Compassion, Bureaucrat Bashing, and Public Administration

: How citizens behave toward public sector workers is crucial for the well-being and performance of workers. Scholars have mainly focused on understanding negative citizen behaviors, such as aggression. We study a positive behavior, namely compassionate behavior. We study real compassionate behavior in the form of writing positive encouragement messages that are distributed to social workers in the field. We test if showing difficulties faced by public sector workers results in citizens writing more encouragement messages. We also test if bureaucrat bashing results in less encouragement messages. Using a preregistered experiment among a representative sample of Canadian citizens (n = 1,264), we find that showing public sector workers’ struggles and imperfections makes citizens almost twice as likely to write an encouragement message. Hence, showing your weakness can be a strength. Bureaucrat bashing, however, has no effect. Results show that citizens can be stimulated to act more positively toward public sector workers.


Evidence for Practice
• Citizens can be stimulated to act more compassionately toward the public sector workers they meet when applying for or receiving public services.We measured compassionate behavior as leaving an encouragement message for workers to lift up their spirits in daily work difficulties.• Showing weakness can be a strength.Showing citizens the difficulties of the day-to-day work of public sector workers, for instance, in the form of stories, increased citizens' compassionate behavior toward public sector workers.In this way, when public sector workers show that they are imperfect and vulnerable, citizens show more compassion toward them.• Bureaucrat bashing is often assumed to negatively affect public services and the public sector workers that provide them.For social workers and similar public sector workers, this fear may be unfounded since it does not make citizens less compassionate toward them.

I
f citizens want to access public services, they must interact with public sector workers such as teachers, nurses, and social workers.During these encounters, citizens behave in various ways toward public sector workers.Citizens can be patient, but they can also be aggressive or patronizing.How citizens behave toward public sector workers is crucial for the well-being and performance of public sector workers (Dubois 2010;Lipsky 2010).For instance, when citizens are aggressive toward public sector workers, this increases the risk of burnout of public sector workers (Hershcovis and Barling 2010;Tummers et al. 2016).However, a positive behavior would be that citizens are compassionate toward public sector workers.Compassionate behaviors happen when someone acts on another's pain or suffering in order to alleviate it (Bloom 2017;Goetz, Keltner, and Simon-Thomas 2010;Singer and Lamm 2009).When a person shows you compassion, it reduces your stress (Eldor 2018).A parent can, for instance, show that they understand the severe workloads of a primary school teacher by helping the teacher in class, or just saying to the teacher that they understand how hard it can be to be a teacher.Compassionate behavior of citizens is beneficial for public sector workers, as it reduces their stress and risk of burnout.It is ultimately also beneficial for the citizens, as they encounter less stressed public sector workers (Eldor 2018).
Studies have shown that experiencing compassion at work has clear benefits, such as that workers have less stress, are more client-oriented, have higher job satisfaction, higher commitment, less burnout, and better overall performance (Choudhary, Ismail, and Hanif 2017;Dutton, Workman, and Hardin 2014;Eldor 2018;Lilius et al. 2008).However, can we promote compassion in the workplace, and if so, how?The answer to this question is not self-evident.Despite compassion being increasingly researched in public administration (for instance Eldor 2018), scholars have neglected how compassionate behavior can be stimulated.
We study two factors that can impact compassionate behavior.First, we analyze whether showing how difficult it can be to be a public sector worker could trigger compassionate behavior in citizens.Many public sector workers encounter problems in their work, such as high workloads, role conflicts, and severe red tape (Lipsky 2010;Scott and Pandey 2000).Reading about the problems someone has could trigger compassionate feelings, and hence compassionate behavior (Kanov et al. 2004;Strauss et al. 2016).Eliciting a feeling of compassion in citizens would then result in these citizens acting more compassionately toward public sector workers.However, citizens' compassionate behavior toward public sector workers can also be discouraged.
We study an important topic in this regard: bureaucrat bashing.Public sector workers are often bashed in the public debate (Goodsell 2004;Marvel 2015;Rölle 2017).Bureaucrat bashing is especially prominent at the moment given the anti-public sector rhetoric, particularly among some populists (Moynihan and Roberts 2021).
The goal of this study is to investigate whether citizens can be stimulated to act compassionately toward public sector workers and to test if bureaucrat bashing harms this compassionate behavior.We answer the following research question: To which extent does bureaucrat bashing and eliciting compassion influence citizens' compassionate behavior toward public sector workers, and does eliciting compassion reduce the effects of bureaucrat bashing?To answer this question, we conducted a preregistered two-step experiment among nationally representative large samples of Canadian citizens.We analyze whether citizens write encouragement messages to social workers.Writing an encouragement message shows a real willingness to help, making it a compassionate behavior (Bloom 2017;Goetz, Keltner, and Simon-Thomas 2010;Singer and Lamm 2009).All positive messages were shared with social workers.
This study provides theoretical and methodological contributions to the literature.Regarding theoretical contributions, it is first unclear whether eliciting compassion leads to actually behaving more compassionately (Reynolds et al. 2019;Welp and Brown 2014).We show that stimulating feelings of compassion among citizens does result in real compassionate behavior, as they write more encouraging messages for public sector workers to lift up their spirits.Similarly, much literature about bureaucrat bashing is available (Garrett et al. 2006;Goodsell 2004;Hubbell 1991), but scholars often do not study the impact of the assumed negative effects on public sector workers.Our second contribution is that we show that the negative effects of bureaucrat bashing are not always found.In our study, bureaucrat bashing had no effect on whether citizens wrote encouragement messages.
Methodologically, both the bureaucrat bashing and compassion literature consists primarily of research investigating perceptions and attitudes toward the public sector or hypothetical behavior, yet includes no studies on how it affects real citizen behavior (Caillier 2018;Caillier 2020;Garrett et al. 2006).We study actual behavior of citizens in the field, thereby adhering to the call to move beyond studying only attitudes or intended behavior (Hansen and Tummers 2020;John 2020;Lonati et al. 2018).

Eliciting Compassion
Compassionate behavior is a specific type of behavior that falls under the umbrella of pro-social behaviors.Pro-social behavior is characterized by actions intended to benefit others than oneself (Resh, Marvel, and Wen 2018).Compassionate behavior is defined through helping behavior (Goetz, Keltner, and Simon-Thomas 2010) and is associated with pro-social behavior (Runyan et al. 2019).
Compassionate behavior is conceptually distinct from feeling compassion because it refers to the behavior resulting from the desire to act on another's pain or suffering in order to alleviate it, rather than the emotion of the state of experienced compassion (Bloom 2017;Goetz, Keltner, and Simon-Thomas 2010;Singer and Lamm 2009).Compassionate behavior refers to acting compassionately.This means doing something to help someone else (Goetz, Keltner, and Simon-Thomas 2010;Singer and Lamm 2009).For instance, acknowledging to a social worker that you understand that they are overworked and that you will wait patiently to be seen.Eliciting compassion is the act of trying to make people feel compassion, such as by pointing out that the worker or profession has a problem (i.e., high burnout rate of social workers).Stimulating compassionate behavior works if people feel compassion (Goetz, Keltner, and Simon-Thomas 2010;Kanov et al. 2004;Lazarus 1991).
How compassion can be elicited can be explained using cognitive appraisal theory.Cognitive appraisal theory states that compassion will be successfully elicited and, in turn, translate into acting compassionately, based on three cognitive appraisal processes, namely (1) someone's deservingness of help; (2) self-relevance of the situation, and (3) self-efficacy (Atkins and Parker 2012;Lazarus 1991).First, deservingness of help refers to the extent a person is responsible for their situation.A citizen could feel compassion toward a police officer being assaulted.However, if the citizen deems that the police officer was responsible for the assault as they used unnecessary force, then the citizen may not feel compassion.Second, self-relevance refers to how much the situation is in line with your personal norms and values (Atkins and Parker 2012).Going back to the police officer, for example, self-relevance would entail questioning whether the citizen finds assault morally OK or not.Third, self-efficacy refers to the costs and benefits of helping.A person would wonder if they can help, at what personal costs, and whether these actions are going to help the other.For instance, the citizen can decide to help the assaulted police officer by being a witness if the citizen thinks this does not cost them too much time or harms their privacy.In short, when you see that someone is in a situation where you deem that person is not responsible, when this problem goes against your values, and when you think the benefits of helping are higher than the costs, you will decide to act compassionately.
Drawing on these three cognitive appraisal processes, we expect that situations triggering these processes will elicit compassion in citizens and, in turn, make citizens behave more compassionately toward the public sector workers they meet.Our first hypothesis, therefore, is: Hypothesis 1: Eliciting compassion will increase compassionate behavior from citizens toward public sector workers.

Bureaucrat Bashing
We also expect that bureaucrat bashing affects compassionate behavior.One source of negativity toward public sector workers is bureaucrat bashing (Goodsell 2004).Bureaucrat bashing is often intertwined with bureaucracy bashing.Bureaucrat bashing refers to the bashing of public sector workers, while bureaucracy bashing refers to bashing public organizations (Goodsell 2004).Scholars distinguish two forms of bureaucrat bashing, namely (1) meaningless bashing and (2) substantive bashing (Caillier 2018;Caillier 2020;Garrett et al. 2006).Meaningless bashing entails generalized criticisms that offer no solutions and that are used to denigrate public sector workers (Caillier 2018(Caillier , 2020).An example is a statement like social workers are inefficient and wasteful of resources.Substantive bashing entails specific criticisms and solutions, such as social workers are underperforming because their caseload is too high.We need more resources for social work programs in schools and increase the number of social workers.
Repeated negative framing of public sector workers contributes to a negative image of public sector workers in society (Hubbell 1991).Bureaucrat bashing statements have even become an integrated norm of, among others, North American culture (McEldowney and Murray 2000).Public sector workers have become a symbol of incompetence within society, and this symbol has been internalized by the public as a stereotype (Hubbell 1991; Van de Walle 2004).Besides, the perpetuated negative image and bashing statement of the "lazy bureaucrat" has become a common characterization in TV entertainment (Lichter, Lichter, and Amundson 2000).A study review of top-10 box office grossing movies from 2000 to 2009 revealed that 91 percent of movies featured at least one government worker character, with depictions tending on the negative side (Pautz and Warnement 2013).
Repeated exposure of bureaucrat bashing could reduce citizens' compassionate behavior toward public sector workers by increasing easy to recall instances of failures over successes as well as relating to a stable negative attitude over time.A meta-analysis indicates the two strongest predictors in attitude to behavior translations are when the attitude is easy to recall and stable over time (Glasman and Albarracin 2006).Being exposed to mainly one-sided information that shapes our attitudes makes instances of it easier to recall and stable over time, and thus stronger predictors of behavior (Glasman and Albarracin 2006).For instance, in societies where bureaucrat bashing is accepted ( Van de Walle 2004), negative exposure of bureaucrat bashing may contribute to easily accessible, stable attitudes over time.
In addition, bureaucrat bashing could go against cognitive appraisal processes (Atkins and Parker 2012) that need to be activated for compassionate behavior.Bureaucrat bashing could mainly affect the first mechanism, which is someone's deservingness of help.Bashing may successfully convince the person that the public sector worker is responsible for their situation and thus not deserving of help (i.e., lazy or incompetent).As bashing is a repeated common rhetoric in society, it can affect the perception of deservingness of help of public sector workers.Furthermore, meaningless bureaucrat bashing may affect the second cognitive appraisal process of self-relevance of the situation because it can use societal norms and values to devalue and blame public sector workers.For instance, if one is bashing social workers for a child not saved, the loss of the life of a child is intuitively against societal and personal norms and values.By attacking workers on societal values and norms (other example, wasting taxpayers money), bashing can create a rhetoric of blaming the workers by going against societal norms and values.This can be especially used during political campaigns (Caillier 2018;Garrett et al. 2006).Finally, if the first two mechanisms are not met, chances are very slim for a person to decide to invest efforts into being compassionate, as the cost-benefit would not be worth it (i.e., the third mechanism of self-efficacy).Hence, we expect that bureaucrat bashing negatively affects citizens' compassionate behavior toward public sector workers.Thus, our second hypothesis is: Hypothesis 2: Bureaucrat bashing will decrease compassionate behavior from citizens toward public sector workers.

The Interaction between Eliciting Compassion and Bureaucrat Bashing
We expect that eliciting compassion will reduce the effects of bureaucrat bashing.Scholars believe that the negative consequences of bureaucrat bashing stem from constantly showing one incomplete side of the story, perpetuating misinformation about job realities (Caillier 2018;Garrett et al. 2006).In other words, bureaucrat bashing creates an incomplete and inaccurate profile of the public sector worker (Garrett et al. 2006;Goodsell 2014).To counteract this, in this study, we elicit compassion by providing information on the daily challenges and difficulties public sector workers face.People exposed to both bureaucrat bashing and eliciting compassion will have more balanced information about public sector workers to process their actions during the cognitive appraisal processes when deciding whether to act compassionately.Thus, we expect that eliciting compassion will decrease the detrimental effects of bureaucrat bashing, leading us to our third study hypothesis.
Hypothesis 3: Eliciting compassion will moderate the effects of bureaucrat bashing on compassionate behavior, in such a way that the negative effects of bureaucrat bashing on compassionate behavior will become weaker when compassion is elicited.
These three hypotheses lead to our theoretical model, which is shown in figure 1.

Method
To test our hypotheses, we used a two-step survey experiment, with (1) a pilot survey (Study 1) to test our manipulations and find those that worked best for eliciting compassion and bureaucrat bashing and (2) an experiment to test the three hypotheses (Study 2).We tested our manipulations in a separate study because although including manipulation checks in experiments is common practice, it raises concerns.Respondents may, for instance, react based on the manipulation check itself instead of the manipulation (Ejelöv and Luke 2020).Therefore, Study 2 did not include manipulation checks, but used the tested manipulations from Study 1.We preregistered our study on the Open Science Framework (OSF) https://osf.io/fqn9a,and our data, analyses, syntax, and supplementary materials are available online at https://osf.io/vef6d/?view_only=93fb149d179145938c410f9b7c516e77.Ethical approval for the study and its procedures was obtained through the ethical committee of the Faculty of Law, Economics, and Governance of Utrecht University.

Case
We test our hypotheses by studying Canadian social workers.Social workers are relevant public sector workers to study.First, many instances of bashing have been documented over the past decades (Peters and Savoie 1995;Stanfield and Beddoe 2013).Second, social workers deal directly with citizens and thus experience both negative and positive behaviors from citizens toward them, such as aggression or politeness.In other words, social workers can experience (un)compassionate behaviors from citizens.Third, negative citizen behaviors toward social workers are prevalent in the form of aggression (Gately and Stabb 2005;Lowe and Korr 2007;Van Heugten 2011).
Experiencing workplace aggression has numerous negative effects such as lower job satisfaction, performance, organizational commitment, psychological and physical well-being, as well as increased stress, fatigue, burnout, and turnover (Hershcovis and Barling 2010;Tummers et al. 2016;Van Heugten 2011).Large numbers of social workers from different domains report experiencing citizen aggression, including verbal aggression, intimidation, property damage, physical assault, and harassment (Gately and Stabb 2005;Lowe and Korr 2007;Van Heugten 2011).In Canada, 44 percent of child welfare social workers reported to have experienced threats or violence on-the-job (Canadian Association of Social Workers 2018).Even though these experiences are not uncommon, we do not imply that the majority of clients act like this.Many clients may behave more positively toward social workers.Receiving acts of compassion in the workplace is associated with an array of positive outcomes such as increased job satisfaction, performance, organizational commitment, and positive emotions at work, with decreased stress, anxiety, and burnout (Choudhary, Ismail, and Hanif 2017;Dutton, Workman, and Hardin 2014;Lilius et al. 2008).Thus, studying compassionate behavior toward social workers is especially relevant because if we can elicit more compassionate behavior, it may even counterbalance the effects of negative behaviors such as workplace aggression, which they experience regularly.

Study 1
Design and Procedure.In order to test our manipulation, we developed 16 vignettes using stories from Canadian social workers.These are shown in Table A1 of Appendix A. We interviewed five social workers and they provided us with real situations they encountered with clients.Based on these stories, we first developed four vignettes that elicit compassion and four control vignettes.These vignettes presented a narrative of a social worker that explained daily tasks and situations of social work.We also constructed four vignettes that display bureaucrat bashing and four control vignettes.Scholars suggest that bashing may function, in part, through a lack of information on job realities or tasks at work (Caillier 2018;Garrett et al. 2006).Therefore, the compassion vignettes and control vignettes contained the same job description to control for a gap of knowledge of job-related tasks in the effect.Control conditions are suitable because-as opposed to developing positive and negative vignettes-they provide a true effect of bureaucrat bashing and compassion (Lonati et al. 2018).In order to not fatigue, bore or reveal our manipulation to our respondents, the respondents were randomized to rate four vignettes.
We intentionally differed the narration and emotional intensity between the vignettes of bashing and compassion.There are no personal stories in the bureaucrat bashing vignettes, since the purpose was to create meaningless bashing vignettes as encountered in political campaigns or popular media.Our goal was not to create conditions equal in emotional response or emotional stimulation.Our research question does not test high negative emotions of bashing compared to emotions of compassion, but rather the effects of meaningless bashing on compassionate behavior.

Measures. Eliciting compassion:
We asked people to rate the extent to which they experience four feelings used in the literature to assess elicited compassion (compassion, sympathy, concern, and moved) (Cronbach α = 0.89) (Galanakis et al. 2016;Reynolds et al. 2019).Participants had to answer a five-point Likert-scale on each emotion after reading a text (0 = not at all, 1 = a little bit, 2 = fairly, 3 = quite a bit, 4 = very much).The scale items are in Tables B1 and B2 of Appendix B.

Bureaucrat bashing:
We asked participants to rate on the same five-point Likert-scale seven questions based on the definition of meaningless bureaucrat bashing (Garrett et al. 2006).Participants rated the degree to which they felt the vignettes criticized social workers, generalized criticism of social workers, and provided concrete solutions (Cronbach α = 0.90).For the full scale, refer to Tables B1 and B2 of Appendix B.
Sample.We used the G*Power program for the calculation of our power, based on a Cohen's d of 0.5.This led us to an estimation of 176 participants with a power of 0.8.As shown in table 1, we collected a representative sample of 283 Canadian citizens in terms of sex, age, and education.
We chose a moderate effect size in our power calculation as we were testing our manipulation check and we wanted a strong manipulation.Our analyses involved the means' comparison between treatment and control, where we expected a salient difference between the conditions.The standard effect size to choose in this case is a Cohen's d of 0.5 (Perugini, Gallucci, and Costantini 2018).

Figure 1 The Hypothesized Effects of Eliciting Compassion and Bureaucrat Bashing on Compassionate Behavior
Results.We conducted a factor analysis for eliciting compassion and for bureaucrat bashing to analyze the factor structure of our measures.For eliciting compassion, one factor explained 75.3 percent of the variance (compassion = 0.896, sympathy = 0.912, moved = 0.866, concern = 0.793).For bureaucrat bashing, two factors were revealed: generalized criticisms and lack of concrete solutions.Generalized criticisms explains 62.6 percent on the variance whereas lack of concrete solutions explains 16.8 percent of the variance (criticized = 0.865, attacked = 0.883, blamed = 0.868, generalized = 0.897, meaningless criticism = 0.825, concrete solutions = 0.944, resolving problem = 0.940).Together, both factors explain 79.4 percent of the variance.Details on the scales and factor analyses are in Tables B1 and B2 of Appendix B.
In each condition, there was at least one pair of vignettes (treatment and control) that had statistically significantly different mean scores.We conducted a two-group analysis of variance (ANOVA) between vignettes to determine statistically significant different pairs.This ensured that our treatment did, indeed, trigger compassion or bashing and the control condition did not.In the compassion conditions, four pairs were statistically significantly different.We selected the pair of eliciting compassion (M = 3.09, SD = 0.85) and control compassion (M = 2.37, SD = 0.90) with the biggest difference in means (F [7, 562] = 4.12, p < .001).Our rating in our control condition suggests that participants did not just choose the available options of emotions.In other words, our results suggest that participants were not primed by choices, as there are statistically significant differences in means in our manipulation check items between our control and treatment conditions.In the bashing conditions, we also selected the pair of bashing (M = 2.45, SD = 0.76) and control (M = 1.71,SD = 1.07) with the biggest mean range (F [7, 548] = 4.70, p < .001).Additional details on the analyses regarding ANOVA are in Tables C1, C2, C3, and C4 of Appendix C. Table 2 demonstrates the vignettes that were selected from Study 1 and that will be used in Study 2.

Study 2
Design.Study 2 tested our hypotheses about the effects of eliciting compassion and bureaucrat bashing on compassionate behavior from citizens toward public sector workers.This study uses the treatment and control conditions identified in Study 1 and in a 2*2 between-subject experiment.We randomized participants in one of the four treatment groups.Participants were instructed to read the vignettes and then fill in a questionnaire.All responses were forced responses, except the open question to leave a message.If participants chose "yes" to leave a message, they were not forced to do so.See figure 2 for the randomization flow and figure 3 for random assignment of participants.

Measures. Compassionate behavior:
We asked participants if they would write an encouragement message for social workers to help  decrease work-related stress and lift up their spirits.Instructions were as follows (wording inspired by Blasco et al. 2016): We are looking to help social workers with their work-related well-being.We, therefore, want to ask you if you want to write an encouragement message for social workers in Canada to lift up their spirits.This encouragement message may help social workers in their working life.No messages are too big or too small.We will gather the messages we receive and share them with social workers via an online platform.Leaving a message is voluntary and up to your discretion.
Compassionate behavior is defined as acting on another's pain in order to alleviate it (Goetz, Keltner, and Simon-Thomas 2010).
Participants who choose to write a message were counted as having acted compassionately-as they decided to take extra time to help a stranger boost up their morale.Thus, the outcome variable is binary (yes-no).We chose to ask participants to write a message, as opposed to how they feel generally about social workers because we aim to study actual behavior rather than a sentiment.We did not want to assess whether participants only felt compassion toward social workers, but rather if they would really act compassionately.
Focusing on actual behavior is a contribution, as it allows us to move beyond discussions surrounding the intention-behavior gap (Webb and Sheeran 2006).
We uploaded the messages for social workers ourselves and they were distributed to real social workers.You can find the website with the messages here: https://compassionforsocialworkers.tumblr.com/.
We chose Tumblr because of its user-friendliness (easy to distribute, read, and access) compared to other social media platforms where an account is needed.A total of 444 messages were left, of which nine were not shared with social workers because they were negative or Sample.We used G*Power for the power calculation, based on a Cohen's d of 0.2.The calculation estimated 787 participants required for a power of 0.8.We outsourced participants' recruitment to an online panel (Lucid).We used a representative sample of 985 Canadian citizens in terms of sex, age, and education.We sampled from all over Canada.We did not stratify regions except for Quebec and the rest of Canada.We stratified our sampling in terms of age, gender, and educational level.Table 3 shows that our sample is comparable to the general population.
Cohen's d differs from Study 1 because their designs and analyses differ (Perugini, Gallucci, and Costantini 2018).In this study, we chose a small effect size.First, this design is comparing four different conditions in an experiment.Examining experiment literature in the domains of compassion and bashing-we did not find enough information to base our expected effect size on previous studies.In addition, we wanted enough power to detect small effect sizes.Finally, we did not expect to find such salient differences between the conditions of the experiment-compared to our manipulation check.Thus, we decided on a small effect size (Perugini, Gallucci, and Costantini 2018).
We ensured our respondents provided quality data and excluded participants based on quality criteria.We excluded the participants who failed 2 out of 3 attention checks, and most of those left gibberish in the messages.We also excluded speeders.In other words, we aimed to exclude participants who skim through experiments primarily for payments.The full list of excluded participants (with reasons for exclusion) can be found in the supplementary materials on OSF.
Randomization Check.We assessed the sample conditions for homogeneity among the demographic variables of sex, age, and educational level.Table 4 shows the differences between the four conditions.The differences are all insignificant showing that randomization was successful.
Results.Hypothesis 1: Hypothesis 1 stated that eliciting compassion would increase compassionate behaviors from citizens to social workers.Eliciting compassion indeed increased compassionate behavior (β = 0.68, SE = 0.19, OR = 1.98, p < .001).People who read the compassion vignette were almost twice as willing to write an encouragement message than people who read the control vignette.An odds ratio of 1.98 is equivalent to a 1.1 in Cohen's d (Chinn 2000).Of all participants exposed to the eliciting compassion condition, 50.8 percent left a message.Of all participants who read the control condition, 39.7 percent left a message.Hypothesis 1 is supported.Below we show two examples of encouragement messages that were written by participants: The work you are doing is very important and you can not only help your clients but our entire society.It may be difficult at time but keep at it because you are doing much more good than you realize at the time.

Social work is indeed demanding but it's necessary service in society. I applaud all of you on your hard work, dedication and contribution to society. Without you, many would be unable to persevere and survive as worthy individuals. I thank you sincerely!
Hypothesis 2: Hypothesis 2 stated that bureaucrat bashing would decrease compassionate behaviors from citizens to social workers.The bureaucrat bashing condition did not significantly affect compassionate behavior (β = 0.35, SE = 0.19, OR = 1.43, p = .057).Of all participants exposed to the bureaucrat bashing condition, 46.6 percent left a message.Of all participants who read the control bureaucrat bashing condition, 43.9 percent left a message.Hypothesis 2 is rejected.
Hypothesis 3: Hypothesis 3 stated that eliciting compassion will moderate the effects of bureaucrat bashing on compassionate behavior, in such a way that the negative effects of bureaucrat bashing on compassionate behaviors will become weaker when compassion is elicited.Our experiment did not find this relationship (β = −0.48,SE = 0.26, OR = 0.64, p = .084).Hypothesis 3 is rejected.

Conclusion and Discussion
We have shown that eliciting compassion among citizens results in more compassionate behaviors from citizens toward public sector workers.Citizens, who were provided with a story about difficulties public sector workers face, were almost twice as likely to write actual encouraging messages to them.We also found that bureaucrat bashing does not decrease citizens' compassionate behavior.Our findings contribute to the literature in three ways.
First, we show that citizens can be stimulated to act compassionately toward public sector workers.This finding contributes to the street-level bureaucracy literature.This literature shows that many public sector workers aim to act compassionately toward citizens, which is related to the citizen-agent narrative (Maynard-Moody, Musheno, and Musheno 2003).Our result indicates that if public sector workers show the difficulties of their jobs, citizens will also put compassion at the center of their decisions.They are more likely to help public sector workers.This finding also contributes to the literature on bureaucratic reputation.In this literature, the focus is often on how to boost the overall reputation of a public organization by showing how well it is performing and avoiding blame by "hiding imperfections" (Lee and Van Ryzin 2020).We elicited compassion by displaying a vulnerable side of public organizations, namely narratives about the problems public sector workers encounter.Our result could mean that the image of public organizations could be improved by showing that public sector workers sometimes struggle in their jobs: they are faced with aggression or high workloads.
When citizens read about this, they experience compassion toward those public sector workers and behave more positively toward them.Yet, based on our current study alone, we can only infer it is a potential mechanism.Future research needs to test this indication and to which extent this vulnerable side of public organization can positively contribute to bureaucratic reputation.
Our second contribution is that, despite the common assumption that bureaucrat bashing has primarily negative effects, our findings do not demonstrate negative effects of bashing on citizens' compassionate behavior toward public sector workers.There are various-mostly normative-studies that highlight that bureaucrat bashing is unfounded (Goodsell 2004(Goodsell , 2014).Yet, bureaucrat bashing is rarely studied empirically (Caillier 2018).To the best of our knowledge, there are only two experimental studies, which yielded mixed findings on bashing's effect on the public (Caillier 2018(Caillier , 2020)).Caillier (2018) found that bashing leads to more negative attitudes of the public.Caillier (2018) suggested that bureaucrat bashing is common and thus, creates memories of it in the minds of citizens, which can be reactivated by other bashing cues.Our results challenge this claim, as bureaucrat bashing has been a prominent part of the Canadian media discourse history, but we find no effect (Campbell and Peters 1988; Peters and Savoie 1995).Caillier's subsequent study (Caillier 2020) also did not replicate his initial results and did not find any effect for bashing.In line with Caillier (2020), our results raise questions about the effects of bureaucrat bashing on the public and their attitudes and behavior toward public sector workers.Furthermore, the effect of bureaucrat bashing was marginally positively significant.In the absence of compassion elicitation, bureaucrat bashing seems to increase compassionate behavior.However, we are cautious with interpretations as the result was not statistically significant.More research is needed to explore this potential effect.
Despite our null effects, our findings are relevant to understand bashing especially in this era of rising populism (Moynihan and Roberts 2021).Politicians using bureaucrat bashing engage in what is known as "attack politics," with an aggressive emphasis on failures over successes and the exploitation of organizational vulnerabilities (Flinders 2011).For instance, politicians exploit the vulnerabilities of the child foster care system by bashing the failed cases by social workers rather than highlighting the multitude of successful cases (Stanfield and Beddoe 2013).This is where our finding becomes relevant.We demonstrate that showing organizational vulnerabilities (i.e., difficulties of public sector workers) prompts compassionate behavior from citizens.A danger associated with attack politics is the politicization and amplification of "bad accountability," which in turn decreases public's confidence in government, services, and workers (Flinders 2011).Showing organizational vulnerabilities based on daily difficulties workers face could help combat bad accountability.Future research is needed, however, to distinguish whether our bureaucrat bashing findings apply to "elite" or career public sector workers as well (Moynihan and Roberts 2021).
A possible explanation for why showing vulnerabilities does increase compassionate behavior toward public sector workers but bashing does not reduce compassionate behavior can be found in the literature on public sector stereotypes (de Boer 2020; Willems 2020) and public service motivation (Schott et al. 2019).Willems ( 2020)  demonstrates that different public sector professions have both positive or negative stereotypes in society, and some of the positive stereotypes include caring, helpful, and dedicated.Social workers are associated with high levels of public service motivation (Vinzant 1998).It is, thus, possible that social workers also resonate within stereotypes of caring, helpful, and dedicated partly because public service motivation refers to the pro-social motivation for a job, such as opportunities to serve society and to aid the population, while simultaneously being less motivated by higher salaries (Lewis and Frank 2002;Wright and Pandey 2008).Bureaucrat bashing effects could, thus, be limited for professions considered high in public service motivation (e.g., social workers) because they trigger positive stereotypes.Since we did not measure public perceptions of social workers, we cannot conclude that our results were not driven by a positive bias toward social workers.However, the neutral vignettes about social workers did not lead to an increase in compassionate behavior-suggesting that a positive bias is unlikely.
To test this, future research should assess the effects of eliciting compassion and bureaucrat bashing on different types of public sector workers.
The third contribution of our study is that our study provides methodological contributions to public administration research.
We measured actual behavior through an experimental approach.We demonstrate a simple, scalable way to measure behavior using written encouragement messages, instead of an attitudes or intended behavior (Hansen and Tummers 2020;John 2020).In addition, to the best of our knowledge, we are also the first to develop a scale to measure meaningless bureaucrat bashing.
This study has important generalizability limitations we must acknowledge.First, we think it is likely that our results generalize to social workers in other countries where they are also exposed to aggression by citizens or similar job hardships.We theorize that for public sector workers in similar service-oriented professions, such as case workers or rehabilitation officers, our findings would generalize because the nature of the profession is highly similar, but we would be hesitant to generalize to more regulation-oriented professions, such as tax officials (see de Boer 2020;Maynard-Moody, Musheno, and Musheno 2003).Future research is needed to test if our findings translate beyond social workers to other sub-groups and beyond the North American cultural context.
Second, we focused on a specific type of compassionate behavior, namely writing encouragement messages.We have argued that writing an encouragement message is an example of compassionate behavior.Compassionate behavior is defined as an act to help alleviate the situation of another person (Goetz, Keltner, and Simon-Thomas 2010).We stressed this clearly in the instructions for our encouragement messages.It required citizens to take their own time to help out a total stranger, making it an act of compassion.This is an action that our participants decided to part-take in or not.They could decide for themselves if they wanted to write a message to help and boost up morale of workers or not.Writing encouragement messages is, however, only one type of compassionate behavior.Future research is needed to test if our results generalize to other forms of compassionate behaviors, such as being patient, understanding, and not showing frustration.Theoretically, if the cognitive appraisal mechanisms (i.e., deservingness to help, self-relevance, and self-efficacy) are met (Atkins and Parker 2012), we expect our findings to have similar effects for other compassionate behaviors.
Third, our experiment was conducted online where there was no "real" interaction with the worker.It could be that face-to-face interactions affect citizens' compassionate behavior.In our study, we do not think that the lack of face-to-face interaction impacted our findings since it was not a prerequisite for our type of compassionate behavior.It may, however, be for other types of compassionate behavior such as being understanding and patient.Future research is needed to understand the role of face-to-face interactions in compassionate behaviors.Using field experiments can be helpful.
It is important to note, however, that our aim was not to create a fix-for-all solution that would eliminate negative behaviors from citizens toward public sector workers in all situations.Rather, our aim was to see if compassionate behavior can be elicited, with the idea that on average citizens could show more compassion to workers.
Our study also has methodological limitations.First, we choose to measure compassionate behavior by typing out an encouragement message.Measuring real encouragement messages like we did has merit since it helps combat the intention-behavior gap (Webb and Sheeran 2006), as opposed to an intention during a hypothetical situation.Intention does not predict behavior well (Rhodes and Dickau 2012;Webb and Sheeran 2006).As a consequence, our dependent variable is limited in terms of mirroring real world compassionate behaviors when citizen and public sector workers meet face-to-face.There are other types of compassionate behaviors that are more realistic in real-life settings that need to be investigated.
Second, we did not examine different dimensions of stimulated compassionate behaviors.We settled to preregister a dichotomous measure over a qualitative one, as a binary measure provides researchers with less degrees of freedom (Wicherts et al. 2016).We did, however, investigate whether the content of messages was supportive or not (i.e., gibberish or nonsupportive).Only nine were not positive.We did not investigate the content of messages in terms of length-as it was not directly relevant to answer our research question.We did not intend to investigate how much effort one puts in, but whether they put effort at all.We acknowledge that our measure of compassionate behavior is limited and future research should adopt a qualitative strategy in understanding the effects of compassionate messages on workers.
Third, social desirability is a serious concern when designing experiments.Overall, we ensured that respondents were guaranteed anonymity, were not under direct observation of the researcher, and received a small compensation.All these design choices help combat social desirability (Levitt and List 2007).Moreover, we have conducted our manipulation checks separately, as a way to combat respondents finding out what the experiment is about (Ejelöv and Luke 2020).Regardless, like in any study, we cannot be completely certain that social desirability was not at play.We think it is, however, unlikely that our results were affected by it as we only found effects in the compassion condition, and not in the other conditions.If social desirability affected leaving a message, we would see significant effects in more or all experimental conditions.However, in terms of payment, we indicated in our instructions that leaving a message was voluntary but we did not explicitly specify leaving an encouragement message would not increase payment.Future research can test payment effects more closely.
Finally, we cannot be sure that our results were not affected by the presence of other prominent emotions.In our manipulation check, we assessed whether compassion was significantly elicited and whether meaningless bureaucrat bashing was perceived.However, future research should investigate how the intensity and interaction of a range of emotions affect (un)compassionate behavior of citizens.
To conclude, our findings urge scholars to rethink the conceptualization of negative consequences of bureaucrat bashing.
Our results also emphasize the importance of showing the vulnerable side of public sector workers.We demonstrate that by showing one's struggles and imperfections, we can stimulate citizens to act compassionately toward public sector workers.Zooming into practical implications, our results suggest to rethink the emphasis on the high-performance side of bureaucratic reputations to stimulate more positive citizens to public sector worker interactions (Lee and Van Ryzin 2020).We encourage building a more "compassionate public administration" and developed the first steps toward it: how to encourage and strengthen interactions with the public and public sector workers through eliciting compassion in citizens.An important next step is to test the effects of compassion from citizens on public sector workers daily working lives, such as their well-being and performance.
Appendix A I am a governmental social worker that works with children between the age of 6 and 17.I provide interventions and accompany children whose parents are at risk to lose custody.We work in disadvantaged areas, with families on the lower end of the socioeconomic spectrum.Our usual working hours are from 9-5, but these are subjected to change depending on how the cases are going (for example, if any emergencies happen in the one of the followed families, understaffing).We organize weekly meetings with the parents and the children, regularly visit the family house, and evaluate when loss of custody may be needed due to child mistreatment.We also provide links with other organizations.Sometimes, parents are looking for help but we cannot response to their request, as they do not fully fit with the mandate of the organization so we try to link them with other organizations that may help.I am a governmental social worker that engages with the elderly population (60 years of age and more).My main tasks consist of assessing clients' social functioning needs, process accommodation requests, aid in coping with loss of autonomy, and provide interventions for grievance.Outside of my general job description, I also have to conduct multidisciplinary meetings to tackle the wide range of issues our clients are facing (physical, emotional, financial, mental aspects of well-being), organize family meetings, write many reports about every case, and details of every decision taken.When a family suffers the loss of an elderly person in our care, we accompany them in mourning and provide coping interventions, information, and referrals.When a client loses too much autonomy or his condition deteriorates too much, we arrange the transfer into specialized care.We work with individuals still living in their home (but that still need assistance) and with individuals in nursing homes.Bureaucrat bashing 1

All Tested Vignettes
Policy is developed at federal, regional, and local levels.To implement social programs and policies in Canada for target population groups specific guidelines must be met.In the past years, social service policies have not been working and governmental social workers have been inefficient and incompetent!In over 100,000 calls received last year in child social service alone, less than half were answered.Social workers work with the most vulnerable people in our society, do not they deserve government services that they can trust?Bureaucrat bashing 2 We need to reimagine our current political economy to keep our country's well-being safe.To do this, we must have meaningful structural and staff change to address core societal issues.Governmental social workers and our system are failing vulnerable children in foster care.Our social workers lack capabilities, confidence, and common sense in good judgment.We must have zero tolerance of state failure.In the upcoming elections, we vow to restore citizens' trust in social work and services to keep our population's well-being safe.Bureaucrat bashing 3 The number of social work vacancies is falling.The number of agency workers in local authorities is dropping.Local elected officials frequently describe social work profession as now at its all-time low "rock bottom".They argue that social work is a failing, demoralized, low-status profession.The misguided nonsense of bureaucratic rules within social work punishes those who need the most help.Too often governmental social workers are poorly trained and not ready for frontline practice when they leave social work education.

Figure 2 Figure 3
Figure 3 Randomization of Participants and Forced Responses of Study 2

Figure 4
Figure 4 Experimental Evidence that Eliciting Compassion Increases Compassionate Behavior.The Y Axis, Ranging from 0 to 60, Shows the Percentage of the Citizens that Left a Message.Each Condition Shows 95 Percent Error Bars

Table 2
Vignettes Selected from Study 1 Condition Vignette Eliciting compassion I am a governmental social worker that works with children between the age of 6 and 17.I provide interventions and accompaniment with children whose parents are at risk to lose custody.Time and time again, we must work outside of normal hours, for which we are never paid.I received a call in the middle of the night, because one child from a case I was working, a 10-year-old girl, attempted suicide.The drug-addicted parents were not available, and I gained that child's trust and support to help her, so I had to show up.But I am not paid for this, it is my own decision to either go along with my values or not.I had worked so hard to get this child to open to me, how can I not be there for an attempted suicide when I know no one else would show up to comfort her?Control compassion I am a governmental social worker that works with individuals with drug addiction.I provide referrals for crisis intervention, create interventions, and link our patients with other organizations for interventions.Overall, I assess and evaluate new clients, monitor the recovery progress, and provide counseling and support during needed moments of treatment period.When someone is referred to us for services, I meet with and interview them, and if possible, interview close people to the individual and go through their medical files.After assessment, I aid in coming up with a treatment plan and link the client with treatment centers if necessary (treatments are also offered by us, but the type of treatment depends on severity of addiction).All treatment plans are based on individual needs, ability, type, and severity of substance abuse problems.I also support clients in creating small realistic goals, how to carry them through, and aid in monitoring progress of these goals.Bureaucrat bashingWe need to reimagine our current political economy to keep our country's well-being safe.To do this, we must have meaningful structural and staff change to address core societal issues.Governmental social workers and our system are failing vulnerable children in foster care.Our social workers lack capabilities, confidence, and common sense in good judgment.We must have zero tolerance of state failure.In the upcoming elections, we vow to restore citizens' trust in social work and services to keep our population's well-being safe.

Table 3
Sample Demographics Study 2 (n = 985) Notes: Education level low = no formal education, primary school, some high school; mid = finished high school, college-CEGEP-technical; high = some university, completed university, graduate studies.

Table 4
Demographic Comparison across Groups

Table 5
Logistic Regression Results (n = 981) governmental social worker that works with children between the age of 6 and 17.I provide interventions and accompany children whose parents are at risk to lose custody.We are faced with trying to meet the need for hunger among young people with little financial means.Often, we must spend our own money as there is not enough funding for projects to improve well-being.Being understaffed we often work outside of normal hours.So, when a child finally starts to confide in us, we cannot say "sorry it's 6:00 PM we have to stop".However, we are not paid for these extra hours.So, if you are working hours are 9-5 and a suicidal teenager comes to your workplace for help, are you going to send her home because you are off the clock?These situations usually go against our values, and we offer our own time and resources to keep helping.Eliciting compassion 2 I am a governmental social worker that works with individuals with drug addiction.I provide referrals for crisis intervention, create interventions, and link our patients with other organizations for interventions.We are constantly working overtime (up to 12 h daily) because of high levels of staff burnout because we are understaffed.We have extremely high caseloads and lack of resources to answer our clients.Unfortunately, just last week, we had to mourn through the death of a client due to overdose.We did not have enough resources to provide him with the extent of mental care he needed.He was suffering of severe depression linked to traumatic past experiences.Our center only had the means to offer basic therapies for depression, and limited staff to answer calls.How do you deal with the death of a close patient for whom you tried everything possible to help?Eliciting compassion 3 I am a governmental social worker that works with the James Bay Cree community in the North.I meet with patients and community officers, coordinate support groups, provide counseling, and evaluate cases.Being in a community in the 'North', resources are limited for the special needs of our patients, and I am the only social worker there.I had a teenager in mental health who was looking to get out of his home and go to supervised housing to get better and stand on his own two feet.However, the only place for mental health accommodation required that the patient speak French.Being in a Cree community, my patient speaks English and Cree.The worker refused to take the patient because of the language barrier.A big difficulty of this job is being faced with situations where the person needs help, and you want to help, but a small bureaucratic rule prevents it, and it is incongruent with our values.Eliciting compassion 4I am a governmental social worker that works with children between the age of 6 and 17.I provide interventions and accompaniment with children whose parents are at risk to lose custody.Time and time again we must work outside of normal hours, for which we are never paid.I received a call in the middle of the night, because one child from a case I was working, a 10-year-old girl, attempted suicide.The drug-addicted parents were not available, and I gained that child's trust and support to help her, so I had to show up.But I am not paid for this, it is my own decision to either go along with my values or not.I had worked so hard to get this child to open to me, how can I not be there for an attempted suicide when I know no one else would show up to comfort her?
governmental social worker that works with individuals with drug addiction.I provide referrals for crisis intervention, create interventions, and link our patients with other organizations for interventions.Overall, I assess and evaluate new clients, monitor the recovery progress, and provide counseling and support during needed moments of treatment period.When someone is referred to us for services, I meet with and interview them, and if possible, interview close people to the individual and go through their medical files.After assessment, I aid in coming up with a treatment plan and link the client with treatment centers if necessary (treatments are also offered by us, but the type of treatment depends on severity of addiction).All treatment plans are based on individual needs, ability, type, and severity of substance abuse problems.I also support clients in creating small realistic goals, how to carry them through, and aid in monitoring progress of these goals.Control compassion 3I am a governmental social worker that works with the James Bay Cree community in the North.I meet with patients and community officers, coordinate support groups, provide counseling, and evaluate cases.Since I do not cover a specific target group, I work with children with behavioral problems to addiction problems in teenagers and adults, to the elderly population.I meet with clients, meet with other community officers, facilitate and coordinate support groups (grief, addiction, coping with loss of autonomy).I link community members with various health services officers, assess social functioning and needs of community members.Since substance abuse is a common problem within the community, I also provide information and referrals for crisis interventions and link with necessary agents and institutions for intervention plans.I assess accommodation requests for individual with mental health issues for supervised housing accommodations (social centers that house individuals to help with mental disorders).

Table C2
Post Hoc Tests (Tukey HSD) for Compassion Vignette Means Comparison The mean difference is significant at the 0.05 level.The vignettes labeled C1 to C4 correspond to eliciting compassion vignettes 1 to 4 and C5 to C8 correspond to control compassion 1 to 4, respectively (order presented in TableA1of Appendix A). *