Understanding street-level bureaucrats’ attitude towards clients: Towards a measurement instrument

Abstract

The attitude of street-level bureaucrats towards their clients has an impact on the decisions they take. Still, such attitudes have not received much scholarly attention, nor are they generally studied in much detail. This article uses Breckler's psychological multicomponent model of attitude to develop a scale to measure street-level bureaucrats' general attitude towards their clients. By means of a test study (N=218) and a replication study (N = 879), the article shows that street-level bureaucrats' attitude towards clients consists of four different components: a cognitive attitude component, a positive affective attitude component, a negative affective attitude component and a behavioural attitude component. It also establishes a conceptual and empirical distinction from related attitudes, such as prosocial motivation, work engagement, bureaucrats’ rule-following identities and self-efficacy, and suggests avenues for application and further validation among different groups of street-level bureaucrats. This instrument opens up opportunities for theory testing and causality testing that surpasses case-specific considerations.

Keywords

Bureaucratic attitudes clients multicomponent model scale development street-level bureaucracy

Introduction

The Weberian model portrays bureaucratic encounters as characterized by impersonality and affective neutrality to prevent bureaucrats’ personal attitudes from permeating these encounters (Katz and Danet, 1973a; Merton, 1940). In reality, however, it is impossible to separate street-level bureaucrats’ attitudinal dispositions and their interactions with clients (Baviskar and Winter, 2017; Lipsky, 1980; Oberfield, 2014). Bureaucrats rely on their attitudinal dispositions to process clients to cope with complex cases, ambiguous rules and resource strain (Lipsky, 1980). By means of these attitudinal dispositions, psychological simplifications of clients emerge that ‘redefine […] the nature of the clientele to be served’ (Lipsky, 1980: 141). As a result, street-level bureaucrats’ attitude towards clients introduces bias to the bureaucratic encounter and its outcomes (e.g. Baviskar and Winter, 2017; Kroeger, 1975; Oberfield, 2014; Prottas, 1979; Stone, 1981), making this attitude key to understanding bureaucratic decision-making, street-level performance and official-client relations.

Attitudes are among the most studied phenomena in survey research in the field of public administration (PA) (Lee et al., 2012) and current developments in this field, such as the emergence of the behavioural PA movement (e.g. Grimmelikhuijsen et al., 2017), triggered a revived interest in this phenomenon. Despite renewed attention for the study of attitudes, the concept of street-level bureaucrats’ attitude towards clients is still plagued by conceptual unclarity and unsound measurement efforts (e.g. see Lee et al., 2012). Poor measurement bears a risk of generating faulty data (DeVellis, 2003) and, subsequently, reaching erroneous conclusions (Lee et al., 2012). Despite multiple calls to improve measurement quality in our field (e.g. Perry, 2012), scale development is still scarce in PA (Grimmelikhuijsen et al., 2017; for examples, see Perry, 1996; Tummers, 2012; Tummers and Musheno, 2015).

The aim of this article is to develop a measurement instrument for street-level bureaucrats’ attitude towards clients, based on material collected among one specific type of bureaucrat. Grounded in psychological attitude theory, such a scale will allow for a multidimensional measurement of attitudes towards clients, at a sufficiently abstract level as to exclude case-specific considerations. The instrument also allows to conceptually distinguish bureaucrats' attitude towards clients from other related attitudes. More in particular, we consider street-level bureaucrats' prosocial motivation, work engagement, rule-following identities and self-efficacy – all attitudes that have been studied in a public work context but that are theoretically different. The use of the psychological multicomponent model of attitudes (see Breckler, 1984) furthermore allows to move beyond current practice to rely on one-dimensional scales when measuring bureaucrats' attitudes towards clients.

Developing a measurement instrument for street-level bureaucrats' general attitude towards clients has merit for multiple reasons. First, it opens up opportunities for the quantification of bureaucrats’ general assessment of clients, aiding scholars to draw inferences that surpass study and context specificities (see Grimmelikhuijsen et al., 2017), which is especially challenging for street-level attitudes (Winter, 2002). Second, a validated measurement instrument of street-level bureaucrats’ client-attitude with sound theoretical foundations enables theory testing and the study of causality in the bureaucratic encounter (see Pandey and Scott, 2002; Perry, 2012).

The results of a series of statistical tests among a specific sample of street-level bureaucrats show that bureaucrats' attitude towards clients consist of four distinct components: a cognitive attitude component, a positive affective attitude component, a negative affective attitude component and a behavioural attitude component. Furthermore, it establishes that this attitude is conceptually and empirically distinct from related attitudes such as prosocial motivation, work engagement or rule-following identities.

In the next section, we position street-level bureaucrats’ general attitude towards clients in the bureaucratic encounter and within the broader literature. We then introduce the concept of attitude and its components as used in social psychological scholarship. This is followed by a description of methods and the sample used to develop and test the new scale. First, a scale is developed following an elaborate item generation phase. Second, the preliminary measurement instrument is tested in a survey. The new scale’s psychometric properties are assessed using exploratory and confirmatory factor analysis (CFA), followed by an assessment of its construct validity. Then, the resulting measurement instrument is tested and cross-validated in a replication study. We conclude with a discussion of the research findings and the applicability of the new scale for the study of bureaucratic encounters.

Why bureaucrats’ attitude towards clients matters

During their interactions with clients, street-level bureaucrats exercise a vast amount of discretion (Hupe and Hill, 2016). Decisions made in this discretionary space can greatly impact citizens’ lives (e.g. Lipsky, 1980; Maynard-Moody and Musheno, 2003). Consequently, how bureaucrats exercise their discretion has always been carefully scrutinized (e.g. Goodsell, 1976; Prottas, 1979). Bureaucrats’ use of discretionary space is based on their categorization of clients (Lipsky, 1980; Jilke and Tummers, 2018), such as based on clients’ ‘worthiness’ or ‘deservingness’ (Maynard-Moody and Musheno, 2003), whether they are deemed ‘desirable’ or ‘undesirable’ (Hasenfeld and Steinmetz, 1981) or their perceived need (Kroeger, 1975). These categorizations subsequently inform the courses of action bureaucrats take (e.g. Maynard-Moody and Musheno, 2003). For instance, Lineberry’s (1977) ‘underclass hypothesis’ posits that ‘clients who are perceived by bureaucrats as less worthy or desirable (e.g. the minorities and poor) are discriminated against in delivery manner and output’ (Goodsell, 1981: 763). Bureaucrats can, however, differ in their classifications and reactions thereto. For example, clients’ perceived need can trigger bureaucrats to put in extra effort to benefit a client’s case (Kroeger, 1975; Maynard-Moody and Musheno, 2003), but can also trigger them to ‘cream’ clients, meaning that they are inclined to favour clients whose cases are most likely to be brought to a successful outcome anyway instead of those who are in most need (Jilke and Tummers, 2018; Lipsky, 1980).

How bureaucrats react to cues like perceived need depends on how they assess and interpret information on clients (e.g. Oberfield, 2014; Raaphorst and Van de Walle, 2017). These interpretations are informed by their general attitude towards clients (e.g. Eagly and Chaiken, 1993). This general attitude is an abstract-level prototype that guides bureaucrats’ assessments and categorizations of clients, as well as their subsequent courses of action (Baviskar and Winter, 2017; Oberfield, 2014). Bureaucrats do not always have very positive views of their clients (Van de Walle and Lahat, 2016), and early studies on bureaucratic encounters have already found that bureaucrats with different general orientations towards clients react differently to the cues they receive from them. For instance, Kroeger (1975: 191) showed that client-oriented bureaucrats use greater discretion to benefit clients and were more prone to manipulation by clients. Stone (1981: 45) argues that bureaucrats with ‘a condemnatory moralistic view of clients’, that is, a negative attitude towards clients, are more inclined to respond punitively to clients. And Billingsley (1964: 406) found that client-oriented bureaucrats are more willing to disregard policy in order to help clients. Despite this early evidence, scholars have mainly studied how bureaucrats categorize clients during decisions, but have looked less into the general attitudes bureaucrats hold towards their clients. Furthermore, current work does not generally distinguish between attitudes towards clients and other related attitudes of bureaucrats. This article attempts to unravel the general attitude of bureaucrats towards clients into its different dimensions and give substantive body to this concept in a street-level context.

What are attitudes? The multicomponent attitude model

To elaborate a scale for measuring street-level bureaucrats' attitude towards clients, we rely on insights from social psychology which treats attitudes as a summary evaluation. Social psychology has a long tradition of studying attitudes. In this field, the attitude concept is commonly understood as an object’s evaluation on a continuum (Ajzen, 2001; Eagly and Chaiken, 1993; Fabrigar et al., 2005; Katz, 1960;). This evaluation is often believed to incorporate the cognitive, affective and behavioural pieces of information that are called into one’s memory upon confrontation with an attitude object (Haddock and Huskinson, 2004; Maio and Haddock, 2009; Olson and Kendrick, 2008). Incorporating these pieces of information is crucial to come to a full understanding of attitudes (Fabrigar et al., 2005; Fazio and Olson, 2003).

A definition of street-level bureaucrats’ attitude towards clients should thus contain three elements: bureaucrats’ general evaluation of clients, an index on which to place this evaluation (e.g. a bipolar continuum) and the pieces of information that constitute this evaluation. To this end, we fuse the attitude definitions by Maio and Haddock (2009) and Petty et al. (1997) and define street-level bureaucrats’ attitude towards clients as: their summary evaluation of clients along a dimension ranging from positive to negative that is based on the bureaucrats' cognitive, affective and behavioural information on clients.

Such a detailed definition of street-level bureaucrats' attitudes is rare in studies of the bureaucratic encounter, where the attitude concept is frequently characterized by conceptual ambiguity: scholars rarely make explicit what they understand by ‘attitude’ (e.g. Baviskar and Winter, 2017; Liou and Cruise, 1994; Scherer and Scherer, 1980; Soydan, 1995; Stone, 1981; Stone and Feldbaum, 1976) or they conceptualize it by referring to related concepts, such as ‘values’ (e.g. Winter, 2002). They also tend to use different concepts to measure similar phenomena (i.e. views of the other party in the bureaucratic encounter), suggesting conceptual overlap and the interchangeable use of concepts. For instance, authors refer to ‘orientations’ (Blau, 1960), ‘assessments’ (Nelson, 1981), ‘perceptions’ (Berman, 1997; Borgatta et al., 1960), ‘views’ (Oberfield, 2012) and ‘attitudes’ (Baviskar and Winter, 2017; Oberfield, 2012; Stone, 1981; Wilson, 1989). Some exceptions among bureaucracy scholars are Wilson (1989) and Oberfield (2012), who both refer to ‘evaluation’ in their attitude conceptualizations, which closely links to attitude conceptualizations in the field of psychology. Furthermore, this attitude has generally been considered to be a one-dimensional construct, mostly tapping into the cognitive dimension of bureaucrats’ assessment of clients (e.g. see Borgatta et al., 1960; Scherer and Scherer, 1980; Winter, 2002).

Pieces of information belonging to the same type are commonly referred to as attitude components (Fazio and Olson, 2003; Maio and Haddock, 2009). Smith (1947) was the first to formally distinguish among three attitude aspects in his use of the attitude concept (Breckler, 1984: 1192). The multicomponent model of attitude took flight in the 1960s and has since been recurrent in attitude theory (Breckler, 1984: 1192; e.g. see Eagly and Chaiken, 1993; Haddock and Huskinson, 2004; Maio and Haddock, 2009; Olson and Kendrick, 2008; Rosenberg and Hovland, 1960). In this study, we use the multicomponent model as depicted by Breckler (1984: 1196). In this model, there are three attitude components: the cognitive, the affective and the behavioural one. The cognitive attitude component refers to the beliefs and attributes that bureaucrats associate with clients (see Fabrigar et al., 2005; Fazio and Olson, 2003). For instance, juvenile probation officers might perceive their clients as disrespecting order and authority. That belief links clients to the attribute ‘disrespect’. Subsequently, this piece of information contributes to a negative assessment of clients.

The affective attitude component refers to bureaucrats’ emotional responses evoked in the bureaucrat by confrontations with the attitude object ‘clients’ (see Breckler, 1984; Haddock and Huskinson, 2004). This confrontation thus acts as a stimulus that brings the feelings and emotions bureaucrats associate with clients to the forefront (see Maio and Haddock, 2009). This affect subsequently informs their general evaluation of clients. An example is provided by Blau (1960) who found that new employees of a public welfare agency experience a reality shock when they realize that clients are dishonest. This ‘shock’ can make the bureaucrat feel threatened. These feelings subsequently pressure the bureaucrat to change his attitude towards clients.

The behavioural attitude component refers to past behavioural responses to clients (see Haddock and Huskinson, 2004; Maio and Haddock, 2009) and is best explained by Bem’s (1972) self-perception theory. Self-perception theory posits that individuals infer their attitudes from their past behaviours, as long as they were not impelled to undertake those behaviours (Bem, 1972). Following this theory, bureaucrats are argued to (subconsciously) reflect on their chosen past behaviours towards clients – which exclude behaviours prescribed by bureaucratic rules and regulations – and infer their attitude towards clients from these observations (Bem, 1972). If these behaviours have been mostly negative, they can inform a bureaucrat of his negative client-attitude (see Eagly and Chaiken, 1993). Thus, a bureaucrat who observes that he is always rude to clients might acquire that he must have a negative attitude towards them from this behavioural piece of information.

For conceptual clarity, it is necessary to distinguish between behaviour as an attitude component and as a consequence of attitude. In PA research, attitudes have mostly been explored as antecedents of behaviours relevant to bureaucratic decision-making (Oberfield, 2012), rather than as an attitude component. Behaviour can both function as an attitudinal indicator and as an attitudinal consequence (Eagly and Chaiken, 1993; Fazio and Olson, 2003). Attitudes itself are unobservable mental constructs (Fazio and Olson, 2003). They become observable by means of the cognitive, affective and behavioural pieces of information linked to the attitude object in the mind of the individual (Himmelfarb, 1993). All information types can subsequently work to elicit behavioural responses (Eagly and Chaiken, 1993). Even though behaviour can both serve as an attitudinal indicator and consequence thereof, behaviours can only be an attitudinal indicator if they ‘relate to the dimension of favourability-unfavourability towards the attitude object’ (Himmelfarb, 1993: 63) and thus harbour evaluative properties.

Methods

Methods of scale development

The efforts to develop a measurement instrument for bureaucrats’ general attitude towards clients were based on the scale development guidelines of DeVellis (2003). We first generated a pool of items for each of the three components. The item-generation procedure consisted of multiple steps: a literature study; a series of qualitative interviews; multiple cognitive interviews with street-level bureaucrats and a brainstorm session with academic experts. Subsequently, this item pool was subjected to a test among a sample of bureaucrats to establish whether the three attitude components existed using exploratory factor analysis (EFA), followed by CFA. This resulted in the inclusion or exclusion of individual items. Finally, the construct validity of the therewith obtained scale was further assessed: after a test of convergent validity, discriminant validity was tested by exploring the relations within the new attitude construct (bureaucrats' general attitude towards clients) and between this new construct and three related constructs (prosocial motivation, work engagement and bureaucrats’ rule-following identities). To further validate the resulting measurement instrument, we conducted a replication study in which discriminant validity was tested using bureaucrats’ rule-following identities and self-efficacy.

Case selection and test sample

For our scale development purposes, we focused on one single type of bureaucrat. Tax bureaucrats are a classic example of street-level bureaucrats (see Finer, 1931). They work in a classic bureaucratic framework yet have ample discretionary space that enables attitudes to affect their work (Aberbach and Christensen, 2007). The main advantage of using the tax administration as a case is that all citizens are its potential clients. This differentiates it from many other street-level bureaucracies whose bureaucrats tend to deal with very specific types of clients, with very specific characteristics, such as poor people or criminals (e.g. Dubois, 2010). The latter could bias the development of a measurement instrument. Another advantage is the rather similar design and tasks of tax administrations across (capitalist) countries (see Campbell, 2005), which facilitates future cross-case comparison.

We develop and test the multicomponent model in the Dutch tax administration, among bureaucrats who have face-to-face interactions with clients. Their task is to audit tax filings of entrepreneurs in the small and medium-sized enterprises (SME) segment (1–50 employees). They typically meet with small entrepreneurs to discuss tax declarations, ask for clarifications and make decisions about deductions or fines, often in confined settings. They experience a range of client actions and reactions ranging from joy, over fear and suspicion, to threats and violence.

The final item pool was tested in a cross-sectional survey of Dutch tax bureaucrats. We selected one of the five Dutch tax regions. All bureaucrats in this tax region who fit our research population were invited to participate (N = 433). Data were collected using an email survey with two reminders, sent to the bureaucrats' work address in the summer of 2015. Seven-point Likert scales ranging from ‘never’ to ‘always’ were used. The response rate was 67.4% (N = 292). We included only those respondents who confirmed that they actually had contact with citizens and had valid replies on the attitude items, resulting in a sample of 218 valid observations. The respondents’ mean age was 52.4, which is representative of the Dutch tax administration’s workforce (Dutch Tax Administration, 2012). Of the respondents, 19.7% were female and 75.2% were male.

Item generation

Item generation for the scale was based on the literature on bureaucratic encounters and in-depth interviews with tax bureaucrats. This literature offers rich descriptions of official-client interactions (e.g. Dubois, 2010; Maynard-Moody and Musheno, 2003; Prottas, 1979). In these studies, scholars, for example, describe critical incidents between bureaucrats and clients (Savaya et al., 2011) or show how bureaucrats cope with clients (Lipsky, 1980). These accounts hold crucial pieces of cognitive, affective and behavioural information relevant to the bureaucratic encounter. For our literature search, we used Google Scholar so both relevant articles and books could be selected. To collect literature, we searched for (combinations of) the following terms: ‘bureaucratic encounter’, ‘civil servant’, ‘official’, ‘public official’, ‘public servant’, ‘front-line official’, ‘street-level bureaucrat’, ‘administrator’, ‘attitude’, ‘orientation’, ‘perception’, ‘evaluation’, ‘interaction’, ‘citizen’, ‘citizen-client’, ‘client’ and ‘the public’. A study was deemed relevant if it described cognitive, affective and/or behavioural elements in the encounter between officials and the public. Studies mostly stemmed from the field of PA, but also studies from the field of psychology were included since the multicomponent model stems from this discipline. Whenever a publication was selected, we checked the publications citing that piece for relevance. Search results were checked until saturation occurred.

To increase the validity of the item generation, we validated and complemented the pieces of information from the literature study with those from 15 interviews with tax bureaucrats who work in the SME segment. In these interviews, bureaucrats were first administered an open-ended measure of the three components – as illustrated by Maio and Haddock (2009): they were asked to name all characteristics, feelings and chosen behaviours that they associate with clients in general. Second, respondents were invited to tell the story of one memorable negative and positive experience with clients. Follow-up questions targeted the clients’ characteristics in those interactions, emotions that they evoked in the bureaucrat, and the courses of action that the respondent took. The interviews were transcribed afterwards.

Pieces of information were coded on whether they stemmed from literature from the PA field or psychology (literature study only); which party in the bureaucratic encounter was evaluated (i.e. the bureaucrat or the client); and the component a piece of information represented. Items were selected based on their recurrence in the literature and the interviews. Each resulting preliminary item pool was refined by means of feedback from three cognitive interviews with street-level tax bureaucrats and a brainstorm session with a group of academics who specialize in survey research and/or street-level bureaucracy. The cognitive interviews inquired about the bureaucrats’ understanding of the items and their applicability to their work context.

These efforts resulted in three initial item pools that contained 15 items for each attitude component (see Appendix 1). Since the aim of this study is to develop a measurement instrument with potential applicability across different types of street-level bureaucrats, all items were constructed as templates. This means that the attitude object the items refer to can be adapted from 'clients' to a term befitting the type of street-level bureaucrat being surveyed (see Tummers, 2012). In the current case, using templates meant that the term ‘clients’ was replaced by ‘taxpayers’. The pieces of information central to each component are described below.

The cognitive attitude component

Central pieces of information to the cognitive component are clients’ honesty and cooperativeness, how clients approach the street-level bureaucrat, and clients’ knowledgeability. Clients’ honesty and cooperation are critical for successful bureaucratic encounters. To decide on the allocation of government benefits and sanctions, bureaucrats often depend on input from clients (Lipsky, 1980); for instance, determining whether clients are eligible for welfare requires that clients honestly report their income. Such (monetary) incentives can cause clients to seek to influence the bureaucrat to alter the outcomes of the bureaucratic encounter to their advantage, for example, by offering bribes or emphasizing their destitution (Hasenfeld and Steinmetz, 1981; Katz and Danet, 1973b). Such tactics can force bureaucrats to put more effort into these cases and are hence more burdensome for them (see Wilson, 1989).

The second recurring element is the way clients approach bureaucrats. Many bureaucratic encounters are non-voluntary (Lipsky, 1980). Citizens do not want to get arrested by the police. They interact with social welfare bureaus because they depend on their services for their livelihood, and they pay taxes because the government obliges them to do so (see Lipsky, 1980). This non-voluntary nature can cause clients to approach bureaucrats with resistance and hostility (e.g. Dubois, 2010; Savaya et al., 2011; Worden, 1989). On the other hand, accounts of bureaucratic encounters also link clients to, for instance, gratitude for the benefits that they receive or the efforts that bureaucrats invest in their cases (Blau, 1960; Lipsky, 1980), and being sympathetic (Finer, 1931), or mild-mannered (Borgatta et al., 1960). The latter types of traits are frequently recurring in descriptions of how clients should behave (e.g. Finer, 1931).

The last piece of information we identified encompasses clients’ knowledgeability. Knowledge is often a requirement for successful bureaucratic encounters; clients applying for welfare must know where to apply, which documents to bring, and how to fill in certain forms (see Dubois, 2010; Gordon, 1975; Lipsky, 1980). Insufficient knowledge can cause clients to give improbable explanations for their situation or make inappropriate appeals to bureaucrats (Katz and Danet, 1973b). Clients can also use their knowledgeability to manipulate the bureaucratic encounter to their advantage (Hasenfeld and Steinmetz, 1981; Prottas, 1979: 40).

The affective attitude component

Because affect has been the focus of much prior research, we used the well-established Positive and Negative Affect Schedule (PANAS) scales by Watson et al. (1988) as the foundation for this attitude component. These scales consist of ‘two 10-item mood scales that comprise the PANAS’ (Watson et al., 1988: 1063). These scales, however, do not encompass the object inducing those moods, whereas we focus on moods specifically evoked by clients. To ensure the content validity of the affective attitude component, we revised the PANAS scales in three steps, based on the literature study, substantive interviews and cognitive interviews.

First, we identified PANAS items that were not suitable for the street-level work context. To identify these poorly fitting items, we administered the Dutch translation of the PANAS scales (Peeters et al., 1996) to the respondents of the cognitive interviews. Items had a poor fit because their meaning was ambiguous to them (e.g. ‘proud’), they did not fit their work context (e.g. ‘guilty’), they showed content overlap (e.g. ‘scared’ and ‘afraid’) or they were not interpreted as affect (e.g. ‘strong’).

Second, to decide which items to discard, we combined these insights from the cognitive interviews with insights from previous research on the PANAS scales: Crawford and Henry (2004) found covariation among the PANAS items, suggesting item redundancy in the PANAS scales (Thompson, 2007). To not jeopardize the validity of the PANAS foundation of the affective attitude component, we ensured that each covarying item group was represented in the final PANAS item selection. This resulted in the selection of the PA items of alert, active, determined and inspired and the NA items of upset, afraid, irritable, ashamed and nervous.

Third, having refined the PANAS scales for our research purposes, we then complemented the remaining PANAS items with observations from the literature study and the substantive interviews. One piece of affective information complementing the PANAS items is a feeling of threat (see Blau, 1960; Savaya et al., 2011). This threat can be either physical or psychological. A feeling of physical threat may result from client aggression; hostility; anger and actual, attempted, or threatened physical violence by clients (Savaya et al., 2011: 65). Psychological threats result from stimuli that affect bureaucrats’ mental well-being, such as feeling inexperienced or unable (Blau, 1960; Lipsky 1980); guilt or embarrassment (Prottas, 1979: 107); or frustrated, unappreciated or uncertain (Savaya et al., 2011). As a form of protection against these threats, bureaucrats might emotionally distance themselves from their clientele and hence become emotionally detached from them (Blau, 1960; Dubois, 2010).

Another piece of affective information is a desire to help that characterizes many street-level bureaucrats upon entering the public sector (Blau, 1960). Bureaucrats have a tendency to show concern for their clients (Stone and Feldbaum, 1976). Consequently, client-encounters can ‘provide workers with their greatest sense of accomplishment’ (Maynard-Moody and Musheno, 2003: 21).

The behavioural attitude component

At the core of the behavioural attitude component are the past behaviours towards clients that street-level bureaucrats engaged in by choice (see Bem, 1972). Pieces of information central to this component are the ‘fair’ treatment of clients, the way bureaucrats approach clients and bureaucrats’ helpfulness.

Despite its recurrence (e.g. Almond and Verba, 1973; Lipsky, 1980), an unequivocal understanding of what is meant by ‘fair treatment’ is yet to emerge. Bureaucracies are traditionally designed to treat clients equally (Gastelaars, 2006); however, many emphasize the need for a personal approach and responsive services (e.g. Almond and Verba, 1973; Brenninkmeijer, 2013). Lipsky (1980: 22) even refers to ‘the apparent unfairness of treating people alike’. Perceived unfairness can motivate bureaucrats to manipulate rules for clients (e.g. Maynard-Moody and Musheno, 2003).

The second piece of behavioural information concerns bureaucrats’ approach to clients. Here, accounts of how bureaucrats should behave are prominent (e.g. Brenninkmeijer, 2013; Gastelaars, 2006). Scholars argue that bureaucrats should be courteous (Nelson, 1981), polite (Finer, 1931), civil (Katz and Danet, 1973a) and respectful (Brenninkmeijer, 2013). Scholars also regularly criticize bureaucrats’ behaviour, accusing them of the inhumane, abusive, unreflective and apathetic treatment of clients (Bartels, 2013: 473; Goodsell, 1981: 763).

The third piece of behavioural information relates to bureaucrats’ helpfulness (e.g. Blau, 1960; Finer, 1931; Wilson, 1989). Many behaviours can be categorized as such, like being accessible and responsive; disseminating information; showing concern and empathy (King and Stivers, 1998: 66); being deliberative (Bartels, 2013); listening to clients attentively and considerately (Almond and Verba, 1973: 43) and ‘making application easy, volunteering unsolicited information, providing prompt responses’ (Hasenfeld and Steinmetz, 1981: 88). On the opposite end of this spectrum, one can find behaviours that obstruct clients’ cases, such as ‘making application difficult, withholding information, evading questions, and applying stigmatizing labels’ (Hasenfeld and Steinmetz, 1981: 88).

Testing and validating the multicomponent model

We now test the multicomponent model by means of EFA and CFA and assess the construct validity of this measure.

Analysing dimensionality: EFA

Principal axis factoring with oblique rotation was used as the extraction method (direct oblimin). Conditions for arriving at a stable, reliable factor solution were present. Because all items were designed to measure one specific component, groups of items were expected to cluster together (see DeVellis, 2003). Upon inspecting the correlation matrix, we omitted six items (see Appendix 1) that had no or almost no correlations above the .3 threshold (Field, 2013).

We used the scree test and the theoretical interpretability of factors to determine the number of factors to retain. This resulted in a four-factor solution that diverged from the hypothesized three-factor solution by its distinction between the positive and negative affective attitude items. This finding is consistent with prior research: the PANAS scales in which the affective component was grounded were found to represent two orthogonal (i.e. distinctive) dimensions of affect (Watson et al., 1988), suggesting that affective items are unipolar in nature. This means that, if affective items are placed on a measurement continuum, its endpoints do not pertain to each other’s opposite – as they would for a bipolar item – but to ‘different degrees of the presence of the same attribute’ (Schwarz, 2008: 43). Thus, if we placed ‘clients make me feel active’ on a continuum, the endpoints would represent ‘very active’ and ‘not active at all’ (unipolar) instead of ‘very active’ and ‘passive’ (bipolar).

To determine the most optimal, parsimonious four-factor solution, we deleted 22 additional items (see Appendix 1) on multiple grounds: they had factor loadings under .4 (Field, 2013), had cross-loadings greater than .3, had low communalities (close to .3 or under) and/or were judged to lack clarity from a theoretical point of view. As a result, we obtained a four-factor model that explained 59.5% of the variance in the remaining 17 items. The Cronbach’s alphas were respectable for all dimensions (see DeVellis, 2003), ranging from .70 to .83. No items could be removed to increase the reliability of the subscales, thus showing no sign of item redundancy. Table 1 shows the EFA results.

Table 1.

Summary of EFA results for the final item pool.

	Rotated factor loadings
	Test study (N = 218)				Replication study (N = 879)
Item	F1	F2	F3	F4	F1	F2	F3	F4
Cognitive attitude component
Taxpayers are manipulative	–.720				.628
Taxpayers are hostile	–.547				.644
Taxpayers are unpredictable	–.575				.570
Taxpayers are stubborn	–.508				.487
Taxpayers are dishonest	–.703				.593
Positive affective attitude component
Taxpayers make me feel alert		.681				.645
Taxpayers make me feel inspired		.674				.682
Taxpayers make me feel determined		.559				.687
Taxpayers make me feel active		.809				.808
Negative affective attitude component
Taxpayers make me feel upset			.585				.530
Taxpayers make me feel afraid			.514				.570
Taxpayers make me feel nervous			.765				.679
Taxpayers make me feel insecure			.705				.646
Taxpayers make me feel uncomfortable			.814				.830
Behavioural attitude component
I explain things to taxpayers				.662				.758
I make taxpayers feel at ease				.710				.739
I help taxpayers				.610				.739
Eigenvalues	4.35	2.66	1.67	1.43	3.65	2.90	1.78	1.58
Percentage of variance	25.59	15.64	9.83	8.44	21.46	17.06	10.49	9.27
α	.77	.77	.83	.70	.73	.80	.79	.79
M	2.86	3.98	1.78	5.35	2.79	4.17	1.81	5.06
SD	.61	.92	.48	.89	.59	.93	.48	1.01

EFA: exploratory factor analysis; SD: standard deviation.

The correlations between the four attitude subscales are listed in Table 2. They show that the components are sufficiently distinctive (e.g. no multicollinearity). The positive and negative affective components are uncorrelated to each other, affirming that the affective items are indeed unipolar. A similar distinction is found between the cognitive and positive affective items.

Table 2.

Intercorrelations between the attitude subscales.

		Cog.	PosAff.	NegAff.	Behav.
Test study (N = 218)
Cognitive att. comp.	Pearson correlation	1
Cognitive att. comp.	Sig. (2-tailed)
Positive affective att. comp.	Pearson correlation	.085	1
Positive affective att. comp.	Sig. (2-tailed)	.212
Negative affective att. comp.	Pearson correlation	.408	–.011	1
Negative affective att. comp.	Sig. (2-tailed)	.000	.875
Behavioural att. comp.	Pearson correlation	–.216	.194	–.306	1
Behavioural att. comp.	Sig. (2-tailed)	.001	.004	.000
Replication study (N = 879)
Cognitive att. comp.	Pearson correlation	1
Cognitive att. comp.	Sig. (2-tailed)
Positive affective att. comp.	Pearson correlation	.107	1
Positive affective att. comp.	Sig. (2-tailed)	.002
Negative affective att. comp.	Pearson correlation	.324	–.068	1
Negative affective att. comp.	Sig. (2-tailed)	.000	.043
Behavioural att. comp.	Pearson correlation	–.112	.229	–.165	1
Behavioural att. comp.	Sig. (2-tailed)	.001	.000	.000

Confirmatory factor analysis

To test the hypothesized structure that resulted from the EFA, we conducted a CFA using AMOS version 24. In the CFA, all the items from one hypothesized component are constrained to freely load on that component but to have zero loadings on the other attitude components (Byrne, 2010).The fit indices of the multicomponent model are: χ²= 217.18 (df = 113, p = .000), the ratio of the minimum discrepancy to degrees of freedom (CMIN/DF) = 1.92, the goodness-of-fit index (GFI) = .90, the comparative fit index (CFI) = .91, the root mean square error of approximation (RMSEA) = .065. The χ² is significant, which means there is no perfect fit between the hypothesized model and the data. However, the χ² assumption of perfect fit has limitations that are offset by other goodness-of-fit indices, such as the CMIN/DF (Byrne, 2010). CMIN/DF values between 1 and 3 represent a good model fit (Vermeeren et al., 2011). The GFI and CFI indicate a good fit if their value is close to 1 (Byrne, 2010). A generally used cut-off point in the social sciences is .9 (Vermeeren et al., 2011). These fit parameters thus support the hypothesized fit of the multicomponent model. The RMSEA value is indicative of reasonable model fit (Byrne, 2010).

Construct validity tests

To further validate the new scale, we assessed the multicomponent model’s construct validity by means of multiple tests that evaluate its convergent validity, discriminant validity and reliability.

Convergent validity

To assess the convergent validity of the multicomponent model, we first examined how well the items measure the component that they are hypothesized to represent. Significant construct loadings are indicative of convergent validity (Vermeeren et al., 2014). The construct loadings range from .565 to .795 (Table 3) and are all significant at the p = 0.001 level, supporting the claim of convergent validity.

Table 3.

Means, standard deviations (SD), standardized estimates, R².

	Test study (N = 218)				Replication study (N = 879)
	M	SD	Stand. regression weights (S.E.)	R²	M	SD	Stand. regression weights (S.E.)	R²
Cognitive component
… are manipulative	2.93	.88	.645 (.060)	.416	2.88	.88	.616 (.031)	.379
… are hostile	2.23	.79	.594 (.055)	.353	2.14	.71	.644 (.025)	.412
… are unpredictable	2.92	.90	.620 (.062)	.384	2.81	.87	.504 (.032)	.254
… are stubborn	3.15	.84	.584 (.059)	.341	2.96	.88	.628 (.031)	.394
… are dishonest	3.08	.81	.705 (.054)	.497	3.16	.89	.589 (.032)	.347
Positive affective component
… make me feel alert	4.19	1.16	.690 (.078)	.476	4.38	1.15	.684 (.038)	.468
… make me feel inspired	3.62	1.18	.668 (.080)	.447	3.78	1.24	.504 (.041)	.254
… make me feel determined	3.97	1.16	.565 (.081)	.319	4.21	1.13	.647 (.037)	.419
… make me feel active	4.16	1.26	.795 (.083)	.632	4.31	1.19	.798 (.038)	.637
Negative affective component
… make me feel upset	1.59	.55	.681 (.035)	.464	1.62	.64	.565 (.022)	.319
… make me feel afraid	1.55	.58	.655 (.037)	.430	1.50	.57	.613 (.347)	.376
… make me feel nervous	1.90	.67	.761 (.042)	.579	1.89	.65	.750 (.021)	.563
… make me feel insecure	1.86	.62	.690 (.040)	.477	1.97	.68	.659 (.023)	.434
… make me feel uncomfortable	2.00	.66	.741 (.041)	.548	2.09	.68	.713 (.022)	.508
Behavioural component
I explain things to taxpayers	5.97	.99	.743 (.073)	.551	5.62	1.07	.766 (.036)	.587
I make taxpayers feel at ease	5.24	1.14	.690 (.084)	.476	5.05	1.23	.749 (.041)	.561
I help taxpayers	4.86	1.22	.571 (.090)	.326	4.50	1.30	.723 (.043)	.523

As a second assessment of convergent validity, we calculated the average variance extracted (AVE) for each component. The AVE is calculated by averaging the squared correlation estimates of the individual items that represent the same component (Farrell and Rudd, 2009). AVEs over .5 indicate that the amount of variance captured by a component is greater than the amount of variance due to measurement error (Fornell and Larcker, 1981: 46). The AVEs are .39 (cognitive component), .47 (positive affective component), .50 (negative affective component) and .45 (behavioural component), somewhat under the .5 threshold. The AVEs are listed in Appendix 2.

Discriminant validity

Discriminant validity is achieved when items representing the same component are more strongly correlated with one another than with items measuring either the other components or different constructs (Farrell and Rudd, 2009). This assumption was first tested by comparing the χ² of the unconstrained measurement model with the χ² of the constrained model, in which the covariance between the four components is constrained to equal 1. The constrained model equals a one-factor representation of the multi-component model. If the unconstrained model has a significantly lower χ² than the constrained model, discriminant validity is established (Bagozzi and Philips, 1982). The χ² of the constrained model is 670.41 (df = 119).¹ The resulting Δχ²(6) is 453.23, which is significant at p < .001 and supportive of the discriminant validity of the multicomponent model.

For the second test of discriminant validity, we again used the AVEs. If the AVEs for any two constructs are both larger than the variance that these two constructs share discriminant validity is established (Farrell and Rudd, 2009: 5). For this test, we assessed the relations within the attitude construct (i.e. between the attitude components only) and between the four attitude components and three similar constructs from the social sciences: prosocial motivation, work engagement and the bureaucrats’ rule-following identities, all measured in the same survey.

Prosocial motivation is defined as ‘the desire to benefit other people’ (Grant et al., 2009: 48). High prosocial motivation has previously been linked to positive perceptions of others in the work environment (Grant et al., 2009). We therefore classify high prosocial motivation as a positive attitude. To measure prosocial motivation, we used the four-item scale developed by Grant et al. (2009). In our study, this instrument exhibited high reliability (α = .87).

Schaufeli et al. (2006: 701) define work engagement as ‘a positive, fulfilling work-related state of mind that is characterized by vigor, dedication, and absorption’. We perceive high work engagement as a positive attitude and measured this construct using the one-factor, nine-item short version of the Utrecht Work Engagement Scale developed by Schaufeli et al. (2006). In our study, the α for this instrument was .95.

Bureaucrats’ rule-following identities refer to ‘their typical views of themselves, and how they should act, vis-a`-vis the rules’ (Oberfield, 2010: 737). Bureaucrats with a strong rule-following mindset are argued to be emotionally detached from their clients and treat like cases equally (Oberfield, 2014: 7). To measure this construct, we used Oberfield’s (2010) measure of workers’ rule-following identities, combined with a self-constructed item and an item from Baker et al. (1973). EFA and reliability analysis showed that the sole item without any reference to rules needed to be omitted. The α of the remaining five items was .78.

Both the within-construct analysis and between-constructs analysis support our claim of discriminant validity: the AVEs for all four components were substantially higher than the variance shared by any combination of attitude components (within-construct analysis). The same applied to the AVEs for all the combinations of attitude components and the three constructs: they were substantially higher than the variance shared by any component and the other constructs (between-construct analysis). The results of this test are presented in Appendix 2.

Reliability

Finally, we turn to the reliability. The R² in Table 3 indicates ‘how consistently the observed variable measures the latent dimension’ (Perry, 1996: 14). R² values range between 32.6 and 63.2%. This finding shows that although there is some variability in the reliabilities, the components generally explain an adequate portion of the variance in their items (see Perry, 1996).

Replication study

A common practice in scale development is to split samples for scale validation purposes (DeVellis, 2003). This practice is problematic for multiple reasons: first, ‘two subsamples are likely to be much more similar than two totally different samples’ (DeVellis, 2003: 99); second, performing EFA and CFA on two subsamples does not allow the cancelling out of effects the omitted items had on respondents’ replies to the final scale items (DeVellis, 2003). For these reasons, we conducted a replication study that followed the same steps as the test study. The replication sample and the results of the cross-validation efforts are discussed below.

Replication sample

The test survey was administered in one of the five Dutch tax regions. The replication survey was conducted in the other four tax regions, in the summer of 2016. The selection- and sample procedures were the same for both studies. The sample for the replication study contained 2257 street-level tax bureaucrats. The response rate was 55.2% (N = 1245). Again, we only included tax bureaucrats who confirmed to be tax auditors with client-contact and valid replies on the attitude items. This resulted in a final sample of 879 street-level tax bureaucrats. The respondents’ mean age was 53.7 and the gender division was 78.8% male versus 21.2% female.

Exploratory- and confirmatory factor analysis

An EFA identified four factors with an Eigen Value over 1, together explaining 58.3% of the variance. Table 1 shows that no cross-loadings above .3 were found and the factor loadings indicate that all items contribute substantially to their respective component (Field, 2013). The Cronbach’s alphas varied between .73 and .80, all well within the respectable range, and showed no signs of item redundancy. The correlations support the distinctness of the components (see Table 2). However, they do imply that a distinction based on the orthogonality of the components is less prominent in this sample.

The CFA supported the hypothesized fit of the model. The fit indices are: χ²= 394.32 (df = 113, p = .000), CMIN/DF = 3.49, GFI = .95, CFI = .93, RMSEA = .053. Only the CMIN/DF is somewhat above the 3.0 threshold. However, since this parameter is affected by sample size authors frequently suggest that values below 5.0 can be deemed reasonable (Marsh and Hocevar, 1985).

Construct validity tests

Significant construct loadings and AVEs over .5 are indicative of convergent validity. Table 3 shows that all construct loadings are significant at the p = 0.001 level, indicative of the multicomponent model’s convergent validity. The AVEs are .36 (cognitive component), .51 (positive affective component), .44 (negative affective component) and .56 (behavioural component), similar to those of the test study.

Discriminant validity is assessed through a comparison of the unconstrained measurement model with the constrained model, an AVE analysis of the within-construct and between-construct shared variance and the reliability of the measured items. The χ² of the unconstrained model was 394.32 (df = 113). The χ² of the constrained model is 2749.69 (df = 119). The resulting Δχ²(6) of 2355.37 is significant at p < .001 and affirms the model’s discriminant validity.²

For the AVE-analysis of discriminant validity, we use one construct from the test survey, that is, workers' rule-following identities³ (α = .78), and introduce one new construct from the social sciences: bureaucrats’ self-efficacy. Self-efficacy refers to the ‘belief in one’s capabilities to organize and execute the courses of action required to produce given attainments’ (Bandura, 1997: 3). We view self-efficacy as a positive attitude towards the self and assess this construct using Chen et al.’s (2001) eight-item new general self-efficacy scale (α = .88).

Appendix 2 shows that the AVE-analysis of both the within-construct and between-construct shared variances endorse the discriminant validity of the multicomponent model: the AVEs for the four components were substantially higher than the variance shared by either any combination of the attitude components or by the components and bureaucrats’ rule-following identities and self-efficacy.

Lastly, the reliabilities of the measured items vary between 25.4 and 63.7% (see Table 3). Although some variability in the reliabilities can be observed between the test study and the replication study, the general tendencies of the replication study’s reliabilities are similar to those of the test study.

Discussion and conclusion

Based on the psychological multicomponent model of attitude, we developed a measurement instrument for street-level bureaucrats’ attitude towards clients that was tested and validated in two surveys among Dutch street-level tax bureaucrats. This study revealed four different attitude components – rather than the expected three – measured by seventeen items: the cognitive attitude component, the positive affective attitude component, the negative affective attitude component and the behavioural attitude component. This measurement instrument demonstrated factorial and construct validity.

In addition to a substantive distinction according to the attitude components, we found that a distinction following the positive and negative connotation of items is inherent to the multicomponent model: the positively and negatively worded affective items were shown to form two orthogonal dimensions of affect, rather than loading on a single affective dimension. The cognitive and behavioural attitude components are also characterized by this positive–negative distinction, although in a different manner: like the affective attitude components, the cognitive component and behavioural attitude component only hold either positively worded or negatively worded items. However, in contrast to the affective items, the positively and negatively worded cognitive and behavioural items did not form separate dimensions of cognition and behaviour in the test study. An explanation therefore can be found in the unipolar versus bipolar nature of the items: where the affective items are unipolar, the cognitive and behavioural items are bipolar in nature. Their bipolarity allows these items to grasp the relevant spectrum of information, regardless of their unidirectional wording. For example, whereas the endpoints of a measurement continuum for the positive affective item ‘clients make me feel active’ are ‘very active’ and ‘not active at all’ (unipolar), the endpoints are ‘very honest’ and ‘very dishonest’ (bipolar) for the cognitive item ‘clients are dishonest’. Thus, one cannot characterize a bureaucrat who does not feel active as feeling passive, whereas a client who is not honest can be classified as being dishonest.

At first sight, the orthogonal distinction between the positive and negative affective items, and the positive affective items and the cognitive items seemingly moves to the background in the replication study: in this study, we do find significant correlations between these components. An explanation therefore can be found in the contextual factors of both studies: the test study had a lower data to respondent ratio than the replication study. A small data to respondent ratio combined with a small sample size can cause distortions to correlation values (DeVellis, 2003). Since EFA and CFA tests of factorial validity in the replication study support the orthogonal nature of the affective items, these results mainly indicate that the four attitude components are indeed related, but distinct components.

This study has some limitations. First, the aim of this study was to develop and validate a measurement instrument for street-level bureaucrats’ attitude towards clients. In general, the fit parameters of the multicomponent model indicate a good model fit. Nevertheless, the model parameters are somewhat ambivalent with regard to the AVE assessment of convergent validity, mainly for the cognitive attitude component. A possible explanation for this observation might be that this AVE application is a relatively conservative measure for assessing convergent validity (Fornell and Larcker, 1981: 46). Nevertheless, the AVEs allude to the added value of further model testing.

Second, for the test study, these parameters were obtained using the same data for both the EFA and CFA. Performing EFA and CFA on the same sample of respondents can inflate model fit indices. CFA parameters obtained under these circumstances do not suffice as a stand-alone test of factorial validity. However, they can be used as an assessment of model validity when their parameters are used for a comparison with CFA parameters derived from an entirely different sample of street-level bureaucrats. Moreover, that the replication study was totally unaffected by distortions from the items omitted in the test study strongly offsets this initial limitation.

Third, there is a risk of survey distortions resulting from social desirability bias and the use of cognitive measures to capture affective and behavioural attitudinal information. Social desirability bias incentives are deemed small in anonymous, self-administered survey research in PA, the more since most topics under inquiry in this academic field are not personally sensitive (Lee et al., 2012). Moreover, given the scale-development purposes of this study, social desirability is less of an issue due to our interest in how the scale items relate to each other, rather than their absolute values – assuming that social desirability bias would affect all four attitude items equally.

Also, cognitive measures may not have full access to an individual’s affective and behavioural information (Breckler, 1984: 1193). Despite this reservation, the measurement instrument developed in this study was specifically designed for use in survey research. Survey research largely confines scholars to using cognitive measures only. Due to this restriction, it is commonly accepted to assess affective and behavioural information with cognitive measures (e.g. Eagly and Chaiken, 1993).

Fourth, the multicomponent model discerns different types of attitude content, but attitudes can also differ in their formation process: attitudes can form through implicit, unconscious or explicit, conscious processes (Olson and Kendrick, 2008: 118). In the literature on street-level bureaucracies, cues and information on clients are often assumed to be processed implicitly (e.g. Jilke and Tummers, 2018). And previous research has shown that attitudes that form through implicit processes tend to be more strongly grounded in an affective attitudinal base (Olson and Kendrick, 2008). Even though surveys are believed to make implicit attitudes explicit through their recall function (see DeVellis, 2003), the next steps to advance the understanding of street-level bureaucrats’ attitude towards clients lie in exploring how this attitude is formed by unravelling the content and process drivers of this attitude.

Lastly, a precondition for any such instrument is its applicability across different types of street-level bureaucrats. The measurement instrument discussed here was mainly developed based on data gathered among street-level tax bureaucrats. Although bureaucrats’ tasks differ depending on the type of bureaucracy they work for, they find common ground in their assessment of clients underlying their client-related activities. Consequently, seemingly different street-level work contexts can be compared due to structural similarities across all types of street-level bureaucracies (Lipsky, 1980). Nevertheless, before claims about applicability across cases can be made, this instrument requires testing beyond the tax administration to verify its external validity.

Despite some limitations, this study has merit for multiple reasons. Knowledge of core PA concepts ‘can be achieved only through more systematic attention to concept development and measurement specification’ (Pandey and Scott, 2002: 578). The current study answers a loudly voiced call to PA scholars to use insights from other fields to advance PA theory and measurement (Grimmelikhuijsen et al., 2017; Perry, 2012).

This article showed that studies exploring street-level bureaucrats’ attitude towards clients often merely touch upon some elements thereof, and that thorough conceptualizations of this construct are scarce. Building on psychological attitude theory, this study brought the cognitive, affective and behavioural attitudinal elements interwoven with the rich descriptions and narratives of the bureaucratic encounter together in one cohesive conceptual framework of street-level bureaucrats’ attitude towards clients. This conceptual framework served as the foundation of a measurement instrument with sound psychometric qualities that showed that the multicomponent model of bureaucrats’ attitude towards clients measures a new, distinctive construct in PA. It also showed that, to fully capture this attitude, four components should be used in future measurements thereof. By means of these merits, the multicomponent assessment of bureaucrats’ attitude towards clients allows scholars to move beyond current practices that mainly tap into the cognitive dimension of this attitude. As a result, this study is exemplary of how introducing psychological insights to PA attitude theory and measurement can advance our understanding of core PA concepts.

In addition to avenues for future testing, a future research agenda should include exploring the applicability of the multicomponent measurement instrument for bureaucrats’ attitude towards clients. A main use of PA research lies in identifying causal relationships (Lee et al., 2012). Examples of causal relations to be explored include those between bureaucrats’ attitude towards clients and a variety of bureaucratic behaviours, such as the coping strategies they employ to deal with pressures inherent to working in a street-level bureaucracy (see Baviskar and Winter, 2017; Tummers and Musheno, 2015). A second avenue lies in the link between bureaucrats’ general client attitude and their categorizations of clients, such as their perceived deservingness (see Jilke and Tummers, 2018).

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded through the Netherlands Organization for Scientific Research, Vidi grant No. 452-11-011.

Notes

References

Aberbach

Christensen

(2007) The challenges of modernizing tax administration: Putting customers first in coercive public organizations. Public Policy and Administration 22(2): 155–182.

Ajzen

(2001) Nature and operation of attitudes. Annual Review of Psychology 52(32): 27–58.

Almond

Verba

(1973) Expectations of treatment by government and the police. In: Katz

Danet

(eds) Bureaucracy and the Public: A Reader in Official-Client Relations, New York, NY: Basic Books, pp. 43–49.

Bagozzi

Philips

(1982) Representing and testing organisational theories: A holistic construct. Administrative Science Quarterly 27(3): 459–489.

Baker

Etzioni

Hansen

, et al. (1973) Tolerance for bureaucratic structure: Theory and measurement. Human Relations 2(6): 775–786.

Bandura

(1997) Self-Efficacy: The Exercise of Control, New York: Freeman.

Bartels

(2013) Public encounters: The history and future of face-to-face contact between public professionals and citizens. Public Administration 91(2): 469–483.

Baviskar

Winter

(2017) Street-level bureaucrats as individual policymakers: The relationship between attitudes and coping behavior toward vulnerable children and youth. International Public Management Journal 20(2): 1–38.

Bem

(1972) Self-perception theory. In: Berkowitz

(ed.) Advances in Experimental Social Psychology Vol. 6, New York, NY: Academic Press, pp. 1–62.

10.

Berman

(1997) Dealing with cynical citizens. Public Administration Review 57(2): 105–112.

11.

Blau

(1960) Orientation towards clients in a public welfare agency. Administrative Science Quarterly 5(3): 341–361.

12.

Billingsley

(1964) Bureaucratic and professional patterns in social casework. Social Service Review 3(4): 400–407.

13.

Borgatta

Fanshel

Meyer

(1960) Social Workers’ Perceptions of Clients: A Study of the Caseload of a Social Agency, New York, NY: Russel Sage Foundation.

14.

Breckler

(1984) Empirical validation of affect, behavior, and cognition as distinct components of attitude. Journal of Personality and Social Psychology 47(6): 1191–1205.

15.

Brenninkmeijer A (2013) Mijn onbegrijpelijke overheid: Verslag van de Nationale Ombudsman over 2012. Available at: www.nationaleombudsman.nl/system/files/jaarverslag-download/Verslag%2B2012_0.pdf (accessed 19 October 2016).

16.

Byrne

(2010) Structural Equation Modeling with AMOS: Basic Concepts, Applications and Programming, 2nd ed. New York, NY: Taylor and Francis Group, LLC.

17.

Campbell JL (2005) Fiscal sociology in an age of globalization: Comparing tax regimes in advanced capitalist countries. Available at: http://sociology.dartmouth.edu/sites/sociology.dartmouth.edu/files/Fiscal%20Sociology%20%26%20Globalization.pdf (accessed 10 March 2018).

18.

Chen

Gully

Eden

(2001) Validation of a new general self-efficacy scale. Organizational Resarch Methods 4(1): 62–83.

19.

Crawford

Henry

(2004) The positive and negative affect schedule (PANAS): Construct validity, measurement properties and normative data in a large non-clinical sample. British Journal of Clinical Psychology 43: 245–265.

20.

DeVellis

(2003) Scale Development: Theory and Applications Vol. 26, 2nd ed. London: Sage.

21.

Dubois

(2010) The Bureaucrat and the Poor, England: Ashgate Publishing Limited.

22.

Dutch Tax Administration (2012) Management report 2011. Available at: www.rijksoverheid.nl/documenten/jaarverslagen/2012/05/16/bijlage-kamerbrief-over-beheerverslag-belastingdienst-2011 (accessed 12 February 2018).

23.

Eagly

Chaiken

(1993) The Psychology of Attitudes, Fort Worth, TX: Harcourt Brace and Company.

24.

Fabrigar

MacDonald

Wegener

(2005) The structure of attitudes. In: Albarracín

Johnson

Zanna

(eds) The Handbook of Attitudes, Mahwah, NJ: Lawrence Erlbaum Associates Publishers, pp. 79–124.

25.

Farrell AM and Rudd JM (2009) Factor analysis and discriminant validity: A brief review of some practical issues. Available at: http://eprints.aston.ac.uk/7644/1/factor_analysis_ANZMAC_2009.pdf (accessed 31 July 2017).

26.

Fazio

Olson

(2003) Attitudes: Foundations, functions, and consequences. In: Hogg

Cooper

(eds) The Sage Handbook of Social Psychology, London: Sage, pp. 139–160.

27.

Field

(2013) Discovering Statistics Using SPSS: And Sex and Drugs and Rock ‘N Roll, 4th ed. Los Angeles, CA: Sage.

28.

Finer

(1931) Officials and the public. Public Administration 9(1): 23–36.

29.

Fornel

Larcker

(1981) Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research 18(1): 39–50.

30.

Gastelaars

(2006) Het echte werk van andere contactambtenaren. Bestuurskunde 3: 48–56.

31.

Goodsell

(1976) Cross-cultural comparison of behavior of postal clerks towards clients. Administrative Science Quarterly 21(1): 140–150.

32.

Goodsell

(1981) Looking once again at human service bureaucracy. The Journal of Politics 43(3): 763–778.

33.

Gordon

(1975) Bureaucratic competence and success in dealing with public bureaucracies. Social Problems 23(2): 197–208.

34.

Grant

Parker

Collins

(2009) Getting credit for proactive behavior: Supervisor reactions depend on what you value and how you feel. Personnel Psychology 62(1): 31–55.

35.

Grimmelikhuijsen

Jilke

Olsen

, et al. (2017) Behavioral public administration: Combining insights from public administration and psychology. Public Administration Review 77(1): 45–56.

36.

Haddock

Huskinson

TLH

(2004) Individual differences in attitude structure. In: Haddock

Maio

(eds) Contemporary Perspectives of the Psychology of Attitudes, New York, NY: Psychology Press, pp. 35–56.

37.

Hasenfeld

Steinmetz

(1981) Client-official encounters in social service agencies. In: Goodsell

(ed.) The Public Encounter: Where State and Citizen Meet, Bloomington: Indiana University Press, pp. 83–101.

38.

Himmelfarb

(1993) The measurement of attitudes. In: Eagly

Chaiken

(eds) Psychology of Attitudes, Forth Worth, TX: Harcourt Brace & Company, pp. 23–88.

39.

Hupe

Hill

(2016) And the rest is implementation: Comparing approaches to what happens in policy processes beyond great expectations. Public Policy and Administration 31(2): 103–121.

40.

Jilke

Tummers

(2018) Which clients are deserving of help? A theoretical model and experimental test. Journal of Public Administration Research and Theory 28(2): 226–238.

41.

Katz

(1960) The functional approach to the study of attitudes. Public Opinion Quarterly 24(2): 163–204.

42.

Katz

Danet

(1973a) Introduction: Bureaucracy as a problem for sociology and society. In: Katz

Danet

(eds) Bureaucracy and the Public: A reader in Official-Client Relations, New York, NY: Basic Books, pp. 3–27.

43.

Katz

Danet

(1973b) Petitions and persuasive appeals: A study of official-client relations. In: Katz

Danet

(eds) Bureaucracy and the Public: A reader in Official-Client Relations, New York, NY: Basic Books, pp. 174–190.

44.

King

Stivers

(1998) Government Is Us: Public Administration in an Anti-Government Era, Thousand Oaks, CA: SAGE Publications.

45.

Kroeger

(1975) Bureaucracy, social exchange, and benefits received in a public assistance agency. Social Problems 23(2): 182–196.

46.

Lee

Benoit-Bryan

Johnson

(2012) Survey research in public administration: Assessing mainstream journals with a total survey error framework. Public Administration Review 72(1): 87–97.

47.

Lineberry

(1977) Equality and Urban Policy: The Distribution of Municipal Public Services, Beverly Hills: Sage Publications.

48.

Liou

Cruise

(1994) Assessing employee attitudes in a community-based aids service organization. Evaluation and the Health Professions 17(3): 273–289.

49.

Lipsky

(1980) Street-Level Bureaucracy: Dilemmas of the Individual in Public Service, New York, NY: Russell Sage Foundation.

50.

Maio

Haddock

(2009) The Psychology of Attitudes and Attitude Change, London: Sage Publications.

51.

Marsh

Hocevar

(1985) Application of confirmatory factor analysis to the study of self-concept: First- and higher order factor models and their invariance across groups. Psychological Bulletin 97(3): 562–582.

52.

Maynard-Moody

Musheno

(2003) Cops, Teachers, Counselors: Stories from the Front Lines of Public Service, Ann Arbor, MI: University of Michigan Press.

53.

Merton

(1940) Bureaucratic structure and personality. Social Forces 18(4): 560–568.

54.

Nelson

(1981) Client evaluations of social programs. In: Goodsell

(ed.) The Public Encounter: Where State and Citizen Meet, Bloomington: Indiana University Press, pp. 23–24.

55.

Oberfield

(2010) Rule following and discretion at government's frontlines: Continuity and change during organization socialization. Journal of Public Administration Research and Theory 20(4): 735–755.

56.

Oberfield

(2012) Socialization and self-selection: How police officers develop their views about using force. Administration & Society 44(6): 702–730.

57.

Oberfield

(2014) Becoming Bureaucrats. Socialization at the Front Lines of Government Service, Philadelphia: University of Pennsylvania Press.

58.

Olson

Kendrick

(2008) Origins of attitudes. In: Crano

Prislin

(eds) Attitudes and Attitude Change, New York, NY; London, UK: Psychology Press, pp. 111–130.

59.

Pandey

Scott

(2002) Red Tape: A review and assessment of concepts and measures. Journal of Public Administration Research and Theory 12(4): 553–580.

60.

Peeters

FPML

Ponds

RWHM

Vermeeren

MTG

(1996) Affectiviteit en zelfbeoordeling van depressie en angst. Tijdschrift Voor Psychiatrie 38(3): 240–250.

61.

Perry

(1996) Measuring public service motivation: An assessment of construct reliability and validity. Journal of Public Administration Research and Theory 6(1): 5–22.

62.

Perry

(2012) How can we improve our science to generate more usable knowledge for public professionals? Public Administration Review 72: 479–482.

63.

Petty

Wegener

Fabrigar

(1997) Attitudes and attitude change. Annual Review of Psychology 48: 609–647.

64.

Prottas

(1979) People-Processing: The Street-Level Bureaucrat in Public Service Bureaucracies, Lexington, MA: Lexington Books.

65.

Raaphorst

Van de Walle

(2017) A signaling perspective on bureaucratic encounters: How public officials interpret signals and cues. Social Policy and Administration. Preprint. DOI: 10.1111/spol.12369.

66.

Rosenberg

Hovland

(1960) Cognitive, affective, and behavioral components of attitudes. In: Rosenberg

Hovland

(eds) Attitude Organization and Change, 3rd ed. New Haven, CT: Yale University Press, pp. 1–14.

67.

Savaya

Gardner

Stange

(2011) Stressful encounters with social work clients: A descriptive account based on critical incidents. Social Work 56(1): 63–71.

68.

Schaufeli

Bakker

Salanova

(2006) The measurement of work engagement with a brief questionnaire: A cross-national study. Educational and Psychological Measurement 66(4): 701–716.

69.

Scherer

(1980) Psychological factors in bureaucratic encounters: Determinants and effects of interactions between officials and clients. In: Singleton

Spurgeon

Stammers

(eds) The Analysis of Social Skill, New York, NY: Plenum Press, pp. 315–328.

70.

Schwarz

(2008) Attitude measurement. In: Crano

Prislin

(eds) Attitudes and Attitude Change, New York, NY; London: Psychology Press, pp. 41–60.

71.

Smith

(1947) The personal setting of public opinions: A study of attitudes towards Russia. Public Opinion Quarterly 11(4): 507–523.

72.

Soydan

(1995) A cross-cultural comparison of how social workers in Sweden and England assess a migrant family. Scandinavian Journal of Social Welfare 4(2): 85–93.

73.

Stone

(1981) Attitudinal tendencies among officials. In: Goodsell

(ed.) The Public Encounter: Where State and Citizen Meet, Bloomington: Indiana University Press, pp. 43–68.

74.

Stone

Feldbaum

(1976) Blame, complacency, and pessimism: Attitudes and problem perceptions among selected street level administrators in two suburban counties. Administration and Society 8(1): 79–106.

75.

Thompson

(2007) Development and validation of an internationally reliable short-form of the Positive and Negative Affect Schedule (PANAS). Journal of Cross-Cultural Psychology 38(2): 227–242.

76.

Tummers

(2012) Policy alienation of public professionals: The construct and its measurement. Public Administration Review 72(4): 516–525.

77.

Tummers LG and Musheno MC (2015) Coping during public service delivery: Developing a measurement instrument for survey research. In: The 2015 EGPA annual conference, Speyer, Germany, 10–12 September. Available at: http://larstummers.com/wp-content/uploads/2012/01/Paper-EGPA-Tummers-Musheno-FINAL.pdf (accessed 31 July 2017).

78.

Van de Walle

Lahat

(2016) Do public officials trust citizens? A welfare state perspective. Social Policy and Administration 51(7): 1450–1469.

79.

Vermeeren

Kuipers

Steijn

(2011) Two faces of the satisfaction mirror. A study of work environment, job satisfaction and customer satisfaction in Dutch municipalities. Review of Public Personnel Administration 31(2): 171–189.

80.

Vermeeren

Kuipers

Steijn

(2014) Does leadership style make a difference? Linking HRM, job satisfaction, and organizational performance. Review of Public Personnel Administration 34(2): 174–195.

81.

Watson

Clark

Tellegen

(1988) Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology 54(6): 1063–1070.

82.

Wilson

(1989) Bureaucracy: What Government Agencies Do and Why They Do It, New York, NY: Basic Books.

83.

Winter SC (2002) Explaining street-level bureaucratic behavior in social and regulatory policies. In: The annual meeting of the American Political Science Association, Boston, MA, 29 August–1 September 2002. Available at: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.482.4957&rep=rep1&type=pdf (accessed 10 January 2018).

84.

Worden

(1989) Situational and attitudinal explanations of police behavior: A theoretical reappraisal and empirical assessment. Law & Society Review 23(4): 667–711.