Latina and black clients are more likely to be recommended for sanctions than white clients. Moreover, Schram et al. (2009) show that two other client characteristics—family size and sanctioning history—may widen these racial disparities. Our study extends Schram et al.’s (2009) work and the RCM with new insights. Specifically, our study expands scholarly understanding of racial or ethnic disparity in public service provision in at least three ways. First, using the Scandinavian welfare state of Denmark as our empirical setting, we provide insights into the applicability and generalizability of both the RCM and Schram et al.’s (2009) findings outside of the US context. Does the RCM predict employment caseworkers’ decisions to sanction clients in a non-US country setting? As we elaborate later on, Denmark is radically different from the United States in terms of public welfare services, racial/ethnic composition, and the historical development of race/ethnicity as a salient policy issue. Second, we consider the role of the employment caseworkers’ ethnicity, gender, and professional work experience as important moderators, by drawing from different streams of behavioral economics (Kahneman and Frederick 2002; Chen, Moskowitz, and Shue 2016), social psychology (Tajfel and Turner 1986; Eagly et al. 2004; Schwartz and Rubel 2005; Rand et al. 2016), and public administration research (Downs 1967; Meier 1993a; Selden 1997; Flyvbjerg 2001; DeHart-Davis, Marlowe, and Pandey 2006; Bradbury and Kellough 2008, 2011). In doing so, we take an important step in unlocking both the micro-foundations of representation on the frontlines of public service and the importance of work experience in minimizing the possible consequences of cognitive biases affecting the distribution of public benefits. Our examination also constitutes a theoretical extension of the RCM. Although Schram et al. (2009) theorize and test how various client markers interact with client race in shaping caseworkers’ sanctioning decisions—and thus extend the RCM with insights into the moderating effects of client characteristics—we examine and expand our knowledge about the moderating effects of caseworker characteristics. By identifying how caseworker characteristics moderate disparity in sanctioning decisions for clients of a different ethnicity, we expand scholarly understanding of when and why welfare programs involving systems for dispensing sanctions may (or may not) produce racial and ethnic disparities. Third, we advance current public administration research by examining disparities in sanctioning activity based on ethnicity using both agency-level administrative data and individual-level experimental data. Study 1 uses nationwide administrative data from the national employment program. We look at three sanctioning indicators at the employment agency-level and test for differences between agencies with different percentages of clients being non-Western immigrants or their descendants. Study 2 uses individual-level experimental data from a national survey among Danish employment caseworkers to build on the findings from Study 1. We present the respondents with a vignette containing a violation of program rules by a client. In the vignette, the client’s ethnicity (majority/minority) is randomly assigned.1 Specifically, client ethnicity was either blinded (unknown to the respondents) or cued as either ethnic Danish or a non-Western ethnic minority (of Middle-Eastern origin). After having read the vignette, the respondents were asked whether they would impose a sanction. The two studies thus complement each other in important ways. Study 1 relates to agency-level sanctioning decisions as they naturally occur in real-life cases and work environments and, therefore, complements Study 2 in terms of ecological validity. Study 2 relates to individual-level sanctioning decisions and allows for causal inferences and, therefore, complements Study 1 in terms of internal validity and findings at the individual employee-level. The article proceeds as follows. First, we describe our application of the RCM in the Danish context, and hypothesize on the nature of the relationship between client ethnicity and a caseworker’s decision to sanction a client. Second, we develop hypotheses on the moderating effects of caseworker characteristics—ethnicity, gender, and work experience–on the decision to sanction. Third, we review the rules and procedures of the Danish unemployment program regarding citizens’ rights and entitlements to unemployment benefits and caseworkers’ decisions to sanction. Fourth, we present the results of our empirical analyses. Finally, we discuss our findings and their implications for both policy and future research. Theory and Hypotheses The Racial Classification Model (RCM) and the Danish Context To understand how client ethnicity may shape employment caseworkers’ decisions to sanction clients, we turn to Soss, Fording, and Schram’s (2008) RCM. In brief, the RCM explains how the race or ethnicity of policy targets (clients) affects the policy implementation decisions of public employees. The RCM represents “a cognitive model of policy decision making built on the necessity of social classification and consequences of group reputation […] specifying how and when racial classifications should affect target group constructions and, hence, officials’ choice regarding policy design and implementation” (Soss, Fording, and Schram 2008, p. 539). Although Soss, Fording, and Schram (2008) discuss the RCM in terms of race, they explicitly state that the RCM may also apply to social classifications defined by race or ethnicity (p. 540). According to Soss, Fording, and Schram (2008), the RCM rests on three premises: (1) To be effective at designing and applying policies to specific target groups, policy actors must rely on salient social classifications; (2) when racial (or ethnic) minorities are salient in a policy context, race or ethnicity is more likely to provide a basis for classifications of targets; and (3) the likelihood of racially or ethnically patterned policy outcomes will be positively associated with the degree of contrast in policy actors’ perceptions of targets. The extent to which policy actors differentiate among target groups and treat them differently is a function of the perceived contrasts among the groups. The RCM has been the basis for empirical inquiry into how the racial characteristics of a population affect social welfare policy choices by legislatures (Soss, Fording, and Schram 2008) as well as disparities faced by African-Americans and Latinas in the United States with respect to sanctions for policy noncompliance (Schram et al. 2009). Relative to Schram et al. (2009), our study covers a different employment policy and a different country, Denmark, with a setting that is radically different from that of the United States in terms of the racial and ethnic majority and minority composition. Ethnic Danes constitute the large majority of the Danish population (88%), whereas non-Western ethnic minorities constitute a much smaller but growing group (1% in 1980, 5% in 2000, and 8% in 2016).2 The four most common countries of origin among non-Western ethnic minorities in Denmark are Turkey, Iraq, Syria, and Lebanon,3 with Islam the most prevalent religion among them (Jacobsen 2012). In seeking to extend the RCM, we apply the model and its underlying assumptions to the Danish context. The Scandinavian countries, including Denmark, are typically viewed as more collectivist and homogenous than the United States, with their preferences for universal and more generous social welfare policies (Esping-Andersen 1990; Caswell, Eskelinen, and Olesen 2011). Scholars root the historical development of the Danish welfare state in a social democratic ideology originating from a strong sense of community and social solidarity among Danes as a homogenous ethnic group (Christiansen et al. 2006), starting in the 1870s and predating the first waves of immigration from the Middle East to Denmark in the 1980s. Although traditionally Denmark has not been marked by the same racial and ethnic cleavages as the United States, the increasing influx of non-Western immigrants since the 1980s has made questions about immigration, national culture, and the societal integration of people of Middle-Eastern origin an important topic in current public and political debates. For example, a recent poll shows that 65 percent of Danes view “refugees, immigration, and integration” as the most important political issue (Redder and Christensen 2016). One particularly salient policy issue concerns the gap in employment rates for non-Western immigrants and their descendants relative to ethnic Danes—and how to minimize this gap (Liebig 2007). In terms of labor market attachment, the employment rates among non-Western ethnic minorities are lower than among ethnic Danes. The employment rates among ethnic Danish males and females in 2014 were 76 and 73 percent, respectively. In comparison, the employment rates among non-Western male and female immigrants were 53 and 46 percent, respectively—and, among non-Western male and female descendants of immigrants, 53 and 52 percent, respectively.4 At the same time, ethnic minorities of non-Western origin are also subject to stereotyping in Denmark about their willingness to work. One preconception is that many non-Western immigrants come to Denmark for the benefits of the Danish welfare system and that more non-Western immigrants than ethnic Danes prefer living off welfare rather than pursuing paid employment. Although academic inquiry into the Danes’ attitudes towards the work ethic of non-Western immigrants is somewhat limited, a 20-year-old survey study shows that 65 percent of Danes agree that “[m]any immigrants come to Denmark to take advantage of our social welfare system” (Gaasholt and Togeby 1995, p. 40)—and scholars suggest that the general attitudes towards immigrants in Denmark have remained fairly stable over time (Togeby 2004; Thomsen 2006). Moreover, comparative research based on General Social Survey data supports the existence of Danish stereotypes of “laziness” and “preference for welfare relative to work” for non-Western immigrants, stereotypes similar to those existing for African-Americans in the United States (Larsen 2012). US researchers have found that this notion of laziness is one of the strongest US racial stereotypes, influencing citizen attitudes towards welfare and deservingness of public assistance (Gilens 1999, see also Hancock 2004; Brown-Iannuzzi et al. 2017).5 The RCM does not argue that the presence or influence of these ethnic stereotypes of a certain group’s willingness to work is a subject about which individual caseworkers would necessarily be cognizant of. Indeed, social psychology suggests that cognitive biases relating to race and ethnicity may be unconscious—and that the individual acting in a biased manner may be unaware of his or her bias (Devine et al. 2002; Amodio and Devine 2006). In the context of a caseworker’s decision to sanction a client who does not comply with the requirements of, say an unemployment program policy, the ethnic stereotypes previously discussed may influence the caseworker’s cognitive classification and perceptions of that client, thereby resulting in systematically different sanctioning behaviors for clients of different ethnicity. These stereotypes may particularly affect how the caseworker perceives and assesses the deservingness of clients of Middle-Eastern origin relative to clients of Danish origin. Stereotypes and group reputations may lead caseworkers to evaluate clients appearing to be non-Western immigrants of Middle-Eastern origin as less deserving of a second chance (or the benefit of the doubt) when they are in policy noncompliance. Because of cognitive biases rooted in ethnicity and ethnic stereotypes, caseworkers may therefore be more likely to recommend sanctions for clients of Middle-Eastern origin than for clients of Danish origin: Hypothesis 1: Caseworkers are more likely to recommend sanctions for ethnic minority clients (here, clients of Middle-Eastern origin) than for ethnic majority clients (here, clients of ethnic Danish origin). In addition to the direct effect of client ethnicity on a caseworker’s decision to sanction, a range of caseworker characteristics could potentially moderate the ways in which client ethnicity affects a caseworker’s sanctioning decisions. Our focus on the caseworker characteristics of ethnicity, gender, and work experience is guided by research that highlights these variables as possible moderators of the effects of client ethnicity on employee decision-making. Both representative bureaucracy research (Bradbury and Kellough 2008; Meier 1993a; Selden 1997) and social psychology theory (Tajfel and Turner 1986) suggest that the caseworkers’ ethnicity can have important implications for the way they understand, identify with, and treat ethnic minority clients. Similarly, insights from social psychology research (Eagly et al. 2004; Schwartz and Rubel 2005; Rand et al. 2016) and research on public administration and public opinion (DeHart-Davis, Marlowe, and Pandey 2006; Crowder-Meyer 2007) suggest that the caseworkers’ gender may moderate the effect of client ethnicity on sanctioning decisions. Finally, insights from behavioral economics research (Kahneman and Frederick 2002; Chen, Moskowitz, and Shue 2016) and management research (Downs 1967; Flyvbjerg 2001) suggest that work experience may mitigate ethnic bias in caseworkers’ decisions to sanction. We believe that our focus on ethnicity, gender, and work experience constitutes a solid foundation for understanding the role of caseworker characteristics in the context of the RCM and the occurrence of racial or ethnic disparity in public services provision. Caseworker Ethnicity Several research perspectives suggest that human beings tend to hold cognitive biases that favor their own social in-group, in turn possibly moderating how client ethnicity shapes caseworkers’ decisions to sanction. As Bradbury and Kellough (2011, p. 160) note, the public administration literature on representative bureaucracy shows that “public organizations with larger proportions of women and/or minorities in decision-making roles are more likely to produce outcomes compatible with the interests of women and/or minorities than similar organizations with fewer women and/or minorities” (e.g., Meier and Stewart 1992; Hindera 1993; Keiser et al. 2002; Wilkins and Keiser 2006). The theory of representative bureaucracy suggests that the demographic characteristics of public employees shape their attitudes, thereby shaping their behavior and decision-making (Meier 1993a). Employees and citizens with similar demographic characteristics (e.g., ethnicity) share common experiences and often values and attitudes (Dolan 2000; Bradbury and Kellough 2008). This shared experience explains why female and/or ethnic minority employees are more likely to make decisions and adopt an advocacy role that favor female and/or ethnic minority clients (Meier 1993a; Selden 1997; Bradbury and Kellough 2008). Social psychological theories elaborate on the similarity-dissimilarity process providing individuals with cognitive biases toward individuals of the same ethnicity as themselves. Similarity/attraction theory (Byrne 1971) and social identity theory (Tajfel and Turner 1986) emphasize how human beings tend to better understand and identify with people exhibiting the same characteristics as themselves. Demographic similarity on a dimension such as ethnicity promotes compatibility, mutual understanding, and identity reinforcement, whereas dissimilarity creates incompatibility and disagreement (Hogg and Abrams 1998). Social identity theory further explains how motives and behavior are affected by “in group”–“out group” stereotyping, group cohesion, and emotional sympathy. The theory posits that individual identities are based on self-categorization into social groups, that is, people classify those who are similar (e.g., in terms of ethnicity) as the in-group, and those they perceive as being different as the out-group (Tajfel and Turner 1986). Individuals who belong to the in-group coalesce and bond based upon shared traits or experiences (Currarini and Mengel 2016). The perceived connection between members of the same in-group (e.g., caseworkers and clients who belong to an ethnic minority group) might inform the basis of an in-group favorability bias or a debiasing of intra-in-group judgments and decision-making. At the minimum, this connection may mitigate the application of ethnic stereotyping. Combining these insights, we theorize that caseworkers have cognitive biases that make them tend to favor clients of their own ethnicity—and that these biases may moderate how client ethnicity shapes their decisions to sanction. Drawing on social identity theory (Tajfel and Turner 1986), we expect that caseworkers tend to engage clients of the same ethnicity as themselves with greater sympathy, understanding, and equity. Because of ethnicity-similarity bias, ethnic stereotyping is less likely to affect decision-making—that is, create a disparity in sanctioning decisions between ethnic minority and ethnic majority clients—when those clients are of the same ethnicity (in-group) as the caseworker. In summary, we expect the propensity to sanction ethnic minority clients more than ethnic majority clients to be smaller for caseworkers who themselves belong to an ethnic minority group: Hypothesis 2a: Caseworkers are less likely to sanction ethnic minority clients more than ethnic majority clients when the caseworker belongs to an ethnic minority group. Caseworker Gender The public administration literature on bureaucratic gender representation focuses mainly on how passive representation leads to active representation for women (Keiser et al. 2002; Wilkins and Keiser 2006; Bradbury and Kellough 2011), that is, the association between the presence of female bureaucrats within public organizations and bureaucratic outcomes consistent with the interests of female citizens. However, other streams of research suggest that caseworkers’ gender may moderate the effect of client ethnicity on caseworkers’ decisions to sanction in other ways. Social role theory suggests that women tend to be more altruistic, benevolent, and compassionate than men. For example, meta-analytical evidence suggests that women, but not men, internalize altruism as their intuitive social response (Rand et al. 2016). Women are more likely to express concern and responsibility for the well-being of others (Beutel and Marini 1995). Similarly, Dehart-Davis, Marlowe, and Pandey (2006) find that women score higher on the “compassion” subdimension of “public service motivation” than their male counterparts. In addition, public opinion research suggests that women are more favorable towards policies promoting equity and provision of social welfare benefits (Shapiro and Mahajan 1986; Crowder-Meyer 2007). Eagly et al. (2004) show that women, more than men, endorse policies that are socially compassionate, traditionally moral, and supportive of equal rights. In a cross-cultural and multi-method study based on 127 samples in 70 countries, Schwartz and Rubel (2005) find that women consistently attribute more importance than men do to both universalism values (understanding, appreciation, tolerance, and protection for the welfare of all people) and benevolence values (preservation and enhancement of the welfare of people with whom one is in personal contact). Combining these insights, we theorize that female caseworkers exhibit social roles and value orientations that make them more likely to engage clients of all ethnicities with greater sympathy, understanding, and equity than male caseworkers. Greater concern for the welfare, well-being, and equitable treatment of others among women may entail that ethnic stereotyping is less likely to shape decision-making—that is, create a disparity in sanctioning decisions between ethnic minority and ethnic majority clients—among female caseworkers. In summary, we expect the propensity to sanction ethnic minority clients more than ethnic majority clients to be smaller for female caseworkers: Hypothesis 2b: Caseworkers are less likely to sanction ethnic minority clients more than ethnic majority clients when the caseworker is female. Work Experience In addition to a caseworker’s ethnicity and gender, we are also interested in the role that work experience plays in caseworkers’ exercise of discretion, i.e., how work experience may moderate the effect of client ethnicity on the decision to sanction. Research suggests that stereotypical beliefs may have stronger effects in situations where the individual finds knowledge about the right course of action (what to do and how to do it) is unclear. In instances where the right action is unclear, Kahneman and colleagues (Kahneman, Slovic, and Tversky 1982; Kahneman 2011) show how bias in an individual’s judgment and decision-making may arise from the application of cognitive heuristics (information-processing shortcuts), such as stereotypes. Kahneman and Frederick (2002) emphasize the role of a process they call attribute substitution, which happens without conscious awareness. According to this theory, stereotypes can be a source of heuristic attributes that people substitute in when judgment of a target attribute is computationally complex. For example, judging whether a client’s behavior in a particular situation warrants sanctioning is more computationally complex than judging their ethnicity. The pre-conscious, intuitive nature of attribute substitution explains how a caseworker’s decisions to sanction may be influenced by client ethnicity, even though the caseworker believes that he or she has made an unbiased evaluation. At the same time, scholars have long known that professional socialization—i.e., the acquisition of not only attitudes and values but also skills and know-how pertaining to a professional subculture—increases over time (Downs 1967). As caseworkers gain work experience, they accumulate increasing knowledge about typical and atypical cases and casework proceedings and practices. They develop and internalize work routines and standard operating procedures that increasingly guide their work activities, discretion, and decision-making. Flyvbjerg (2001) describes such accumulation of work experience as a learning process through which a novice may gradually become a competent performer and even an expert over time. The accumulation of work experience and the development of expertise might lead to professional norms or standards serving as a guide to action, replacing individual caseworker attitudes, motives, or unconscious beliefs, such as stereotypes. Jensen and Vestergaard (2017) find that experience can moderate (suppress) the effect of motives with respect to predicting behavior. They argue that as people accumulate experiences, they rely on those experiences to formulate judgments about the best way to act in a situation. In the process, their personal motives or attitudes become less important. We, therefore, theorize that ethnic stereotypes are more likely to shape caseworkers’ decision-making when the caseworker has less work experience with employment casework. Moreover, Chen, Moskowitz, and Shue (2016) offer robust empirical evidence for the importance of work experience in mitigating biased decision-making. Across three high-stakes field settings (asylum, loan, and officiating decisions), they show how people tend to underestimate the likelihood of sequential streaks occurring by chance. This underestimation leads to negatively auto-correlated decisions that result in error. Due to misperception of random processes, people tend to believe that streaks of good or bad cases are unlikely to occur by chance (Rabin and Vayanos 2010). Therefore, decision makers may approach a new case with the subconscious belief that the case is likely to be positive if they considered the previous case negative, and vice versa. Decisions are affected by prior decisions in an inverse relationship, that is, decisions are negatively auto-correlated. Importantly, Chen, Moskowitz, and Shue (2016) find that the negative autocorrelation (bias) is less strong among more experienced decision makers. Similar to the way that work experience diminishes this particular form of cognitive bias in decision-making, we expect that work experience may mitigate ethnic bias in caseworkers’ sanctioning decisions. In line with Flyvbjerg (2001), we expect that an experienced performer or expert is less likely than a novice to resort to a (simplified) decision-making process based on ethnic cues. In summary, we expect that caseworkers’ work experience moderates the influence of client ethnicity on caseworkers’ decision to sanction in the following way: Hypothesis 2c: Caseworkers are less likely to sanction ethnic minority clients more than ethnic majority clients when the caseworker has more work experience. Empirical Setting We test the hypotheses using Danish employment agencies as our empirical setting. Denmark comprises 98 municipalities, each with its own employment agency. All unemployed citizens of working age and fit to work, but who are not members of an unemployment insurance fund (typically individuals having only a secondary education or lower), are eligible for a monthly means-tested unemployment cash benefit. To receive the cash benefit, unemployed citizens have to register at their local employment agency. Each citizen is then assigned a caseworker, who is responsible for helping that citizen obtain employment. A client’s entitlement to an unemployment benefit is conditioned by compliance with a set of rules and procedures. For example, all clients are expected to regularly consult with their caseworker (every month for the first half year, then every third month), actively seek employment, and be willing to accept employment offers. All clients must accept and attend employability-increasing activities that their local employment agency arranges, such as on-the-job training. The caseworkers have the authority to sanction clients for noncompliance with these rules and procedures. For example, a client may be assigned to an unpaid on-the-job training program in a private firm or public organization. If the client fails to show up without reasonable cause, the caseworker has the discretionary authority to reduce or terminate the client’s unemployment benefits until the client either requalifies for assistance or provides evidence of reasonable cause (see the Online Appendix for a visual presentation of the sanctioning procedure). Study 1: Administrative Data Our administrative data comprises Danish employment agency-level records for the first quarter of 2016. In the data, we have measures capturing the percentage of non-Western immigrants or their descendants who were cash benefit recipients, as well as three agency-level sanctioning indicators: (1) the average number of sanctions per cash benefit recipient; (2) the percentage of cash benefit recipients receiving a sanction; and (3) the average number of sanctions per cash benefit recipient receiving a sanction. We obtained the records from jobindsats.dk, an online platform hosted by the Danish Agency for Labour Market and Recruitment.6 We combine these data with administrative data from 2015 on the characteristics of the municipality where each employment agency is geographically located. All the data were available from Statistic Denmark’s population register and from noegletal.dk, an online platform hosted by the Danish Ministry of Social Affairs and the Interior. The dataset gives us access to the following municipal characteristics: number of residents, number of residents of Danish origin, percentage of residents living in urban agglomeration, percentage of residents aged 25–64 without vocational training, percentage of residents aged 25–64 with a higher education, and the unemployment rate (the average number of full-time unemployed who are fit to work per 100 residents aged 17–64). We include these variables in our analysis as controls. Table 1 gives the descriptive statistics. Table 1. Employment Agency Characteristics Employment Agency Characteristics  Mean  SD  Min  Max  N  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .37  .10  0  .62  98  Average number of sanctions per cash benefit recipient  .11  .10  0  .68  98  Percentage of cash benefit recipients receiving sanction  .06  .03  0  .22  98  Average number of sanctions per cash benefit recipient receiving sanction  1.69  .60  0  4.46  98  Number of municipal residents (1000s)  58.16  69.38  1.82  589.70  98  Number of municipal residents of Danish origin (1000s)  51.05  55.21  1.72  450.35  98  Percentage of municipal residents living in urban agglomeration  83.37  12.72  34.7  100  98  Percentage of municipal residents aged 25–64 without vocational training  20.59  4.95  8.2  31.4  98  Percentage of municipal residents aged 25–64 with a higher education  25.10  8.96  14.1  52  98  Municipal unemployment rate  3.99  1.10  2  8.4  98  Employment Agency Characteristics  Mean  SD  Min  Max  N  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .37  .10  0  .62  98  Average number of sanctions per cash benefit recipient  .11  .10  0  .68  98  Percentage of cash benefit recipients receiving sanction  .06  .03  0  .22  98  Average number of sanctions per cash benefit recipient receiving sanction  1.69  .60  0  4.46  98  Number of municipal residents (1000s)  58.16  69.38  1.82  589.70  98  Number of municipal residents of Danish origin (1000s)  51.05  55.21  1.72  450.35  98  Percentage of municipal residents living in urban agglomeration  83.37  12.72  34.7  100  98  Percentage of municipal residents aged 25–64 without vocational training  20.59  4.95  8.2  31.4  98  Percentage of municipal residents aged 25–64 with a higher education  25.10  8.96  14.1  52  98  Municipal unemployment rate  3.99  1.10  2  8.4  98  View Large Table 1. Employment Agency Characteristics Employment Agency Characteristics  Mean  SD  Min  Max  N  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .37  .10  0  .62  98  Average number of sanctions per cash benefit recipient  .11  .10  0  .68  98  Percentage of cash benefit recipients receiving sanction  .06  .03  0  .22  98  Average number of sanctions per cash benefit recipient receiving sanction  1.69  .60  0  4.46  98  Number of municipal residents (1000s)  58.16  69.38  1.82  589.70  98  Number of municipal residents of Danish origin (1000s)  51.05  55.21  1.72  450.35  98  Percentage of municipal residents living in urban agglomeration  83.37  12.72  34.7  100  98  Percentage of municipal residents aged 25–64 without vocational training  20.59  4.95  8.2  31.4  98  Percentage of municipal residents aged 25–64 with a higher education  25.10  8.96  14.1  52  98  Municipal unemployment rate  3.99  1.10  2  8.4  98  Employment Agency Characteristics  Mean  SD  Min  Max  N  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .37  .10  0  .62  98  Average number of sanctions per cash benefit recipient  .11  .10  0  .68  98  Percentage of cash benefit recipients receiving sanction  .06  .03  0  .22  98  Average number of sanctions per cash benefit recipient receiving sanction  1.69  .60  0  4.46  98  Number of municipal residents (1000s)  58.16  69.38  1.82  589.70  98  Number of municipal residents of Danish origin (1000s)  51.05  55.21  1.72  450.35  98  Percentage of municipal residents living in urban agglomeration  83.37  12.72  34.7  100  98  Percentage of municipal residents aged 25–64 without vocational training  20.59  4.95  8.2  31.4  98  Percentage of municipal residents aged 25–64 with a higher education  25.10  8.96  14.1  52  98  Municipal unemployment rate  3.99  1.10  2  8.4  98  View Large Results We analyze the administrative data using multivariate OLS regression with robust standard errors. We run three models in which we regress the percentage of cash benefit recipients who are non-Western immigrants or their descendants on the average number of sanctions per cash benefit recipient (model 1), the percentage of cash benefit recipients receiving a sanction (model 2), and the average number of sanctions per cash benefit recipient receiving a sanction (model 3). All three models include a set of statistical control variables reflecting the municipal characteristics retrieved from Statistic Denmark’s population register and noegletal.dk. Table 2 shows the results of the analyses. Table 2. OLS Regression, Administrative Data Independent variables  Model 1  Model 2  Model 3  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .255* (.124)  .089† (.047)  1.92* (.948)  Municipal characteristics        Number of residents (1000s)  .001 (.001)  .001 (.001)  −.008 (.008)  Number of residents of Danish origin (1000s)  −.001 (.002)  −.001 (.001)  .011 (.011)  Percentage of residents living in urban agglomeration  −.002 (.001)  −.001 (.001)  −.004 (.004)  Percentage of residents aged 25–64 without vocational training  .017* (.008)  .005* (.002)  .084 (.056)  Percentage of residents aged 25–64 with a higher education  .004 (.003)  .001 (.001)  .023 (.203)  Unemployment rate  −.033* (.014)  −.009* (.004)  −.202** (.072)  R2  .17  .16  .16  N  98  98  98  Independent variables  Model 1  Model 2  Model 3  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .255* (.124)  .089† (.047)  1.92* (.948)  Municipal characteristics        Number of residents (1000s)  .001 (.001)  .001 (.001)  −.008 (.008)  Number of residents of Danish origin (1000s)  −.001 (.002)  −.001 (.001)  .011 (.011)  Percentage of residents living in urban agglomeration  −.002 (.001)  −.001 (.001)  −.004 (.004)  Percentage of residents aged 25–64 without vocational training  .017* (.008)  .005* (.002)  .084 (.056)  Percentage of residents aged 25–64 with a higher education  .004 (.003)  .001 (.001)  .023 (.203)  Unemployment rate  −.033* (.014)  −.009* (.004)  −.202** (.072)  R2  .17  .16  .16  N  98  98  98  Notes: Entries are coefficients with robust standard errors (in parentheses). †p ≤ .1; *p ≤ .05; **p ≤ .01. View Large Table 2. OLS Regression, Administrative Data Independent variables  Model 1  Model 2  Model 3  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .255* (.124)  .089† (.047)  1.92* (.948)  Municipal characteristics        Number of residents (1000s)  .001 (.001)  .001 (.001)  −.008 (.008)  Number of residents of Danish origin (1000s)  −.001 (.002)  −.001 (.001)  .011 (.011)  Percentage of residents living in urban agglomeration  −.002 (.001)  −.001 (.001)  −.004 (.004)  Percentage of residents aged 25–64 without vocational training  .017* (.008)  .005* (.002)  .084 (.056)  Percentage of residents aged 25–64 with a higher education  .004 (.003)  .001 (.001)  .023 (.203)  Unemployment rate  −.033* (.014)  −.009* (.004)  −.202** (.072)  R2  .17  .16  .16  N  98  98  98  Independent variables  Model 1  Model 2  Model 3  Percentage of non-Western immigrants or their descendants as cash benefit recipients  .255* (.124)  .089† (.047)  1.92* (.948)  Municipal characteristics        Number of residents (1000s)  .001 (.001)  .001 (.001)  −.008 (.008)  Number of residents of Danish origin (1000s)  −.001 (.002)  −.001 (.001)  .011 (.011)  Percentage of residents living in urban agglomeration  −.002 (.001)  −.001 (.001)  −.004 (.004)  Percentage of residents aged 25–64 without vocational training  .017* (.008)  .005* (.002)  .084 (.056)  Percentage of residents aged 25–64 with a higher education  .004 (.003)  .001 (.001)  .023 (.203)  Unemployment rate  −.033* (.014)  −.009* (.004)  −.202** (.072)  R2  .17  .16  .16  N  98  98  98  Notes: Entries are coefficients with robust standard errors (in parentheses). †p ≤ .1; *p ≤ .05; **p ≤ .01. View Large The analyses suggest that employment agencies in which a larger percentage of cash benefit recipients are non-Western immigrants or their descendants impose a greater overall number of sanctions and dispense them with greater frequency. In model 1, we find that employment agencies in which a greater percentage of the cash benefit recipients are non-Western immigrants or their descendants are characterized by a greater average number of sanctions per cash benefit recipient (β = .255, p ≤ .05). Similarly, models 2 and 3 show, respectively, that the percentage of non-Western immigrants or their descendants receiving cash benefits is positively associated with a higher percentage of cash benefit recipients receiving a sanction (β = .089, p ≤ .10) and a greater number of sanctions per sanctioned cash benefit recipient (β = 1.92, p ≤ .05). The results are thus consistent with the notion that caseworkers are more likely to recommend sanctions for ethnic minority clients than for ethnic majority clients (Hypothesis 1). As the administrative data reflects the real world of caseworker decision-making, the findings are characterized by high ecological validity. Moreover, the nationwide nature of the data eliminates the risk of bias caused by missing observations or nonresponse. However, the analyses of the administrative data are marked by one key limitation: a modest potential for causal inferences. Although the analyses indicate that ethnic minority clients are sanctioned more than ethnic majority clients, we recognize that other factors than ethnic bias among caseworkers may possibly explain this finding. For example, the agency-level differences in sanctioning activity that we observe could be the result of systematic differences in client behavior (i.e., clients of Middle-Eastern origin exhibiting more sanctionable behavior than clients of Danish origin). To better understand how ethnic bias shapes caseworkers’ decisions to sanction clients, we complement the analyses of the administrative data with a survey experiment that relates to individual-level sanctioning decisions and allows for a causal interpretation of the results. Study 2: Survey Experiment Our experimental data originates from a national e-survey among Danish employment caseworkers. To build our sample frame of employment agency caseworkers, we sent a request to employment agency managers from all 98 Danish municipalities for the contact information of each of their local employment caseworkers and for permission to allow us to invite the caseworkers to participate in a research project involving a survey. Managers from 63 municipalities responded, with the email addresses of a total of 1,335 caseworkers and permission from their supervisors to request their participation in our study. We conducted the survey experiment in April 2016. We received complete survey responses from 497 caseworkers (≈37%). In line with Danish guidelines for ethics in social sciences research, the respondents were instructed that while their responses would be used for research purposes, their individual responses were anonymous and that the researchers would comply with standard confidentiality procedures. Furthermore, we made clear that participation in our study was voluntary and that respondents could choose not to participate or to stop at any time. Table 3 shows a demographic profile of the survey respondents. Table 3. Respondent Characteristics Respondent characteristics  Mean  SD  Min  Max  N  Gender (female)  .81  .39  0  1  497  Age (years)  43.21  11.16  21  69  497  Ethnicity (non-Dane)  .09  .26  0  1  497  Education: Lower secondary  .01  .11  0  1  497      : Secondary  .02  .15  0  1  497      : Youth vocational training  .08  .28  0  1  497      : Higher, short-cycle  .05  .23  0  1  497      : Higher, medium-cycle  .73  .44  0  1  497      : Higher, long-cycle  .09  .29  0  1  497  Primary client group: Cash benefit recipients  .57  .50  0  1  497  Tenure (years)  4.73  4.94  0  35  497  Work experience, caseworker (years)  8.62  7.78  0  38  497  Respondent characteristics  Mean  SD  Min  Max  N  Gender (female)  .81  .39  0  1  497  Age (years)  43.21  11.16  21  69  497  Ethnicity (non-Dane)  .09  .26  0  1  497  Education: Lower secondary  .01  .11  0  1  497      : Secondary  .02  .15  0  1  497      : Youth vocational training  .08  .28  0  1  497      : Higher, short-cycle  .05  .23  0  1  497      : Higher, medium-cycle  .73  .44  0  1  497      : Higher, long-cycle  .09  .29  0  1  497  Primary client group: Cash benefit recipients  .57  .50  0  1  497  Tenure (years)  4.73  4.94  0  35  497  Work experience, caseworker (years)  8.62  7.78  0  38  497  View Large Table 3. Respondent Characteristics Respondent characteristics  Mean  SD  Min  Max  N  Gender (female)  .81  .39  0  1  497  Age (years)  43.21  11.16  21  69  497  Ethnicity (non-Dane)  .09  .26  0  1  497  Education: Lower secondary  .01  .11  0  1  497      : Secondary  .02  .15  0  1  497      : Youth vocational training  .08  .28  0  1  497      : Higher, short-cycle  .05  .23  0  1  497      : Higher, medium-cycle  .73  .44  0  1  497      : Higher, long-cycle  .09  .29  0  1  497  Primary client group: Cash benefit recipients  .57  .50  0  1  497  Tenure (years)  4.73  4.94  0  35  497  Work experience, caseworker (years)  8.62  7.78  0  38  497  Respondent characteristics  Mean  SD  Min  Max  N  Gender (female)  .81  .39  0  1  497  Age (years)  43.21  11.16  21  69  497  Ethnicity (non-Dane)  .09  .26  0  1  497  Education: Lower secondary  .01  .11  0  1  497      : Secondary  .02  .15  0  1  497      : Youth vocational training  .08  .28  0  1  497      : Higher, short-cycle  .05  .23  0  1  497      : Higher, medium-cycle  .73  .44  0  1  497      : Higher, long-cycle  .09  .29  0  1  497  Primary client group: Cash benefit recipients  .57  .50  0  1  497  Tenure (years)  4.73  4.94  0  35  497  Work experience, caseworker (years)  8.62  7.78  0  38  497  View Large While the sampling procedure and response rate suggest some caution in extrapolating from this sample, the respondents appear relatively diverse and suit the needs of our experimental design (see Auginis and Bradley 2014). The sampling procedure—while imperfect—nonetheless provides us with a population of actual caseworkers. Experimental Design The design involves a vignette portraying an unemployed male cash benefit recipient who is assigned to an unpaid on-the-job training program in a public organization. The vignette explains that the client did not show up for the program on some occasions—without notifying the employer—and that the client is unable to document having been ill (his stated reason for being absent). The caseworkers were asked to decide, based on the vignette, whether to impose a sanction on the client or not.7 Each caseworker respondent received one of three versions of the vignette. For capturing the influence of client ethnicity on caseworker responses, we randomly assigned the client described in the vignette (1) a “Middle-Eastern-sounding” name, (2) an “ethnic Danish-sounding” name, or (3) a blinded name. In all other respects, all respondents received identical vignette information, including some biographical information to add to the contextual realism of the case (Auginis and Bradley 2014). Similar “name manipulation” procedures are used in other studies examining the effects of client race in the area of unemployment (Bertrand and Mullainathan 2004; Schram et al. 2009). We use the name “Mohammed” as the ethnic minority of Middle-Eastern origin-sounding name, and “Peter” as the ethnic Danish-sounding name. We selected those names from administrative data from Statistic Denmark (2016), data suggesting that Peter is the most common male first name in Denmark and Mohammed the most common male first name among non-Western immigrants and their descendants in Denmark (Sahin and Dreyer 2015) and among Muslims in Denmark (Meldgaard 2005). Their ethnic connotations notwithstanding, neither name is associated with any known social stigma. For example, the name Brian is also fairly common in Denmark but is associated with the stereotype of an ethnic Danish working-class troublemaker. The version of the vignette involving a blinded name simply showed a black box where the name should be, signaling that we had anonymized the client’s name. The random assignment process resulted in a statistically equal distribution of client names to respondents (t-tests): 169 received the Mohammed treatment (34%), 173 received the Peter treatment (35%), and 155 received a blinded name (31%). The Vignette The vignette was inspired by case descriptions in Beer et al. (2008) and Ejrnæs and Monrad (2012). For contextual realism, two experienced unemployment caseworkers and an expert in immigrants and multicultural relations were involved in developing the vignette. The full text is as follows (translated from Danish to English with the assistance of a professional language editor): “[Peter/Mohammed/”blinded”] is a 31-year-old male. He is fit to work and receives unemployment cash benefits. Peter left Folkeskolen [Danish compulsory education] after 8th grade and has not received any further education. He has about five years of work experience from different minor jobs. He has been dismissed from all of them due to absenteeism. He has received cash benefits or been self-supporting on a rotating basis between these jobs. He has been unemployed and received cash benefits since he was fired from his latest job five months ago. Peter was diagnosed with depression two years ago and received treatment with antidepressants for a period of six months. Peter started on-the-job training in a public organization in January 2016. During the start-up phase, Peter was informed in writing about the consequences for his cash benefits should he fail to attend the training without reasonable cause, and about what he could do to regain his cash benefits entitlement in such a case. Peter’s time sheet from the on-the-job training for February 2016 shows that he has been absent five days without informing the employer. In a consultation, Peter has attributed his absences to illness caused by migraines. However, he did not consult a general practitioner. He has not been sanctioned previously.” The vignette allows us to test whether caseworkers may be more likely to recommend sanctions for clients appearing to be an ethnic minority (of a Middle-Eastern origin) than for clients of Danish ethnic origin. The vignette portrays a situation in which a caseworker would have substantial discretion to decide whether to impose a sanction. With the blinded version of the vignette as an additional baseline, we expect that the Middle-Eastern-sounding name (Mohammed) may cue negative stereotypes related to unemployed clients of Middle-Eastern origin—in turn resulting in a greater sanctioning propensity among recipients of the “Mohammed” version of the vignette than among recipients of the “Peter” or “blinded” versions. Following the vignette was a question asking the respondents to indicate the extent to which they would impose a sanction on this client. Responses were measured on an 11-point scale, anchored at 0 (“definitely not sanction”) and at 10 (“definitely sanction”). All scale points were labeled with a numeral. The mean score was 7.28, with a standard deviation of 2.88.8 As an experimental manipulation check, at a later stage of the survey the respondents were asked to recall the name of the welfare recipient (response categories: “Anonymized,” “Peter,” “Michael,” “Mohammed,” “Ahmed,” “do not remember”). The “correct” name (the name in the vignette they received) was reported by 78 percent of the caseworkers. The number of false or “do not remember” responses was distributed equally across the three experimental conditions. We reran all model specifications with inclusion of a manipulation check dummy variable (“correct” response or not), and we tested all model specifications on a subsample comprising only respondents with “correct” responses. Neither of these robustness checks resulted in any qualitative changes of the findings. Analysis and Results One-way analysis of variance tests show balance across the three experimental groups in the distributions of gender, age, ethnicity, education, primary client group, tenure, and work experience (at p < .1), indicating that the randomization process worked. Moreover, one-way analysis of variance shows a significant difference in sanctioning recommendation across the three experimental conditions (p = .04). Caseworkers assigned to the “Mohammed” vignette reported a greater inclination to sanction the client than caseworkers assigned to the Peter or the “blinded” vignette. The mean sanctioning score was 7.73 for Mohammed, 7.06 for Peter, and 7.02 for the version where the name was blinded. T-tests show that the sanctioning score for Mohammed is significantly higher than the other two sanctioning scores (p = .03 and p = .02). While this result confirms Hypothesis 1, we also performed multivariate OLS regression analysis that minimizes the risk of bias due to potential imbalances in caseworker characteristics across the experimental conditions. The regression model includes two binary variables representing the Mohammed and the Peter treatment (the “blinded” version serves as the reference category). We control for the caseworker characteristics in Table 1 (gender, age, ethnicity, education, primary client group, tenure, and work experience). We estimate robust standard errors that allow for the presence of heteroscedasticity. Figure 1 illustrates the key findings: the coefficients for our experimental conditions (the underlying regression model appear in the Online Appendix, Table A-1). The point estimates (X-axis) show the differences in sanctioning recommendations among the caseworkers assigned to the Peter and Mohammed treatments relative to those assigned to the blinded vignette (Y-axis). The horizontal lines signify 95 percent confidence bands. Figure 1. View largeDownload slide Plot of the Peter and Mohammed Coefficients, Vignette Experiment Figure 1. View largeDownload slide Plot of the Peter and Mohammed Coefficients, Vignette Experiment In support of Hypothesis 1, we see that caseworkers receiving the Mohammed treatment report being more likely to sanction the client than those assigned to the blinded version of the vignette (β = .824, p = .01). The effect translates to an increase of .29 of a standard deviation in sanctioning inclination, which is comparable to an effect size of d ≈ .29, more than a “small” effect as defined by Cohen (1988). A likelihood-ratio test reveals that caseworkers receiving the Mohammed treatment are also more likely to sanction the client than those assigned to the Peter treatment (p = .03) (d ≈ .23). Assignment of the Peter treatment does not result in a statistically different sanctioning recommendation relative to assignment of the blinded version of the vignette. Moderation Analyses We examine the moderating effects of caseworker ethnicity, gender, and work experience using the survey experimental data. In particular, we employ multivariate OLS regression analysis with robust standard errors and test how assignment of experimental condition (Peter, Mohammed, or “blinded”) interact with each of the three caseworker characteristics in predicting caseworker sanctioning responses. We thus perform three regression analyses. Each regression includes interaction terms in which the binary variables representing the Mohammed and the Peter treatments are multiplied, respectively, by one of the three caseworker characteristics (gender, ethnicity, or work experience). The analysis provides a partial confirmation of our hypotheses related to the moderating effects of caseworker characteristics (the full models appear in the Online Appendix, Table A-2). Neither caseworker ethnicity nor gender appears to moderate the effect of assignment to the Mohammed treatment or the Peter treatment relative to the “blinded” version of the vignette (Hypotheses 2a and 2b). However, as anticipated, we find that assignment to the Mohammed treatment relative to the “blinded” version of the vignette results in a lesser propensity to sanction the client among caseworkers with more work experience (Hypothesis 2c). This finding supports our theory-based expectation that work experience suppresses or reduces reliance on ethnic stereotypes, thereby lessening bias in frontline employee decision-making. Figure 2 illustrates this finding. We compute and plot the marginal effects of assignment to the Mohammed treatment relative to the “blinded” version of the vignette (Y-axis) at different levels of caseworker work experience (X-axis). Figure 2. View largeDownload slide Plot of the Effect of the Mohammed Treatment Relative to the Blinded Version of the Vignette at Different Levels of Caseworker Work Experience Figure 2. View largeDownload slide Plot of the Effect of the Mohammed Treatment Relative to the Blinded Version of the Vignette at Different Levels of Caseworker Work Experience As Figure 2 shows, the statistical difference in sanctioning response between recipients of the Mohammed treatment and those of the “blinded” version of the vignettes goes towards zero as the work experience of the caseworker increases. Caseworker assignment to the Mohammed treatment relative to the “blinded” version of the vignette is associated with a higher sanctioning propensity for caseworkers with 12 years or less of work experience as a caseworker (≈75% of the sample respondents). In contrast, assignment to the Mohammed treatment relative to the “blinded” version of the vignette does not appear to influence the sanctioning recommendations of caseworkers with more than 12 years of work experience (≈25% of the sample respondents). Caseworker work experience is naturally correlated with caseworker age. Importantly, however, similar interaction term estimates across models without and with control for age (see the Online Appendix, Table A-2) diminishes the concern that the moderation finding is driven by a generational cohort effect rather than by work experience.9 Discussion and Conclusion In many public welfare programs, public employees have discretionary authority to decide when and, if so, whether welfare recipients should be penalized for failure to comply with policy requirements. In terms of formal rules and prescriptions, such welfare sanctions should be imposed as responses to the behavior of the individual client. Client characteristics, such as race or ethnicity, should not influence a caseworker’s decisions to sanction. This article, however, provides evidence that client characteristics do affect casework sanctioning decisions. Drawing on the RCM (Soss, Schram, and Fording 2008) for understanding race and ethnic biases in employee decision-making, we examine and show how client ethnicity influences sanctioning behavior in the Danish unemployment program. Drawing on the RCM, we theorized that ethnic stereotypes may shape how public employees perceive and assess the deservingness of clients of different ethnicity, and thus affect their propensity for imposing a sanction in response to client noncompliance with unemployment policy requirements. We hypothesized that employment caseworkers in Denmark are more likely to sanction a client of Middle-Eastern descent than one of Danish origin. Our empirical findings are consistent with this theoretical expectation. Examining correlates of sanctioning activity at the agency-level, our administrative data indicate that employment agencies serving a larger population of non-Western immigrants or their descendants impose a larger overall number of sanctions and dispense them to clients with greater frequency. Addressing concerns about internal validity and a need for analyses at the individual employee-level, our survey experimental data show that employment caseworkers are more likely to recommend sanctions for a client named Mohammed than for a comparable client named Peter or an unnamed client. Research suggests a proliferation of ethnic stereotypes in Denmark, implying that immigrants and people of Middle-Eastern origin in Denmark are occasionally typecast as “taking advantage” of the social welfare system (Gaasholt and Togeby 1995; Larsen 2012). In line with the RCM, such ethnic stereotypes may explain the causal mechanism underlying our empirical finding, that is, that caseworkers use discretion in a biased manner that is punitive when the client appears Middle-Eastern, not “Danish.” In addition to the direct effect of the client’s ethnicity on a caseworker’s likelihood of administering a sanction, this article extends the RCM by considering the moderating effects of three caseworker characteristics. Both theoretically and empirically, we considered the role of the employment caseworker’s ethnicity, gender, and professional work experience as important moderator variables. We find no evidence that caseworker ethnicity or gender moderates the effect of a client’s ethnicity on a caseworker’s decision to sanction. Some caution is advised when interpreting these results. The null finding for caseworker ethnicity may be a product of limited empirical variance in caseworkers’ ethnicity (91% are ethnic Danes). Moreover, a full test of the gender moderation hypothesis would require a 2 × 3 factorial design (manipulation of both client gender and ethnicity). As the client in our vignette is a male, we do not reject the possibility that female caseworkers may be less likely than male caseworkers to sanction female Middle-Eastern origin clients more than female Danish origin clients. Nevertheless, Schram et al. (2009) found similar null results in their study of caseworkers in the United States. They found no evidence that white case managers differ from their non-white peers in their sanction decisions (see also Soss, Fording, and Schram 2011b: pp. 251–261). While our null findings are inconsistent with the theoretical rationales of representative bureaucracy (Meier 1993a), social identity theory (Tajfel and Turner 1986), and theories related to group identity and in-group biases and loyalties (Currarini and Mengel 2016), the results nonetheless match the logic and underpinnings of the RCM. As Schram et al. (2009, p. 415) note: “Racial classifications operate in implicit ways—without conscious racism—to generate racial disparities (Quillian 2008). Race matters in more subtle ways than overt hostility or loyalty; race is built into the cognitive processes that provide the foundation for decisions about how target groups should be treated in welfare policy settings.” Thus, caseworker attitudes related to ethnic stereotypes might be more important in driving behaviors, actions, or decisions than shared (or unshared) ethnic characteristics. This insight has important implications for our understanding of representation in public organizations. The effects of representation found in studies at the organizational level may be a function of dynamic, group processes that may not be explained by simple aggregation of individual employee actions based on in-group bias or favoritism. Work experience constituted the only caseworker characteristic that had the hypothesized moderation effect. Theoretically, heuristics are thought most important for caseworkers when a high level of uncertainty exists in a particular case. As people gain professional work experience, they obtain professional knowledge. Over time, they rely more on knowledge gained through experience—and less on stereotypes—in their situational decision-making and navigating uncertainty. The result is greater consistency in how they treat clients and process cases under conditions of uncertainty, resulting in less biased outcomes. By using our vignette, we found that the professional work experience of a caseworker mitigates bias against an individual presumably of Middle-Eastern origin. This finding is consistent with studies suggesting that professional work experience mitigates biased decision-making (Chen, Moskowitz, and Shue 2016) and suppresses the importance of motives in predicting behavior (Jensen and Vestergaard 2017). In summary, we find that welfare sanctions are not imposed exclusively as responses to client behavior. Sanctions are also imposed as responses to client ethnicity—and the caseworkers’ professional work experience may alleviate this relationship. In line with Schram et al. (2009), the results reinforce the conclusion that policy choices not only reflect but also create the elements that underpin racial or ethnic disparity in the provision of public goods and services. Our findings contest the idea of race- and ethnic-neutral policies, instead showing that real-world social welfare systems operate in ways in which the imposing of sanctions may reflect racial and ethnic stereotypes and (re)produce racial and ethnic inequities. In times where sanctioning practices have become an increasingly integral part of most social welfare programs (Pavetti, Derr and Hesketh 2003; Hood and Peters 2004; Kettl 2005), our findings call for the attention of both policy makers and scholars. Policy makers need to know, and factor into their decision-making, that while present welfare arrangements may appear ostensibly race- and ethnic-neutral, the employees carry them out in ways that allow for racial and ethnic stereotyping, thus marking official decision-making by cognitive biases against certain races or ethnic groups. Similarly, scholars should seek to provide further insights into how and why disciplinary welfare practices produce racial and ethnic disadvantages and inequity. More specifically, we believe that our moderation result for work experience warrants future research inquiry. Managerial transformation of employee demographic characteristics, such as gender and ethnicity, is infeasible. However, professional training and development initiatives could target and cultivate the formation of the specific elements of knowledge and know-how that come with professional work experience and mitigate the influence of client ethnicity on decisions to sanction or other types of decisions. This perspective is in line with recent hermeneutic research suggesting that biases and prejudices can be corrected through dialogical inquiry (Metselaar, Meynen, and Widdershoven 2016) and research demonstrating how training can effectively debias decision makers over the long term (Dhami, Schlottmann, and Waldmann 2013; Morewedge et al. 2015). Although identifying bias in public service provision is critical, a next step for future research should be to identify the precise knowledge components that come with increasing work experience and that diminish cognitive bias in decision-making at the frontlines of public service delivery. Supplementary Material Supplementary data are available at Journal of Public Administration Research and Theory online. Acknowledgements We thank the editor and three anonymous reviewers for helpful comments and suggestions. We also thank Isak K. Houe for his assistance in generating the administrative data set. A previous version of this article was presented at the 14th Biennial Public Management Research Conference (PMRC), American University, Washington, D.C., June 8–10, 2017; at the 2nd Conference on Street-level Research in the Employment and Social Policy Area, Aalborg University Copenhagen Campus, Copenhagen, Denmark, June 21–22, 2017; and at the 77th Annual Meeting of the Academy of Management (AOM), The Public and Nonprofit Division (PNP), Atlanta, GA, August 4–5, 2017. We thank conference participants for useful feedback and suggestions. References ALS Research. 2014. Rapport: Myter om minoriteter i københavn . Copenhagen, Denmark: Als research. Amodio, David M., and Patricia G. Devine. 2006. Stereotyping and evaluation in implicit race bias: Evidence for independent constructs and unique effects on behavior. Journal of Personality and Social Psychology  91: 652– 661. Google Scholar CrossRef Search ADS PubMed  Atkins, Danielle N., and Vicky M. Wilkins. 2014. Going beyond reading, writing, and arithmetic: The effects of teacher representation on teen pregnancy rates. Journal of Public Administration Research and Theory  4: 771– 90. Auginis, Herman, and Kyle J. Bradley. 2014. Best practice recommendations for designing and implementing experimental vignette methodology studies. Organizational Research Methods  17: 351– 71. Google Scholar CrossRef Search ADS   Beer, Frederikke, Søren C. Winter, Mette H. Skou, Mads V. Stigaard, Annemette C. Henriksen, and Nina Friisberg. 2008. Statslig og kommunal Beskæftigelsesindsats . SFI Rapport 08:19. Copenhagen: SFI—The Danish National Centre for Social Research. 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   Beutel, Ann M., and Maragret M. Marini. 1995. Gender and values. American Sociological Review  60: 436– 448. Google Scholar CrossRef Search ADS   Blau, Peter M. 1955. The dynamics of bureaucracy: A study of interpersonal relations in two government agencies . Chicago, IL: University of Chicago Press. Bradbury, Mark, and J. Edward Kellough. 2008. Representative bureaucracy: Exploring the potential for active representation in local government. Journal of Public Administration Research and Theory  18: 697– 714. Google Scholar CrossRef Search ADS   Bradbury, Mark, and J. Edward Kellough. 2011. Representative bureaucracy: Assessing the evidence on active representation. The American Review of Public Administration  41: 157– 67. Google Scholar CrossRef Search ADS   Brodkin, Evelyn Z. 2011. Policy work: Street-level organizations under new managerialism. Journal of Public Administration Research and Theory  21( suppl 2): i253– i277. Google Scholar CrossRef Search ADS   Brown-Iannuzzi, Jazmin L., Ron Dotsch, Erin Cooley, and B. Keith Payne. 2017. The relationship between mental representations of welfare recipients and attitudes toward welfare. Psychological Science  (Epub ahead of print). doi: 10.1177/0956797616674999 Bursell, Moa. 2012. Name change and destigmatization among middle eastern immigrants in sweden. Ethnic and Racial Studies  35: 471– 487. Google Scholar CrossRef Search ADS   Byrne, Donn. 1971. The attraction paradigm . New York, NY: Academic Press. Caswell, Dorte, Henrik L. Andersen, Mathilde Høybye-Mortensen, Anna M. Markussen, and Sille L. Thuesen. 2011. Når kassen smækkes i. Analyser af økonomiske sanktioner over for kontanthjælpsmodtagere . Copenhagen: AKF. Caswell, Dorte, Leena Eskelinen, and Søren P. Olesen. 2011. Identity work and client resistance underneath the canopy of employment policy. Qualitative Social Work  12: 8– 23. Google Scholar CrossRef Search ADS   Chen, Daniel L., Tobias J. Moskowitz, and Kelly Shue. 2016. Decision making under the gambler’s fallacy: Evidence from asylum judges, loan officers, and baseball umpires. The Quarterly Journal of Economics  131: 1181– 1242. Google Scholar CrossRef Search ADS   Christiansen, Niels F., Klaus Petersen, Nils Edling, and Per Haave. 2006. The nordic model of welfare: A historical reappraissal . Copenhagen, Denmark: Museum Tusculanum Press. Cohen, Jacob. 1988. Statistical power analysis for the behavioral sciences . Hillsdale, NJ: Lawrence Erlbaum. Crowder-Meyer, Melody. 2007. Gender differences in policy preferences and priorities . Paper presented at the annual meeting of the Midwest Political Science Association, Chicago, IL, April 12, 2007. Currarini, Sergio, and Friederike Mengel. 2016. Identity, homophily and in-group bias. European Economic Review  90: 40– 55. Google Scholar CrossRef Search ADS   Dhami, Mandeep, Anne Schlottmann, and Michael R. Waldmann. 2013. Judgment and decision making as a skill: Learning, development, and evolution . Cambridge, UK: Cambridge Univ. Press. Danziger, Sandra K., and Kristin S. Seefeldt. 2003. Barriers to employment and the ‘hard to serve’: Implications for services, sanctions, and time limits. Social Policy and Society  2: 151– 160. Google Scholar CrossRef Search ADS   DeHart-Davis, Leisha, Justin Marlowe, and Sanjay K. Pandey. 2006. Gender dimensions of public service motivation. Public Administration Review  66: 837– 887. Google Scholar CrossRef Search ADS   Devine, Patricia G., E. Ashby Plant, David M. Amodio, Eddie Harmon-Jones, and Stephanie L. Vance. 2002. The regulation of explicit and implicit race bias: The role of motivations to respond without prejudice. Journal of Personality and Social Psychology  82: 835. Google Scholar CrossRef Search ADS PubMed  Dolan, Julie. 2000. The senior executive service: Gender, attitudes, and representative bureaucracy. Journal of Public Administration Research and Theory  10: 513– 30. Google Scholar CrossRef Search ADS   Downs, Anthony. 1967. Inside bureaucracy . Boston: Little, Brown and Company. Google Scholar CrossRef Search ADS   Eagly, Alice H., Amanda B. Diekman, Mary C. Johannesen-Schmidt, and Anne M. Koenig. 2004. gender gaps in sociopolitical attititudes: A social psychological analysis. Journal of Personality and Social Psychology  87: 796– 816. Google Scholar CrossRef Search ADS PubMed  Ejrnæs, Morten, and Merete Monrad. 2012. Vignetmetoden: Sociologiske metode og redskab til faglig udvikling . Copenhagen: Akademisk Forlag. Esping-Andersen, Gøsta. 1990. The three worlds of welfare capitalism. Princeton, NJ: Princeton University Press. Flyvbjerg, Bent. 2001. Making social science matter: Why social inquiry fails and how it can succeed again . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Frederickson, H. George. 1990. Public administration and social equity. Public Administration Review  50: 228– 237. Google Scholar CrossRef Search ADS   Gaasholt, Øystein, and Lise Togeby. 1995. I syv sind: Danskernes holdninger til flygtninge og indvandrere . Aarhus, Denmark: Forlaget Politica. Gilens, Martin. 1999. Why americans hate welfare: Race, media, and the politics of antipoverty policy . Chicago, IL: Univ. of Chicago Press. Google Scholar CrossRef Search ADS   Grissom, Jason A., Jill Nicholson-Crotty, and Sean Nicholson-Crotty. 2009. Race, region, and representative bureaucracy. Public Administration Review  69: 911– 19. Google Scholar CrossRef Search ADS   Hancock, Ange-Marie. 2004. The politics of disgust: The public identity of the welfare queen . New York: New York Univ. Press. Hasenfeld, Yeheskel, Toorjo Ghose, and Kadyce Larson. 2004. The logic of sanctioning welfare recipients: An empirical assessment. Social Service Review  78: 304– 19. Google Scholar CrossRef Search ADS   Hindera, John J. 1993. Representative bureaucracy: Further evidence of active representation in the EEOC district offices. Journal of Public Administration Research and Theory  3: 415– 29. Hogg, Michael A., and Dominic Abrams. 1998. Social identifications: A social psychology of intergroup relations and group processes . London: Routledge. Hood, Christopher, and Guy Peters. 2004. The middle aging of new public management: Into the age of paradox? Journal of Public Administration Research and Theory  14: 267– 82. Google Scholar CrossRef Search ADS   Hupe, Peter, and Michael Hill. 2007. Street-level bureaucracy and public accountability. Public Administration  85: 279– 99. Google Scholar CrossRef Search ADS   Jacobsen, Brian A. 2012. Muslims in Denmark: A critical evaluation of estimations. In Islam in Denmark: The challenge of diversity , ed. Jørgen S. Nielsen, 31– 56. Plymouth, UK: Lexington Books. Jensen, Ulrich T., and Christian F. Vestergaard. 2017. Public service motivation and public service behaviors: Testing the moderating effect of tenure. Journal of Public Administration Research and Theory  27: 52– 67. Google Scholar CrossRef Search ADS   Johnson, Brian D. 2014. Judges on trial: A reexamination of judicial race and gender. Effects across modes of conviction. Criminal Justice Policy Review  25: 159– 84. Google Scholar CrossRef Search ADS   Kahneman, Daniel. 2011. Thinking, fast and slow . New York: Farrar, Straus and Giroux. Kahneman, Daniel, and Shane Frederick. 2002. Representativeness revisited: Attribute substitution in intuitive judgment. In Heuristics and biases: The psychology of intuitive judgment , eds. Thomas Gilovich, Dale Griffin, and Daniel Kahneman, 49– 81. Cambridge, UK: Cambridge Univ. Press. Google Scholar CrossRef Search ADS   Kahneman, Daniel, Amos Tversky, and Paul Slovic. 1982. Judgment under uncertainty: Heuristics & biases . Cambridge, UK: Cambridge University Press. Google Scholar CrossRef Search ADS   Keiser, Lael R., Vicky M. Wilkins, Kenneth J. Meier, and Catherine A. Holland. 2002. Lipstick and logarithms: Gender, institutional context, and representative bureaucracy. American Political Science Review  96: 553– 563. Google Scholar CrossRef Search ADS   Kettl, Donald F. 2005. The global public management revolution (2e) . Washington, DC: Brookings Institution Press. Krislov, Samuel J. 1974. Representative bureaucracy . Englewood Cliffs, NJ: Prentice Hall. Larsen, Christian A. 2012. Fordomme over for etniske minoriteter i Danmark, Sverige, Storbritannien og USA. Politica  44: 488– 504. Liebig, Thomas. 2007. The labour market integration of immigrants in denmark . OECD Social, Employment, and Migration Working Papers No. 50. Paris: OECD. Lipsky, Michael. 1980. Street-level bureaucracy: Dilemmas of the individual in public services . New York: Russell Sage Foundation Publications. Meier, Kenneth J. 1993a. Representative bureaucracy: A theoretical and empirical exposition. In Research in public administration , ed. James L. Perry,vol. 2, 1– 35. New Greenwich, CT: JAI Press. Meier, Kenneth J. 1993b. Latinos and representative bureaucracy testing the thompson and henderson hypotheses. Journal of Public Administration Research and Theory  3: 393– 414. Meier, Kenneth J., and Joseph Stewart. 1992. The impact of representative bureaucracies: Educational systems and public policies. The American Review of Public Administration  22: 157– 71. Google Scholar CrossRef Search ADS   Meldgaard, Eva V. 2005. Muslimske fornavne i Danmark. http://nfi.ku.dk/publikationer/webpublikationer/muslimske_fornavne (accessed October 7, 2016). Merton, Robert K. 1957. Social theory and social structure . Glencoe, IL: The Free Press. Metselaar, Suzanne, Gerben Meynen, and Guy Widdershoven. 2016. Reconsidering bias: A hermeneutic perspective. The American Journal of Bioethics  16: 33– 35. Google Scholar CrossRef Search ADS PubMed  Morewedge, Carey K., Haewon Yoon, Irene Scopelliti, Carl W. Symborski, James H. Korris, and Karim S. Kassam. 2015. Debiasing decisions: Improved decision making with a single training intervention. Policy Insights form the Behavioral and Brain Sciences  2: 129– 140. Google Scholar CrossRef Search ADS   Okonofua, Jason A., and Jennifer L. Eberhardt. 2015. Two strikes: Race and the disciplining of young students. Psychological Science  26: 617– 24. Google Scholar CrossRef Search ADS PubMed  Olson, Jeremiah C. 2016. Race and punishment in american prisons. Journal of Public Administration Research and Theory  26: 758– 68. Google Scholar CrossRef Search ADS   Pager, Devah. 2007. Marked: Race, crime, and finding work in an era of mass incarceration . Chicago, IL: Univ. of Chicago Press. Google Scholar CrossRef Search ADS   Pavetti, LaDonna, Michelle K. Derr, and Heather Hesketh. 2003. Review of sanction policies and research studies: final literature review . Washington, DC: US Department of Health and Human Services. https://aspe.hhs.gov/sites/default/files/pdf/72646/full-report.pdf (accessed January 3, 2017). Pickerill, J. Mitchell, Clayton Mosher, and Travis Pratt. 2009. Search and seizure, racial profiling, and traffic stops: A disparate impact framework. Law & Policy  31: 1– 30. Google Scholar CrossRef Search ADS   Quillian, Lincoln. 2008. Does unconscious racism exist. Social Psychology Quarterly  71: 6– 11. Google Scholar CrossRef Search ADS   Rabin, Matthew, and Dmitri Vayanos. 2010. The gambler’s and hot-hand fallacies: Theory and applications. Review of Economic Studies  77: 730– 78. Google Scholar CrossRef Search ADS   Rand, David G., Victoria L. Brescoll, Jim A. C. Everett, Valerio Capraro, and Hélène Barcelo. 2016. Social heuristics and social roles: Intuition favors altruism for women but not for men. Journal of Experimental Psychology  145: 389– 96. Google Scholar CrossRef Search ADS PubMed  Redder, Gitte, and Allan Christensen. 2016. Danskerne: Flygtninge og integration er vores allerstørste problem. Ugebrevet A4 (January 4, 2016). http://www.ugebreveta4.dk/danskerne-flygtninge-og-integration-er-vores-allerst_20333.aspx (accessed January 4, 2017). Sahin, Sultan, and Pernille Dreyer. 2015. Hvem er fatma og muhammed – og hvad mener de om Danmark? Berlingske, 8 August 2015. http://www.b.dk/nationalt/hvem-er-fatma-og-muhammed-og-hvad-mener-de-om-danmark (accessed October 7, 2016). Schram, Sanford F. 2006. Welfare discipline: Discourse, governance, and globalization . Philadelphia, PA: Temple Univ. Press. Schram, Sanford F., Joe Soss, Richard C. Fording, and Lina Houser. 2009. Deciding to discipline: race, choice, and punishment at the frontlines of welfare reform. American Sociological Review  74: 398– 422. Google Scholar CrossRef Search ADS   Schwartz, Shalom H., and Tammy Rubel. 2005. Sex differences in value priorities: cross-cultural and multimethod studies. Journal of Personality and Social Psychology  89: 1010– 1028. Google Scholar CrossRef Search ADS PubMed  Selden, Sally C. 1997. Representative bureaucracy: examining the linkage between passive and active representation in the farmers home administration. The American Review of Public Administration  27: 22– 42. Google Scholar CrossRef Search ADS   Selden, Sally C., Jeffrey Brudney, and J. Edward Kellough. 1998. Bureaucracy as a representative institution: toward a reconciliation of bureaucratic government and democratic theory. American Journal of Political Science  42: 719– 744. Shapiro, Robert Y., and Harpreet Mahajan. 1986. Gender differences in policy preferences: a summary of trends from the 1960s to the 1980s. Public Opinion Quarterly  50: 42– 61. Google Scholar CrossRef Search ADS   Soss, Joe, Richard C. Fording, and Sanford F. Schram. 2008. The color of devolution: the politics of local punishment in an era of neo-liberal welfare reform. American Journal of Political Science  52: 536– 53. Google Scholar CrossRef Search ADS   Soss, Joe, Richard C. Fording, and Sanford F. Schram. 2011a. The organization of discipline: from performance management to perversity and punishment. Journal of Public Administration Research and Theory  21( suppl 2): i203– i232. Google Scholar CrossRef Search ADS   Soss, Joe, Richard C. Fording, and Sanford F. Schram. 2011b. Diciplining the poor: neoliberal paternalsm and the persistent power of race . Chicago, IL: Univ. of Chicago Press. Statistics Denmark. 2015. Indvandrere i Danmark 2015 [Immigrants in Denmark 2015] . Copenhagen: Danmarks Statistik. Statistics Denmark. 2016. Navne i hele befolkningen [Names of the Total Population]. http://www.dst.dk/da/Statistik/emner/navne/navne-i-hele-befolkningen (accessed October 7, 2016). Sunshine, Jason, and Tom R. Tyler. 2003. The role of procedural justice and legitimacy in shaping public support for policing. Law & Society Review  37: 513– 548. Google Scholar CrossRef Search ADS   Tajfel, Henri, and John C. Turner. 1986. The social identity theory of intergroup behavior. In Psychology of intergroup relations , 2nd ed., eds. Stephen Worschel and William G. Austin, 7– 24. Chicago, IL: Nelson Hall. Thomsen, Jens Peter F. 2006. Konflikten om de nye danskere: Om danskernes holdninger til etniske minoriteters kultur og rettigheder . Copenhagen, Denmark: Akademisk Forlag. Togeby, Lise. 2004. Man har et standpunkt…: Om stabilitet og forandring i befolkningens holdninger . Magtudredningen. Aarhus, Denmark: Aarhus Universitetsforlag. Tyler, Tom R. 2003. Procedural justice, legitimacy, and the effective rule of law. Crime and Justice  30: 283– 357. Google Scholar CrossRef Search ADS   Wilkins, Vicky M., and Lael R. Keiser. 2006. Linking active and passive representation for gender: The case of child support. Journal of Public Administration Research and Theory  16: 87– 102. Google Scholar CrossRef Search ADS   Wilkins, Vicky M., and Brian N. Williams. 2008. Black or blue: Racial profiling and representative bureaucracy. Public Administration Review  68: 654– 64. Google Scholar CrossRef Search ADS   Footnotes 1 For the purpose of this study, “ethnic minorities” refers to non-Western immigrants having settled in Denmark and their descendants (the “second” or “third” generation). Statistics Denmark classifies an individual as a native Dane, or immigrant (“first-generation immigrant”), or a descendant of immigrants (“second-generation immigrant”) (Statistics Denmark 2015, pp. 11–12). A person is classified as native Danish if at least one of the person’s parents was born in Denmark and holds Danish citizenship. Individuals born outside of Denmark are considered first-generation immigrants, whereas individuals with parents born outside of Denmark are classified as descendants (second-generation immigrants). We also follow Statistic Denmark in terms of classifying non-Western immigrants and their descendants as those individuals not originating in the EU-28, Iceland, Norway, Switzerland, the European micro-states, the United States, Canada, Australia, or New Zealand (Statistics Denmark 2015, 11–12). 2 Figures based on authors’ calculations, using population data (FOLK2) from Statistics Denmark, first quarter of 2016. Available at www.statistikbanken.dk. 3 Ranking based on authors’ calculations, using population data (FOLK1) from Statistics Denmark, first quarter of 2016. Available at www.statistikbanken.dk. 4 Figures based on authors’ calculations, using the latest available unemployment data (RAS1F) from Statistics Denmark. Available at www.statistikbanken.dk. 5 Furthermore, non-Western immigrants are occasionally stereotyped as having a penchant for petty crime and lower levels of educational attainment (ALS Research 2014), as well as being unwilling to accept or live by the democratic values of the countries in which they settle (Bursell 2012)—group reputations that may also affect perceptions of deservingness with respect to public assistance. 6 The administrative data contain information on the share of cash benefit recipients who are non-Western immigrants or their descendants for each employment agency but do not allow us to focus especially on those of Middle-Eastern origin. However, analyses conducted by Statistics Denmark show that a larger percentage of non-Western immigrants than ethnic Danes live of public support (cash or other benefits). The countries of origin most prevalent among those non-Western immigrants are Turkey, Syria, Iraq, and Lebanon (Statistics Denmark 2016). 7 The survey experimental findings have a modest claim to ecological validity. The caseworkers’ responses are based on a description of a hypothetical client in a survey. This situation is clearly different from actual casework. For example, the respondents did not interact with a real person, and they were asked to base their responses on only the information that the vignette provided. In real life, caseworkers can request or seek additional data, and face-to-face client meetings are likely to inform decision-making. However, Study 1 complements Study 2 in terms of ecological validity (similar to the way that Study 2 complements Study 1 in terms of internal validity). Although the administrative data analysis does not offer the same potential for causal inference as the survey experimental analysis, it allows us to motivate the survey experiment with data reflecting sanctioning decisions as they naturally occur in real-life cases and work environments. 8 A skewness-kurtosis test for normality shows that the score distribution is slightly left-skewed (−.93, p < .001) but has a kurtosis similar to that of a standard normal distribution (2.84, p = .50). As a robustness check, because of the left-skewed distribution, we have re-analyzed the survey experimental data using quantile (median) regression instead of OLS. Although the median regression estimates are slightly smaller and accompanied by larger standard errors, the estimates do not change the substantive findings. 9 As a robustness test, we also estimated the work experience interaction model with the inclusion of an index variable capturing the caseworkers’ general perceptions of clients’ efforts and motivations for obtaining employment. In theory, the moderation effect could be driven by more negative perceptions of clients among less experienced caseworkers (rather than by work experience per se). We constructed the index measure using three Likert-scale items from the caseworker survey: “Almost all clients are sincerely very much motivated to obtain employment”; “Many clients are at times unreasonable or troublesome (reversed)”; and “A fair share of clients do not, in reality, wish to work (reversed).” Cronbach’s alpha was .79. As expected, a more positive perception of clients’ efforts and motivations is negatively associated with decisions to sanction (β = −.557; p < .001). More importantly, however, including the index variable in the model does not qualitatively change the work experience interaction term estimate (β = −.083, p = .039)—thus suggesting that our work experience moderation result is not driven (or biased) by general perceptions of the clients’ efforts and motivations for obtaining employment. A second concern relates to the dynamics of policy contexts. Although the use of sanctions has been an integral part of the Danish unemployment regime since the late 1980s, the cash benefit program has been marked by a gradually increasing emphasis on sanctioning clients for noncompliance (Caswell et al. 2011). Given this trajectory, could the moderating effect of work experience be the result of a policy socialization process in which the caseworkers’ sanctioning norms are shaped by the unemployment policy operating at the point in time when they entered the profession? In the attempt to consider this concern, we examined the correlation between work experience and overall sanctioning propensity—for the full sample and by treatment group. Our expectation is that said policy socialization process would be reflected in a lesser overall propensity to sanction among the more experienced caseworkers. However, the analyses do not support this notion. To the contrary, we find weak evidence that the more experienced caseworkers are more likely to recommend sanction than their less experienced peers (p = .09). © 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: Apr 11, 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