- Split View
-
Views
-
Cite
Cite
Steffen Eckhard, Laurin Friedrich, Linguistic Features of Public Service Encounters: How Spoken Administrative Language Affects Citizen Satisfaction, Journal of Public Administration Research and Theory, Volume 34, Issue 1, January 2024, Pages 122–135, https://doi.org/10.1093/jopart/muac052
- Share Icon Share
Abstract
Spoken administrative language is a critical element in the relationship between citizens and the state, especially when it comes to face-to-face interactions between officials and citizens during the delivery of public services. But preceding work offers little insights into the verbal features of street-level bureaucracy. Drawing on communication studies, we argue that administrative language differs along both a relational and an informational linguistic component. To test the consequentiality of this theory, we design a factorial survey experiment with a representative sample of 1,402 German citizens. Participants evaluated audio recordings of a hypothetical service encounter where we systematically varied the language used by the official and the service decision, measuring participants’ service satisfaction as the main outcome. Based on regression analysis, we find that relational elements of administrative language improve citizen satisfaction, independent of the service outcome, but that the effect does not hold for the informational component. These findings emphasize the importance of relational communication in citizen-state interactions, which tends to be neglected in public administration theory and practice.
Introduction
Although public service encounters—face-to-face interactions between citizens and public officials during public service provision (Bartels, 2013)—are at the centre of any personal interaction between the state administration and its citizens, we still know relatively little about their dynamics and consequences. One key insight, shown in studies, is that the characteristics of communication during the encounters, such as the duration of the conversation, are of central importance for the service outcome (Bruhn and Ekström, 2017; Hand and Catlaw, 2019; Raaphorst, Groeneveld, and Van de Walle, 2018). Recently, there have been further attempts to explore the black box of spoken administrative language by identifying linguistic elements that are important for how citizens experience an encounter (Eckhard, et al., 2022; Holzinger, 2020; Kasdorp and Schakel, 2022; Scheibelhofer, Holzinger, and Draxl, 2021). But we do not know yet whether and to what extent differences in spoken administrative language influence citizen perceptions. In contributing toward closing this gap, this paper asks how variation in administrative language affects citizen satisfaction with a public service encounter.
Reviewing the state of the art in communication studies and street-level bureaucracy, we present a taxonomy of how citizens perceive administrative language (Eckhard et al., 2022). The theory posits that frontline communication has an informational and a relational component and that each of these consists of two dimensions: On the one hand, administrative language serves citizen information needs depending on the comprehensibility of administrative speech acts and the extent of reification, such as elaborations on the regulatory origin or intention of a given rule. On the other hand, any verbal communication also encodes a relational message between sender and recipient, represented by two linguistic dimensions: emotionality and complaisance (helpfulness).
As yet, confirmatory evidence on the consequentiality of administrative language does not exist. In particular, it remains unclear whether the proposed components and dimensions of administrative language are truly relevant to citizen perceptions. To test these propositions, we report findings from a factorial survey experiment (James et al., 2017b) designed to explore how variation in administrative language affects citizen satisfaction. On this basis, we focus on the realm of welfare service delivery where bureaucrat–client interactions often have a substantial impact on peoples’ lives. We confronted survey participants with audio recordings of a hypothetical public service encounter where a citizen applies for a welfare service in relation to the COVID-19 crisis. We varied systematically administrative speech acts by de-emphasizing (control group) or emphasizing dimensions of the taxonomy (three treatment groups with either informational, relational, or informational and relational language); and by also varying the service decision (positive or negative). We find firstly that participants’ assessments of the experimental treatment are in line with the varying values of the independent variable, demonstrating that the experimental manipulation worked. Secondly, by asking respondents about their satisfaction with the encounter, we find that the relational dimensions and all dimensions together increase citizen satisfaction, whilst no effect is found for the informational dimensions alone. This effect prevails independently of the service decision as experimental subjects reported higher satisfaction ratings in groups with a negative and groups with a positive service decision.
With this study, we demonstrate for the first time how the substantive content of spoken administrative language affects citizen perceptions, i.e. their satisfaction. We also demonstrate that this effect is independent from the service outcome. These findings validate our previously proposed taxonomy of administrative language (Eckhard et al., 2022). Furthermore, speaking to earlier studies that explore the black box of administrative language (Holzinger, 2020; Kasdorp and Schakel, 2022; Scheibelhofer et al., 2021), we highlight the importance of the relational elements of bureaucratic communication. Strikingly, this is opposed to the dominant focus among practitioners, who, in general, address primarily informational linguistic dimensions when considering improvements in communication.
Our findings also contribute more broadly to studies on street-level bureaucracy, which have identified communicative practices in public service encounters as influential for bureaucratic actions (Nielsen, 2007; Raaphorst and Loyens, 2020; Raaphorst et al., 2018) but have so far neglected to analyze the factors determining the citizen assessment of the interaction (de Boer, 2020; Hansen, 2021; Nielsen, Nielsen, and Bisgaard, 2021). Given the relevance of spoken language, our work may also help to better understand how agencies may influence positively citizen experiences in the encounter. This, in turn, is fundamental to the legitimacy of public institutions as citizens “care as much [ . . . ] about the process of their interactions with the state as they do about the outcome.” (Moynihan, Herd, and Harvey, 2014, p. 47; see also, Christensen and Lægreid, 2005; Hansen, 2021; Kumlin and Rothstein, 2005). And measurement of real administrative language through our taxonomy may allow us to establishing which citizen groups spoken language represents an administrative burden in accessing essential state services (Herd and Moynihan, 2019), which again is key to better understanding the role of language and encounters in bias and discrimination.
The paper is structured as follows: we begin by situating our research in public administration literature; we then discuss the theoretical concept of administrative language, including by reporting our own interview data for illustration; after presenting the survey experiment and its findings, we offer a discussion and conclusion.
The Relevance of Administrative Language in Public Service Provision
Street-level bureaucrats are at the centre of the state’s actions in policy implementation. These frontline civil servants specify the concrete application of services to the individual cases of citizens, for which policies can only provide abstract and vague guidelines (Hupe and Hill, 2007; Lipsky, [1980] 2010; Maynard-Moody and Musheno, 2000). Officials are consequently equipped with a considerable amount of discretionary power and are expected to deploy this discretion towards effective, fair, and responsive public service delivery (Frederickson and Hart, 1985; Jensen and Pedersen, 2017; Keiser, 1999; Lipsky, [1980] 2010; Tummers and Bekkers, 2014).
Citizens’ perceptions of these encounters are consequential in their assessment of the state’s authority (Christensen and Lægreid, 2005). The manner in which an encounter proceeds is relevant beyond the factual service outcome, as citizens also judge state institutions based on their experiences of interactional justice (Döring, 2022, p. 801). Several studies show that public officials’ behavior toward citizens may contribute to legitimizing or delegitimizing the entire politico-administrative system (Denhardt and Denhardt, 2007; Hansen, 2021; Kampen, Van De Walle, and Bouckaert, 2006; Rothstein, 2009; Van Ryzin et al., 2004; Yackee and Lowery, 2005).
Yet, administrative action at the frontline often seems to fail the “promise of modern bureaucracy”; that is, to generate fair and equal public service delivery (Andersen and Guul, 2019, p. 442). Lipsky ([1980] 2010) was among the first to assert that the nature of their work would prevent street-level bureaucrats “from coming even close to the ideal conception of their jobs” (p. xiv). Instead, he claimed that their daily actions are characterized by “favoritism, stereotyping, and routinizing” (ibid.). Several scholars have taken up this argument to study biases in public service provision (Raaphorst and Groeneveld, 2019, p. 117). They show how individual patterns of discrimination emerge from bureaucratic face-to-face encounters, eventually causing unequal treatment due to citizens’ gender, ethnicity, age, or social status (Blessett et al., 2019; Cepiku and Mastrodascio, 2021). Representative bureaucracy literature holds that such (unconscious) discriminatory practices prevent the equal distribution of public services across societal groups (Grimmelikhuijsen et al., 2017; Guul, 2018; Kingsley, 1944; McCrea, 2021; Meier, 1993; Vinopal, 2017).
Verbal communication is at the core of most interpersonal citizen–state interactions. Existing studies reveal that the outcome of public service provision emerges in an “interactive or reflexive process mediated by sociolinguistic practices” performed by citizens and bureaucrats during an encounter (Picciotto, 2007, p. 11, cf. Bartels, 2015; Bruhn and Ekström, 2017; Hand and Catlaw, 2019; Holzinger, 2020; Kasdorp and Schakel, 2022; Picciotto, 2007; Raaphorst and Loyens, 2020). The findings from Nielsen (2007) and Raaphorst and colleagues (2018) even hint that the more consensual and direct the bureaucratic client conversation is, the more favorable the administrative outcome for citizens.
Communicative practices may therefore have systematic effects on equal service delivery, including the representation of the interests of marginalized groups. A recent study by Holzinger (2020), for instance, shows how migrants experience patterns of “linguistic discrimination” when accessing benefits (p. 1804). Based on their poor language skills, street-level bureaucrats consider them as “illegitimate speakers” (p. 1801) and therefore provide inferior services compared to native-speaking applicants.
However, despite such initial insights, preceding literature has focused rarely on the concrete features of verbal administrative language and how these relate to the citizen’s view of the encounter.
How Bureaucrats’ Spoken Administrative Language Affects Citizen Perceptions
In this study, we employ a rather broad notion of spoken communication in an encounter. This follows Bartels (2013, p. 476), who argued that the “in-between is not simply a communicative void for the neutral transmission of information but, instead, a multifaceted process of interwoven situated performances which enables or disadvantages the actual abilities of public professionals and citizens to make claims, influence decisions, and understand each other.” Acknowledging such a broader perspective, we understand spoken administrative language in an encounter as the distinctive practices of verbal interpersonal communication upon which bureaucrats rely during a conversation with citizens. This implies that bureaucratic speech acts may differ depending on their structure or patterns of wording, but also in the space that they give to citizens to bring in their own concerns and perspective.
Furthermore, we use the term “citizen perceptions” when referring to the extent to which administrative language is positively or negatively perceived by citizens and thus contributes to their overall satisfaction or dissatisfaction with the encounter. Citizen satisfaction is a commonly used concept when studies aim at assessing the quality of service provision (Andrews and Van de Walle, 2013; Chan et al., 2020; Wu and Jung, 2016). We expect that there are certain styles of administrative communication that lead to a more positive citizen evaluation—independent of the content and outcome of service delivery.
To answer which forms of spoken administrative language are influential for positive or negative citizen perceptions, we follow a foundational distinction made in human communication and psychology: Watzlawick and colleagues (2011 [1967]) asserted that language varies regarding the information it conveys (such as comprehensibility) and what relational message it encodes (such as hierarchy between sender and recipient). Their influential work is often viewed as one of the “beginnings of communication theory” (Lutterer, 2007, p. 1022). For many subsequent communication scholars, their two-fold taxonomy has been the basis to study the meaning of verbal interaction beyond the mere exchange of information (Burgoon and Hale, 1984; Dillard, Solomon, and Palmer, 1999). Meanwhile, the two dimensions find use in various specific research streams, such as the linguistic evidentiality of speech acts (Söderqvist, 2020), the classification of conversational goals (Yeomans, Schweitzer, and Brooks, 2022), negotiations (Adair, Brett, Okumura, and Taylor, 2007) or communication between organizations and their members (Einwiller, Ruppel, and Stranzl, 2021; Gray and Laidlaw, 2004). It is plausible, therefore, that these two basic components also matter for administrative communication in public service encounters.
First, citizens usually have a demand for information on technical or legal issues when venturing onto the “unfamiliar turf” of bureaucracy (Döring, 2021, p. 1159): They have to identify and match existing services to their personal situation and also need to engage in administrative procedures to insert their personal concerns into bureaucratic case processing. How elaborate frontline officials are in satisfying these information needs is the first component of administrative language that we assume to be notable to citizen clients.
Second, as in any human communication, conversations during public service encounters encode a relational message between sender and recipient. In most cases, citizens are naturally in an inferior position toward civil servants since the latter represents the governing state authority and additionally decide about access to the service clients seek to obtain (Maynard-Moody and Portillo, 2010, p. 258). Thus, to what extent front-line officials emphasize or downplay this hierarchical relationship is the second component of administrative language that should be crucial for citizen perceptions.
To identify concrete communicative practices of administrative language, we undertook an explorative interview-based study among frontline civil servants (Eckhard et al., 2022). Communicative practices are concrete speech acts that can be observed at the level of spoken communication. We conducted 64 expert interviews with local public officials who interact directly with citizens on a daily basis, or have done so in previous positions.1 These interviewees were asked to report what observations they had made regarding the impact of variation in their spoken communication on positive or negative citizen perceptions. As a result, we were able to cluster the resulting communicative practices in four dimensions, two of these linked to the informational component of language, and two linked to the relational component.
Figure 1 summarizes the resulting taxonomy. In addition to the two basic components (the informational and relational component), it distinguishes four dimensions of administrative language (comprehensibility, reification, emotionality, and complaisance) and seven concrete communication practices. The four dimensions build the basic factors of the concept; they provide a deep understanding of how frontline communication varies and how it should be related to citizen perceptions of the public service encounter. Meanwhile, the seven communicative practices build the lowest and most concrete level of the concept as they map the linguistic elements of communication which can be directly observed and measured. They are each associated with one of the four dimensions and as such constitute their indicators.
We describe next each of the four dimensions of the taxonomy, including unpublished utterances and exemplifying quotes by interview participants to disclose how we interpreted the interview results and identified the elements of the taxonomy.2
Comprehensibility
Public service encounters constitute typical expert-layperson communication situations (Lewalter, Geyer, and Neubauer, 2014, p. 161), where little common professional knowledge of bureaucracy exists (Luttermann, 2017; Müller, 2017). Relative to public servants, citizens are, in general, unfamiliar with the laws, regulations, and administrative procedures upon which the public services they seek access are based. Comprehensible administrative communication aims to help citizens understand how exactly the service in question applies to their specific life situation. The following quote by a jobcentre employee is representative of this:
Once you have established the facts correctly—if you know how people live, what kind of income they have, what are their financial circumstances, what is their family situation—then you can advise people extensively and explain, this is the legal situation, and you may get this or that (ID: 53).
To achieve this goal, interviewees noted the importance of simple and plain administrative language that makes complex administrative procedures and regulations understandable. Several of the interviewees remarked that they rely explicitly on everyday language when talking about administrative issues with clients while avoiding technical terminology “from inside the administration” (ID: 22). Many of them had experienced that technical language use is difficult to understand: “I speak as I think. I do not talk ‘officialese’. That is always received well by citizens. They understand me” (ID: 56).
Overall, the expectation is that citizens gain a more favorable impression of service encounters if officials communicate in such a simple and plain style.
Reification
The second dimension is reification. Under this term, we subsume communicative practices that provide administrative actions and decisions with a plausible statement of reason. This dimension emphasizes that to satisfy citizens’ information needs, officials should also establish transparency in the service procedure and decision. If citizens perceive that the bureaucrat determines the final outcome of state action without providing justification, they will have difficulties in accepting this. As Kumlin and Rothstein (2005, p. 349) write, discretionary decisions are then likely to be accompanied by “suspicions of cheating, arbitrariness, and discrimination.” This intuitive association is also confirmed by insights from communication studies and psychology, where the knowledge-deficit model implies that administrative decisions should be more acceptable if their causal origin is not obscured by complex technical information (Einsiedel, 2000; Weisberg et al. 2018).
Two concrete communicative practices could be identified from interview statements. Firstly, interviewees consistently stated the need to reify the executive role of the public administration as a policy implementer. Because of their position at the interface of the state and citizens, front-line officials personify the government benefits and sanctions citizens experience. It is therefore intuitive that citizen clients may perceive the acting official, who represents the state, as at least partly responsible for the service decision. Reifications on the role of the front-line official within the rule-bound system may therefore help to clarify the legal origin of a decision. Exemplifying this, the employee of a citizen office told us:
The argument is that we have a legal basis, which we have been given by the legislator and then we have to stick to it just as prescribed [...]. This is just the way things are. At the end, we are only an executive body” (ID: 53).
Another interviewee noted that he tries to “talk about the facts” that he describes to his clients as “something that I cannot change but that is derived from this or that law.” (ID: 65).
The second communicative practice concerns reifications of the motivation and intentions of administrative practices. Interviewees explained that when citizens experience the essence of public service delivery and how it affects their lives, it is often not directly clear to them which logic a procedure follows and why. Even if citizens are not satisfied with the service outcome, officials explained that they tend to be more sympathetic once they realize that administrative actions arise out of a certain purpose. In case citizens have the possibility to “look behind the whole thing” and to understand “how it came about, how it happened,” administrative outcomes will be more satisfactory for them, as one interviewee explained (ID: 4). As an example, the employee of a jobcentre told how she usually raises acceptance for the fact that the basic income granted by her agency is strictly means-tested by explaining to citizens:
Our task is to grant substantial benefits to people in need to make a living. That does not mean that we make everybody rich. That means that we help people temporarily avoid serious adversity (ID: 44).
Overall, the expectation is that such reifying administrative speech acts, those that explain the executive role of the administration and the motivation of administrative practices should leave a more favorable impression among citizen clients.
Emotionality
Next are the two dimensions associated with the relational component of an administrative language. The first dimension is emotionality, an aspect also reported as relevant in a range of studies on public administration. For instance, literature on emotional labor generally highlights that some professions need to actively manage their emotional expressions as part of their occupation (Hochschild, 1983). A number of scholars have pointed out that the bureaucrats’ emotional engagement towards citizens gives an impression of high quality in public service delivery (Clayton et al., 2015; Guy and Lee, 2015; Guy, Newman, and Mastracci, 2014; Hsieh, 2014). It is plausible that citizens who sense that a frontline official is emotionally committed to their concerns will be more appreciative of the encounter and its outcomes.
When it comes to communicative practices, firstly we identified from interviews that casual communication is important. In particular, interviewees reported consistently that they seek consciously to endow citizen conversations with a personal touch, such as including elements of small talk into the conversation to create a relaxed atmosphere. But casualness also serves as a strategy to moderate conflict, as interviewees report, because service interactions are often accompanied by high emotional tension or stress. A jobcentre employee emphasized the relevance to begin an encounter with a short “chit chat” (ID: 30). Another public employee charged with processing residence registrations confirmed: “When I see, for instance, that someone has moved from Hamburg to Wernau then I say: Oh, Hamburg is a beautiful city. So you try to get into a conversation” (ID: 56).
Casual communicative practices thus contribute to a positive citizen perception. They endow the interaction with a personal touch and show, as a jobcentre employee remarked aptly, that “we are still just somehow humans and still have personal sensitivities and stories that we carry with us.” (ID: 52). This in turn should be interpreted by citizens as a personal commitment to their individual case, demonstrating that an official’s engagement goes beyond formal duties.
The second emotional communicative practice is empathic communication. Most interviewees referred to communicative practices that express appreciation and understanding of a citizen’s life situation. Social services in particular are often provided to citizens who are less privileged and face existential problems. In such situations, as one interviewee said, “. . . you have to find the right words [ . . . ] You need to be compassionate and empathic” (ID: 30). One employee of a migration office told that she usually says to her clients: “I fully understand your distress and your fears or your interest” (ID: 27).
Frontline officials are also often required to enforce restrictive service decisions. In such cases, empathic communication serves to alleviate negative client perceptions. For instance, one interviewee, working in a social service office, explained: “Of course, I can understand the clients’ situation if their claims have been rejected [ . . . ] and why they are angry then [ . . . ] and of course you should make them aware of this” (ID: 34). Many officials argue that such communicative practices create acceptance among clients who will then feel that they and their concerns are taken more seriously and appreciated.
Complaisance
As the last dimension of the taxonomy, complaisance captures how officials communicate on a continuum “from the language of helping to the language of control and threat” (Brodkin, 1997, p. 14). Street-level bureaucrats are typically performing either of two fundamental government actions: service-producing tasks designed to help and support citizens in certain life situations; or regulatory tasks that include the enforcement of restrictions and prohibitions (Jensen, 2018, p. 1130 ff.). One problem is that citizens frequently cannot distinguish the two “roles” and extrapolate previous encounters (de Boer, 2020). A jobcentre employee, for instance, said in an interview that “there is often a hesitation to visit the jobcentre, because people are afraid” (ID: 23). Complaisance implies that officials establish a sense of availability and helpfulness through their communication which may help to mitigate negative attitudes among citizens toward public authorities.
The first communicative practice mentioned by interviewees refers to speech acts that impart the supportive nature of state action in dissociation from its restrictive side. Interviewees reported that they directly respond to citizens who are insecure and anxious about encountering a state institution by offering help and support. A citizen office employee, for example, reported how she dealt with a client who was afraid of losing his flat:
And then he tells me, ‘[…] I’m already three months in arrears with my rent, the landlord is exerting pressure and wants to kick me out of the flat, and then I’ll have to sleep under the bridge from tomorrow onwards.’ […] Then I can take the pressure off a little bit when I say ‘I’ll try to show you a possibility to help so that life goes on’ (ID: 2).
As the second communicative practice of complaisance, interviewees consistently indicated that it was important to impart availability towards their clients. Interviewees said that it makes a huge difference for citizen perceptions whether agencies and officials aspire to accessibility and availability, or whether they hide behind thick walls so that getting in touch with them is a burden per se. The employee of a jobcentre, for instance, noted that for citizens “nothing is as bad as applying for some service or handing in documents and not getting any response.” (ID: 52). Imparting availability, in this sense, reduces the power asymmetry in citizen-state interactions. A social office employee explained:
I would say that this is central and this is also what my clients mirror to me […] if the clients have a direct contact person who is tangible which makes it easier to deal with things, I think it is important that clients feel appreciated, against this background, it is important to be available for the client (ID: 34).
Overall, the dimension of complaisance highlights that citizen perceptions can be more favorable if administrative speech acts convey a willingness on the side of the state to individually support them, whilst also indicating accessibility.
Research Design and Hypotheses
This paper aims to study the consequentiality of variation in spoken administrative language for citizen perceptions. So far, there is not as yet any robust and systematic empirical evidence beyond the anecdotal interview statements reported above. We therefore undertook a factorial survey experiment (Auspurg and Hinz, 2015, p. 9 ff.) aimed at verifying to what extent the proposed components and dimensions of administrative language are relevant to citizen perceptions. Experimental designs offer an a priori balancing of potentially confounding variables while their integration into a survey allows for an application to a real-world setting. Thus, survey experiments promise to provide results characterized by both high internal and external validity which is why they have become a common approach in public administration research (James et al., 2017b, p. 10). The method is therefore suitable to assess the causal relationship between spoken administrative language and citizen perceptions.
As the main dependent variable, we focus on citizen satisfaction. This decision follows earlier studies that used the concept as a measure for studying citizen perceptions of service delivery irrespective of the service domains (Andrews and Van de Walle, 2013; Chan et al., 2020). The main independent variable is administrative language along the dimensions of the taxonomy. Following the structure of the taxonomy, we subsequently specify three hypotheses at the level of the informational and relational components of language that is scrutinized as part of the experiment. This implies that when we speak about highly informational language, we mean language that is pronounced in terms of both comprehensibility and reification. Likewise, when we refer to highly relational language, we mean language that is pronounced in terms of emotionality and complaisance at the same time. The downside of this procedure is that we cannot assess individual dimensions, but may only observe the overall impact of each of the three groups. Below, we discuss this in greater detail. The first two resulting hypotheses read:
H1:Citizens service satisfaction is higher if frontline officials use highly informational language, compared to a setup with less informational language.
H2:Citizens service satisfaction is higher if frontline officials use highly relational language, compared to a setup with less relational language.
We further expect that the combination of highly relational and highly informational language serves as a distinct pattern of communication. Such spoken communication should then satisfy citizens’ communicative needs in the best possible way, causing the most substantive effect compared to the setup where none of the linguistic dimensions are pronounced.3 Hypothesis three therefore reads:
H3:Citizens service satisfaction is higher if frontline officials use a combination of highly relational and informational language, compared to a setup with less informational and relational language, or with either highly informational or highly relational language.
The experiment also includes control variables for a range of alternative explanations. First and most importantly, a plausible expectation is that client satisfaction is highly influenced by the service decision outcome. We therefore include the service decision as an additional treatment (positive/ negative service decision) in the experiment. The expectation is that we should be able to confirm the above hypotheses under both conditions.
Following the existing literature, we also include measures for stable attitudes towards the state, such as institutional trust and federal government support (Kampen et al., 2006; Van de Walle and Bouckaert, 2003). In line with the expectancy-disconfirmation model (Van Ryzin, 2004; Zhang et al., 2022), we include a measure for citizens’ prior expectations.4 Finally, given that citizens’ frequent utilization of a service category has shown to be influential (Grunow and Strüngmann, 2008; Van de Walle, 2004), we include a measure capturing whether regular receipt of social benefits prevails. Lastly, participants’ gender, age, education, and the political party affiliation were included to inspect the structure of the experimental sample and the composition of the experimental groups.
Experimental Setup
We designed the experiment in the context of a welfare service, being a policy field that builds substantially on the interpersonal interaction between bureaucrats and citizens. Welfare administrations are charged with providing individual help to citizens in certain life situations qualifying as archetypal “people-processing bureaucracies” (Lipsky, [1980] 2010, p. 105). Personal encounters with beneficiaries where frontline workers “convey the messages of the system” to the people (Sandfort, Kalil, and Gottschalk, 1999, p. 81) are therefore a core component of their daily tasks (Dubois, 2010, p. 2; Hasenfeld and Steinmetz, 1981, p. 84 f.; Rothstein, 1985, p. 151 f.). As the dimensions of administrative language were derived from interviews with German public servants, we also employ this first empirical test in the context of Germany. Overall our focus on a German welfare service domain limits the general application of our findings regarding other welfare regimes (Esping-Andersen, 1990) and more confrontational types of encounters, as discussed in greater detail below.
In the experimental setting, participants evaluate a hypothetical public service encounter where elements of spoken administrative language are systematically varied according to the taxonomy presented above.5 Survey participants visited a website, where they listened to an audio recording and subsequently answered questions. In the recording, participants were asked to imagine a situation in which they visited a local social office to clarify their eligibility for a COVID-19 related social benefit in the form of a rent allowance. The audio recording then entailed a fictional conversation with a narrator (introduction and citizen perspective, male voice) and a civil servant who answered in the direct language (female voice). For instance, the narrator voice said “Once you have outlined your request, the official explains to you under which circumstances you may receive the additional payments to your rent.” Respondents then could hear the official responding “Well, it is regulated as follows: [ . . . ].”
In these conversations, we varied systematically the components of the bureaucratic language contained in our theory as vignette treatments. More precisely, for each of the seven communicative practices, there is a position in the hypothetical conversation where this practice was either emphasized or de-emphasized. To give an example, we operationalized the treatment “highly relational language” by including language that contained more casual and empathic communicative practices (emotionality) and more explicit statements of state support and availability (complaisance). We applied the same procedure for “highly informational” language. For the H3 treatment, “highly informational and relational language,” all communicative practices were emphasized. Overall, this entails four vignettes: One for the control group (no treatment), and one each for the three language treatments. Given that we also included the service decision as a control variable (statement at the end of the conversation that the applicant was likely eligible for the service or likely ineligible), we obtain a total of eight experimental groups: Four with a positive and four with a negative service decision. Therefore, while we operationalized and varied all four dimensions and seven communicative practices in the experimental encounter, the empirical assessment of their effects on citizen satisfaction is examined only at the level of the informational and relational components.
We included a manipulation check to confirm that the participants’ assessments of the experimental language treatments were in line with our expectations (Meier and Funk, 2017, p. 47). In the survey, we therefore added four questions to determine whether participants indeed perceived highly pronounced relational administrative speech acts as more personal and helpful (dimensions emotionality and complaisance). And we measured how well they understood the regulatory context and content of the speech (dimensions comprehensibility and reification) as a consequence of highly pronounced informational language (see Supplementary Appendix 5 for survey questions).
Factorial survey experiments always entail a certain artificial character. Results may be biased due to experimenter demand effects or social desirability (Auspurg and Hinz, 2015, p. 73). Thus, we installed several mechanisms to increase the authenticity of the experiment. First, we disguised the real intention of the experiment by providing only broad information on the project.6 Second, we realized a pure between-subject design with respondents only listening to one conversation (Charness, Gneezy, and Kuhn, 2012, p. 1). Third, a female voice was selected to embody the frontline worker as several interviewees from jobcentres and social offices suggested that frontline officials in local welfare administrations in Germany are predominantly female.7 Last, the service occasion was highly realistic given the COVID-19 pandemic at the time. As a strictly means-tested social assistance program addressing needy populations, the concrete scenario of a hypothetical rent allowance was highly plausible given the German welfare state regime (Seeleib-Kaiser, 2016).
Prior to listening to the audio file, respondents had to pass an instructional manipulation check to ensure that they read survey questions and items carefully (Oppenheimer, Meyvis, and Davidenko, 2009). A second check was placed right after the experimental encounter. It was followed by the measure of the main dependent variable “citizen satisfaction” that we introduced through the question All in all, how satisfied are you with the bureaucratic encounter? with possible answers ranging from one to ten (1 = not satisfied at all, 5 = partly satisfied, 10 = very satisfied). Subsequently, the more specific perception measures serving as manipulation checks followed. We placed the remaining control variables at the end of the questionnaire (see Supplementary Appendix 5 for all questions).8
Sampling and Experimental Group Assignment
The sample of survey participants is representative of the German population and was recruited from an online access panel administered by Respondi (N = 1,402) in January 2021.9 As each participant evaluated only one vignette, the experimental groups are congruent with the vignettes so that eight experimental groups were formed. The sampling procedure was fully randomized to ensure that the group assignment was independent of potential background variables. Therefore, each of the 1,402 participants was randomly assigned to one of the eight groups/ audio recordings.10Figure 2 provides an illustration of this procedure and the final distribution. Because of the randomized procedure, the group size is uneven. However, this is unproblematic because the group composition turned out to be highly similar regarding the participants’ socio-structural characteristics and potentially confounding variables (see Supplementary Appendix 8). Some attrition occurred as we had to remove 130 respondents from the analysis because the time they spent listening to the experimental encounter fell below the length of the audio recording, indicating that they had not listened to it completely. The number of cases for the analysis is therefore limited to 1,272. However, this should not create any biases since these dropouts turned out to be well-balanced across the experimental groups (see Supplementary Appendix 8) and did not exhibit any significant differences across the language treatments (see Supplementary Appendix 12).
Results
Table 1 reports how participants’ specific assessments of administrative language match the values of the independent variables, which we included as a manipulation check. Findings show that the different groups differ in how they perceived the communication in line with the treatment to which they were exposed; participants presented with an encounter that entailed highly pronounced informational language exhibited significantly higher perception measures than those in the control group. They assess the bureaucrat’s utterances as more comprehensible (perceived comprehensibility) and show higher understanding of her actions and decisions (perceived reification). The same effect prevails for relational administrative language that combines the dimensions of emotionality and complaisance: Here, participants report significantly higher ratings when evaluating how important the citizen’s individual situation was to the bureaucrat during the experimental encounter (perceived emotionality) and how well the official advocated for his or her personal concerns (perceived complaisance). We therefore conclude that each component of administrative language has a distinct effect on citizen perceptions as intended by the experimental manipulation.
Independent Variables/Treatment . | Dependent Variables . | |||
---|---|---|---|---|
Perceived Comprehensibility . | Perceived Reification . | Perceived Emotionality . | Perceived Complaisance . | |
Administrative language a | ||||
Highly informational | 1.08 [0.00] (0.12) | 0.64 [0.00] (0.19) | 0.71 [0.00] (0.20) | 0.74 [0.00] (0.19) |
Highly relational | 0.58 [0.00] (0.17) | 1.18 [0.00] (0.18) | 2.07 [0.00] (0.19) | 2.18 [0.00] (0.18) |
Adjusted R2 | 0.06/ 0.02 | 0.02/ 0.06 | 0.02/ 0.16 | 0.02/ 0.18 |
N | 646/ 659 | 650/ 662 | 648/ 661 | 649/ 661 |
Independent Variables/Treatment . | Dependent Variables . | |||
---|---|---|---|---|
Perceived Comprehensibility . | Perceived Reification . | Perceived Emotionality . | Perceived Complaisance . | |
Administrative language a | ||||
Highly informational | 1.08 [0.00] (0.12) | 0.64 [0.00] (0.19) | 0.71 [0.00] (0.20) | 0.74 [0.00] (0.19) |
Highly relational | 0.58 [0.00] (0.17) | 1.18 [0.00] (0.18) | 2.07 [0.00] (0.19) | 2.18 [0.00] (0.18) |
Adjusted R2 | 0.06/ 0.02 | 0.02/ 0.06 | 0.02/ 0.16 | 0.02/ 0.18 |
N | 646/ 659 | 650/ 662 | 648/ 661 | 649/ 661 |
Dependent variables scaled from 1 to 10.
a The category control groups builds the baseline for highly informational and relational language, respectively.
Unstandardized regression coefficients depicted, p-values in squared brackets, SE in parentheses.
Independent Variables/Treatment . | Dependent Variables . | |||
---|---|---|---|---|
Perceived Comprehensibility . | Perceived Reification . | Perceived Emotionality . | Perceived Complaisance . | |
Administrative language a | ||||
Highly informational | 1.08 [0.00] (0.12) | 0.64 [0.00] (0.19) | 0.71 [0.00] (0.20) | 0.74 [0.00] (0.19) |
Highly relational | 0.58 [0.00] (0.17) | 1.18 [0.00] (0.18) | 2.07 [0.00] (0.19) | 2.18 [0.00] (0.18) |
Adjusted R2 | 0.06/ 0.02 | 0.02/ 0.06 | 0.02/ 0.16 | 0.02/ 0.18 |
N | 646/ 659 | 650/ 662 | 648/ 661 | 649/ 661 |
Independent Variables/Treatment . | Dependent Variables . | |||
---|---|---|---|---|
Perceived Comprehensibility . | Perceived Reification . | Perceived Emotionality . | Perceived Complaisance . | |
Administrative language a | ||||
Highly informational | 1.08 [0.00] (0.12) | 0.64 [0.00] (0.19) | 0.71 [0.00] (0.20) | 0.74 [0.00] (0.19) |
Highly relational | 0.58 [0.00] (0.17) | 1.18 [0.00] (0.18) | 2.07 [0.00] (0.19) | 2.18 [0.00] (0.18) |
Adjusted R2 | 0.06/ 0.02 | 0.02/ 0.06 | 0.02/ 0.16 | 0.02/ 0.18 |
N | 646/ 659 | 650/ 662 | 648/ 661 | 649/ 661 |
Dependent variables scaled from 1 to 10.
a The category control groups builds the baseline for highly informational and relational language, respectively.
Unstandardized regression coefficients depicted, p-values in squared brackets, SE in parentheses.
The manipulation check also yields that the perceptions of the two components of administrative language are not fully independent of one another. As shown in Table 1, an increase in a highly relational language is also associated with an increase in perceived comprehensibility and reification, although to a much smaller extent. Likewise, an increase in a highly informational language is associated with an increase in perceptions of emotionality and complaisance.11 Overall, while the manipulation check confirms that the experimental treatment worked, it also indicates that changes in either component of administrative language influence citizen perceptions more broadly than expected for each treatment individually.
Our main findings on the satisfaction ratings reported by participants are summarized in Figure 3. The descriptive illustration (see also Appendix 9) shows the distribution of mean satisfaction values among the eight experimental groups. The figure gives a first indication that the differences in the treatments caused systematic differences between groups: Firstly, as expected, satisfaction values are systematically higher for all groups with a positive service decision compared to negative outcomes. Secondly, for each four groups with the same outcome decision, respondents from the control group reported lower satisfaction values than any of the other groups with language treatment. Thirdly, for each four groups with the same outcome decision, respondents from the group with full language treatment show the highest satisfaction values (in line with hypothesis 3). In fact, satisfaction values are on average around 1.5 points higher than in the control groups, which is a sizable difference.
Overall, the increase in satisfaction values between groups seems to be driven more by the relational dimensions of administrative language than the informational dimensions. For informational language, the difference between treatment and non-treatment is 0.34 units on average. For relational language, the difference is 1.25 units on average. A simple t-test confirms that compared to the control groups, only the groups with relational language and both informational and relational language are significantly different, but not the groups with informational treatment alone (Appendix 10). This indicates that the data yield support for hypothesis 2, but not for hypothesis 1. An additional observation is that the mean satisfaction value for highly relational and informational language and negative service decision is higher (6.88) than with less pronounced language and a positive service decision in the control group (6.25). In our experiment, speech even seems to substitute for a negative service decision.
Next, Table 2 reports the results of the OLS regression analysis which includes potentially confounding variables.12 Findings confirm that in comparison to the control groups, the groups with relational language treatments yield statistically significant effects, but not the groups with the informational language treatment. We therefore maintain hypothesis 2 but the findings do not allow us to maintain hypothesis 1. Given the more pronounced effect for the two groups with full language treatment, findings also support hypothesis 3.13,14
Independent Variables . | Binary Model . | Full Model . |
---|---|---|
Administrative language a | ||
Highly informational | 0.26 [0.15] (0.18) | 0.25 [0.15] (0.17) |
Highly relational | 1.25 [0.00] (0.18) | 1.27 [0.00] (0.17) |
Highly informational and relational | 1.55 [0.00] (0.18) | 1.59 [0.00] (0.17) |
Positive service decision | 0.95 [0.00] (0.12) | |
Social benefit receipt | 0.33 [0.04] (0.16) | |
Recent encounter evaluation | 0.15 [0.00] (0.02) | |
Institutional trust | 0.26 [0.00] (0.03) | |
Federal government support | −0.15 [0.26] (0.13) | |
Adjusted R2 | 0.07 | 0.21 |
N | 1,271 | 1,230 |
Independent Variables . | Binary Model . | Full Model . |
---|---|---|
Administrative language a | ||
Highly informational | 0.26 [0.15] (0.18) | 0.25 [0.15] (0.17) |
Highly relational | 1.25 [0.00] (0.18) | 1.27 [0.00] (0.17) |
Highly informational and relational | 1.55 [0.00] (0.18) | 1.59 [0.00] (0.17) |
Positive service decision | 0.95 [0.00] (0.12) | |
Social benefit receipt | 0.33 [0.04] (0.16) | |
Recent encounter evaluation | 0.15 [0.00] (0.02) | |
Institutional trust | 0.26 [0.00] (0.03) | |
Federal government support | −0.15 [0.26] (0.13) | |
Adjusted R2 | 0.07 | 0.21 |
N | 1,271 | 1,230 |
Dependent variable citizen satisfaction with the public service encounter scaled from 1 to 10.
a The category control group builds the baseline for all other categories.
Unstandardized regression coefficients depicted, p-values in squared brackets, SE in parentheses.
Independent Variables . | Binary Model . | Full Model . |
---|---|---|
Administrative language a | ||
Highly informational | 0.26 [0.15] (0.18) | 0.25 [0.15] (0.17) |
Highly relational | 1.25 [0.00] (0.18) | 1.27 [0.00] (0.17) |
Highly informational and relational | 1.55 [0.00] (0.18) | 1.59 [0.00] (0.17) |
Positive service decision | 0.95 [0.00] (0.12) | |
Social benefit receipt | 0.33 [0.04] (0.16) | |
Recent encounter evaluation | 0.15 [0.00] (0.02) | |
Institutional trust | 0.26 [0.00] (0.03) | |
Federal government support | −0.15 [0.26] (0.13) | |
Adjusted R2 | 0.07 | 0.21 |
N | 1,271 | 1,230 |
Independent Variables . | Binary Model . | Full Model . |
---|---|---|
Administrative language a | ||
Highly informational | 0.26 [0.15] (0.18) | 0.25 [0.15] (0.17) |
Highly relational | 1.25 [0.00] (0.18) | 1.27 [0.00] (0.17) |
Highly informational and relational | 1.55 [0.00] (0.18) | 1.59 [0.00] (0.17) |
Positive service decision | 0.95 [0.00] (0.12) | |
Social benefit receipt | 0.33 [0.04] (0.16) | |
Recent encounter evaluation | 0.15 [0.00] (0.02) | |
Institutional trust | 0.26 [0.00] (0.03) | |
Federal government support | −0.15 [0.26] (0.13) | |
Adjusted R2 | 0.07 | 0.21 |
N | 1,271 | 1,230 |
Dependent variable citizen satisfaction with the public service encounter scaled from 1 to 10.
a The category control group builds the baseline for all other categories.
Unstandardized regression coefficients depicted, p-values in squared brackets, SE in parentheses.
Note that the strength and the significance of the regression coefficient stay nearly unchanged once the service decision and the other controls are included in the model which becomes clear as we compare the coefficient of all dimensions-categories across the binary and the full model. Likewise, as the standard error is only moderate in size we can be confident that the effect of administrative language is not spurious.
Discussion
As discussed, we took several steps in designing the experiment to ensure that it would deliver valid results. Nonetheless, some limitations remain. The first limitation links to the way in which we operationalized the communicative practices when we developed the dialogue. It is possible that some specific formulations or notions in the dialogue drive the results. At the same time, the manipulation check confirms that participants perceived variation in language as intended. Therefore, it remains plausible that our results are indeed caused by the different communication styles.
We should further consider as a possible limitation that we employed the voice of a female frontline official in the experimental encounter. We chose a female voice to offer a realistic scenario, given that there are on average more women employed in German local welfare administrations. However, research on voice-pitch suggests that there are significant differences between the auditive characteristics of language used by women and men which in turn may have an impact on the perception of speakers (Dietrich, Hayes, and O’Brien, 2019). Citizens might therefore assess administrative communication differently once they are interacting with a male frontline worker and for which our design cannot account.
Another limitation is that we did not study variation in administrative language at the level of the four dimensions of the taxonomy, but only at the level of the informational and relational components. This had practical reasons because an increase in the number of treatment conditions would also have required a much larger sample size.15 Future studies could replicate the experiment at the level of dimensions. From a conceptual perspective, however, we expect that not all logically possible combinations of dimensions do exist in practice. For instance, it is implausible to assume that an official imparting support and availability in his or her communication is not also empathic.
Moreover, despite our best efforts to make the experimental encounter as authentic as possible, we need to bear in mind that the setting is somewhat static while in reality, each face-to-face interaction constitutes a “circular process” where the participants’ communicative reactions are in a reciprocal fashion interwoven with one another (Bartels, 2015, p. 10). As such, public service encounters exhibit a highly situational character. There might be characteristics of frontline communication which emanate from the situation at hand and therefore also influence the citizen view on the encounter but are impossible to capture in an experimental setting where each participant evaluates an audio recording and does not take part actively in the communication. Furthermore, the real-world implications of a conversation with a bureaucrat are less perceptible for citizens in an artificial context. This might explain why participant perceptions of reifying language in our manipulation check were somewhat less pronounced than expected, as discussed above. The artificial context might also explain why we did not find highly informational language to have a more pronounced impact on satisfaction. Our findings therefore serve as a first confirmation of the causal link between administrative language and citizen satisfaction, but more research is needed to document the precise nuances.
Last, the specific empirical focus of the experiment necessarily compromises the external validity of our findings. On the one hand, our research covers Germany, a country that traditionally seeks to “encourage citizens to get involved in government directly” (Huxley et al., 2016, p. 397). Communicative practices and their impact on citizens may be somehow different in other countries with other administrative traditions (Kuhlmann and Wollmann, 2019).
Furthermore, we are looking at one particular (fictional) welfare service that determines a rather unique context for the interaction between citizens and bureaucrats. A different communication between citizens and public servants takes place in the welfare office then, for instance, during a police traffic check where officials are rather concerned about law enforcement than the individual support of citizens (Jensen, 2018). Apart from this, the importance of a given service for the recipients should be also relevant. In areas such as public education, health, or welfare, where encounters with the state have substantial consequences for people’s lives, empathic and personal communication is probably more salient than in the case of passport or residence registration matters where citizens might especially value plain and understandable language to keep the encounter as fast and uncomplicated as possible. Furthermore, some frontline officials might work in a context that does not allow them to deploy the full range of the communication styles in a meaningful way. For instance, imparting availability presupposes that frontline workers have the resources to be present and available for their clients.
Conclusion
In this paper, we asked about the consequentiality of variation in administrative language for citizen perceptions, in particular their satisfaction with a public service encounter. Bearing in mind some limitations linked to the experimental method, we were able to show that variation in language used by officials during an encounter has an independent effect on citizen perceptions. With the experimental method, we were also able to pin down the nature of this effect; we found that administrative speech acts that are more pronounced in their informational nature (highly comprehensible and reifying) have no significant impact on citizen satisfaction values. By contrast, communication styles that are more pronounced in terms of their relational component, or in both their relational and informational component have a significant effect on citizen satisfaction. This effect is independent of whether the service decision is positive or negative. Our findings should apply to a broad range of service encounters in the welfare domain, but more research is needed to test whether they are also relevant for other state functions (such as more regulatory encounters).
Our findings are of significant bearing for the practice and study of bureaucratic encounters, which appears to be emerging slowly as a distinct research subject (Bartels, 2015; Bruhn and Ekström, 2017; Hand and Catlaw, 2019; Kasdorp and Schakel, 2022; Raaphorst and Loyens, 2020). In validating our earlier taxonomy of administrative language (Eckhard et al., 2022), we show that the content of spoken administrative language matters and is consequential for citizen perceptions. Of particular relevance is the finding that bureaucrats’ relational communication matters most for citizen evaluations. This is opposed to the dominant focus among practitioners who focus mostly on the informational dimensions of language when considering how to improve citizen communication (e.g. Müller, 2014). The insights from this study suggest that both practitioners and researchers should devote more attention to the relational aspects of citizen–state interactions.
The findings we have presented are also relevant to other public administration literature that have service encounters as a component of their causal chain, such as literature on street-level bureaucracy, representative bureaucracy, citizen trust, and administrative burden. Street-level bureaucracy literature has so far neglected the question as to which factors drive the citizens views on their encounters with bureaucrats (de Boer, 2020; Hansen, 2021; Nielsen et al., 2021). Representative bureaucracy literature often observes distributed effects of service decisions (who gets what in society) without observing citizen interactions directly (Grimmelikhuijsen et al., 2017; Kingsley, 1944; Meier, 1993; but see, Guul, 2018; McCrea, 2021; Vinopal, 2017). Citizen trust literature may benefit by recognizing communicative features of the encounter as a factor for the long-term relationship between citizens and the state (Denhardt and Denhardt, 2007; Hansen, 2021; Kampen et al., 2006; Rothstein, 2009; Van Ryzin et al., 2004; Yackee and Lowery, 2005). And administrative burden literature, which explicitly acknowledges the relevance of written administrative language as a barrier to access, might benefit by also considering the role of spoken language (e.g. Herd and Moynihan, 2019). The latter is particularly important because whenever citizens do not understand the content of a document, such as a form or a letter, their only alternative route for access is to make a personal appointment or take up the phone.
Our recommendation for future research is straightforward: as the findings from the experiment indicate that language use matters for the citizen assessment, it should be promising to observe empirically real dialogues taking place during public service encounters. Such examinations could also build directly on the taxonomy used in this paper, which has proved valuable when measuring and testing assumptions about administrative language use in real-world settings.
Acknowledgements
We are grateful for helpful comments and suggestions from the editor Mary K. Feeney and three anonymous reviewers. Previous versions of this paper have been presented at the IRSPM conference 2022, where this paper won the Christopher Pollitt prize for best conference paper, the annual meeting of the section “Policy Analysis and Public Administration Research” of the German Political Science Association (DVPW) 2021, the workshop “Street-Level Bureaucrats in Times of Change” organized by Tanja Klenk, Nissim Cohen, Maayan Davidovitz, and Sarah Cardaun in October 2021 as well as at a miniconference on “Social Equity and Public Administration: The Behavioral Perspective” organized by Kenneth J. Meier, Sebastian Jilke, and K. Jurée Capers in April 2021. We received instructive and valuable feedback during all of these occasions which we gratefully acknowledge. We also thank Alexa Lenz for providing helpful advice on the experimental design and the data analysis. Moreover, we benefitted from the support of our colleagues at the interdisciplinary research project “Inequality in Street-Level Bureaucracy: Linguistic Analysis of Public Service Encounters” at the University of Konstanz, which the paper is a product of. Lastly and most importantly, we would like to express our highest gratitude to Hannah Riemann for her outstanding vocal performance as “the social office employee”.
Conflict of Interest
The authors have no conflict of interest to declare.
Funding
This article is a product of the research project “Inequality in Street-Level Bureaucracy: Linguistic Analysis of Public Service Encounters” at the University of Konstanz which is funded by the German Research Foundation (DFG—Deutsche Forschungsgemeinschaft) under Germany‘s Excellence Strategy—EXC-2035/1 – 390681379.
Data availability statement
The data underlying this article and the R-code used for the analysis are available at https://doi.org/10.7910/DVN/JBTY6V. Online Appendix available via the supplementary data section on the publication webpage: https://academic.oup.com/jpart/advance-article/doi/10.1093/jopart/muac052/6891178.
References
Footnotes
A total of 64 interviews were conducted in Germany between April and November 2020. The sample consisted of different types of street-level bureaucrats delivering a broad range of public services. By this, we aimed at identifying general patterns of frontline communication that are not exclusively limited to some specific service domains. Interviewees were employed in local social service offices (Sozialämter), jobcentres (Jobcenter), migration offices (Ausländerbehörden/ Integrationsbüros), citizen offices (Bürgerämter/Bürgerbüros), building and environment offices (Bau- und Umweltbehörden) and staff units dealing with internal and external communication. Interviews followed a semi-structured questionnaire to extract authentic and vivid accounts from the frontlines of public administration (Bogner and Menz, 2009; Brinkmann, 2018; Honer, 1994). Although we did our best to avoid biases, we should emphasize that interviews may have been affected by the COVID-19 pandemic at the time and there may also be selection biases in that only those officials participated who were inclined to reflect on their communication patterns. At the same time, it is important to consider that the interviews did not aim at establishing whether and to what extent frontline public servants deploy certain communication style but to explore the variety of how such communication could look like. Following best practices in explorative research methods, our exploration ended once the point of saturation had been reached (Saunders et al., 2018).
All quotes translated by authors. Every quote is marked with the corresponding ID of an interviewee. The IDs are indicated in the interview list in Supplementary Appendix 1 so that the quotes can be linked with interviewee positions.
Note that with hypothesis 3 we do not imply that the effects of the informational and the relational components should add up or interact.
Since our experimental design includes a fictional public service, participants were barely able to form certain expectations towards it. Therefore, we decided to rely on the evaluations of their last real-world bureaucratic encounter as a proxy for their expectations. Eventually, these evaluations rest on different kinds of public service encounters our study participants have gone through.
Supplementary Appendix 4 includes the experimental service occasion as presented to the participants and the transcripts for all complete audio-recordings with all possible factor combinations (vignette universe).
Respondents were only told that the research project dealt with interactions between public officials and citizens, but not specifically with language or communication.
However, there is no comparative data on local welfare administrations in Germany available that allows us to verify this notion. The decision to employ a female voice comes with a limitation as discussed below.
Questions on gender, age, and regular receipt of social benefits had to be placed at the beginning as they were needed to steer the representativeness of the experimental sample.
Respondi hosts an online access panel consisting of roughly 340,000 people who are paid for participating in surveys. Relying on access panel data has become increasingly common in social sciences as the quality of data has been shown to be high in terms of reliability and not significantly different from traditional ways of sampling such as phone or door-to-door surveys (for instance, James, 2011). For detailed information on how Respondi manages its access panel and keeps the reliability and trustworthiness of its members high, see https://www.respondi.com/access-panel.
Often researchers deploy additional randomization techniques to avoid the experimental groups being just by chance not alike, such as block random allocation or cluster randomization (James, Jilke, and Van Ryzin, 2017a, p. 75). We refrained from such procedures as the sampling process lasted roughly one week and was associated with fixed quotations for certain groups of participants to ensure that the sample was representative. Alternative sampling procedures would have risked that the group formation would have hampered a functioning randomization.
Table 1 also shows that the effect on perceived reification is less pronounced than for emotionality and complaisance; whereas the expectation was that it should be higher. But note that reification captures the extent to which an official’s information provision causes clients to develop a “better understand” of the official’s behavior and decisions. We expect that since our experimental setting did not include longer question-answer sequences, there was not enough input for participants to perceive the distinctive impact of reifying communication in a more pronounced manner. Other than that, the value is significantly higher than in the groups without treatment, which confirms that the manipulation worked.
For the demographic variables (gender, age, education, and political party affiliation) we did not find statistically significant effects on the participants’ evaluations of the experimental encounter.
With H3 we postulated the combined effect of highly informational and relational language to be more pronounced than the effect of either of these components alone. Comparing the significance and strength of the regression coefficients across these categories therefore seemed an appropriate analytical procedure to test this hypothesis. To inspect the combined effect more closely, an alternative strategy would be to check the interaction between highly informational and highly relational language. However, we employed only a small fraction of the vignette universe—eight combinations out of 256 possible combinations—so the data analysis structure does not allow for the inclusion of interaction terms. Below, we discuss the limitations of our experimental design in more detail.
One-way analyses of variance (ANOVA) for each of the experimental language treatments further confirm the findings from the regression models (Appendix 11).
As the experimental setting entails eight dichotomous factors the complete vignette universe would cover 256 possible factor combinations. Therefore, the required sample size to test the effects of all communicative practices would have been quite large.