Introduction

Street-level bureaucrats (SLBs) (for example, welfare workers, police officers, educators) interact directly and on a regular basis with citizens, and exercise discretionary power when delivering public services (Lipsky, 1980; Maynard-Moody and Musheno, 2003). Such discretion at the frontline is requested to motivate civil servants to enforce policy rules (Tummers, 2011; Tummers and Bekkers, 2014; Thomann et al, 2018) and tailor their implementation decisions to local political contexts and individual circumstances (Lipsky, 1980; May and Winter, 2007). However, it may also lead to less desirable effects, such as gaps between the legislator’s intention and the way policy is delivered (Hupe and Buffat, 2014), or unequal treatment of citizens’ demands (Meyers et al, 1998; Pedersen et al, 2018; Thomann and Rapp, 2018). For instance, when implementing policy tools (for example, granting a disability benefit), frontline welfare workers may use their leeway to prioritise some citizens over others, and justify their discriminatory behaviour by arguing that some citizens (for example, those who are vulnerable, meritorious or worthy) deserve more help than others (Van Oorschot, 2000; Jilke and Tummers, 2018). Such ‘deservingness cues’ and behavioural decision biases are probably legitimised if they resonate with the personal preferences of SLBs (May and Winter, 2007; Dubois, 2010; Raaphorst and Van de Walle, 2018), their moral dispositions (Zacka, 2017), their professional norms (Evans and Harris, 2004), or if they reproduce dominant social stereotypes about different policy beneficiaries (for example, Harrits and Moller, 2014; Kallio and Kouvo, 2015; Einstein and Glick, 2017; Thomann and Rapp, 2018).

These factors also affect the way SLBs process the available information when assessing the policy beneficiaries. This is most likely to take place when they have to tackle highly complex and abundant information in a limited time frame (Brodkin, 2006; 2011; Keiser, 2009). SLBs then tend to develop their own filters to process information, based on personal values and experiences, ideology, adherence to agency goals, background, and so on (Wood and Vetlitz, 2007), thus mechanically focusing on specific pieces of information and neglecting the others. In some cases, such partial information processing may impede the consideration of relevant information and result in less effective decisions (Wood and Vetlitz, 2007). This eventually leads to a disjuncture between the targeted policy goals and the actual implementation practices (Hasenfeld, 2010).