Policy implementation, street-level bureaucracy and the importance of discretion 2014 Accepted article for Public Management Review Special issue 'Street-level bureaucracy and government performance' Lars Tummers and Victor Bekkers To be cited as: Tummers, L.G. & Bekkers, V.J.J.M. (2014). Policy implementation, street-level bureaucracy and the importance of discretion. Public Management Review, 16(4), 527-547. Dept. of Public Administration Erasmus University Rotterdam P.O. Box 1738 NL-3000 DR Rotterdam [email protected], [email protected] Abstract Street-level bureaucrats implementing public policies have a certain degree of autonomy – or discretion – in their work. Following Lipsky, discretion has received wide attention in the policy implementation literature. However, scholars have not developed theoretical frameworks regarding the effects of discretion, which were then tested these using large samples. This study therefore develops a theoretical framework regarding two main effects of discretion: client meaningfulness and willingness to implement. The relationships are tested using a survey among 1,300 healthcare professionals implementing a new policy. The results underscore the importance of discretion. Implications of the findings and a future research agenda is shown. Key words:      Discretion Public policy Policy implementation Street-level bureaucracy Quantitative analysis 2 1 Introduction In his book “Street-level bureaucracy: Dilemmas of the individual in public services”, Michael Lipsky (1980) analysed the behaviour of front-line staff in policy delivery agencies. Lipsky refers to these frontline workers as “street-level bureaucrats”. These are public employees who interact directly with citizens and have substantial discretion in the execution of their work (1980:3). Examples are teachers, police officers, general practitioners and social workers. These street-level bureaucrats implement public policies. However, street-level bureaucrats have to respond to citizens with only a limited amount of information or time to make a decision. Moreover, very often the rules the street-level bureaucrats have follow do not correspond to the specific situation of the involved citizen. In response, street bureaucrats develop coping mechanisms. They can do that because they have a certain degree of discretion – or autonomy - in their work (1980:14). Following the work of Lipsky, the concept of discretion has received wide attention in the policy implementation literature (Brodkin, 1997; Buffat, 2011; Hill & Hupe, 2009; Sandfort, 2000; Tummers et al., 2009; Vinzant et al., 1998). However, scholars have not yet developed theoretical frameworks regarding the effects of discretion, which were subsequently tested using large scale quantitative approaches (Hill & Hupe, 2009; O'Toole, 2000). This study aims fill this gap by developing a theoretical framework regarding two effects of discretion. The first effect, which is often noted, is that a certain amount of discretion can increase the meaningfulness of a policy for clients (Palumbo et al., 1984). An example can clarify this. A teacher could adapt the teaching method to the particular circumstances of the pupil, such as his/her problems with long-term reading, but swiftness when discussing the material in groups. The teacher could devote more attention to the pupil’s reading difficulties, thereby providing a more balanced development. More in general, it is argued that when street-level bureaucrats have a certain degree of discretion, this will make the policy more meaningful for the clients. Client meaningfulness can thus be considered a potential effect of discretion. Here, we note that client meaningfulness is highly related to concepts such as client utility or usefulness. Furthermore, it can be argued that providing street-level bureaucrats discretion increases their willingness to implement the policy (Meyers & Vorsanger, 2003; Sandfort, 2000). Tummers (2011) showed this effect while studying 'policy alienation', a new concept for understanding the problems of 3 street-level bureaucrats with new policies. One mechanism underlying this relationship between discretion and willingness to implement seems to be that a certain amount of discretion increases the (perceived) meaningfulness for clients, which in turn enhances their willingness to implement this policy (Hill & Hupe, 2009; Lipsky, 1980). This is expected as street-level bureaucrats want to make a difference to their clients’ lives when implementing a policy (Maynard-Moody & Musheno, 2000). Hence, when street-level bureaucrats perceive that they have discretion, they feel that they are better able to help client (more perceived client meaningfulness), which in turn increases their willingness to implement the policy. This is known as a mediation effect. This mediation effect is often implicitly argued, has yet to be studied empirically. Based on this rationale the central research question is: To what extent does discretion influence client meaningfulness and willingness to implement public policies, and does client meaningfulness mediate the discretion-willingness relationship? This brings us to the outline of this article. We will firstly develop a theoretical framework (Section 2), outlining the relationships between discretion, client meaningfulness, and willingness to implement. Section 3 describes the operationalization of the concepts and research design, which is based on a Dutch nationwide survey among 1.300 psychologists, psychiatrists and psychotherapists implementing a new reimbursement policy. The results section shows descriptive statistics and discusses the hypotheses. We conclude by discussing the contribution of this article to policy implementation literature with a particular emphasis on the importance discretion of street-level bureaucrats. 2 2.1 Theoretical framework Background on discretion This article focuses on the discretion of street-level bureaucrats during policy implementation. Due to the abundance of literature and the intrinsic difficulties with the discretion concept (such as the different interpretations attached to as well as criticisms of these interpretations) we will provide only a short overview of the term discretion (for elaborate overviews, see Evans, 2010; Hill & Hupe, 2009; Lipsky, 1980; Maynard-Moody & Portillo, 2010; Meyers & Vorsanger, 2003; Saetren, 2005; Winter, 2007). For a recent critique on discretion, see Maynard-Moody and Musheno (2012). 4 Evans (2010) has noted that for employees, discretion can be seen as the extent of freedom he or she can exercise in a specific context. Related to this, Davis (1969:4) states “a public officer has discretion whenever the effective limits on his power leave him free to make a choice among possible courses of action or inaction” (see also Vinzant et al., 1998). Lipsky (1980) focuses more specifically on discretion of street-level bureaucrats. He views discretion as the freedom that street-level bureaucrats have in determining the sort, quantity and quality of sanctions and rewards during policy implementation (see also Hill & Hupe, 2009; Tummers, 2012). We then define discretion as the perceived freedom of street-level bureaucrats in making choices concerning the sort, quantity, and quality of sanctions and rewards on offer when implementing a policy; for instance, to what extent experience policemen that they themselves decide whether to give an on-the-spot fine? To what extent feel teachers they can decide what and how to teach students about the development of mankind, i.e. evolution or creationism (Berkman & Plutzer, 2010)? As can be seen from the previous paragraph, we focus on experienced discretion. This is based on Lewin’s notion (1936) that people behave on the basis of their perceptions of reality, not on the basis of reality itself (Thomas Theorem). Street-level bureaucrats may experience different levels of discretion within the same policy because for example because a) they possess more knowledge on (loopholes) in the rules, b) their organization operationalized the policy somewhat differently, c) they have a better relationship with their manager which enables them to adjust themselves to circumstances, or d) the personality of the street-level bureaucrat is more rule-following or rebellious (Brehm & Hamilton, 1996; Prottas, 1979). In both top-down and bottom-up approaches of policy implementation, the notion of discretion is important (DeLeon & DeLeon, 2002; Hill & Hupe, 2009). From a top-down perspective, discretion is often not welcomed (Davis, 1969; Polsky, 1993). Discretion is primarily seen as a possibility that street-level bureaucrats use to pursue their own, private goals. This can influence the policy program to be implemented in a negative way, which undermines the effectiveness and democratic legitimacy of a program (Brehm & Gates, 1999). In order to deal with this issue, control mechanisms are often put in place in order to achieve compliance. In the bottom-up perspective discretion is assessed differently. Discretion is seen as inevitable in order to deploy general rules, regulations and norms in specific situations, which helps to improve the effectiveness of policy programs and the democratic support for the program. Moreover, given the 5 limited time, money and other resources available and the large amount of rules, regulations and norms that have to be implemented, it is important that street level bureaucrats are able to prioritize what rules to apply, given the specific circumstances in which they operates in (Brodkin, 1997; Maynard-Moody & Musheno, 2000; Maynard-Moody & Portillo, 2010). From a top-down and bottom-up perspective it can be argued that discretion has a different meaning for citizens as a client. In the top-down perspective discretion could possibly harm the position of a citizen, because private considerations and interpretations of the goals of policy program by the street-level bureaucrat prevent that citizens are threatened in equally. In the bottom-perspective discretion will help to strengthen the value/meaningfulness of a policy for clients, as policy programs can be targeted to their specific situation. Hence, from a bottom-up perspective discretion might increase the client meaningfulness that is, the value of the policy for clients (Barrick et al., 2012; Brodkin, 1997; D. R. May et al., 2004; Maynard-Moody & Musheno, 2003; Tummers, 2011). Client meaningfulness can be defined as the perception of street-level bureaucrats that their implementing a policy has value for their own clients. Client meaningfulness is therefore about the perception of the street-level bureaucrat that a policy is valuable for client (a client may not feel the same way). For instance, a social worker might feel that when he/she implements a policy focused on getting clients back to work, this indeed helps the client to get employed and improves the quality of life for this client. Granting street-level bureaucrats discretion during policy implementation can increase client meaningfulness as many situations street-level bureaucrats face are too complicated to be reduced to programmatic formats. Discretion makes it possible to adapt the policy to meet the local needs of the citizens/clients, increasing the meaningfulness of the policy to clients. It seems that discretion could positively also affect the street-level bureaucrats’ willingness to implement the policy. Willingness to implement is defined as a positive behavioural intention of the street-level bureaucrat towards the implementation the policy (Ajzen, 1991; Metselaar, 1997). Hence, the street-level bureaucrat aims to put effort in implementing this policy: he/she tries to make it work. Policy implementation literature, especially the studies rooted in the bottom-up perspective, suggests that an important factor in this willingness of street-level bureaucrats is the extent to which organizations are willing and able to delegate decision-making authority to the frontline (Meier & O'Toole, 2002). This influence may be particularly pronounced in professionals whose expectations of discretion and autonomy contradict notions of bureaucratic control (Freidson, 2001). 6 To conclude, it seems that discretion can have various effects. In this article, we specifically examine two possible positive effects of discretion: enhanced client meaningfulness for clients and more willingness to implement the policy. These effects are chosen given their dominant role in the policy implementation debate (Ewalt & Jennings, 2004; Riccucci, 2005; Simon, 1987; Tummers et al., 2012). 2.2 The effects on discretion on client meaningfulness and willingness to implement Given the arguments stated previously, we firstly expect that when street-level bureaucrats experience high discretion, this positively influences their perception of client meaningfulness. Sandfort (2000) illustrates this by describing a case in U.S. public welfare system (Work First contracters). Regardless of the specifies of the local office, street-level bureaucrats are given the same resources to carry out their tasks: standardized forms, policy manuals, complex computer programmes, etc. Such structures cause the street-level bureaucrats to be isolated from other professionals and unable to adapt existing practices to altering demands. Hence, it reduces their discretion and this could result in less client meaningfulness. We will study this same process using a quantitative approach, bringing us to the first hypothesis. H1: When street-level bureaucrats experience more discretion, this positively influences their experienced client meaningfulness of the policy Next, we expect that when street-level bureaucrats feel that they have enough discretion, this positively influences their willingness to implement a policy. Maynard-Moody and Portillo (2010:259) note, “Street-level workers rely on their discretion to manage the physical and emotional demands of their jobs. They also rely on their discretion to claim some small successes and redeem some satisfaction.” Examining this more generally, the mechanism linking discretion to willingness to implement can be traced back to the human relations movement (McGregor, 1960). One of the central tenets of this movement is that employees have a right to give input into decisions that affect their lives. Employees enjoy carrying out decisions they have helped create. As such, the human relations movement argues that when employees experience discretion during their work, this will positively influence several job indicators by fulfilling intrinsic employee needs. Next to this, self-determination theory (Deci & Ryan, 2004) argues that three psychological needs must be fulfilled to foster 7 motivation: competence, relatedness and autonomy. In short, they argue that when people perceive to have autonomy, they are more motivated to perform. H2: When street-level bureaucrats experience more discretion, this positively and directly influences their willingness to implement the policy Furthermore, we expect that when street-level bureaucrats experience more discretion, this positively influences their client meaningfulness, which in turn positively influences their willingness to implement a policy. Hence, client meaningfulness could influence the willingness to implement a policy. This is expected as street-level bureaucrats want to make a difference to their clients’ lives when implementing a policy. May and Winter (2009) found that if frontline workers perceive the instruments at their disposal for implementing a policy as ineffective, in terms of delivering to clients, this is likely to add to their frustrations. They do not see how their implementation of the policy helps their clients so wonder why they should implement it. Technically speaking we expect a mediation effect to occur (Zhao et al., 2010). Mediation is the effect of an independent variable (here: discretion) on a dependent variable (willingness to implement) via a mediator variable (client meaningfulness). Hence, besides hypothesizing the direct effect of discretion on willingness to implement, we expect that part of this effect is caused by increasing client meaningfulness. This can be considered a partially mediated effect: part of the effect of discretion on willingness to implement is mediated by client meaningfulness. Full mediation is not expected. Some of the influence of discretion on willingness to implement is explained by factors other than increasing client meaningfulness, i.e. peoples intrinsic need for autonomy in their work (Wagner III, 1994). H3: The positive influence of discretion on willingness to implement is partially mediated by the level of client meaningfulness This mediation effect can be related to established job design theories like the job characteristics model of Hackman & Oldham (1980). Hackman & Oldham note that autonomy (related to discretion) is one of the core job characteristics, enhancing experienced responsibility for outcomes. This influences 8 critical psychological states, such as experienced meaningfulness of work (related to client meaningfulness). In turn, experienced meaningfulness of work fosters individual and organizational outcomes, such as high internal motivation (related to willingness to implement). Hence, important similarities between their line of reasoning and ours can be found. An important different is that we focus on the level of policy implementation instead of the general job level. Based on these three hypotheses, a theoretical framework is constructed, shown in Figure 1. Figure 1 Proposed theoretical framework regarding two main effects of discretion. + Discretion 3 3.1 + Client meaningfulness + Willingness to implement Methods Case To test the theoretical framework, we undertook a survey of Dutch mental healthcare professionals implementing a new reimbursement policy (Diagnosis Related Groups). First, a short overview of this policy is provided. In January 2008, the Dutch government introduced Diagnoses Related Groups (DRGs, in Dutch DiagnoseBehandelingCombinaties, or DBC’s) in mental healthcare. The DRGs are part of the new Law Health Market Organization. The DRGs can be seen as the introduction of regulated competition into Dutch healthcare, a move in line with New Public Management (NPM) ideas. More specifically, it can be seen as a shift to greater competition and more efficient resource use (Hood, 1991:5). The system of DRGs was developed as a means of determining the level of financial exchange for mental healthcare provision. The DRG-policy differs significantly from the former method in which each medical action resulted in a financial claim; this meant that the more sessions a professional caregiver (a psychologist, psychiatrist or psychotherapist) had with a patient, the more 9 recompense could be claimed. This former system was considered inefficient by some (Kimberly et al., 2009). The DRG-policy changed the situation by stipulating a standard rate for each disorder. For instance, for a mild depression, the mental healthcare gets a standard rate and can treat the patient (direct and indirect time) between 250-800 minutes. The DRG policy these professionals have to implement is related more to service management than to service delivery. However, this policy does have effects on service delivery. Professionals have to work in a more ‘evidence-based’ way and are required to account for their cost declarations in terms of the mental health DSM (Diagnostic Statistical Manual) classification system. As a result it becomes harder to use practices that are difficult to standardize and evaluate, such as psychodynamic treatments. Discretion regarding the length of treatment is arguably also increasingly limited. Whereas, in the former system, each medical action resulted in a payment, under the DRG policy a standard rate is determined for each disorder meaning it has become more difficult to adjust the treatment to the specific patient needs. Hence, the number of treatments for a patient is often limited due to the DRGpolicy, thereby changing service delivery. It is interesting to study how much discretion street-level bureaucrats really experienced during implementing this policy, and what effects this has. We noted that we focus on experienced discretion. Even within the same policy, some streetlevel bureaucrats will perceive more discretion than others. Indeed, in the open answers of the survey we witnessed that some respondents felt that they had substantial discretion when implementing this policy, while others felt very limited. Illustrative quotes from different respondents are (all from open answers in the survey, which is reported next): “The DRG-policy does not force me into a certain choices. I examine the funding scheme of the treatment only ‘in second instance’” “I do my work first and foremost according to professional standards and hereafter just attach a DRG-label which I think fits but best.” “With the DRG-policy, I am being forced into a straitjacket.” “You are bound by the rules. so that's a harness.” 10 3.2 Sampling and response Our sampling frame consisted of 5,199 professionals who were members of two nationwide mental health care associations (Dutch Association of Psychologists, in Dutch ‘Nederlands Instituut van Psychologen’, NIP, and the Netherlands Association for Psychiatry, in Dutch ‘Nederlandse Vereniging voor Psychiatrie’, NVvP). These were all the members of those associations, who could in principle, be working with the DRG policy. Using an email and two reminders, we received 1,317 returns of our questionnaire: a response of 25%. Our sampling frame consisted of high status professionals: psychiatrists, psychologists and psychotherapists. Most research analyzing discretion focuses on traditional street-level bureaucrats, such as welfare workers and police officers (Maynard-Moody & Portillo, 2010). However, these mental healthcare professionals are a specific group of highly trained professionals, which traditionally, due to their professional training, have substantial autonomy. Furthermore, they also have to implement governmental policies (in this case, Diagnosis Related Groups). Hence, it seems worthwhile to analyse such professional groups using the theoretical lens of street-level bureaucracy (see also Hupe & Hill, 2007). Of the valid respondents, 36% were men and 64% women, which is consistent with Dutch averages for mental healthcare professionals, where 69% of the workforce are female (Palm et al., 2008). The respondents’ ages ranged from 23 to 91 years (M = 48), which is slightly older then the Dutch national average for mental healthcare professionals (M = 44). Hence, respondents mean age and gender-distribution are quite similar to those of the overall mental healthcare sector. To rule-out a possible non-response bias, we conducted non-response research where we contacted the nonresponders for their reasons for non-participation. Common reasons for not participating were: lack of time, retirement, change of occupation, or not working with the DRG policy. Some organizations, including some hospitals, were not yet working with this policy. The large number of respondents, their characteristics in terms of gender and age, and the results of the non-response research indicate that our respondents are quite a good representation of the population. 3.3 Measures This section reports the measurement of the variables. Unless stated otherwise, the measures were formatted using five-point Likert scales, ranging from strongly agree to strongly disagree. For the items 11 tapping discretion, client meaningfulness. and willingness to implement, we used templates. Templates allow the researcher to specify an item by replacing general phrases with more specific ones that better fit the research context (DeVellis, 2003). For example, instead of stating ‘the policy’ or ‘professionals’, the researcher can rephrase these items using the specific policy and group of professionals being examined. Here, ‘the DRG policy’ and ‘healthcare professionals’ replaced the template terms. Items are therefore easier for professionals to understand, since items are better tailored to their context and this, in turn, increases reliability and content validity (DeVellis, 2003:62). All items are shown in the Appendix 2. Discretion Discretion concerns the perceived freedom of the implementer in terms of the type, quantity and quality of sanctions and rewards delivered (Lipsky, 1980). The scale is based on the validated measurement instrument of policy alienation, specifically the dimension ‘operational powerlessness’ (Tummers, 2012). Three items were used based on confirmatory factor analysis (see Section 4). Cronbach’s alpha = .78. Client meaningfulness. Client meaningfulness (or meaninglessness) was also conceptualized as a dimension of policy alienation (Tummers, 2012). It refers to the perception of professionals about the benefits of implementing the DRG policy for their own clients. For instance, do they perceive that they are really helping their patients by implementing this policy? Three items were used based on confirmatory factor analysis. Cronbach’s alpha = .77. Willingness to implement Willingness to implement was measured using Metselaar’s (1997) four-item scale. All items were used based on confirmatory factor analysis. Cronbach’s alpha = .83. Control variables Commonly used individual characteristics were included: gender, age and management position (yes/no). We also distinguish between psychiatrists and others, because the former belong to a 12 medical profession. Psychologists and psychotherapists are non-medical professionals, which possibly influenced their perceptions. 3.4 Statistical methods used We used Confirmatory Factor Analysis (CFA) followed by Structural Equation Modeling (SEM). The CFA and SEM techniques are often used in psychology research, but quite new to most public administration scholars (but see for instance Wright et al., 2012). We therefore discuss a number of the analyses’ characteristics in detail. CFA is a technique used to test the factor structure of latent constructs based on theory and prior research experience. This is appropriate in our case given that prior analyses have already explored the variables discretion, client meaningfulness and willingness to implement. CFA has several advantages over exploratory factor analysis, such as more stringent psychometric criteria for accepting models, thereby improving validity and reliability (Brown, 2006). Using CFA a measurement model is specified. The measurement model specifies the number of factors and shows how the indicators (items) relate to the various factors (Brown, 2006:51). Hence, it shows for instance how the items asked to measure discretion relate to the latent construct of discretion. This measurement model is a precursor for the SEM-analysis. In the SEM-analysis, a structural model is constructed showing how the various latent factors relate to each other. For instance, it shows how discretion is related to willingness to implement. In the SEM-analysis a total model can be tested where variables can be both dependent and independent. This is an advantage over regression analyses. Given that we hypothesize that client meaningfulness is both dependent (influenced by discretion) and independent (influencing willingness to implement), this was appropriate for our model. For mediation models, as is our model, SEM is preferred over regression analysis (Zhao et al., 2010). The latent variable program Mplus was used for the analyses (Muthén & Muthén, 1998-2010). Mplus is suited for handling non-normally distributed data, which is often the case when employing surveys. As our data were (mildly) non-normally distributed, this was an advantage. Robust Maximum Likelihood was used, which works well in these circumstances (Brown, 2006:379). 3.5 Measurement model Before analyzing the structural model (See Section 4), the measurement model is analysed. 13 Based on the analyses for the measurement model, some modifications were made to improve the model. The only modifications were to delete a number of items for the latent factors: three for discretion, one for client meaningfulness, one for willingness to implement. This was based on theoretical grounds, fit of item content with definition of concept/latent factor, and the minimization of the Akaike Information Criterion (AIC). This fit index can be used to compare competing models. As suggested we selected the model with the lowest AIC, thereby taking into account theoretically plausibility (Schreiber et al., 2006). More specifics about the measurement model are described in Appendix 1. 4 4.1 Results Descriptive statistics Table 1 shows the means and variances/covariances for all items used. A number of interesting results can be seen. First, many street-level bureaucrats are psychiatrists and these often occupy management positions. Next, the means for discretion are quite low, showing that the street-level bureaucrats do not feel that they have a lot of autonomy in this policy. We also found low scores for willingness to implement and even lower scores for client meaningfulness. Hence, in general the street-level bureaucrats were quite negative about this policy. The covariances for the items linked via our hypotheses are in the anticipated direction. For example, items regarding willingness to implement are positively related to discretion. 14 Table 1 Mean and variance/covariance matrix (variances on the diagonal) Discretion Mean 1 2 Client meaningfulness 3 1 2 3 Willingness to implement 1 2 3 Control variables 4 Gender Age Psychiatrist Mng. position Discretion Discretion 1 2,54 1.07 Discretion 2 2,78 0.69 1.32 Discretion 3 3,01 0.49 0.74 1.05 Meaningfulness 1 1,77 0.17 0.24 0.19 0.57 Meaningfulness 2 1,81 0.15 0.21 0.18 0.49 0.63 Meaningfulness 3 2,04 0.16 0.21 0.20 0.36 0.36 1.06 Willingness 1 1.93 0.23 0.34 0.25 0.34 0.35 0.32 0.74 Willingness 2 2.55 0.23 0.30 0.28 0.29 0.29 0.24 0.51 1.10 Willingness 3 2.27 0.22 0.32 0.23 0.28 0.30 0.28 0.58 0.59 0.85 Willingness 4 2.63 0.17 0.30 0.27 0.21 0.24 0.22 0.41 0.51 0.47 1.01 Gender (female) 64% 0.04 0.01 0.04 0.04 0.04 0.06 0.04 0.06 0.05 0.07 0.24 Age 47.94 -0.23 0.20 0.10 -0.77 -1.04 -0.91 -0.48 -1.74 -0.71 -1.13 -1.66 114.55 Psychiatrist 42% -0.04 -0.06 -0.06 -0.02 -0.03 -0.01 0.02 -0.01 0.03 -0.01 -0.06 0.96 0.25 Managing position 44% -0.03 -0.04 -0.06 -0.04 -0.06 -0.05 -0.05 -0.06 -0.04 -0.07 -0.06 1.14 0.09 Client meaningfulness Willingness to implement Control variables 0.25 4.2 Structural model The structural equation model is shown in Figure 2. Table 2 shows the results, including control variables. First, an effect of discretion on client meaningfulness was found (standardized coefficient .33, p<.01). Hence, we do not reject Hypothesis 1. Second, the empirical tests show a cascading effect from discretion to willingness to implement through the mediating variable client meaningfulness. The effect (standardized coefficient) of discretion on client meaningfulness was .33 (p<.01), while the effect from client meaningfulness on willingness to implement was .49 (p<.01). The total indirect effect was therefore .16 (33*.49, p<.01). Based on this, we do not reject hypothesis 3. Furthermore, the direct effect was also significant (β=.27, p<.01), thus hypothesis 2 is not rejected. The total effect of discretion on willingness to implement is the sum of its direct and indirect effects: .27+.16=.43. This means that – all other things being equal – when the perceived discretion of the street-level bureaucrat increases by 1, the willingness to implement increases by .43. As there is both a direct and an indirect significant effect there is evidence of partial mediation, which was also hypothesized. This (partially mediated) model proved to be a good fit of the data: RMSEA = .04 (criterion ≤ .08), CFI = .97 (criterion ≥ .90), TLI = .96 (criterion ≥ .90). To shed more light on the mediating mechanisms we conducted additional SEM analyses to test the validity of two alternative models: a model without mediation and a model with full mediation. The model without mediation did not fit as adequately as the partially mediated model, given that the AIC was higher compared to the partially mediated model, and the fit indexes showed a worse fit. The fully mediated model also had a higher AIC, and worse scores on RMSEA, CFI and TLI than the partially mediated model, although differences are small. We used bootstrapping to test the indirect effect of discretion on willingness to implement via client meaningfulness. Bootstrapping is the preferred method for testing mediated effects (Preacher & Hayes, 2004; Zhao et al., 2010). It presents estimates and confidence intervals so that we can test the significance of the mediation effect. The 99% confidence interval for the standardized indirect effect 1 (which was .16) is between .11 and .22, meaning the indirect effect is not equal to 0 (p<.01). Hence, a 1 Bootstrap 5000 times, Maximum Likelihood estimation is used as Robust Maximum Likelihood is not available for bootstrapping. mediation effect is clearly present here. In the discussion and conclusion, we discuss the implications of these results for both theory and practice. Figure 2 Structural equation model for relationships between discretion, client meaningfulness and willingness to implement (control variables not shown) .28 Discretion .33 Client meaningfulness (R2=.14) .49 Willingness to implement (R2=.45) Table 2 Results From Structural Equation Modelling Model Meaningfulness for Meaningfulness Willingness to Willingness to clients for clients implement implement (standardized (unstandardized (standardized (unstandardized scores) scores) scores) scores) Gender NS NS NS NS Age -.092 -.006 NS NS Managing position NS NS .144 .212 Psychiatrist NS NS NS NS Discretion .330 .334 .278 .302 Meaningfulness for clients - - .491 .527 - - .162 .176 .135 - .446 - Control variables Direct influences Indirect influence Discretion via meaningfulness for clients R2 Note: NS = Not significant. All shown scores are significant at p<.01 17 5 Conclusion The central goal of this article is to understand the mechanisms at work between discretion, client meaningfulness and willingness to implement. Based on a literature review, a theoretical model was constructed linking discretion, client meaningfulness and willingness to implement. This model was tested in a survey of 1,317 mental healthcare professionals implementing a new policy. The model worked adequately in that discretion, together with conventional control variables, indeed partly 2 explained client meaningfulness (R =14%). Furthermore, willingness to implement was indeed 2 explained by discretion, client meaningfulness and the control variables (R =45%). Fit criteria were very good for the measurement model and the structural model, thereby strengthening the reliability and validity of the study. As such, we can conclude that the approach worked satisfactorily and adds to the literature on street-level bureaucracy. Having reached this conclusion, we can summarize the results, highlight limitations, and develop a future research agenda on discretion. We found that the discretion of street-level bureaucrats influences the willingness to implement in two ways. First, discretion influences client meaningfulness because street-level bureaucrats are more able to tailor their decisions and the procedures they have to follow to the specific situations and needs of their clients. Hence, discretion gives street-level bureaucrats the possibility to apply their own judgments when dealing with the needs and wishes of citizens. Our results strengthen the claim made by several authors: that discretion could indeed have positive effects for clients (Handler, 1990; P. J. May & Winter, 2009). At the same time, the positive effect that discretion has on the bureaucrat’s perception of client meaningfulness can be seen as a condition for the second effect: more willingness to implement the policy. When street-level bureaucrats perceive that their work is meaningful for his/her clients, this strongly influences their willingness to implement it. This is in line with the notion that street-level bureaucrats want to make a difference to their clients’ lives (Maynard-Moody & Musheno, 2003). Furthermore, the results also point to another, more autonomous, effect that discretion directly influences willingness to implement; hence, discretion is inherently valued by bureaucrats. The results have interesting implications for the theory and practice of policy implementation. From a theoretical point of view, it contributes to the long lasting discussion about the validity of a more top-down or bottom-up perspective on policy implementation. Discretion indeed seems to have a positive effect on the effectiveness of policy programs, as it reduces resistance. At the same time it 18 adds to the legitimacy of the policy implementation process, because it enables street-level bureaucrats to meet the needs and wishes of citizens (in the eyes of the street-level bureaucrats). These implications of the findings are strengthened by the large scale quantitative analysis and sophisticated techniques. The arguments that are put forward in the bottom-up perspective on the positive role that discretion plays in the effectiveness and democratic legitimacy of public policy programs are being confirmed. For practitioners, it is important to note that when drafting policy program it can be beneficial to give the implementing street-level bureaucrats some (perceived) freedom to adjust the policy program in order to be effective and legitimate. This has also important consequences for the role of performance and risk management in the implementation of these programs. The central role that detailed performance indicators and risk reduction rules play in the implementation process often leads to a broad variety of detailed norms and guidelines that the street-level bureaucrats involved must obey (Power, 1997). Next to this, the results show that client meaningfulness, in itself, proved to be very important, something which is not often mentioned in the street-level bureaucracy literature or in more general management literature, which focus often on influence, autonomy and discretion (Green, 2008; McGregor, 1960; Sowa & Selden, 2003; Spence Laschinger et al., 2001). For instance, Judson (1991) argues that providing employees with influence is the most powerful lever in gaining acceptance for a change. However, given the results of this study, we urge practitioners and scholars to also consider the perceived meaningfulness of the policy for clients, rather than to restrict their focus on discretion and influence aspects. This brings us to the limitations and future research suggestions. First, the results found could be dependent on this research context. This study addresses high status professionals: psychologists, psychiatrists, and psychotherapists. Furthermore, the specific policy context (DRG-policy, focused on cost-cutting and transparency) could influence the results. It would be interesting to conduct studies using the same theoretical model which focus on other groups of street-level bureaucrats who have other types of professional training or who are a part of government service bureaucracy. Related to this, an interesting venue for research would be to analyze cases which are more directly related to service delivery and less to service management. Here, stronger effects of discretion on client meaningfulness could be found. Furthermore, it would be worthwhile to analyze the developed model 19 in a situation where there was in general high discretion, client meaninglessness and willingness to implement, contrary to the case analyzed. Are the effects of discretion and client meaningfulness also important in such rather different policy contexts? Secondly, further research could use multiple sources to measure the indicators, and measure new effects of discretion. It would be worthwhile to measure client meaningfulness by asking the clients themselves. Furthermore, other indicators could be linked to discretion, such as objective indicators like the percentage of people getting a job when implementing reintegration policies. Does granting street-level bureaucrats discretion in such a policy heighten the ‘success’ of such a policy? Linked to this, we should note that we have looked at only two possible positive effects of discretion. We have largely ignored its negative side, such as discrimination of clients or the ways discretion can break public trust (Sandfort, 2000). Thirdly, future research could investigate other factors influencing client meaningfulness and willingness to implement, including other control variables. Scholars could, for instance, examine the influence of organizational factors such as the level of trust between professionals and management, incentive systems which promote or stymie implementing a policy or the way the policy has been implemented (top-down, bottom-up) within an organization. Next to this, personality characteristics could be taken into account, such as optimism, self-efficacy beliefs and locus of control. To conclude, this study provides important insights that help to understand the effects of granting street-level bureaucrats discretion in their work. It underscores the importance of studying discretion. Embracing and further researching this should prove to be a timely and productive endeavour for both researchers and practitioners alike. 20 References Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179-211. Barrick, M., Mount, M., & Li, N. (2012). The theory of purposeful work behavior: The role of personality, job characteristics, and experienced meaningfulness. Academy of Management Review, 38(1), 132-153. Berkman, M., & Plutzer, E. (2010). Evolution, creationism, and the battle to control america's classrooms. Cambridge: Cambridge Univ Press. Brehm, J., & Gates, S. (1999). Working, shirking, and sabotage: Bureaucratic response to a democratic public. Michigan: University of Michigan Press. Brehm, J., & Hamilton, J. T. (1996). Noncompliance in environmental reporting: Are violators ignorant, or evasive, of the law? American Journal of Political Science, 40(2), 444-477. Brodkin, E. Z. (1997). Inside the welfare contract: Discretion and accountability in state welfare administration. The Social Service Review, 71(1), 1-33. Brown, T. A. (2006). Confirmatory factor analysis for applied research. London: The Guilford Press. Buffat, A. (2011). Pouvoir discrétionnaire et redevabilité de la bureaucratie de guichet : Les taxateurs d'une caisse de chômage comme acteurs de mise en oeuvre . Lausanne: Université de Lausanne. Davis, K. C. (1969). Discretionary justice: A preliminary inquiry. Baton Rouge, LA: Louisiana State University Press. Deci, E. L., & Ryan, R. M. (2004). Handbook of self-determination research. Rochester: Univ of Rochester Pr. DeLeon, P., & DeLeon, L. (2002). What ever happened to policy implementation? an alternative approach. Journal of Public Administration Research and Theory, 12(4), 467. DeVellis, R. F. (2003). Scale development: Theory and applications. Thousand Oaks: Sage. Evans, T. (2010). Professional discretion in welfare services: Beyond street-level bureaucracy. London: Ashgate. Ewalt, J. A. G., & Jennings, E. T. (2004). Administration, governance, and policy tools in welfare policy implementation. Public Administration Review, 64(4), 449-462. 21 Freidson, E. (2001). Professionalism: The third logic. Cambridge: Cambridge University Press. Green, F. (2008). Work effort and worker well-being in the age of affluence. In R. Burke, & C. L. Cooper (Eds.), Effects of working hours and work addiction: Strategies for dealing with them (pp. 115-136). Elsevier: London. Hackman, J. R., & Oldham, G. R. (1980). Work redesign. Reading, MA: Addison Wesley. Handler, J. F. (1990). Law and the search for community. Philadelphia: University of Pennsylvania Press. Hill, M., & Hupe, P. (2009). Implementing public policy (2nd ed.). Thousand Oaks: Sage. Hood, C. (1991). A public management for all seasons. Public Administration, 19(1), 3-19. Hooper, D., Coughlan, J., & Mullen, M. (2008). Structural equation modelling: Guidelines for determining model fit. Electronic Journal of Business Research Methods, 6(1), 53-60. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55. Hupe, P., & Hill, M. (2007). Street-level bureaucracy and public accountability. Public Administration, 85(2), 279-299. Judson, A. S. (1991). Changing behavior in organization: Minimizing resistance to change. Cambridge, MA: Basil Blackwell. Kimberly, J. R., De Pouvourville, G., & Thomas, A. D. A. (2009). The globalization of managerial innovation in health care. Cambridge: Cambridge University Press. Kline, R. B. (2010). Principles and practice of structural equation modeling. London: The Guilford Press. Lance, C. E., Dawson, B., Birkelbach, D., & Hoffman, B. J. (2010). Method effects, measurement error, and substantive conclusions. Organizational Research Methods, 13(3), 435-455. Lewin, K. (1936). Principles of topological psychology. New York: McGraw-Hill. Lipsky, M. (1980). Street-level bureaucracy. New York: Russell Sage Foundation. 22 May, D. R., Gilson, R. L., & Harter, L. M. (2004). The psychological conditions of meaningfulness, safety and availability and the engagement of the human spirit at work. Journal of Occupational and Organizational Psychology, 77(1), 11-37. May, P. J., & Winter, S. C. (2009). Politicians, managers, and street-level bureaucrats: Influences on policy implementation. Journal of Public Administration Research and Theory, 19(3), 453. Maynard-Moody, S., & Musheno, M. (2000). State agent or citizen agent: Two narratives of discretion. Journal of Public Administration Research and Theory, 10(2), 329. Maynard‐Moody, S., & Musheno, M. (2012). Social equities and inequities in practice: Street‐Level workers as agents and pragmatists. Public Administration Review, 72(s1), 16-23. Maynard-Moody, S., & Musheno, M. C. (2003). Cops, teachers, counselors: Stories from the front lines of public service. University of Michigan: University of Michigan Press. Maynard-Moody, S., & Portillo, S. (2010). Street-level bureaucracy theory. In R. Durant (Ed.), Oxford handbook of american bureaucracy (pp. 252-277). Oxford: Oxford University Press. McGregor, D. (1960). The human side of enterprise. New York: Wiley. Meier, K. J., & O'Toole, L. J. (2002). Public management and organizational performance: The effect of managerial quality. Journal of Policy Analysis and Management, 21(4), 629-643. Metselaar, E. E. (1997). Assessing the willingness to change: Construction and validation of the DINAMO. (Doctoral dissertation, Free University of Amsterdam). Meyers, M. K., & Vorsanger, S. (2003). Street-level bureaucrats and the implementation of public policy. In B. Guy Peters, & J. Pierre (Eds.), Handbook of public administration (pp. 245–254). London: Sage. Muthén, L., & Muthén, B. (1998-2010). Mplus user's guide (Sixth ed.). Los Angeles, CA: Muthén & Muthén. O'Toole, L. J. (2000). Research on policy implementation: Assessment and prospects. Journal of Public Administration Research and Theory, 10(2), 263-288. Palm, I., Leffers, F., Emons, T., Van Egmond, V., & Zeegers, S. (2008). De GGz ontwricht: Een praktijkonderzoek naar de gevolgen van het nieuwe zorgstelsel in de geestelijke gezondheidszorg. Den Haag: SP. Palumbo, D. J., Maynard-Moody, S., & Wright, P. (1984). Measuring degrees of successful implementation. Evaluation Review, 8(1), 45-74. 23 Podsakoff, P. M., & Organ, D. W. (1986). Self-reports in organizational research: Problems and prospects. Journal of Management, 12(4), 531-544. Polsky, A. J. (1993). The rise of the therapeutic state. Princeton: Princeton University Press. Power, M. (1997). The audit society: Rituals of verification. Oxford: Oxford University Press. Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior Research Methods, 36(4), 717-731. Prottas, J. M. (1979). People processing: The street-level bureaucrat in public service bureaucracies. Lexington, MA: Lexington Books. Riccucci, N. M. (2005). How management matters: Street-level bureaucrats and welfare reform. Georgetown: Georgetown University Press. Saetren, H. (2005). Facts and myths about research on public policy implementation: Out‐of‐Fashion, allegedly dead, but still very much alive and relevant. Policy Studies Journal, 33(4), 559-582. Sandfort, J. R. (2000). Moving beyond discretion and outcomes: Examining public management from the front lines of the welfare system. Journal of Public Administration Research and Theory, 10(4), 729-756. Schreiber, J. B., Nora, A., Stage, F. K., Barlow, E. A., & King, J. (2006). Reporting structural equation modeling and confirmatory factor analysis results: A review. The Journal of Educational Research, 99(6), 323-338. Simon, W. H. (1987). Ethical discretion in lawyering. Harvard Law Review, 101(6), 1083. Sowa, J. E., & Selden, S. C. (2003). Administrative discretion and active representation: An expansion of the theory of representative bureaucracy. Public Administration Review, 63(6), 700-710. Spence Laschinger, H. K., Finegan, J., & Shamian, J. (2001). The impact of workplace empowerment, organizational trust on staff nurses' work satisfaction and organizational commitment. Health Care Management Review, 26(3), 7. Tummers, L. G. (2011). Explaining the willingness of public professionals to implement new policies: A policy alienation framework. International Review of Administrative Sciences, 77(3), 555-581. Tummers, L. G. (2012). Policy alienation of public professionals: The construct and its measurement. Public Administration Review, 72(4), 516-525. 24 Tummers, L. G., Bekkers, V. J. J. M., & Steijn, A. J. (2009). Policy alienation of public professionals: Application in a new public management context. Public Management Review, 11(5), 685-706. Tummers, L. G., Steijn, A. J., & Bekkers, V. J. J. M. (2012). Explaining willingness of public professionals to implement public policies: Content, context, and personality characteristics. Public Administration, 90(3), 716-736. Van de Schoot, R., Lugtig, P., & Hox, J. (2012). A checklist for testing measurement invariance. European Journal of Developmental Psychology, 9(4), 37-41. Vinzant, J. C., Denhardt, J. V., & Crothers, L. (1998). Street-level leadership: Discretion and legitimacy in front-line public service. Washington: Georgetown University Press. Wagner III, J. A. (1994). Participation's effects on performance and satisfaction: A reconsideration of research evidence. Academy of Management Review, 19(2), 312-330. Winter, S. C. (2007). Implementation perspectives, status and reconsideration. In B. Guy Peters, & J. Pierre (Eds.), The handbook of public administration. concise paperback edition (pp. 131-141). New York: Sage. Wright, B. E., Moynihan, D. P., & Pandey, S. K. (2012). Pulling the levers: Transformational leadership, public service motivation, and mission valence. Public Administration Review, 72(2), 206-215. Zhao, X., Lynch, J. G., & Chen, Q. (2010). Reconsidering baron and kenny: Myths and truths about mediation analysis. Journal of Consumer Research, 37(2), 197-206. 25 Appendix 1: Measurement model This Appendix describes some additional reliability and validity checks on the measurement model. Several authors suggest reporting RMSEA, TLI and CFI statistics when describing model fit (Schreiber et al., 2006; Van de Schoot et al., 2012). The Root Mean Square Error Of Approximation (RMSEA) – a widely recommended fit index which tests the absolute fit of the model – was .048. This indicates good fit as Hu and Bentler (1999) suggest that values ≤ .06 indicate good fit (≤.08 average fit). The TuckerLewis Index (TLI) is a comparative fit index that compares the fit of the model with the baseline model. The TLI here was .98, which is considered excellent (≥ .90, better ≥ .95). The Comparative Fit Index is also a comparative fit index and was .98 in our final model showing good fit (≥ .90, better ≥ .95). Note that – based on the recommendations of Hooper et al. (2008) - we have not used correlated error terms. In the final model each item loaded significantly onto its appropriate latent variable. For instance, an item tapping discretion loaded onto the variable discretion. The values of the standardized factor loadings were all relatively high (min. .51, max. .91, average .75). This shows evidence of convergent validity: items that tap the same latent construct are related to each other (Kline, 2010). We should also discuss the possibility of common method variance. Self-reported data based on a single application of a questionnaire can result in inflated relationships between variables due to common method variance, i.e. variance that is due to the measurement method rather than the constructs themselves (Podsakoff & Organ, 1986). Although a recent study showed that “in contrast to conventional wisdom, common method effects do not appear to be so large as to pose a serious threat to organizational research” (Lance et al., 2010:450), we conducted a test to analyse whether common method bias was a major concern. We compared the three-factor structure (discretion, client meaningfulness and willingness to implement) with a one-factor model. The fit indices show that the one-factor model had a much poorer fit than the three factor model. The AIC was higher, and the RMSEA (.16), CFI (.58) and TLI (.54) indicated much poorer fit. Hence, common method variance does not seem to be a major problem in this study. 26 Appendix 2: Items used for the scales As indicated in the main text, we used templates to specify the policy. Templates allow the researcher to specify an item by replacing general phrases with more specific ones that better fit the research context. Template words are underlined. The templates are in this case: Policy: DRG-policy Clients: Patients Professionals: Healthcare professionals Organization: Institution Note: Item 4-5 (client meaningfulness) and Item 1-3 (discretion) are not used in the final model as they negatively influenced fit-indices in the Confirmatory Factor Analysis. Client meaningfulness 1. 2. 3. 4. 5. The policy is harmful for my clients privacy With the policy I can better solve the problems of my clients The policy is contributing to the welfare of my clients Because of the policy, I can help clients more efficiently than before I think that the policy is ultimately favourable for my clients Discretion 1. 2. 3. 4. 5. 6. I have freedom to decide how to use the policy While working with the policy, I can be in keeping with the client’s needs [not used] Working with the policy feels like a harness in which I cannot easily move [not used] When I work with the policy, I have to adhere to tight procedures While working with the policy, I cannot sufficiently tailor it to the needs of my clients While working with the policy, I can make my own judgments Willingness to implement 1. 2. 3. 4. I intend to try to convince employees of the benefits the policy will bring I intend to put effort into achieving the goals of the policy I intend to reduce resistance among employees regarding the policy I intend to make time to implement the policy 27