Responding to online complaints in webcare by public organizations: the impact on continuance intention and reputation

Reputation in the public sector

Although the differences “only” lie in the type of organization, the literature on corporate reputation and public sector or bureaucratic reputation are rooted in different fields. Research on public sector reputation has its foundations in public administration, focusing mainly on substantive organizational behavior and decision-making as reputation management “tools” (Carpenter and Krause, 2012; Wæraas and Maor, 2014). In contrast, corporate reputation literature has a stronger focus on strategic communication and branding as a means to influence public perceptions (Barnett and Pollock, 2012). Corporate reputation has a narrower orientation than public sector reputation and tends to focus on profit-making and rivalry with competitors (Overman et al., 2020). This competitive focus is however less relevant for most public sector organizations, as they often do not operate in competitive environments, despite new public management reforms (Waeraas and Byrkjeflot, 2012). Simultaneously, citizens have other expectations regarding public sector organizations that are not included in the corporate definition, e.g. related to fairness, justice and democratic input (Overman et al., 2020; Waeraas and Byrkjeflot, 2012).

The dominant definition of reputation in the field of public sector organizations is that of Carpenter and Krause: “a set of beliefs about an organization's capacities, intentions, history, and mission that are embedded in a network of multiple audiences” (2012, p. 26). They distinguish between multiple dimensions or forms of reputation in a public sector context: organizations can be evaluated based on their performative (competency), moral (compassion and flexibility), legal-procedural (rule-following behavior) and technical/professional (skills and capacities) dimensions (Carpenter and Krause, 2012). Although these reputational dimensions are not necessarily unique for public sector organizations (one might use “compassion” as a criterion to evaluate a company too), the different context in which public sector organizations operate requires a broader, multidimensional operationalization (Overman et al., 2020). Building on the work by Carpenter and Krause, measures that reflect the several dimensions have been developed to assess bureaucratic reputation (e.g. Lee and van Ryzin, 2019; Overman et al., 2020).

Social media have gained importance in the formation of organizational reputation (Etter et al., 2019). Reputations on social media are expected to be shaped differently than in the traditional media landscape, but communication directors in the public sector struggle to find their organization's way in the new media landscape (Jacobs and Wonneberger, 2019). Etter et al. (2019) expect a more active role for audiences in the evaluation of organizations, different manners of expressing their opinions about the organization, and a more dynamic and heterogenic way in which reputations are formed online. This creates specific challenges for organizations that aim to use social media in their reputation management. More specifically, the challenges identified concern message and content control, the accessibility of the organization and its members, responsiveness to complainers and engagement with audiences (Chaudri et al., 2021). In particular, accessibility and responsiveness are relevant in the context of webcare and reputation management. Social media provide organizations with the opportunity to respond quickly to online complaints, but organizations also need to carefully evaluate the information that they provide (for instance, taking GDPR requirements into account or providing factual information, based on the consultation of other experts within the organization). They simultaneously need to take the affordances of social media into account, e.g. the ability for messages to amplify quickly across multiple audiences (Albu and Etter, 2016). The demand for responsiveness might also create an “always-on environment” and can cause more profound reflections on organizational priorities, like how responding to online complaints relates to other organizational activities that might impact the reputation (Chaudri et al., 2021).

Online complaint handling by public organizations

Webcare has been introduced to interact with social media users on a variety of topics, both from the organizational and user's interest (Javornik et al., 2020; Van Noort et al., 2014). It can be defined as: “The act of engaging in online interactions with (complaining) consumers, by actively searching the web to address consumer feedback (e.g. questions, concerns and complaints)” (Van Noort and Willemsen, 2012, p. 133). Public organizations use webcare to understand sentiments within specific groups of citizens, in which reputation management and anticipating to questions and needs of citizens also play a role (Edwards and De Kool, 2015). Differences in service recovery expectations between for-profit and public sector organizations might be rooted in the relations that users have with both organization types. Citizens often have no choice but to accept services or “products” from public sector organizations, e.g. passports or social services (Waeraas and Byrkjeflot, 2012). In addition to that, some public organizations, like tax agencies, also have service-related tasks, providing advice and assistance (Waeraas and Byrkjeflot, 2012).

The use of conversational human voice in online complaint handling

The organization's tone of voice matters in webcare. For private organizations it has been shown that a CHV, described as “an engaging and natural style of organizational communication as perceived by an organization's publics based on interactions between individuals in the organization and individuals in public” (Kelleher, 2009, p. 177), is often positively associated with a range of outcomes, such as complaint handling satisfaction (Javornik et al., 2020) and reputation (Dijkmans et al., 2015). Initially, the tone of voice was conceptualized by the perception of CHV by the public, measured by means of an instrument with 11 items (Kelleher and Miller, 2006; Kelleher, 2009). Other researchers distinguished three tactics that can be adopted in organizational messages to obtain a certain amount of CHV: (1) message personalization (e.g. greeting the user personally and signing the organizational response with the employee's name), (2) informal speech (e.g. using informal vocabulary such as contractions and visual elements like emoticons) and (3) invitational rhetoric (e.g. acknowledging the complainer's feedback and showing sympathy for the situation) (Liebrecht et al., 2021; Van Noort et al., 2014). Especially the application of multiple linguistic elements to suggest CHV leads to higher perceptions of CHV (Liebrecht et al., 2021).

Response strategies in online complaint handling

Originating from the field of crisis communication, webcare studies recognize two general response strategies in complaint handling: accommodative and defensive. Accommodative strategies are characterized by taking responsibility and trying to meet the needs of the complainer, e.g. by offering compensation or an apology. The denial of responsibility and blame-shifting to other actors is characteristic to defensive response strategies (Huibers and Verhoeven, 2014; Javornik et al., 2020; Weitzl and Hutzinger, 2017). Accommodative response strategies appear more frequently in (private organizations) webcare (van Hooijdonk and Liebrecht, 2021). They are associated with more positive effects than defensive responses or no response at all (e.g. van Noort et al., 2014). By using an accommodative response strategy, organizations show they put the interests of complainers first, express their compassion with the situation at hand and try to solve their problem, instead of denying the situation or putting their own interests first (Huibers and Verhoeven, 2014).

Little is known, however, about the impact of an organization's response strategy on the effectiveness of the webcare response's tone of voice, since these concepts have been studied separately in previous research. As CHV is associated with improved relationships and organizational perceptions of customers (Dijkmans et al., 2015; Crijns et al., 2017), it can be argued that this tone of voice is more effective in a webcare response with an accommodative strategy in which organizations mark their compassion with the complainer's situation, as opposed to a defensive response in which organizations distance themselves from the complainer (van Noort et al., 2014). We therefore expect a moderating effect of the organization's response strategy, meaning that:

Perceived justice as an explanation

Reading the online interactions between social media users and organizations leads to evaluations of how well the complaint have been handled (Blodgett et al., 1997; Bacile et al., 2018; Ghosh and Mandal, 2020). First, the outcomes of the service recovery can be evaluated. Distributive justice refers to the perceived fairness of the solution, remedy or compensation, also an apology (Blodgett et al., 1997; Javornik et al., 2020). Second, people can also evaluate the procedure of the complaint handling, the organization's flexibility of the procedures and the fairness of the policy (i.e. procedural justice). Third, the evaluation can focus on how well the complainer has been treated during the complaint handling: the perceptions of interactional justice. We expect that the use of CHV in webcare will lead to a halo effect: the readers of the conversation will rely on “global affect” (Leuthesser et al., 1995) based on CHV, rather than distinguishing between different aspects of the conversation, such as the substantive solution provided for the problem or the procedure that had been followed. For that reason, we expect that CHV will positively affect all justice perceptions, and, subsequently, organizational reputation. Following the line of reasoning underlying the first hypothesis, we argue that the organization's response strategy will moderate this effect:

User's involvement in online complaint handling

Webcare not only serves a goal for (individual) customer care but also for reputation management (Van Noort et al., 2014) since complaint handling via social media is also visible to a much larger audience than the complainer him or herself (Albu and Etter, 2016). Observers who read social media messages of others presumably have a low situational involvement. In contrast, complainers are highly involved in the situation and face the consequences of complaint handling. Given the halo effect alluded to earlier, one would probably expect positive effects on all types of users. Interestingly, empirical findings indicate that CHV has an opposite effect on those who have a high involvement in the situation. The use of CHV was found to have a positive effect on purchase intentions in situations of low involvement, but consumers in a high situational involvement setting preferred a more corporate tone of voice, which is also associated with competent organizational behavior (Barcelos et al., 2018; Javornik et al., 2020). Given the lower situational involvement of observers, we expect a positive relation between CHV and reputation for observers, which was already hypothesized in the first part of hypothesis 1. Given the higher situational involvement of complainers, we expect that:

For observers, CHV's positive effect on complaint handling satisfaction can be explained by interactional justice. Javornik et al. (2020) found that the direct effect of CHV on procedural and distributive justice was negative, but the indirect effect, with interactional justice as a mediator between CHV and both other justice perceptions, was positive. This indicates that interactional justice plays an important role for observers in the relationship between CHV and complainer satisfaction.

For complainers, with a higher situational involvement, we expect that the outcome (distributive justice) and the procedure of complaint handling (procedural justice) are most important, and therefore mediate the relationship between tone of voice in webcare and organizational reputation. As a corporate tone of voice can contribute to the perception of a competent and professional organization (Decock et al., 2021; Javornik et al., 2020), we expect that the negative association between CHV and reputation can be explained by procedural and distributive justice for complainers.

Repeated exposure to CHV in webcare

The effects of CHV in webcare on people's perceptions have been examined extensively in previous research (Liebrecht et al., 2021). Little is known, however, about the robustness of these effects although Dijkmans et al. (2015) found that people's exposure to the organization's social media activities positively impacts organizational reputation over time. This relation was mediated by people's perception of CHV. The organization's tone of voice can reinforce the effects on reputation in the long term, which we also expect in the current webcare context.

We are also interested in a potential behavioral consequence of webcare: the extent to which users expect to (re)approach the organization via social media after exposure to the organization's webcare (Bhattacherjee, 2001). Since the success of webcare services is dependent on the extent to which people use it to contact the organization (Guo et al., 2016), and as we assume that positive experiences will stimulate repetitive behavior, we expect that repeated exposure to CHV positively affects continuance intention.

Method

Design. We conducted an experimental study conforming to a 2 (Tone of voice: CHV vs corporate) × 2 (Response strategy: defensive vs accommodative) between-subjects design. This yielded four experimental conditions of one (fictitious) webcare conversation of a public transport organization. Participants were randomly assigned to one condition. The dependent variables reputation and continuance intention and the mediating variables concerning the three justice perceptions of the study were measured by means of a questionnaire.

Participants. Participants were recruited via Dynata, a leading recruitment agency (www.dynata.com) and confirmed to our criteria to generate a representative sample of the Dutch population with regard to their age, gender and education. The final dataset consisted of 424 Dutch adults [1] (51.9% female, 47.6% male, 0.5% other) throughout the whole country. The average age of the participants was approximately 48 years (M = 48.29, SD = 16.99) and ranged from 18 to 87 years. The highest obtained degree of education can be classified in three equally sized groups (secondary school, senior secondary vocational education, higher vocational or university education). We conducted randomization checks by comparing participants assigned to one of the four conditions on several background variables. No significant differences were found.

Stimuli. The stimuli were based on a real Twitter conversation (in Dutch) between a passenger and a public transport organization; the Dutch railways (NS). Although the NS is legally speaking a company, the function that it fulfills in society – public transport – can be considered as a public service (Koppenjan, 2012) and its communication therefore belongs to the field of public sector communication (Luoma-aho and Canel, 2020). In addition to that, if people want to use public transport on the main railroad system, they have no choice but to use the services of the NS, as the NS is currently the exclusive exploiter of the main railroad system in the Netherlands (Authority for Consumers and Markets (ACM), n.d.; Koppenjan, 2012). From the perspective of citizens, the NS thus does not operate in a competitive environment.

The conversation started with the passenger's complaint about a train that did not arrive on an early morning. The railways responded to the complaint by asking on which station this happened. Subsequently, the passenger replied with the name of the city concerned, upon which the railways responded again. In total, the conversation consisted of four messages: two passenger messages and two webcare responses. Both webcare responses contained the experimental manipulations on response strategy and tone of voice.

Response strategy consisted of two levels: defensive strategy or accommodative strategy. The defensive strategy typically contains two sub strategies: denial and justification (Van Hooijdonk and Liebrecht, 2021). A denial strategy was present in the first webcare response in the conversation (i.e. “it seems unlikely that the delay was not communicated”). A justification strategy was implemented in the organization's second response where the railways explained the delay was caused by work activities, consequently shifting the responsibility to another party (i.e. “the external party did not communicate the work activities to us in time”).

Accommodative responses refer to the acknowledgement and acceptance of a dissatisfying event caused by organizations. Two highly accommodative actions were selected for the study: an apology and a corrective action (Van Hooijdonk and Liebrecht, 2021). In the first webcare response, the railways apologized for the service failure (i.e. “apologies for the delay”). In the second message, the railways used the corrective action strategy by referring to a financial compensation for the dissatisfying event (i.e. “via this hyperlink, you can request a refund for the delay”).

Tone of voice consisted of the options CHV and corporate. CHV was operationalized based on the classification of Van Noort et al. (2014) who distinguished three tactics that can be used to establish a sense of CHV in webcare communication: message personalization, informal language and invitational rhetoric. Based on an integrative literature review, Liebrecht et al. (2021) presented a taxonomy of concrete, linguistic elements to operationalize these three tactics that we used for the manipulations in the current study. Six linguistic elements were added to the CHV condition: two linguistic elements per main category. Two elements of message personalization were added to the first webcare response: a personal greeting (i.e. “Hi Nicky”) and a personal signature of the webcare employee (“^RW”). The company also showed sympathy to the passenger, which is a linguistic element of invitational rhetoric (i.e. “it is reasonable this annoys you”). The second webcare response contained two elements of informal language: an interjection (i.e. “oh my”) and a non-verbal cue (i.e. “☹”). The organization closed the conversation with an expression of well-wishing (i.e. “have a nice day anyway!”), which is another element of the invitational rhetoric category. These six CHV elements were absent in the webcare conversations with a corporate tone of voice. In Figure 1, two examples of the experimental stimuli are shown.

Instrumentation. The dependent and mediating variables were measured through a questionnaire. All items were assessed on a 7-point Likert scale (1 = totally disagree, 7 = totally agree).

The organization's reputation was measured before (t = 0) and after the participants were exposed to the experimental stimuli (t = 1). We adopted the standardized 5-item scale proposed by Lee and Van Ryzin (2019) that contains the domains of (1) performance (i.e. the NS is a well-run organization), (2) morality (i.e. the NS maintains high ethical standards), (3) technical ability (i.e. the NS bases its decisions on evidence) and (4) procedure (i.e. the NS treats people fairly). This builds upon the aspects of public sector organizations reputation identified by Carpenter and Krause (2012) and adds a “general domain” that covers personal attitudes towards organizations (i.e. Overall, the NS has a good reputation), which we included as well. We added two items of their 10-item scale (which is the 5-item scale, with 5 additional items) to adequately reflect aspects of the organization that are typical for the NS and this context, as a technical organization that provides services to citizens and operates in the public eye: I believe what the NS says and the NS has the skills to deal with complex situations. The items were translated to Dutch and showed a good internal consistency (α = 0.95, M = 4.75, SD = 1.16).

The validated scales developed by Ghosh and Mandal (2020) were used to measure the three justice types that serve as mediating variables in the current study. Concerning distributive justice, we translated Ghosh and Mandal's four items into Dutch and adjusted them to the current webcare context, e.g. The complaint handling by the NS was fair. The internal consistency of the items was good (α = 0.78, M = 4.42, SD = 1.16).

Procedural justice was measured with the two items of Ghosh and Mandal's (2020) scale, e.g. The length of the time taken to resolve the complaint was longer than necessary (R) (r = 0.196, p < 0.001, M = 4.65, SD = 1.09).

Regarding interactional justice, four items were used (Ghosh and Mandal, 2020) and adopted to the current webcare context, e.g. The webcare employee was appropriately concerned about the complaint. The items showed a good internal consistency (α = 0.83, M = 4.40, SD = 1.30).

Procedure. After approval from the Ethics committee of our university (2021-CC-13045), participants were recruited via agency Dynata, which offered the participants small incentives. They read a brief introduction on the topic and were informed about the confidentiality and independency of the study. After agreeing on the informed consent, participants first answered the pre-reputation measures. To distract the participants' attention from the main organization of the study in the pre-measure of reputation, the reputation of two other transport related organizations were also assessed. Thereafter, participants' demographic information, and social media and webcare experience were asked. In order to ensure the participants' engagement in the study and, thus, the reliability of their answers in the questionnaire, an attention check was included.

Next, the participants were randomly assigned to one of the four experimental conditions, followed by the measures on the three justice types, the post-measure of reputation and the manipulation checks. Subsequently, a debriefing was shown in which it was stated the webcare conversation was fictitious and the organization was not involved in the study. Also, the participants were informed about the follow-up invitation they would receive within one or two weeks to participate in the second part of the study. The whole procedure took about 7 min (M = 412.75 s, SD = 480.35).

Results

Before testing our hypotheses, we examined whether the manipulation of the response strategies and tone of voice were successful. The manipulation of response strategy was measured with The NS tries to accommodate to the needs of the traveler and assessed on a 7-point Likert scale (1 = totally disagree, 7 = totally agree). An independent samples t-test showed a clear distinction between the response strategies: the participants indicated that the organization tried to meet the needs of the traveler more in the accommodative condition than the defensive condition (M = 5.41, SD = 1.21 versus M = 3.22, SD = 1.74, t(201) = −10.44, p < 0.001).

The operationalization of tone of voice, however, was not distinctive between the initial corporate and CHV condition. The perceived CHV, measured with the item The NS communicated in a human way (derived from the original CHV-measure of Kelleher, 2009), was assessed as equally high in the corporate and CHV condition (M = 4.90, SD = 1.33 versus M = 4.84, SD = 1.51, t(201) = 0.315, p = 0.753). Thus, the manipulation of tone of voice that was based on theory did not correspond with the participants' perceptions of the webcare conversations. Given the relatively high median of 5, participants presumably considered the presence of a webcare response itself already a human way of communicating (compared to a lack of a response), regardless the tone of voice. In order to respect the participants' perceptions of CHV, we decided to redistribute the participants by means of a median split to test the hypotheses. Participants who assessed the perception of CHV of the webcare response with a 5 or higher, were classified in the perceived CHV condition (n = 275), whereas participants with a lower score were classified in the perceived corporate tone of voice condition (n = 149). We again compared participants assigned to one of the four conditions on the background variables and found the distribution of the participants over the four conditions was still equal.

Hypotheses testing. A moderated mediation analysis was conducted using Hayes' PROCESS Model 8 v3.5 (Hayes and Little, 2018) in SPSS 24. In Table 1, the descriptive statistics of the dependent variable and the mediator variables for all four experimental conditions can be found.

The moderated mediation analysis contained the perception of tone of voice as independent variable, response strategy as moderating variable, reputation as dependent variable and the three justice perceptions as mediating variables. The findings are visualized in Figure 2 and show that the data confirm H1a that the usage of CHV in webcare is positively associated with organizational reputation. In contrast to our expectations (1b), the response strategy did not affect this relation. Furthermore, in accordance with H2a, the data show that the positive association between CHV and organizational reputation is mediated by (1) distributive and (2) procedural justice, but contradictory to our hypothesis (3) interactional justice did not serve as a mediator. Again, response strategy did not moderate these relations, rejecting H2b [2].

The question remains, however, whether the participant's involvement in the situation affects the effectiveness of CHV. We therefore conducted the second study.

Method

Design. We conducted an experimental study conforming to a 2 (Tone of voice: CHV vs corporate) × 2 (User's involvement: complainer vs observer) between-subjects design. This yielded four experimental conditions of one webcare conversation of the same public transport organization; participants were assigned to one of the experimental conditions that contained the similar tone of voice condition as they assessed in study 1. The user's involvement condition was assigned randomly.

Participants. For this follow-up study, the same participants of study 1 were recruited via agency Dynata. The questionnaire was opened by 246 respondents, of whom 203 finished the whole survey. The final dataset of 203 Dutch adults had similar characteristics as those in the study 1 sample. The distribution with regard to gender was equal (50.2% female, 49.8% male), the average age of the participants was again approximately 48 years (M = 47.90, SD = 15.68, range 18–78 years) and their education level again conformed to the three distinguished education levels. None of the participants experienced any negative events between the two measurements that could impact the outcomes of the study.

Randomization checks. To ensure a consistent parallel with study 1, we again redistributed the participants based on their perception of CHV in the experimental webcare conversation [3]. The subsequent randomization checks revealed that the participants in the four conditions were comparable concerning their background variables as no significant differences were found.

Stimuli. Similar to study 1, the experimental stimuli consisted of a (fictitious) webcare conversation on Twitter between a complainer and the public transport organization. The traveler approached the organization with a complaint concerning unclear announcements at the station about a railroad switch. The organization responded by verifying the location, subsequently followed by an answer of the traveler. Thereafter, the organization explained the nature of the cause and asked for additional information. In contrast to study 1, the response strategy was held constant. Both webcare responses contained an accommodative response strategy consisting of an apology and a corrective action.

In order to systematically compare the current study's findings with the first study's findings, tone of voice was operationalized in a similar way as in study 1. In the CHV condition, the first webcare tweet contained a personal greeting, a signature of the webcare employee and an expression of sympathy. The second webcare tweet contained an interjection, a sad smiley and an acknowledgement. These elements were absent in the corporate tone of voice condition.

The involvement of the user was operationalized by providing the participant with a brief description of a scenario. Participants in the complainer condition were asked to imagine they traveled by public transport themselves and experienced the announcement failures at the station. In the CHV condition, the personal greeting to the traveler was also personalized. In the observer condition, it was only asked to imagine the participant would see the following webcare conversation between traveler Beau van den Broek and the public transport organization in a Twitter timeline, which was similar to study 1 (see Figure 3).

Instrumentation. The same measures were used as in study 1 and assessed on a 7-point Likert scale (1 = totally disagree, 7 = totally agree): reputation (7 items, α = 0.95, M = 4.77, SD = 1.16), distributive justice perception (4 items, α = 0.83, M = 4.45, SD = 1.23), procedural justice perception (2 items, r = 0.18, p = 0.010, M = 4.67, SD = 1.09) and interactional justice perception (4 items, α = 0.83, M = 4.57, SD = 1.26).

Procedure. The respondents who participated in study 1 received an invitation via Dynata one week later to participate in this follow-up study. After a brief introduction on the topic and agreement with the informed consent, participants were randomly assigned to one of the involvement roles, but the tone of voice in the stimuli was similar to the tone of voice they assessed in study 1. Next to the demographic, control and attention questions similar to study 1, it was also investigated whether the Dutch railways dominated the news in the past days (which turned out not to be the case) and whether the participant engaged in a webcare conversation with the organization after the first experiment. Similar to study 1, participation in study 2 also took 7 min on average (M = 400.14 s, SD = 1036.07); participants received Dynata credits for their participation.

Results

The descriptive statistics of the dependent variable and the mediator variables for all four experimental conditions are shown in Table 2.

To test the hypotheses, we again conducted a moderated mediation analysis using Hayes' PROCESS Model 8 v3.5 (Hayes and Little, 2018) in SPSS 24. The analysis contained the perception of tone of voice as independent variable, user's involvement as moderating variable, reputation as dependent variable and the three justice perceptions as mediating variables. The findings are visualized in Figure 4.

Generally, tone of voice indeed positively affected the evaluation of the organization's reputation, but the expected negative effect of CHV on reputation for particularly complainers was not found (rejecting H3). Furthermore, the data did not confirm the expectations formulated in H4a and H4b of the moderating role of user's involvement on the mediation of justice perceptions on reputation. The user's role of observer or complainer did not impact perceptions on either of the three justice types.

Method

Finally, we examined the effects of repeated exposure to CHV in webcare responses. H5 states that organizational reputation will be higher after repeated exposure to CHV in webcare, compared to a corporate tone of voice. During the research, organizational reputation was measured three times: in study 1 prior to exposure to the webcare conversation (t = 0), afterward (t = 1) and one week later in study 2 (t = 2). The items used to measure organizational reputations are reported in study 1.

In a similar vein, H6 describes the effects of repeated exposure to CHV on continuance intention. This variable was measured after exposure to the webcare conversation in the first study (t = 1) and similarly in the second study (t = 2). Continuance intention was assessed by participants after exposure to the experimental stimuli of study 1 (t = 1) and of study 2 (t = 2). The variable was measured with two items based on Bhattacherjee (2001): I intend to approach the NS more often via social media and I would recommend others to approach the NS via social media. The correlation between the two items was mediocre (Study 1: r = 0.566, p < 0.001, M = 3.62, SD = 1.46; Study 2: r = 0.59, p < 0.001, M = 3.85, SD = 1.39).

Results

In order to test the hypotheses, the data of the participants were used who participated in both studies (N = 203). We maintained study 2's median split on the perceived CHV condition (n = 144) and perceived corporate voice condition (n = 59) to execute the statistical analyses.

Reputation. A mixed ANOVA with perceived tone of voice as a between-subjects factor and reputation (at three intervals: t = 0; t = 1; t = 2) as a within-subjects factor was conducted to assess whether repeated exposure to the organization's tone of voice led to higher reputation scores (H5) [4]. A marginally significant main effect of time on reputation was found, F(1.86,373.39) = 2.927, p = 0.059. The pre-measure of the organization's reputation resulted in an average score of 4.79 (SD = 1.18), after the first exposure of a webcare conversation it was 4.72 (SD = 1.23) and after the second exposure 4.77 (SD = 1.16). Strikingly, the contrasts between the three intervals revealed that the assessed reputation after the first exposure (t = 1) was significantly lower than the organization's initial reputation (t = 0), F(1,201) = 7.407, p = 0.007. No significant difference was found between the assessed reputation after exposure in study 1 (t = 1) and after exposure in study 2 (t = 2), F(1,201) = 0.305, p = 0.58).

Subsequently, a significant interaction effect was found, indicating the perceived tone of voice and interval time impacts the organization's reputation, F(1.86,373.39) = 5.41, p = 0.006. The lower reputation score after the first exposure (t = 1), seems to be caused by the respondents who perceived a corporate tone of voice. They assigned lower reputation scores after reading the first webcare conversation (t = 1) compared to participants who perceived CHV, F(1,201) = 5.514, p = 0.02 (Corporate: M = 3.81, SD = 1.27, CHV: M = 5.09, SD = 1.01). The perception of CHV did not impact the assessed reputation score after exposure to the second webcare conversation (t = 2), F(1,201) = 1.450, p = 0.23 (Corporate: M = 3.78, SD = 1.13, CHV: M = 5.18, SD = 0.89).

Continuance Intention. A mixed ANOVA was conducted with perceived tone of voice as between-subjects factor and continuance intention (2 intervals: t = 1; t = 2) as within-subjects factor to assess the effects of the organization's tone of voice in webcare on users' continuance intentions (H6). Given that our repeated measures variable (continuance intention) has only two levels, we could assume sphericity. The within-subjects comparison revealed a significant effect of time on continuance intention, F(1,201) = 6.647, p = 0.011. Participants voiced a higher continuance intention after the second exposure to the webcare conversation (t = 1: M = 3.62, SD = 1.46, t = 2: M = 3.85, SD = 1.39). The marginally significant interaction effect of time and perceived tone of voice, F(1,201) = 3.787, p = 0.053, suggests this long-term effect on continuance intention was especially true for participants who perceived higher levels of CHV in webcare (Corporate: M = 3.42, SD = 1.48, CHV: M = 4.03, SD = 1.39).

Implications for theory and practice

Our finding that CHV positively affects organizational reputation is in line with previous findings that relate this tone of voice to several positive outcomes for the organization (Javornik et al., 2020; Liebrecht et al., 2021), but it was not yet confirmed for a public sector context. Justice perceptions, especially people's evaluation of the complaint handling's procedure, explain this effect. In addition to the confirmation of these short-term effects of CHV, our study also revealed long-term implications. The adoption of CHV in webcare seems to be fruitful for public sector organizations, since the reputation remained stable over time and people's intention to (re)approach the organization via social media increased after multiple exposures. The latter finding has not been investigated before, not in a for-profit context either. This underlines the importance of perceived CHV for organizational reputation as a means for reputation management (Dijkmans et al., 2015; Van Noort et al., 2014). This is a rather attractive option for public organizations, as the message content of responses to online complaints can be fully controlled by the organization (Albu and Etter, 2016), which is not the case for journalistic media attention.

If we reflect on this finding from a broader perspective, one of the implications is that CHV could help to build a reputational buffer. In their study on media coverage of corporations, Jonkman et al. (2020, p. 291) found that “people who hold more positive attitudes toward a company are less susceptible to the media effects of (negative) future coverage of the company”. Their findings give reason for public organizations to include webcare in their reputation management. Given their public nature, public sector organizations are often subject to media scrutiny (Luoma-aho and Canel, 2020). Adequate webcare using CHV could therefore probably help to build a reputational buffer which can protect against the consequences of negative news.

Limitations and future research

The findings of these studies contribute to our insights in the reputational and customer care perceptions of tone of voice in webcare. A complicating factor in the use of CHV in webcare to deliberately obtain positive organizational outcomes is that CHV is often operationalized and applied in terms of linguistic elements intended to provide the user a “sense of CHV” (Liebrecht et al., 2021; van Noort et al., 2014). Yet, our findings suggest that people's perception of CHV in webcare is of utmost importance. The very presence of a conversation might have been interpreted as “human” rather than the specific linguistic elements. Although we theoretically grounded and carefully pretested our stimuli, many participants perceived the corporate tone of voice as human too. This might be a consequence of our one-item manipulation check (“The NS communicated in a human way”), so future studies need to assess the success of the manipulation in a multi-dimensional way, e.g. also including “perceived willingness to converse” and the experienced level of “openness to dialogue” (Dijkmans et al., 2015, p. 637). The inclusion of a non-response condition or using an even stronger manipulation (which might subsequently affect the validity of the experiment) could therefore be considered for future, similar studies. A further suggestion is to assess the factors that, according to participants, contribute to their experience of an organization “communicating in a human way.” Many studies take the linguistic and conversational aspects as a starting point to establish a sense of CHV (Liebrecht et al., 2021), but a more qualitative or explorative approach could shed light on other (interfering) factors that are related to the experience of a conversation with a human. This is even more urgent nowadays as people increasingly become accustomed to CHV in online conversations, so how can we explain differences in experienced CHV? Another limitation of the current study is that a real and well-known organization was used towards which participants could have pre-existing attitudes, which might affect the findings.

Interestingly, we did not find substantial differences between observers and complainers. Two tentative explanations can be formulated. First, the design itself may have caused the lack of an effect, since the participants were asked to put themselves in the position of the complainer, while they did not face the problem in their real life. A second explanation is the degree to which observers could put themselves in the position of the complainer, even without explicitly asking them to do that. The situation in the experiment could happen to everyone who travels by train. Given the large number of participants that does so at least a couple of times per year, they could probably easily imagine the situation. The implication of this finding is that if CHV affects reputation, irrespective of the involvement of the user (observer or complainer), this is even more a reason for organizations to engage in webcare with CHV, as it serves reputational goals beyond mere customer care.

In sum, both studies show the practical and theoretical relevance of applying CHV in online complaint handling, not only by companies but also by organizations in the public sector. CHV has reputational benefits and serves the goal of customer care. With the challenges that public sector organizations experience in their online communication (Jacobs and Wonneberger, 2019), these studies could support them in explaining and applying the benefits of specific approaches to online communication. Finally, with this study, we also aimed to pave the way for a stronger focus on the public sector in online strategic communication research, as there are many ongoing academic challenges in this field.