Public servant stereotypes: It is not (at) all about being lazy, greedy and corrupt

Correspondence Jurgen Willems, Institute for Public Management & Governance, Vienna University of Economics and Business, Welthandelsplatz 1, Vienna 1020, Austria. Email: jurgen.willems@wu.ac.at Abstract As stereotypes strongly influence social interactions, this study explores the stereotypical associations regarding public servants, and about various professions in the public sector as well as the for-profit and nonprofit sectors. This leads to a better understanding of the theoretical and practical challenges, such as citizen behaviour towards public servants, attractiveness of and political decisions about public service jobs. With a mixed-method analysis of cognitive associations (7,470 associations by 415 respondents for 12 professions), the defining epithets of public servants are clarified, along with their positive or negative connotation. Despite the strongest associations for public servants being positive (caring, helpful and dedicated), as an overall category, it has a less positive connotation compared to some specific professions typical in the public sector (nurse, firefighter and police). However, cognitive associations are substantially more positive for public servants compared to politicians, lawyers and salesmen.

public sector are characterized by strong stereotypical elements, which might influence the attractiveness of particular jobs, rather than the attractiveness of the sector as a whole (Van de Walle 2004).
Stereotypes are cognitive schemes that people have which in an abstract way summarize the typifying characteristics of groups of people in society (Hilton and von Hippel 1996). These cognitive schemes are at the basis of processing information about actual people, and in turn influence decisions on how to behave in relation to these people (Bodenhausen 1990; Kunda and Thagard 1996). As a result, stereotypes have been dealt with from various perspectives in various areas of social sciences, focusing on gender, race, nationalities and/or age groups (Steele and Aronson 1995;Eckes 2002;Ellemers 2018;Tresh et al. 2019). Also, in the literature on professional image and individual reputations, stereotypes are conceptualized and analysed to better understand human interactions among and with employees, subordinates, leaders, customers and citizens (Vigoda 2000;Roberts 2005;Uhlmann et al. 2013).
Moreover, when stereotypes are persistently negative and do not resemble reality, the target group might be dealt with unfairly or unequally (Steele and Aronson 1995;Hoobler et al. 2009;Tresh et al. 2019).
As a consequence, in the context of public servants, it is necessary to get a better insight into the detailed content of these stereotypes-that is, in the summarizing associations made about them-and about the extent to which these associations are negative, neutral or positive. In doing so, cognitive schemes are documented, which are at the basis of the behaviour towards public servants. As a result, rather than assuming that certain associations are consistently made about public servants, it remains relevant to document the content and negative or positive connotation of stereotypes held by citizens. From a theoretical perspective, it can provide an insight into potential sources of stereotypical association and the extent to which job categories and professions are perceived as clearly distinct. This can in turn be relevant for assessing the generalizability of context dependency of theoretical consideration on a variety of public servant studies. Moreover, when public servants confirm certain stereotypical traits with their actual behaviour, it is inherent to the associative process that other stereotypical associations will be drawn about them, regardless of their factual accuracy (Ashton and Esses 1999;Diekmanf et al. 2002;Ben-Zeev et al. 2005;Spencer et al. 2016). An insight into co-associated characteristics about public servants can thus clarify what might be additionally assumed about public servants if their behaviour conforms to a selection of stereotypical traits (Bodenhausen 1990).
Therefore, with a sample of 7,470 direct associations for 12 professions collected from 415 US respondents, this study clarifies several of these job stereotypes by testing the distinctiveness of the overall public servant stereotype from more profession-related stereotypes, such as firefighter, police officer or nurse. Similarity and distinctiveness are evaluated both for the content of stereotypes and for positive or negative connotations. Moreover, in the empirical analysis, this comparison is made between the overall, summarizing public servant category on the one hand, and on the other, a range of specific professions that are found in the public as well as in the for-profit and/or nonprofit sectors (e.g., teacher, professor, lawyer, banker, social entrepreneur). This cross-sector comparison additionally allows evaluation of the overall and specific sub-categories of the public servant stereotype against a broader spectrum of professions, in turn providing a richer insight into the public servant stereotype.
From a theoretical perspective, this study builds on insights from stereotype-focused theories. In particular, the representativeness heuristic is applied in the public sector context, given its crucial role in stereotype thinking. This heuristic has been dealt with extensively in dual process theory (Tversky and Kahneman 1983;Bodenhausen 1990) and ecological rationality theory (Gigerenzer 1991;Hertwig and Gigerenzer 1999;Smith 2003;Gigerenzer and Brighton 2009). Both theories start from the use of heuristics in human cognitive processes to form perceptions about people, and thus also about public servants. However, both theories have different inherent assumptions on the usefulness and/or harmfulness of heuristic-based thinking (Gigerenzer 1991;Gigerenzer and Brighton 2009).
This study does not aim to contribute to the discussion on this divide between these theoretical perspectives, but builds on the common denominator in both theories, in order to obtain a holistic view of stereotype thinking about public servants and some related, but more specific, professions. By combining both the theoretical perspectives on stereotypes, and public servant stereotypes in particular, a deeper insight can be gained into the common associations that are made for public servants, and for specific professions.
From a practical perspective, this study provides important empirical evidence to a range of contemporary challenges in public services, such as managing citizen behaviour towards public servants, attractiveness of public service jobs and political decisions about public service jobs. For example, the empirical approach in this study verifies several recommendations made by Van de Walle (2004), who describes how a negative perception towards the public sector (e.g., due to persistent red tape and bureaucracy), amplified by archetypical representations of public servant professions in popular media and movies, might negatively affect the attractiveness of particular public servant jobs. Confirming significant differences between professional stereotypes and sector-related stereotypes would endorse the practical recommendation that public jobs can and should be promoted based on profession characteristics, rather than sector characteristics.

| THEORETICAL BACKGROUND
Stereotypes are shared social cognitions about the typifying characteristics of social entities, such as particular groups in society (LaViolette and Silvert 1951). They have often been studied from a general societal perspective, by focusing, for example, on ethnic and minority characteristics (e.g., Ashton and Esses 1999), or on gender-specific characteristics in the overall workforce (e.g., Diekmanf et al. 2002;Leach et al. 2017).

| Professional stereotypes
With respect to professional stereotypes, the 'ideal worker' stereotype has been studied to develop a better understanding of stereotype thinking within and across organizations (Roberts 2005). For example, when people in their role as employees conform to certain stereotypical characteristics in their behaviour, it triggers associative cognitive processes among their colleagues, supervisors, customers or citizens, which in turn initiate associative cognitive processes in which other stereotypical characteristics might be attributed to those specific employees (Uhlmann et al. 2013). Hence, stereotypes can work in both a positive and a negative way and play a major role in building professional reputations (Roberts 2005;Coffman 2014;Dumas and Sanchez-Burks 2015).
From the perspective of the perceiver, stereotypes-or at least the underlying associative processes-are an inherent human way of processing and structuring new information about a person. Rather than remembering the exact pieces of information, and only those pieces of information, new experiences are compared and framed within existing mental frameworks, enabling the human mind to make quick and efficient decisions (Bodenhausen 1990;Smith 2003;Todd and Brighton 2016). Hence, stereotypes and their use for quick decisions are a particular type of decision heuristics, which has been studied extensively in dual process theory and in ecological rationality theory.
For example, the representativeness heuristic evaluates the extent to which new information about a person is similar to an existing stereotypical image (Tversky and Kahneman 1983;Bodenhausen 1990). When information about a person confirms this stereotype, the representativeness heuristic enables the classification of (the information about) this person with little mental effort, within a broader, existing mental framework (Feldman 1981;Hertwig and Gigerenzer 1999). However, due to this cognitive framing process, the probability is high that other stereotypic characteristics are attributed to the specific person, even in the absence of any initial information on these characteristics (Bodenhausen 1990).
Academic and practitioner debates have focused intensively on several normative aspects of stereotype-based thinking. One of these aspects is when stereotypical associations are consistently negative and persistent through time, for example about race, gender or national origins, and when such stereotypical associations lead to unfair or unequal treatment of certain groups in society. However, the origin of strong persistent associations is not necessarily the inability of the human brain to reconsider whether a single person has the negative characteristics associated with certain stereotypes (Kunda and Thagard 1996), but the strong social influence from others in setting reference frameworks for stereotype thinking (Arkes and Tetlock 2004). As a result, persistent negative stereotypes can have substantial disadvantages for the respective target groups.
For instance, the holders of strong stereotypical associations about certain groups might perceive the defining characteristic of these groups-which they do not have themselves-as more negative and might assume stronger correlations with (other) negative traits (Hilton and von Hippel 1996;Levy et al. 1998). Moreover, members of the groups for which negative stereotypes exist adjust their own behaviour when confronted with self-group prejudices and/or feel the need for acting in conformity with the stereotypes about the group, resulting in self-fulfilling prophesies, where in-group members consciously or unconsciously adjust their behaviour to the stereotype (Steele and Aronson 1995;Hilton and von Hippel 1996). This can lead to lower levels of performance in a professional context (Steele and Aronson 1995;Kunda and Thagard 1996;Eckes 2002;Tresh et al. 2019).
In the specific context of public servants, negative consequences of inaccurate stereotypes relate to how citizens interact with and evaluate the performance and professionalism of public servants. Moreover, when negative associations are made about people based on the sector of employment, people might face difficulties changing jobs to other sectors, and/or public sector organizations might not be able to attract good employees, given the reputational loss they might face as a result of public sector employment (Peiffer et al. 2018).

| Sources of stereotypical associations about public servants
Stereotypes about public servants might thus be fuelled by the overall social construction in society of the role of public services and servants, rather than the sum of personal and real experiences. In popular media and movies, public servants (who are not the protagonist of a story) are often represented as lazy, risk-averse and/or corrupt Particular traits of (groups of) public servants might be at the origin of stereotypical associations that citizens conceive about them. For example, a substantial body of literature comparing public servants with employees in the for-profit and nonprofit sectors has shown that public servants can differ substantially with respect to perceptions about job content, work-life balance, job performance, job satisfaction and motivation (Goulet and Frank 2002;Ben-Ner et al. 2011;Willems 2014). These differences from other sectors might be the result of self-selection by prosocial, motivated people in this sector, or by socialization when working with like-minded colleagues (Becker and Connor 2005;De Cooman et al. 2011). As a result, the well-documented differences with respect to motivation-for example, in the literature on public service motivation-can shed light on commonly made associations about public servants as a result of common traits (Kim et al. 2013;Schott et al. 2019).
However, associations made about public servants might also originate from other related entities and in particular from the public sector in general. For instance, strong perceptions about the public sector being bureaucratic, slow and inefficient (Pandey et al. 2007;Marvel 2016) might spill over to the perceptions about the people working in this sector. A substantial body of literature has documented citizens' trust in various levels of government and public administration (Albrecht and Travaglione 2003;Morgeson and Petrescu 2011). The inherent relatedness of public servants-working in and for these organizations and institutions-can thus be a source of similar types of associations. For example, on an abstract level, attitudes towards democracy explain people's trust in the government and in public sector reforms (Christensen and Laegreid 2003), while on a more concrete level, trust and attitudes towards particular services and their public nature explain citizen behaviour in particular contexts, such as public versus private transport (Beirão and Cabral 2007), e-government (Kolsaker and Lee-Kelley 2008), and windpower policies (Ek 2005). As a result, these organizational and political entities might be a source of stereotypical associations about public servants too. The public sector is often typified as being bureaucratic, slow and inefficient (Pandey et al. 2007;Marvel 2016), which might, in turn, influence the overall perceptions of citizens about public servants and about the political systems in which they operate (Vigoda 2000).

| METHOD
In order to gain an insight into (1) the cognitive associations made about public servants and some specific professions inside and outside the public sector, and (2) the negativity or positivity of these associations (i.e., the connotation), a mixed-method approach was adopted, in which the associations for a set of 12 professions were analysed.
This set was chosen for a broad representation of professions and occupations across all three sectors, that is, the public, for-profit and nonprofit. This enables comparison of the overall public servant category not only with its subcategories, but also with professions outside the public sector. The professions analysed in this study are the following: firefighter, social entrepreneur, police officer, teacher, lawyer, nurse, banker, politician, professor, salesman and volunteer. As a twelfth category, public servant is the main reference category for this study. Despite the fact that some of the jobs are particular types of the public servant category, the category was included as an overarching category that enables straightforward comparison with more specific professions.

| Survey design and variables
To explore the stereotypical characteristics of the various professions, word associations for these professions were solicited in a survey. For this inductive design, the study relied on a similar design adopted by Rozin et al. (2012), which explores and tests masculinity as a stereotypical characteristic of meat. In such association questions, respondents are asked to think of words they associate with a particular focal word, which were the 12 different categories in the context of this study. Respondents could, in addition, indicate whether they consider the associated words to be negative (coded as −1), neutral (coded as 0) or positive (coded as 1).
Given the research question of this study, this open associative approach has an important advantage compared to other methods that are often used to study stereotypical thinking, such as the stereotype content model and implicit association tests. In these methods, concrete words have been selected and are provided by the researcher to respondents, in order to assess quantitatively the way differences between respondents and/or cognitive entities (such as gender and race) result in different associative strengths for the words and dimensions provided by the researchers. In the method applied in this study, no a priori assumptions are made about the words and dimensions that matter to evaluate professional stereotypes. As a result, the open associative method of this study enables exploration of what these relevant words and dimensions are, from a content perspective too. This is important, in particular, for the specific research question of this study, and for the exploration of job stereotypes, as no extensive literature exists on most important stereotypical associations and dimensions for specific job categories. Moreover, due to the connotation score assigned by the respondents, an initial quantitative assessment can also be made, enabling comparisons between respondents and for different cognitive entities (i.e., professional categories in the context of this study).
Each respondent received a random selection of six out of 12 professions, to provide three associations per profession. These six questions were incorporated into a larger survey, which had a double purpose: the association questions for this study were alternated with Likert-scale constructs for an unrelated study that focused on pre-testing new and unrelated survey constructs. Likert scales and word association questions were alternated with each other, and the order of the survey questions was also randomized, resulting in a random sequence of Likert-scale questions and word association questions (with Qualtrics survey randomization functions). Hence, both types of questions served as distraction questions for each other, which was done to reduce a consistent bias in the data of Likert scales having an effect on word associations, or vice versa. Moreover, the alternation and randomization reduces potential biases with respect to survey acquiescence (Billiet and Davidov 2008). Demographic and control variables probed for gender, marital status, having children, age and political preference.

| Participants and data structure
Respondents in the US were addressed through the Qualtrics Panel service, and a respondent fee of $5 for about 15 minutes was paid (including administration costs). In total, 415 respondents completed the survey, and were each given a random selection of six out of the 12 categories. Each respondent could provide up to three words per profession. This resulted in a two-level data structure, of 7,470 associated words (accompanied with a code of −1, 0 or 1) for 12 categories, nested in 415 respondents. Descriptive statistics are given in Table 1. It should be noted that this sample is not representative of all US citizens. Nevertheless, it can also be assumed that the willingness of respondents to participate in an online questionnaire is not (too specifically) related to particular personal characteristics that in turn are highly related to completely different opinions on professions and sectors (at least these variables as control variables in the main analysis do not signal major differences). However, this shortcoming should be the topic of further and more large-scale verification in follow-up studies. All data are OpenAccess available at: https://osf.io/xky93/.

| Analysis
The data were analysed in two steps. First, a qualitative exploration of the words helped in understanding the specific epithets that typify professions and the overall public servant category. Word clouds per category offer an insight into their strongest epithets, while a Sankey plot analysis focused on the content relatedness of categories.
Second, in a multilevel multinomial regression analysis, the three categorical word codes (−1, 0 or 1) were explained based on profession-as a categorical variable-and on individual control variables. This means that the dependent variable is the three-category variable 'negative-neutral-positive'. The main independent variable is the dummy coded categorization of the professions. The public servant category is used as the reference category; so in fact, two times 11 hypotheses are tested, where each time the null-hypothesis states: This profession (e.g., firefighter) has the same proportion of positive (or negative) associations as the public servant category, with the neutral associations as reference category for the dependent variable. For each of these hypotheses, a significant deviation means that relatively more or less positive and/or negative words are associated with the profession. Due to the multilevel clustering of associations per respondent, this analysis also controls for respondent-specific effects, like his/her overall mental framework, or the fact that his/her associations on the first random category seen might have influenced further word associations. However, in this context, I also refer to the other actions taken, such as the random order and selection of professions per respondent and altering the association questions with other survey questions (Billiet and Davidov 2008).  (Willems 2020a). The advantage of using word clouds as a basis for a first exploratory analysis is that the size of words is related to their relative occurrence in the total set of associated words. As a result, the most common words can be identified easily, which therefore are the strongest stereotypical associations. Moreover, the extent to which a single epithet defines the category, compared to a set of defining words, can also easily be inferred. For example, the category banker is mainly associated with money, and nurse with caring, as these associations occurred much more frequently than all other associations. Similarly, the F I G U R E 1 Word cloud per category for which associations were made (available on Figshare: Willems, 2020a) category salesman is strongly associated with pushy, teacher with caring, and politician with liar. Other categories seem to be associated with a set of words, rather than with a single word. This potentially indicates that for these categories, stereotypical associations are less outspoken with a single characteristic and/or have a more diverse set of common associations. The strongest associations for public servant are caring and helpful. Other associations are also explored in more detail in the subsequent steps. Table 2 indicates the 15 most frequently associated words for the overall dataset, with frequencies per category.

| Content analysis of stereotypical associations
Frequencies are also given on whether words are overall seen as positive, neutral or negative, and an average score, further denoted as S, is given to indicate the negativity or positivity of associated words (−1 means 'consistently negative word'; 1 means 'consistently positive word'). This score is calculated as the average of the word codes (−1, 0 or 1) that all respondents gave when they named that word for any of the professions. Overall, the five most often-

| Content associations across categories
As public servant is a sector-related category, and the others are profession-related categories, some of the categories are completely or partially sub-categories of the overall public servant category. Therefore, associations of the profession-related categories might to a varying extent overlap with the sector-related public servant category. A Sankey graph in Figure 2 visualizes such content overlapping across categories (Willems 2020b). At the left side of the graph, the public servant category is represented. The grey streams towards the job categories on the right give an indication of the proportion of overlap with respect to most occurring words. For this plot, the overlap between public servants and the other job categories is represented by the words caring, helpful, and dedicated (as positive words: dark grey), and lazy, greedy and corrupt (as negative words: light grey). As a result, two main elements are shown with this graph.
First, the vertical length of the category bars on the right-hand side give a proportional insight into the extent to which similar words are associated between the public servant category and the other job categories. For example, politician, nurse and volunteer have relatively longer bars compared to professor, salesman and social entrepreneur.
This means that for politician, nurse and volunteer a higher proportion of the words associated with these profes- (caring, helpful and dedicated). Also for firefighter and teacher, the content overlap is mainly based on positive associations, while it is more mixed for professor, police officer and social entrepreneur. In contrast, the overlap for banker, lawyer and salesman is based mainly on negative associations and is therefore similar to the politician category.
In sum, from a content perspective, there is thus a large relatedness of the public servant category, not only with the categories of teacher and nurse, but also with volunteer. Potentially, the higher levels of prosocial and public ser-   Moreover, it is interesting to observe that the banker stereotype, at least with respect to negativity and positivity, is similar to the public servant profession, as no significant differences occurred in either part of the analysis.

| Positiveness versus negativeness of combined stereotypical associations
However, the Sankey plot in Figure 2 shows that overlap with respect to content is quite minimal and is based on negative associations.

| Content of the public servant stereotype and potential spill-over effects
This study does not suggest consistent negativity with respect to associations made for the public servant stereotype. Its strongest epithets are caring, helpful and dedicated. Building on theories that focus on the representativeness heuristic (LaViolette and Silvert 1951;Bodenhausen 1990;Smith 2003), there is thus a higher likelihood that positive characteristics such as caring, helpful and dedicated are associated with people for whom the information is given that they are public servants. However, with the highest frequencies only ranging up to 41 for the strongest characteristics caring (41), helpful (30) and dedicated (15), it seems that the public servant stereotype is not typified to the same extent by a single defining epithet such as caring in the case of nurse (127), volunteer (99), or teacher (70); brave in the case of firefighter (74); or pushy in the case of a salesman (67). This is perhaps due to its being an overarching sector classification that involves several, more specific and very diverse job categories. Given its more holistic and overarching nature, a more heterogeneous set of associations seems to be made by citizens. Against the background of the representativeness heuristic, these results suggest that public servant, as a cognitive entity, is only to a limited extent representative of many public service jobs. In contrast, the representativeness heuristic might be much more applicable to specific job categories, such as nurse or police officer. This means that from a theoretical perspective, the extent to which a cognitive entity unambiguously represents a specific and homogeneous group of people and/or profession needs to be taken into account more explicitly in future research. Moreover, further testing Abbreviations: Estimation in MLwiN, Rasbash et al. (2015); regression estimates are given, standard errors in parentheses.
can verify whether job stereotype characteristics spill over into perceptions about individual persons. In particular, the reasons and context in which such spill-over effects exist are discussed at length in the literature on stereotypes, as it is increasingly argued that the mental attribution of stereotype characteristics to concrete individuals is rather a matter of conscious social norming, rather than unconscious cognitive bias (Kunda and Thagard 1996;Arkes and Tetlock 2004).
The higher prosocial and/or distinct motivation that public servants have might explain the specific epithets such as caring, helpful and dedicated. This has been researched extensively in the field of public service motivation and related concepts (Perry and Wise 1990;Wright et al. 2012;Schott et al. 2019). In particular, the high content-wise relatedness with volunteer ( Figure 2) suggests that public servants are overall, and without public sector and/or political framing, seen as prosocial and altruistic. Hence, the results in this study verify to some extent the relatedness between public service employment and a higher likelihood of volunteering or acting prosocially, even outside the context of one's professional employment (Borzaga and Tortia 2006;Leisink et al. 2018;Schott et al. 2019). As a consequence, these results contribute-in an exploratory and preliminary way-to theoretical insights on prosocial and public sector motivation in the public administration literature. While a vast number of contributions have focused on public service motivation as a latent and internally focused concept to study the behaviour of (potential) public service employees, the results herein suggest that it also has a noticeable and externally perceived element. In other words, the prosocial element of public service motivation is also visible from an outside perspective on public servants. Consequently, public service motivation as a concept is probably relevant not only to understand the behaviour of public service motivated people, but also to understand the behaviour of others in interaction with people who display public service motivated behaviour.

| Politicalness and sector as sources of cognitive job associations
In order to gain an insight into the overall and non-primed stereotypical associations, this study focused on the exploration of first-order associations, which means that as little as possible framing information was given when respondents were asked to make cognitive associations. However, in order to better understand how stereotypes work in the human cognitive process, and to be able to influence them, it is also relevant to further explore potential cognitive sources of stereotypical associations (Feldman 1981;Hertwig and Gigerenzer 1999). For example, a public servant is considered, at least from a positivity and negativity evaluation, similar to a banker, and is considered significantly more positive than politicians, salesmen and lawyers. As a consequence, these results show that there is no widespread public servant bias when comparing this overall category with more specific professions outside the public sector. Nevertheless, the overarching public servant stereotype is less positive than some specific professions such as firefighter, police officer, teacher, nurse and professor. Hence, when building on earlier findings about a consistently negative public service reputation (Marvel 2016), this negative public service reputation might spill over only to a small extent to the individuals working in the sector. From a theoretical perspective, future studies can thus focus on the relatedness of citizens' attitudes towards public servants and the public sector in general, and test whether these are truly distinct perceptional entities. Further, it can be verified whether these attitudes explain spillover effects between sector and servant stereotypical associations.
Moreover, the public servant category is clearly more positive than some iconic stereotypes such as politicians and salesmen. In particular, the mainly negative association overlapping with public servants, and the stronger negative connotations for the politician stereotype, deserve further attention. Given that the public sector is inherently associated with the political field, negativity about the sector might be related to, or even the result of, negative stereotypes about politicians rather than public servants. A substantial body of literature has focused on the role of political affiliations of public servants in their decision-making (Ebinger et al. 2019;Shaw and Eichbaum 2020). The results herein open avenues for further exploration of how such political affiliation influences perceptions of public servants. As a consequence, explicit political framing might be tested as a moderating concept to better understand negative perceptions of the public sector and public servants.

| LIMITATIONS AND FURTHER RESEARCH
This study has at least three limitations. First, it is based on a convenience sample of 415 respondents in one particular country (the US). The sample has, for example, a substantial over-representation of women (66.7 per cent).
Despite the fact that gender as a control variable does not show significant differences for the number of positive or negative associations, future research should-based on large and representative samples-hypothesize and test whether particular job categories are perceived differently based not only on the respondents' gender, but also on other characteristics. This will allow the evaluation of the extent to which stereotypes are broadly shared and deviation from general stereotypical beliefs. Moreover, as institutional logics, cultures and perceptions of the public sector and its employees are very diverse across countries, the results of this study are hard to generalize to other countries (Miao et al. 2019). In addition, the motivation of people to work in the public sector might also be very different across countries (Kim et al. 2013;Miao et al. 2019). As a result, stereotypes can and should be documented, explored and compared within and across national and regional contexts.
Second, despite some advantages of the associative method in documenting and exploring stereotypes, it has to be noted that certain cognitive processes might have influenced the concrete associations made. For example, as we asked respondents to make associations for several professions in one single questionnaire, earlier associations for one profession might have triggered particular associations for another profession. As a result, some anchor bias (Gehlbach and Barge 2012) and order or acquiescence effects (Billiet and Davidov 2008) could have influenced individual cognitive processes in such a way that associations within this dataset might be relatively more similar to each other, compared to a situation when multiple associations would not have been asked in a single questionnaire. Nevertheless, the randomization of both question order and sequence with other in-between constructs should have eliminated a structural bias in the overall dataset. In addition, for the quantitative analysis of the data, the nested data structure in the multilevel regression analysis adjusts for within-respondent similarity of associations across categories. As a result, at this stage, structural biases have been avoided in the data collection design and data analysis. However, it is not fully clear whether and how cognitive processes influence the overall association process for these job stereotypes.
Third, this study focuses only on job stereotypes. However, stereotypical thinking is often not limited to a single type of category and might interact with other cognitive categories too (Eckes 2002). Therefore, it is relevant to test whether these findings hold when other elements are also included. For example, the interaction of job stereotypes with race, gender and nationality stereotypes can be explored and tested in further studies. This would create additional knowledge of the alleviating or aggravating effects when particular categories are combined.

| PRACTICAL RECOMMENDATIONS
The empirical analysis in this study supports two main practical recommendations. First, the results confirm the recommendation of Van de Walle (2004) that a focus on a particular profession, rather than on the overall public servant category, is related to more positive associations and therefore might be perceived as more attractive. In other words, attracting people to particular professions can perhaps be made more successful by avoiding associations with the overall public servant category.
Second, associations regarding the public servant are still significantly more positive than those regarding some for-profit professions and politicians. This is perhaps a perceived advantage of public servants in general over other job types and can potentially be used by public managers and policy-makers to value the contribution of public servants for citizens and for themselves. For example, Vogel and Willems (2020) find that having public service employees reflect on their societal impact results in higher perceived well-being, willingness to recommend their job, and intention to continue in that job. The results in this study suggest that citizens are well aware of some strong and positive epithets of public servants and public service professions, which can be relied on to make the underestimated value of public servants to society more explicit.
Despite the strongest associations made for public servants being positive (caring, helpful and dedicated) as an overall category, it has a less positive connotation compared to some specific professions such as firefighter, police officer, teacher and nurse. In contrast, it has a more positive connotation compared to other categories such as lawyer, salesman and politician. However, the overall content-wise overlap between public servant and politician is based mainly on negative associations. In sum, the data herein do not support a consistently negative perception among citizens about public servants, contrary to what might often be expected from political discourse, the media or even a certain rhetoric in scientific literature on public servants.