The effect of attributions and failure severity on consumer complain behaviors in sharing economy

Abstract

Sharing economy has become a prominent business model that has been discussed in many previous studies, yet its consumer complaint behavior has not been sufficiently explored. Consumers’ tolerance and overly positive review toward Partners have dominated studies in post-failure evaluation and behavior. This study was conducted to examine the effects of failure attributions, i.e. locus of causality and controllability, and severity of failure on consumers’ direct and indirect complaint behavior, i.e. rating and negative word of mouth. Data were collected through a laboratory experiment with 280 participants in Indonesia. Findings revealed that at all levels of controllability and severity of failure, consumers’ intentions to give low rates were higher when failures were attributed to Platforms, while there were no differences related to intentions to engage in negative word of mouth. Results also showed the interaction of attribution of controllability and severity of failure toward consumers’ intention to give Partners low rates. These results contributed to research related to consumer complaint behavior in sharing economy, particularly negative word of mouth that still needed to be discussed adequately.

Keywords:

• sharing economy
• attribution
• severity of failure
• complaint behavior
• digital platforms
• negative word of mouth

1. Introduction

Emerging in more than a decade, sharing economy has become a business model that changes transaction patterns and behavior of business people (Eckhardt et al., 2019) also succeeded in disrupting various economic sectors (Geissinger et al., 2020; Ly & Tan, 2020). Sharing economy (SE) has become a part of the lives of most people. The business model allows consumers to enjoy various conveniences and “generosity” in consuming the desired value through a “renting than owning” mechanism supported by the internet technology revolution, which allows various parties to meet through digital platforms (Ganapati & Reddick, 2018). In terms of business growth, SE-based service revenue was predicted to increase to USD 335 billion in 2025 (PWC, 2015b), which the prediction was supported by the fact revealed in PWC’s survey that 72% of respondents that have used SE services stated that they will continue using SE services (PWC, 2015a). The emergence of this business model was triggered by a series of events and changes related to the conditions and behavior of the world’s economic community, including the world financial crisis, which reduced people’s purchasing power, the development and dissemination of information technology, especially the internet, shifts in trends and patterns of consumer consumption, and increasing public awareness of sustainability issues (Bardhi & Eckhardt, 2012; Kathan et al., 2016; Kumar, 2018; Lawson et al., 2016; Niezgoda & Kowalska, 2020).

Belk (2014) defines SE as the acquisition or distribution of resources coordinated by a group of people to obtain compensation or fees. SE is also known by various terms that refer to similar business practices, including collaborative consumption (Botsman & Rogers, 2015), access-based consumption (Bardhi & Eckhardt, 2012) and the gig economy (Ferrell et al., 2017), which all of them at least have something in common regarding the use of technology by Platforms that connect assets owners as Partners with consumers which allows consumers to access these assets temporarily. Compared to conventional business models, there are some dissimilarities at the marketing level in SE related to the actors involved, the value creation process, to the value generated, which then raises some interesting theoretical gaps to study (Eckhardt et al., 2019). In the conventional business model, consumers only interact with firms as service providers. While in SE, consumers interact with two entities in the consumption process, namely Platform companies (e.g. Uber) as collaborators or enablers and Partners (e.g. drivers) as asset owners as well as service providers for consumers. These three entities then act as value generators, where the behavior of each entity and the interactions between them will affect the value obtained by other parties (Eckhardt et al., 2019), that then affect consumer expectations and evaluation criteria for service quality (Mallargé et al., 2019).

Partners are direct service provider for consumers but not organic parts of the Platforms, instead they are independent entities that flexibly provide access to their private-owned assets to consumers through Platform mediation in order to earn monetary income (Burtch et al., 2018; Zervas et al., 2017). Unlike service providers that are employees of firms, Partners are perceived as not having the capacity or capability on par with firms’ employees (Suri et al., 2019). The partners’ roles, positions and conditions then influence the consumer’s perception and tolerance of the deficiencies that may occur during the services (Eckhardt et al., 2019). Some studies show that consumers are more tolerant and show more positive responses after experiencing service failures in SE services (Berg et al., 2020; Osman et al., 2019; Pera et al., 2019). This tolerance leads to consumers’ reluctance to raise complaints to Platforms, which has the potential to reduce the effectiveness of the rating mechanism as means of building trust between Partners and consumers through reflecting service quality (Basili & Rossi, 2020; Meijerink & Schoenmakers, 2020).

Several studies were dedicated in response to those specific characteristics of SE. However, research in the service failure context still needs to be considered, given that not much research has been conducted on consumers’ behavior and what factors that influence them. This study contributes to a better understanding of consumers’ direct and indirect behavioral intentions after experiencing a service failure in SE services, and what factor that differs the direct and indirect complaint behavioral intentions. Service failures in SE potentially occur in consumer interactions with assets and services consumed, platforms as mediators through digital services, partners as service providers for consumers, and even interactions with fellow consumers (Schaefers et al., 2016). Failures to deliver services at any point of interaction will affect consumer evaluations of the total service and consumer attributions to the causes of service failure (Suri et al., 2019). Based on the attribution theory, Suri et al. found that consumers’ forgiveness is affected by their perception toward controllability and locus of causality of the failure incidents, that due to the triadic relationship in SE, consumers were more forgiving when the incidents were perceived as caused by Partners than Platforms and less controllable than high controllable. However, we suggest that perceptions of severity of failure will affect the influence of those attributions evaluation toward consumers’ complaint behavior. The highly positive tendency of consumers’ evaluation and behavior after service failure is also a unique characteristic of SE services, where several studies have described some factors that influence this phenomenon, but research on the effect of failure attribution and severity of failure on this positive bias is still needed. Therefore, this study will elaborate the influence of consumers’ perception of locus of causality, controllability and severity of failure toward consumers’ intention to give direct complaint using rating mechanism and indirect complaint, i.e. negative word of mouth (NWOM) in Indonesia. As many previous research were conducted in the context of developed countries, not many in the context of Asian developing countries. Indonesia represents an Asian culture with a collectivist society, in which an interdependent self-construal norm influences consumers’ cognitive evaluation and choice of behavior (Markus & Kitayama, 1991; Walker et al., 2005). People with interdependent self-construal norm tend to precipitate harmony with higher levels of forgiveness (Sinha & Lu, 2016; Walker et al., 2005) differ from those with independent self-construal norm mostly adopted in American and Western European culture (Markus & Kitayama, 1991).

This paper is organized as follows: a literature review related to sharing economy, attribution theory, severity of failure and consumer complaint behavior followed by hypothesis development. The next section will consist of methodology, results, discussion, conclusion, managerial implication and limitations and further research.

2. Theoretical background and hypotheses

2.1. Sharing economy

SE can be defined as a mechanism for providing goods and services on a demand basis involving more than one Partner as service providers and consumer through empowering underutilized assets facilitated by Platforms, which tend to be motivated by financial factors (Basili & Rossi, 2020; Eckhardt & Bardhi, 2016; Gaber & Elsamadicy, 2021). The SE business model generally uses a technology-based digital platform that brings together multisided demands between Partners and consumers in an exchange transaction (Ganapati & Reddick, 2018) through a matching mechanism (Dellaert, 2019). It is an economic phenomenon based on collaborative consumption in which there is a process of sharing or joint use of goods and services as an alternative to maximizing the use of underutilized resources (Sung et al., 2018) and spare labor (Ganapati & Reddick, 2018), through the provision of non-permanent access to tangible and intangible assets through digital platforms (Eckhardt et al., 2019). SE platforms offer a flexible requirements and work arrangements, which enable individuals whom have limited capacity and capability even discriminated in conventional platforms to participate as service providers, that may increase the economic efficiency and personal income distribution in a country (Calderón-Milán et al., 2020).

Apart from doing the matchmaking process, Platform has number of other functions as an intermediary, namely bridging the communication between Partners and consumers, expanding the reach area for both Partners and consumers, creating flexibility, building trust between all parties, and managing transactions (Sutherland & Jarrahi, 2018). Partners who are consumers of the platform and producers for consumers play a crucial role through their involvement in the value creation process by taking over several marketing functions previously carried out by the company, up to promotion, communication, to handling complaints, resulting in the large role of Partners in shaping consumer perceptions of service quality (Eckhardt et al., 2019). Consumers expect to get economical, social and hedonic values from SE-based services (Ratilla et al., 2021). Therefore, Partners’ failure to provide one or more of these values leads to consumer dissatisfaction and service failure

2.2. Consumer attributions

Attribution theory explains how individuals analyze and look for causes of events they experience, which will then influence the individual’s responses and actions in the future (Schmitt, 2015; Weiner, 2011). Individuals will interpret and use the information as a basis for their behavior or response to the events they experience. According to attribution theory, a continuous learning process that is formed in individual social interactions will develop subjective meanings that lead to the appointment of a party as the source of the cause (locus of causality), as well as the perception of the level of controllability and the potential stability of the cause as three dimensions of consumers’ attribution (Weiner, 1985b; Weiner & Graham, 1984). Locus of causality defined as “who is attributed to be the cause of the failure”, while controllability determines how likely the failure can be prevented and stability reflects how likely the failure to be permanent (Weiner, 1985a, 2000). Attribution of stability is considered not relevant in SE services due to the nature of the transaction and interaction, in which consumers are less likely to be served by the same partner more than once as a consequence of either algorithm mechanism (e.g. ridesourcing services) or consumer’s power to choose other partners (e.g. accommodation services).

Different from service failures in conventional services which generally consumers will only attribute the cause to firms and not to the individual service providers who are parts of the company, when service failures occur in an SE-based transaction, the consumer conflicts with two independent entities that have the potential to cause the failure, Platform and Partner. Related to differences in possession of both tangible and intangible resources, consumers perceive that Partners have a lower ability to control or prevent failure than Platforms. Combined together, consumers’ attribution of locus of causality and controllability will affect consumers’ intention to forgive and complain, either by directly communicating the failure to Platforms or indirectly by acting through negative word of mouth (NWOM) (Furunes & Mkono, 2019; Ruth et al., 2002; Suri et al., 2019; Tronvoll, 2012; Van Vaerenbergh et al., 2014; Watson & Spence, 2007).

2.3. Severity of failure

Severity of failure describes the amount of loss suffered by consumers as a result of service failure (Hess et al., 2003; Sengupta et al., 2014), both tangible, in the form of financial losses or reduced product value, and intangible, including discomfort and negative emotions such as worries and time loss (Smith et al., 1999). Consumers who perceive a high severity of failure in service failure conditions will perceive significant losses, give negative evaluations, show dissatisfaction, tend to avoid future or long-term relationships as well as interactions, and have the potential to form negative word of mouth regarding service providers (Del Río-Lanza et al., 2009; Kalamas et al., 2008). High severity of failure will trigger consumers to evaluate the causes and consequences of service failures and be more involved in problem solving efforts (Sengupta et al., 2014).

Consumers’ perception of the severity of failure affects consumers’ coping mechanism. When consumers perceive losses to be severe, consumers tend to adopt problem focused coping, where consumers tend to form judgments, responses, and negative behavioral intentions through expressing complaints to service providers. Conversely, emotional-focused coping will be used when the loss is perceived to be less severe, where consumers try to manage the negative emotions form more positive service judgments and behavioral intentions (Gabbott et al., 2011; Lazarus & Folkman, 1984).

Previous research showed that if a high level of loss occurs in the initial service experience consumed by consumers, consumers will be motivated to form NWOM (Balaji et al., 2016), especially to family and acquaintances (Swanson & Hsu, 2011). However, in another study, the severity level was found to only affect consumer behavior in submitting complaints to firms and did not affect NWOM (Jayasimha & Billore, 2016). To the best of our knowledge, no research has been found that elaborates on the role of consumers’ perception of the severity of failure on consumer complaint intentions in SE services, not to mention its role on the effect of consumers’ failure attribution toward consumers’ complaint intention.

2.4. Consumer complaint behavior

Filing a complaint is a post-service failure behavior resulting from a negative evaluation of the interaction or value that is outside the consumer’s tolerance zone (Parasuraman et al., 1985; Zeithaml et al., 1993). In other words, complaints are consequences of service providers’ inability to provide value within consumers’ zone of expectations (Tronvoll, 2007). Complaints may be defined as expressions of consumers’ dissatisfaction from services consumed directly or indirectly, means to convey emotions, mechanisms to achieve intrapersonal or interpersonal goals, or a combination of the three (Kowalski, 1996). So complaints are based on the perception of consumer dissatisfaction, no matter how small the dissatisfaction is, although dissatisfaction does not always lead to complaint behavior (Singh, 1988). Not only dissatisfaction, complaints may also be triggered by emotional factors—negative emotions mediate the relationship between consumer evaluation and consumer intentions to complain (Oliver, 1993)—and personality traits, including, among others, perceptions of injustice consumers for the service or behavior they receive and the desire to attribute mistakes to certain parties (Tronvoll, 2007). These factors trigger consumers to file complaints against service providers for several purposes, including channeling negative emotions, obtaining compensation, helping service providers improve quality, or embodying altruistic motives to provide information to other consumers (Wirtz, 2017).

Complaints can be manifested in behavioral responses such as in the form of exits, voicing complaints directly to service providers, negative word of mouth and filing complaints through third-party institutions (Singh, 1988), as well as non-behavioral responses, such as when consumers do not give negative evaluations and actions (Singh & Pandya, 1991). While Tronvoll (2012) sorts complaint behavior into no complaining, communication complaining (e.g. voicing negative evaluations to service providers), and action complaining (e.g. spreading NWOM).

Consumers’ complaints become important to service providers or firms in a way that they are informed and have the opportunity to manage consumers’ dissatisfaction so that it will not become a stumbling block that has the potential to reduce company performance in the long term (Mattila, 2006; Tax et al., 1998). Non-complaining behavior may harm firms not only by failing to get the opportunity to accommodate and improve consumers’ dissatisfaction but also consumers’ choice not to voice their complaints has the potential to encourage consumers to spread NWOM (Bunker & Bradley, 2007; Nyer, 1997; Singh & Wilkes, 1996), especially when dissatisfaction tends to be high (Singh & Wilkes, 1996).

NWOM contains complaints or comments that discredit a particular brand or product as a reaction to an unpleasant experience related to that brand or product that leads to dissatisfaction (Kimmel & Audrain-Pontevia, 2010; Richins, 1984). Consumers engage in NWOM to reduce anxiety, get advice, get revenge, and provide information to other parties to avoid the same service failure (Jayasimha & Billore, 2016; Sundaram et al., 1998). NWOM may also be a way to attract the attention of service providers in order to provide a solution, express consumer feelings, and encourage service providers to make improvements (Verhagen et al., 2013).

NWOM concerns service provider as it has a greater influence on potential consumer decision-making than positive WOM (Balaji et al., 2016). NWOM has the potential to have a greater negative impact on firms than exit behavior (Singh, 1990). As a communication and social sharing mechanism, NWOM can change the attitude of other individuals as recipients of NWOM towards certain products or brands (Ardyan et al., 2021; Augusto & Torres, 2018). Likewise, on the other hand, the content of NWOM tends to be subjective and prone to bias in its proponents, which then may increase negative perceptions of the product or brand, i.e. fact bias (consumers do not have complete facts), memory bias, source attribution bias, motivational bias and audience bias (Richins, 1984).

2.5. Hypotheses

Consumer awareness that Partners, as service providers in SE, are not professionals (Mallargé et al., 2019) but “ordinary people” like consumers (Eckhardt et al., 2019) raises consumer tolerance towards Partners when service failure occurs. Dissimilar to failures in conventional services which service failures often lead to negative responses as consequences of consumer dissatisfaction, consumers of SE services tend to elicit more positive responses both at the affective and behavior level (Osman et al., 2019; Pera et al., 2019; Suri et al., 2019). The positive bias of SE consumers’ complaint behavior, primarily related to rating and negative reviews of Partners, has been exposed in several previous studies (Berg et al., 2020; Bridges & Vásquez, 2018; Bulchand-Gidumal & Melián-González, 2020). SE consumers’ negative ratings and reviews tend not to be based on objectivity and rationality (Fradkin et al., 2021; Zervas et al., 2020). Some factors have been found to be the determinant of this positive bias, including short social distance, conformity of norms, style, and intensity of interaction between consumers and partners to low consumer expectations for SE services (Bridges & Vásquez, 2018; Meijerink & Schoenmakers, 2020; Mody et al., 2020; Yannopoulou, 2013). Furthermore, consumers were found to be more forgiving when the failure is attributed to Partners than Platform but only when consumers perceived the failure is highly controllable, and empathy plays a significant role only when the failure is attributed to Partners (Suri et al., 2019). Empathy toward partners leads to consumers’ reluctance to do things that may potentially give negative consequences for Partners, including giving a low rate that may bring punishment from Platforms to Partners. Therefore, aligned with the higher empathy and intention to forgive Partners, consumers will be more unwilling to give a low rate to Partners than Platforms.

H1.

Consumers’ intention to give low rate is higher when the failure is attributed to Platforms than Partners but only when the failure is less controllable

The failure severity will also affect how consumers deal with service failure conditions. When the severity of failure is perceived to be low, consumers tend to control the emotions that arise due to their discomfort, leading to more positive evaluations and behavioral intentions. Meanwhile, when consumers perceive their loss to be more severe, they will make judgments that tend to be negative using cognitive and rational analyses that generate responses and behaviors aimed to overcome their discomfort and prevent additional losses or the same potential losses in the future (Gabbott et al., 2011; Lazarus & Folkman, 1987). The higher the severity of the failure, the harder consumers to forgive (Riek & Mania, 2012; Wade & Worthington, 2003) and the more consumers attribute blame and responsibility toward the perpetrator (Coombs, 1995; Laufer et al., 2005). In other words, high severity of failure is expected to increase consumers’ tendency to resolve their discomfort rationally through direct complaint behavior, regardless of to whom the failure is attributed. Instead, when the failure severity is low, consumers tend to forgive to cope with their relatively minor dissatisfaction and discomfort. In this context of SE, consumers will forgive Partners more than Platforms.

H2.

Consumers’ intention to give low rate is higher when the failure is attributed to Platforms than Partners but only when the failure is less severe

NWOM allows consumers to distribute their negative emotions and dissatisfaction after service failure (Wang & Wu, 2013; Wetzer et al., 2007). In contrast to ratings, NWOM is a complaint mechanism that does not directly affect either Platforms or the Partners, so consumers will not feel reluctant to engage in NWOM behavior, even in failures caused by Partners, considering that the Partner will not acknowledge nor receive negative consequences. By engaging in NWOM, consumers will not feel guilty toward Partners or hurt their closeness nor empathy for Partners, knowing that even NWOM may influence other people to have negative perceptions toward Platforms but it will not directly harm Partners who actually caused the failure.

H3.

Consumers’ NWOM intention does not differ whether the failure is attributed to Platforms or Partners, regardless of the level of controllability and severity of failure

The positive bias of consumer evaluations has become a prominent topic in SE context, where we found high ratings and favorable evaluations toward Partners on leading digital platforms (Bulchand-Gidumal & Melián-González, 2020; Zervas et al., 2020) very few low ratings and negative reviews (Hu et al., 2009). A positive bias occurs when consumers who experience service failure or negative experiences with Partners are reluctant to submit low rates and instead keep submitting high rates or not at all (Berg et al., 2020; Fradkin et al., 2021). Previous studies have demonstrated the role of consumers’ perception of controllability and severity of failure on consumer complaint behavior (Folkes et al., 1987; Hess, 2008; Van Vaerenbergh et al., 2014; Weun et al., 2004). Attributions of controllability were found to play the most important role in determining consumer response after service failures, surpassing attributions of locus of causality and stability (Suri et al., 2019; Van Vaerenbergh et al., 2014). When consumers attribute failures to Partners, perceptions of controllability will also affect consumers’ responses, including complaint behavior.

After experiencing a service failure, initially, consumers will evaluate the results, the extent to which the services received have deviated from consumer expectations, and how severe the loss is (Sugathan et al., 2017). Increased failure severity will encourage consumers to (Folkes et al., 1987; Grégoire & Fisher, 2008; Jayasimha & Billore, 2016) conduct further evaluations, one of which is evaluating the attribution of failures, including controllability, in which a high perception of controllability will increase the consumer’s intention to file a complaint to seek redress, warn other consumers and avoid the recurrence of the same failure in the future (Folkes et al., 1987; Grégoire & Fisher, 2008; Jayasimha & Billore, 2016). However, when consumers’ perceptions of controllability and severity of failure are low, SE consumers tend to ignore the inconveniences and losses they experience and do not file complaints.

Regarding to filing complaints directly through ratings in SE services, consumer evaluation of service failures attributed to Partners will be more complicated, where perceptions of controllability will be influenced by perceptions of severity of failure (Laufer et al., 2005). Suri et al. (2019) found that in SE services, low controllability significantly increased consumers’ empathy and forgiveness toward Partners more than toward Platform, whereas the perception of high controllability will reduce the consideration of partners’ independent status with limited resources toward consumer tolerance. Severity of failure also affects the affective and behavioral responses of consumers that when consumers perceive the severity of failure to be low, consumers will forgive more easily and bring up more positive behavior (McCullough et al., 1998; Tsarenko & Tojib, 2012). Due to consumer tolerance which leads to the positive bias of post-service failure evaluations, it takes a high controllability and severity of failure to overcome tolerance and potential guilt of consumers towards Partners to encourage consumers to give low rates to Partners.

H4.

When the failure is attributed to Partners, consumers’ intention to give a low rate is higher when the failure is highly controllable than when it is less controllable, but only when the severity of the failure is high.

H5.

When the failure is attributed to Partners, consumers’ intention to give a low rate is higher when the severity of the failure is high than when it is low, but only when the failure is highly controllable.

Differ with NWOM, which Partners will receive no direct negative consequences, the role of cognitive evaluation of consumers’ NWOM intentions will be simpler. Perceived controllability will affect the consumer’s NWOM intention, as well as the consumer’s perception of the severity of failure (Choi & Mattila, 2008; Kalamas et al., 2008; Swanson & Hsu, 2011; Van Vaerenbergh et al., 2014). When the service failures are attributed to Partners, consumers’ perceptions of controllability and severity of failure will each positively affect NWOM intentions.

H6.

When the failure is attributed to Partners, consumers’ NWOM intention is higher when the failure is highly controllable than when it is less controllable

H7.

When the failure is attributed to Partners, consumers’ NWOM intention is higher when the severity of failure is high than when it is low

3. Methods

3.1. Data collection procedure

This study adopted a scenario-based experimental design on ridesourcing services to examine differences in consumers’ intention to submit low rates and engage in negative word of mouth based on their perceptions of locus of causality, controllability and severity of the failure incidents. Experiment is commonly employed in social sciences such as psychology and marketing (Oh et al., 2004). This method involves the manipulation of one or more independent variables to observe their impact on the dependent variables (Malhotra, 2006). It provides researchers with the ability to control extraneous variables, particularly in laboratory settings, enabling them to precisely examine causal relationships (Fong et al., 2016). In the context of service failure, experiment offers a more suitable approach for mitigating memory bias, in which participants are exposed to scenarios of failure incidents, allowing them to provide targeted responses (Smith et al., 1999). Memory bias is a crucial factor that may distort findings, especially related to failure incidents that occurred in the past, where consumers may have difficulties to accurately recall the exact conditions that prevailed at that moment.

Two hundred and eighty participants (47% male) who used ridesourcing services more than three times in 2 months (49% > 7 times) were recruited. All participants voluntarily participated and signed the informed consent form before the experiment started. Participants were randomly assigned in a 2 (locus of causality: Platforms vs Partners) × 2 (controllability: high vs low) × 2 (severity of failure: high vs low) between-subject laboratory experiment design, where participants were divided into eight groups, each was presented with a different pre-tested scenario. The scenario described a service failure incident in a ridesourcing setting, wherein the participant was asked to visualize her/himself experiencing the incident then rate their intention to give low rate and spread negative word of mouth. Due to the limited research related to consumers’ rating intention, this study uses a one-item scale that directly asked participants to rate their intention to give low rate. Meanwhile, intention to spread negative word of mouth was measured using a three-item scale adopted from Wan (2013) and Jayasimha and Billore (2016).

3.2. Manipulation checks and measurement

Manipulation check procedures were performed on participants. Using 7-point Likert scale items, participants were asked to determine whether participants attribute the incident in the scenario to Platform or Partner (1 = Platform, 7 = Partner), whether the cause of the incident was highly or lowly controllable (1 = low, 7 = high) and whether the severity of failure is high or low (1 = low, 7 = high). Results indicated a significant difference in locus of causality, wherein participants assigned to failure incident attributed to Platform blamed the Platform, while participants assigned to failure incident attributed to Partner blamed the Partner (M_Platform = 2.11 vs M_Partner = 6.02; t(140) = −54.205, p < 005). Likewise, significant differences were also found related to participant’s perceived controllability (M_High = 5.97 vs M_Low = 2.13; t(140) = 52.901, p < 005) and severity of failure (M_High = 6.37 vs M_Low = 2.29; t(140) = 51.198, p < 005) (Table 1).

Table 1. Results of manipulation checks

Manipulated variables	Category	N	Means	ANOVA
				F	Sig
Locus of Causality	Platforms Partners	140 140	2.11 6.02	2938.188	0.000
Controllability	High Low	140 140	5.97 2.14	2798.526	0,000
Severity of Failure	High Low	140 140	6.38 2.27	2967.449	0.000

Intention to submit low rates was measured with single-item question as indicator, while negative word of mouth intention was measured using three questions as indicators. Then, reliability and validity tests for three measurement items of negative word of mouth intention were conducted. Result indicated that all items were reliable (α = 0.853) and valid (FL > 0.7) to measure negative word of mouth intention (Table 2).

Table 2. Results for measurement items

Dependent variable	Indicators	Factor loadings	Cronbach alpha
Negative Word of Mouth	NWOM 1 NWOM 2 NWOM 3	0.916 0.905 0.813	0.853

3.3. Results

3.3.1. Main and interaction effects

General Linear Method was used to identify the main and interaction effects of participants’ perceived locus of causality, controllability and severity of failure on participants’ intention in giving low rate and spreading negative word of mouth. Results indicated that participants’ perceived locus of causality only affected participants’ intention to give low rates. While controllability and severity of failure independently affect both participants’ intention to submit low rates and engage in negative word of mouth. Furthermore, interaction effects were found between the two attributions and severity of failure on participants’ intention to submit low rates; however, interaction effect on participants’ intention to engage in NWOM only found between controllability and severity of failure (Table 3).

Table 3. Main and interaction effects of independent variables on dependent variables

Dependent variables	Effects	Independent variables	F	Sig	Partial η^2
Submitting Low Rates	Main	Locus of Causality (LoC) Controllability (Cr) Severity of Failure (SoF)	161.783 95.051 61.034	0.000 0.000 0.000	0.373 0.259 0.183
	Interaction	LOC x Cr LOC x SoF Cr x SoF LoC x Cr x SoF	13.544 6.341 19.847 0.204	0.000 0.012 0.000 0.652	0.047 0.023 0.068 0.001
Negative Word of Mouth	Main	Locus of Causality (LoC) Controllability (Cr) Severity of Failure (SoF)	3.725 82.640 213.902	0.055 0.000 0.000	0.014 0.233 0.440
	Interaction	LOC x Cr LOC x SoF Cr x SoF LoC x Cr x SoF	0.309 0.376 9.531 2.802	0.579 0.540 0.002 0.095	0.001 0.001 0.034 0.010

ANOVA results confirmed the effects of participants’ perceived locus of causality, controllability and severity of failure toward complaint intentions, in which significant differences were found in participants’ intention to submit low rates (F = 51.115, p < 0.05) and engage in negative word of mouth (F = 44.769, p < 0.05) formed in eight groups of experiment (Table 4).

Table 4. Test of differences among eight groups of participants

Dependent variables	ANOVA
	F	Sig
Submitting Low Rates	51.115	0.000
Negative Word of Mouth	44.769	0.000

3.3.2. Hypotheses testing

Next, we did the hypothesis testing using ANOVA to determine the differences between groups of participants related to the intention to give low rate and spread negative word of mouth. Testing hypotheses 1, participants’ intention to give low rate is higher when the failure incident was attributed to Platform than Partner at all levels of controllability (High Controllability: M_Platform = 4.157 vs M_Partner = 2.342; F = 85.000, p < 005; Low Controllability: M_Platform = 4.157 vs M_Partner = 2.342; F = 85.000, p < 005) and severity of failure (High Severity: M_Platform = 3.986 vs M_Partner = 2.300; F = 67.384, p < 005; Low Severity: M_Platform = 2.843 vs M_Partner = 1.714; F = 44.081, p < 005); hence, H1 and H2 were not supported. However, no significant difference found in participants’ intention to spread negative word of mouth regardless the level of controllability (High Controllability: M_Platform = 3.662 vs M_Partner = 3.533; F = 0.445, p > 005; Low Controllability: M_Platform = 2.862 vs M_Partner = 2.629; F = 2.074, p > 005) and severity of failure (High Severity: M_Platform = 3.919 vs M_Partner = 3.795; F = 0.570, p > 005; Low Severity: M_Platform = 2.605 vs M_Partner = 2.367; F = 2.832, p > 005), hence H3 was fully supported.

Next hypotheses elaborate the interaction effect of consumers’ perception of controllability and severity of failure that attributed to Partners on consumers’ intention to submit low rates toward Partners. After experiencing service failures that caused by Partners, consumers’ intention to give low rate is higher when the controllability was perceived high than low but only when consumers perceived the loss was highly severe (High Severity: M_HighControl = 2.857 vs M_LowControl = 1.743; F = 30.172, p < 005; Low Severity: M_HighControl = 1.829 vs M_LowControl = 1.600; F = 1.312, p > 005), this results give supports to H4. The intention to give low rates was also found higher when the severity is high that low but only when the controllability was high (High Controllability: M_HighSeverity = 2.857 vs M_Lowseverity = 1.829; F = 23.639, p < 005; Low Controllability: M_HighSeverity = 1.743 vs M_Lowseverity = 1.600; F = 0.564, p > 005), which supported H5.

Results on hypothesis related to NWOM showed that, consumers’ intention to engage in NWOM was higher when the severity of failure was perceived high regardless the level of controllability (High Controllability: M_HighSeverity = 4.315 vs M_Lowseverity = 2.752; F = 77.762, p < 005; Low Controllability: M_HighSeverity = 3.276 vs M_Lowseverity = 1.982; F = 49.168, p < 005) also when the failures were perceived as highly controllable regardless the level of severity (High Severity: M_HighControl = 4.315 vs M_LowControl = 3.277; F = 34.902, p < 005; Low Severity: M_HighControl = 2.752 vs M_LowControl = 1.982; F = 17.150, p < 005); therefore, H6 and H7 were supported.

4. Discussion

The study elaborates on the effect of attributions and severity of failures on consumers’ direct and indirect complaint intention, which complements previous SE service-based studies. The study provides empirical supports to the implication of attribution theory on consumer complaint behavior and the role of failure severity in consumers’ cognitive evaluation on failure attributions. The results not only support a positive bias in consumer evaluations toward Partners, which potentially brings negative consequences for Platforms, but also determine the role of cognitive factor dynamics to that positive bias.

In line with previous studies, attribution of locus of causality was found to have an effect on SE consumer complaint intentions, both directly through the rating mechanism and indirectly through NWOM, where consumers’ intentions to give low rates and NWOM are higher when service failures are attributed to Platforms compared to Partners. Differ from findings from Suri et al. (2019), results in this study reveal that consumers are more forgiving when failures are attributed to Platforms than Partners at all levels of controllability, as their intentions to complain directly through ratings are higher when failures are attributed to Platforms regardless the level of controllability. Likewise, consumers are also more motivated to give low rates when failures are attributed to Platforms, whether the severity of failures is high or low. In other words, despite the level of controllability and severity of failure, consumers’ intention to submit low rates toward Platform is higher than Partners.

In addition to considering negative consequences for Partners, this reluctance reflects consumers’ preference of complaint behavior, that in the conventional setting when consumers perceive that their bad experiences are caused by firms instead of direct service providers, consumers prefer to file complaints directly against the company in the hope of getting a solution and clearer outcomes (Balaji et al., 2016; Kim et al., 2003). Consumers may also consider that low rates will not decrease Platforms’ well-being but will restore consumer well-being after the failures, as well as provide constructive inputs so that Platform can use to improve the service quality in the future (Singh, 1988).

These results also show that Indonesian consumers tend to be more permissive and forgiving toward Partners, so even in a highly controllable failure, consumers are still reluctant to give direct negative evaluation. A high level of forgiveness is identical to collectivist societies like Indonesia, where individuals have interdependent self-construal, which tend to be more forgiving (Sinha & Lu, 2016; Takaku et al., 2001). This context may account for the disparities of findings compared to Suri et al. (2019) which focused on consumers within individualistic societies characterized by independent self-construal.

Furthermore, 49% of participants used ridesourcing services more than seven times within the last 2 months. Therefore, the reluctance can be the result of repeated consumption on certain platforms in which they are served by different partners (pseudo-relationship). Satisfying experience from previous consumptions and awareness that it will be unlikely that the same partner will serve them in the future may increase consumers’ forgiveness and empathy regardless of the controllability level (McCullough et al., 1998; Sinha & Lu, 2016), as well as reduce negative emotions (Mody et al., 2020).

In contrast, this study reveals that the effect of the attribution of locus of causality on consumers’ intention to engage in NWOM is insignificant. Regardless of the controllability level and severity of failure, consumers’ intention to engage in NWOM when the failures are attributed to Partners is not lower than when failures are attributed to Platforms. Dissimilar with rating and other direct complaint behavior of consumer SE that has been reported to be biased in previous studies (Berg et al., 2020; Bridges & Vásquez, 2018; Bulchand-Gidumal & Melián-González, 2020), NWOM may be perceived to be a “save” choice of post-failure behavior, in a way that NWOM will not affect either Platforms’ or Partners’ well-being, at least not directly nor in the short term. In other words, engaging in NWOM will not harm either consumers’ relationship with Partners or bring up guilty feelings in consumers. By doing NWOM, consumers are not trying to seek redress related to their failures but more to channel their negative emotions that they are reluctant to channel through direct complaint mechanisms, including rating.

Scooping deeper into the phenomenon of consumers’ tolerance toward Partners as direct service providers in SE services, this study examines the role of controllability attribution and severity of failure on consumers’ intention to complain against Partners. The results show significant interaction effect between consumers’ perception of controllability and failure severity on their intention to submit low rates post-service failures attributed to Partners, both of which are crucial to their cognitive evaluation process (Lazarus, 1991; Ruth et al., 2002). High controllable failures encourage consumers to submit low rates for Partners more than low controllable but only when consumers suffer from high failure severity. Differently stated, when the severity of failure is low, consumers are reluctant to give Partners low rates even when they perceive that failures are supposed to be highly controllable by Partners. Also, consumers’ intention to give low rates is higher when the failure severity is high than low but only if the controllability is high. Differently stated, when failures are perceived to be less controllable by Partners, consumers are still reluctant to give Partners low rates even when they perceive the severity of failure is high.

When the severity of failure is low, consumers tend to simplify their cognitive process, disregard the failure attributions and determine their complaint intention based only on the outcome evaluation (Sugathan et al., 2017). Furthermore, when the severity of failure is high, consumers will cope their losses by engaging in a more rational cognitive approach (Gabbott et al., 2011) and put more effort in analyzing failure attributions including controllability, which controllability will affect consumers’ choice and intention of complaint behavior in order to strive for solutions, revenge and reconciliation (Bradfield & Aquino, 1999; Tsarenko & Tojib, 2012). On the other hand, failures that occur within high controllable factors indicate that partners do not have sufficient concern or effort to provide consumers with good services and ignore consumers’ well-being, which will overcome consumers’ tolerance toward Partners (Suri et al., 2019). In contrast, failures that occur within low controllable factors, consumers tend to tolerate the failures considering their awareness of Partners’ limited resources, tangible or intangible (Suri et al., 2019), norms congruity and also their social closeness with Partners (Mallargé et al., 2019; Pera et al., 2019; Shuqair et al., 2021). Moreover, consumers’ perception that Partners are not professionals may lower consumers’ expectation of the service quality which then also widen their zone of tolerance (Mallargé et al., 2019).

Nevertheless, consumers’ intention to engage in NWOM in failures attributed to Partners is higher in high controllable than low controllable incidents, regardless the level of failure severity. Vice versa, consumers’ intention to engage in NWOM in failures attributed to Partners is higher when they perceive a high severity than low severity of failure, regardless the level of controllability. These findings demonstrated a less consideration of consumers related to their intention to engage in indirect complaint behavior.

5. Conclusion

The triadic parties involved in SE services make locus of causality as a crucial attribution factor in determining consumers’ complaint behavior in SE services. Consistent with previous findings, this study shows the effect of locus of causality on consumers’ intention to submit low rates as a form of direct complaint behavior that consumers’ intention is higher in incidents attributed to Platforms than Partners. Nevertheless, this study also reveals that locus of causality does not affect consumers’ intention to engage in NWOM as an indirect complaint behavior that consumers’ intention to engage in NWOM is consistent in incidents attributed to Platforms and Partners.

This study not only supports the previous studies that enhanced the positive bias of consumers’ evaluation of SE services but it also reveals the prominent consumers’ zone of tolerance that even in high controllable failures, consumers’ intention to give low rates to Platforms are higher than to Partners. Highlighting consumers’ permissive behavior toward Partners, this study shows that consumers are most motivated to give low rates in high controllable and severe failures. However, this study shows that the influences of controllability and failure severity on the effect of locus of causality only found significant on consumers’ direct complaint behavior, i.e. submitting low rates but not on indirect complaint behavior, i.e. engaging in NWOM. These findings lend support to the role of consumers’ underlying perception related to Partners’ limited capacity and the consequences for Partners from each direct and indirect complaint mechanisms in determining consumers’ complaint behavior in SE services.

6. Managerial implication

The study shows a latent peril that can potentially disrupt Platforms’ performance in the long term. The reluctance of consumers to communicate the failures they experience due to Partners’ negligence to Platforms will close Platforms’ opportunity to take corrective steps and make efforts to restore consumer satisfaction (Singh, 1988). Platforms should be able to identify factors that cause failure with high controllability and severity of failure to improve the quality of Partners’ services in the future. Furthermore, Platforms should also be aware that consumers’ hesitation to directly voice their complaint may lead to the suppression of negative emotions stemming from unsatisfactory experience following failures that they may channel through indirect complaint mechanisms that are undetected by Platform, including NWOM (Balaji et al., 2016). Nonetheless, it does not imply that consumers that voice their complaints by submitting ratings will not engage in NWOM, as consumers may engage in multiple complaint behavior based on their expected benefits and available resources (Tronvoll, 2012). Hence, Platforms should encourage consumers to provide ratings objectively by ensuring that consumers will receive adequate responses, solutions, even compensation, not only to enhance Platform awareness regarding service failures but also to decrease consumers’ NWOM intentions by restoring consumers’ satisfaction and diminishing consumers’ negative emotion (Kim et al., 2003; Parlamis & Posthuma, 2012).

7. Limitations and further research

This study is not without limitations. This study only emphasizes the intention of giving Partners low rates, which represents direct complaint behavior. Further research can examine the effect of failure attribution on other direct complaint behaviors, such as intentions to write negative reviews, considering that negative reviews require greater effort than just submitting ratings. Currently, NWOM can be spread directly and through social media, which also has the opportunity to have a significant effect not only for Platforms but also for Partners if the NWOM goes viral. This study only accounts for consumers’ intention to spread NWOM directly, future research can examine the intention of SE consumers to channel their complaints through NWOM on social media. This study employs a laboratory experiment method that emphasizes on internal validity, and future research may be executed by employing other data collecting methods, such as survey, to provide external validity.

Correction

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Appendix

Appendix A: Scenarios

Locus of Causality : Platforms
High Controllability vs High Severity of Failure
Imagine you are scheduled to take the Final Examination at 08.00 am. You ordered an online motorcyele cab to take you to campus at 07.00, expecting to arrive 20 minutes before the exam started. But the Application provided a diver whose position was far from your location, so it took 25 minutes to pick you up. You did not cancel the order, fearing it would be difficult to get another driver. As a result, you arrived at the campus at 08.15 and were late for the exam.
High Controllability vs Low Severity of Failure
Imagine you are scheduled to attend a class at 08.00 in the morning. You ordered an online motorcycle cab to take you to campus at 07.00, expecting to arrive at campus 20 minutes before the exam started. But the Application provided a driver whose position was far from your location, so it took 25 minutes to pick you up. You did not cancel the order, for fear that it would be difficult to get another driver. Even so, you made it to campus at 7:50 a.m., 10 minutes before class started.
Low Controllability vs High Severity of Failure
Imagine you are scheduled to take the Final Exam at 08.00 am. You ordered an online motorcycle cab to take you to campus at 07.00, expecting to arrive at campus 20 minutes before class started. But due to conditions on a weekday morning and rain, you got a driver 25 minutes later. As a result, you arrived on campus at 08.15 and were late for your exams.
Low Controllability vs Low Severity of Failure
Imagine you are scheduled to attend a class at 08.00 am. You ordered an online motorcycle cab to take you to campus at 07.00, expecting to arrive at campus 20 minutes before class started. But due to conditions on a weekday morning and rain, you got a driver 25 minutes later. Even so, you made it to campus at 7:50 a.m., 10 minutes before class started.
Locus of Causality : Partners
High Controllability vs High Severity of Failure
Imagine you are scheduled to take the final exam at 08.00 am. You ordered an online motorcycle cab to take you to campus at 07.00, expecting to arrive at campus 20 minutes before the exam. But the driver did not go by the road from the map and said he would take a shortcut. However, it turned out that the driver did not know precisely the intended shortcut, causing him to get lost. As a result, you arrived at campus at 08.15 and were late for the exam.
High Controllability vs Low Severity of Failure
Imagine you are scheduled to attend a class at 08.00 a.m. You ordered an online motorcycle cab to take you to campus at 07.00, expecting to arrive at campus 20 minutes before the exam. But the driver did not go by the road from the map, and said he would take a shortcut. However, it turned out that the driver did not know precisely the intended shortcut, causing him to get lost. Even so, you made it to campus at 7:50 a.m., 10 minutes before the class started.
Low Controllability vs High Severity of Failure
Imagine that you are scheduled to take the final exam at 08.00 am. You ordered an online motorcycle cab to take you to campus at 07.00, expecting to arrive at campus 20 minutes before class started. The driver went directly to your location, but due to the morning conditions and the unusual traffic jams, the driver arrived to pick you up 25 minutes later. As a result, you arrived on campus at 08.15 and were late for your exams.
Low Controllability vs Low Severity of Failure
Imagine you are scheduled to attend a class at 08.00 a.m. You ordered an online motorcycle cab to take you to campus at 07.00, expecting to arrive at campus 20 minutes before class started. The driver went directly to your location, but due to the morning conditions and the unusual traffic jams, the driver arrived to pick you up 25 minutes later. Even so, you made it to campus at 7:50 a.m., 10 minutes before class started.

Appendix B: Questions for Manipulation Check of Independent Variables (7-point scale) Original questions were in Indonesian

Independent Variables	Manipulation Check Questions
Locus of Causality	In you opinion, the failure was caused by . Application Partners 1 2 3 4 5 6 7
Controllability	In your opinion, the failure was supposed to be. Very unpreventable Very preventable 1 2 3 4 5 6 7
Severity of Failure	The failure caused you damage and inconvenience … Very minor Very major 1 2 3 4 5 6 7

Appendix C: Indicators for Dependent Variables (temp-point scale, strongly disagree/strongly disagree) Original questions were in Indonesian

Dependent Variables	Codes	Indicators
Submitting Low Rates	SLR 1	On a rating scale of 1–5, I will rate the driver/Application 1–2.
Negative Word of Mouth	NWOM 1	I will tell my relatives and family about the failure of this service
	NWOM 2	I will tell my relatives and family about the source of the failure of the service
	NWOM 3	I will warn my relatives and family not to use this service