et al., 2020, p. 1211). That is, SDCs are able to sense the driving environment and operate without human intervention using AI-based systems (Sestino et al., 2022) .

SDCs have the potential to significantly improve the safety and efficiency of mobility (Park et al., 2021) , including by reducing users’ travel times, traffic congestion, accident rates, and fuel consumption and emissions (Herrmann et al., 2018) . However, as SDC adoption remains at the (pre-) introduction stage of the product life cycle (PLC), consumer motivations, engagement and behaviors relating to this technology remain tenuous, thus warranting further investigation.

Several attempts have been made to understand the factors influencing SDC adoption. For example, existing studies have identified a key role of consumer-perceived SDC usefulness, ease of use and trust (e.g. Zhang et al., 2020) . However, as these works have focused on SDC-related purchase decision-making, little remains known regarding consumers’ ongoing SDC engagement, meriting further exploration. Specifically, further insight into consumers’ SDC engagement, defined as a consumer’s (e.g. cognitive, emotional and/or behavioral) resource investment in his/her SDC interactions (Hollebeek et al., 2019) , is expected to clarify their SDC-related dynamics, thus offering vital insight to theoreticians and SDC manufacturers.

Thus, while the literature suggests that consumer-perceived SDC safety and control represent major adoption-related concerns (Ha et al., 2020) , we expect these variables to exert an ongoing effect on consumers’ SDC engagement, which – however – remains tenuous to date, exposing an important research gap. That is, while perceived safety is pivotal at any stage of a consumer’s relationship with his/her vehicle (Osswald et al., 2012) , its role for SDCs necessitates further investigation. Moreover, while traditional vehicles offer elevated user control and responsibility (Schneble and Shaw, 2021) , SDCs, by minimizing driver intervention (Lee et al., 2019) , are expected to see lower user-perceived control (vs conventional cars), thus potentially altering their engagement with these vehicles. Therefore, while consumers are traditionally accustomed to being in control of their driving, SDCs’ autonomous nature requires consumers to cede their vehicle-related control to an automated system (Baccarella et al., 2020) that we anticipate to impact their SDC engagement, as explored in this study.

To investigate these issues, we adopt a theory of planned behavior (TPB)-informed perspective that links individuals’ attitudes to their object-related behavior and engagement (Ajzen, 1991; Bitter et al., 2014) . The theory proposes a key role of perceived behavioral control (PBC), the extent to which an individual believes he/she has the resources, ability and opportunity to perform a behavior, in impacting his/her subsequent intent to perform the behavior (Ajzen, 1991) . These issues remain nebulous in the SDC-based engagement context, thus offering a suitable lens to explore the outlined issues. Moreover, we expect PBC to be impacted by the individual’s perceived SDC safety (Roy et al., 2018) , or the extent to which SDC “drivers or passengers . . . feel relaxed, safe and comfortable while driving” (Xu et al., 2018, p. 323) that also merits further exploration. An under-explored SDC-based tension, therefore exists: On the one hand, SDCs may feel safer than traditional cars (e.g. given their reduced proclivity for errors/accidents; Hollebeek et al., 2021) , leading users to potentially prefer driving an SDC. However, on the other hand, consumers may feel uncomfortable relinquishing driving-related control (Ha et al., 2020) , potentially yielding their preference for traditional cars (vs SDCs). In this study, we further explore, and elucidate, this tension.

This paper makes the following contributions to the SDC and consumer engagement (CE) literature. First, deploying a TPB-informed perspective, we empirically investigate the relationship of consumer-perceived SDC-related behavioral control and engagement on their intent to purchase an SDC. The findings suggest a positive relationship between consumers’ PBC and their SDC purchase intent. Specifically, consumers perceiving greater SDC-related control are more inclined to purchase an SDC, echoing Hollebeek et al.’s (2022a) findings for consumers’ video-game engagement. The results also substantiate engagement’s mediating role in the association of PBC and their SDC purchase intent. Consumers’ elevated behavioral control, thus, yields higher SDC engagement, in turn raising their SDC purchase intent, corroborating prior SDC-related findings (e.g. Hollebeek et al., 2022b) . Moreover, the findings suggest that, even prior to a product’s mainstream market introduction, consumers may already engage with it, thus offering further novel insight. These findings are useful to SDC manufacturers, given their vested interest in SDC adoption and engagement.

Second, we assess the potentially moderating role of consumer-perceived safety in the association of PBC and SDC engagement. We hypothesize that the association of consumerbehavioral control and engagement is mitigated depending on whether consumers perceive SDCs to be safer (vs less safe). We use perceived SDC safety as a moderator because userperceived safety is vital in high-risk purchases, including SDCs (Nunes et al., 2018) . The findings show that perceived safety exerts a negative moderating effect on the association of consumer-perceived behavioral SDC control and SDC engagement. That is, consumers perceiving an SDC to afford them higher (vs lower) driving-related control tend to display higher SDC engagement. However, this association is weaker for consumers who perceive SDCs to be safer (vs less safe). That is, the role of PBC in generating engagement is more important for consumers who perceive SDCs to be less safe (vs safer), revealing pertinent implications. The findings, thus, reveal a key interplay between SDC-related safety and PBC, particularly in terms of raising SDC-related behavioral control for those perceiving SDCs as less safe.

The paper is structured as follows. Section 2 reviews the relevant literature, followed by our hypothesis development in Section 3. Section 4 outlines the deployed methodology, followed by a summary of the results in Section 5. In Section 6, we outline key implications that arise from our work.

Extending the theory of reasoned action (Fishbein and Ajzen, 1975) , the TPB posits that individuals’ behavior is influenced by their attitude toward a behavior, subjective norms and PBC (Ajzen, 1991) . While consumer attitudes refer to their relatively permanent disposition toward an object (Raimondo and Farace, 2013) , subjective norms reflect an individual’s perceived social pressure to (not) engage in a behavior (Fu et al., 2010) . We argue that, for SDCs, consumers’ attitudes and subjective norms are still in the formation phase, given SDCs’ current developmental/introductory PLC stage. For example, SDC-related social pressure is expected to be limited to date, as SDCs are yet to enter mainstream consumer markets globally (Poo and Dalziel, 2016) , leaving people with relatively low SDC-related knowledge and opinions to date.

However, we posit that PBC, defined as a user’s experienced level of control over an action or behavior (e.g. driving an SDC), may, indeed, affect the individual’s behavior, or behavioral intention (Chen and Yan, 2019) . Relatedly, while prior authors have investigated consumers’ SDC adoption from a TPB perspective (Gkartzonikas et al., 2022; Jing et al., 2019) , these have tended to focus on consumers’ SDC adoption, yielding limited insight into consumers’ ongoing SDC engagement, in particular vis-a-vis their perceived SDC-related behavioral control and safety. We next review the key CE literature.

While CE has gained prominence in the literature (Menidjel et al., 2022; Naqvi et al., 2021a) , its conceptualization remains debated. For example, while Brodie et al. (2011, p. 258) conceptualize CE as a consumer’s “psychological state, which occurs by virtue of interactive [consumer] experiences with a focal agent/object,” Hollebeek et al. (2019, p. 166) define it as a consumer’s “investment of operant [i.e. cognitive, emotional, behavioral] . . . and operand (e.g., equipment-based) resources . . . in [his/her] brand interactions.” Despite this dissent, most definitions concur regarding engagement’s interactive conceptual core (Hollebeek and Belk, 2021) .

Second, CE is typically viewed as a multidimensional concept comprising cognitive, emotional and/or behavioral components (e.g. Munaro et al., 2021) . Here, cognitive engagement reflects a consumer’s “level of brand-related thought processing and elaboration in a . . . brand interaction;” emotional engagement represents a consumer’s “degree of positive brand-related affect in a . . . brand interaction;” behavioral engagement reflects a consumer’s “level of energy, effort, and time spent on a brand in a . . . brand interaction” (Hollebeek et al., 2014, p. 154; Harrigan et al., 2018) .

Third, while the engagement subject is made explicit in the CE concept (i.e. the consumer), its engagement object is left implicit (Hollebeek, 2011) . The latter can, therefore, comprise different objects, including a brand, firm or product (e.g. an SDC; Vivek et al., 2014) . While published work offers insight into consumers’ engagement with particular products or brands (e.g. Hollebeek et al., 2022a) , little remains known regarding their engagement with SDCs (Clark and Feng, 2017) , as therefore examined in this paper. We next develop our research hypotheses.

PBC, or the extent to which an individual believes he/she has the resources, ability and opportunity to perform a behavior, is a critical determinant of consumers’ behavioral intentions (Ajzen, 1991) . Under high behavioral control, consumers feel their behavior has an extensive effect on their environment (Parkinson et al., 2017) . For example, those who believe to have high behavioral control may seek to turn around a firm’s unethical behavior. These consumers also tend to speak positively about the brand, have favorable brand attitudes, exhibit greater positive (and fewer negative) brand-related emotions and tend to pay more for the brand. Conversely, those who feel they have low behavioral control are more doubtful that their actions will have an impact on their environment (Ajzen, 1991) . For example, in less competitive markets, buyer bargaining power tends to remain low, as many other buyers are available that can disempower individuals or give rise to their less positive emotions (De Ruyter and Semeijn, 2002) .

In the TPB, individuals are predicted to be more likely to perform a behavior if they believe they have the ability to perform it well (Hollebeek et al., 2022a) . For SDCs, PBC is expected to be primarily linked to the individual’s perceived ability to control the autonomous vehicle (J€orling et al., 2019) . For example, if consumers lament their loss of driving-related control (vs traditional cars), their PBC is likely lowered, rendering them less likely to (intend to) use SDCs (Chen and Yan, 2019) . It is, then, expected that consumers perceiving higher SDC-related control are likely to display a greater intent to purchase an SDC (Gkartzonikas et al., 2022; Jing et al., 2019) . We hypothesize the following:

H1. A positive association exists between consumers’ perceived SDC-related behavioral control and their SDC-related intent to purchase an SDC.

While prior research has typically focused on PBC as a direct determinant of consumers’ behavioral intent (Yang, 2012) , we propose this association may be mediated by CE, as follows: A consumer’s PBC is likely to impact (i.e. raise) the individual’s resource investment (i.e. engagement) in his/her SDC-related interactions (Pansari and Kumar, 2017) . That is, the higher one’s perceived SDC-related control, the higher his/her expected SDC engagement, in turn impacting (i.e. enhancing) the individual’s intent to purchase an SDC (Villagra et al., 2021) . We posit the following:

H2. CE mediates the association of PBC and their SDC-related intent to purchase an SDC.

Perceived safety, or the extent to which an individual feels safe in a situation (e.g. while driving an SDC; Blut et al., 2021) , has been found to play a critical role in shaping consumer behavior (Roy et al., 2018) . Perceived safe products/brands are, therefore, conducive to the formation of positive attitudes to, satisfaction with and elevated behavioral intent toward the object and vice versa (e.g. Featherman et al., 2021; Zhang et al., 2019) . For SDCs, perceived safety is pivotal in affecting consumers’ product acceptance and choice (Park et al., 2021) .

As noted, SDCs were developed to free up their users from driving, reducing their driving burden (Sparrow and Howard, 2017) . These cars drive autonomously, while independently executing driving functions. When using SDCs, consumers may feel concerned about the car’s safety, given their lacking mobility-related autonomy and control (Baccarella et al., 2020) . In particular, autonomous cars without a steering wheel, pedals or a gear box are likely to elicit consumers’ perceived lack of SDC-related control, potentially raising their SDCrelated safety concerns (Meyer-Waarden and Cloarec, 2022; Xu et al., 2018) . Moreover, consumers may question SDC safety under failing connectivity (Shetty et al., 2021) , maintaining their SDC engagement at potentially modest levels (Nunes et al., 2018) .

We posit that the association of PBC and engagement will pan out differently for consumers perceiving SDCs to feature high (vs low) safety. For example, those that perceive SDCs as unsafe may make mental reference to SDC testing resulting in incidents (e.g. a woman being killed in an SDC-related accident in Arizona; McCausland, 2019). These individuals may be also concerned about the efficiency of an SDC’s driving/safety features, including its automatic emergency braking, traffic sign recognition, lane-keeping capability, hazard detection and avoidance, etc. (Baccarella et al., 2020) , affecting the association of their perceived SDC-related control and engagement. Conversely, consumers who feel SDCs are relatively safe to use exhibit significantly fewer concerns that may, likewise, impact the association of their SDC-related PBC and engagement. Therefore, the effect of SDC-related PBC on engagement is mitigated, depending on whether consumers perceive SDCs as safer (vs less safe). We postulate the following:

H3. Perceived safety significantly moderates PBC’s positive effect on their SDC engagement.

The National Highway Traffic Safety Administration (NHTSA) defines five levels of SDC automation (Anderson et al., 2016) . Levels 0–2 cover no automation (i.e. drivers completely control the car’s functions), to combined-function automation that – despite automating some driving functions (e.g. steering/acceleration) – still require drivers to remain attentive in resuming control at any time. Levels 3–4 represent limited and full self-driving automation, respectively. As Levels 0–2 are already deployed in conventional cars worldwide, we designed our experimental scenarios based on Vehicle Levels 3–4 that are at the (pre-) introductory stage in the PLC (Zhao et al., 2018) .

Our scenarios describe a situation where the consumer is using a level 3–4 SDC that controls and executes the vehicle’s driving and safety assessment functions from start to finish, without any human intervention (Baccarella et al., 2020; J€orling et al., 2019; see Appendix 1) . We used scenario-based methodology to examine consumers’ SDC-related engagement because it, first, reduces potential bias (Park et al., 2014) , while also offering elevated internal validity by allowing researchers to control for the effect of other extraneous factors impacting the results (King and Auschaitrakul, 2021) . Second, it is difficult to capture consumers’ SDC interactions in real-world settings, as SDCs are not yet widely available (Payre et al., 2021) , yielding its widespread adoption in SDC- and marketing research (e.g. J€orling et al., 2019) . Correspondingly, none of our participants had used an SDC to date, suggesting the suitability of adopting a hypothetical scenario in this study.

Our scenarios were designed based on validated literature-based SDC scenarios (Baccarella et al., 2020; J€orling et al., 2019) . To validate the scenarios, several colleagues were invited to assess, and check, the respective scenarios’ content for clarity and relevance. Based on their feedback, necessary adjustments were made. Moreover, a pretest was conducted with 36 respondents that indicated that the scenarios sounded realistic, and that their content was clear, concise and understandable (Huang et al., 2020; Liao, 2007) . We deployed the pretest-based scenario validation as a manipulation check validating the interpretation of our scenarios for adoption in the main data collection.

The data were collected by using a self-administered questionnaire distributed through Facebook in January–February 2022. The survey was first pretested with a sample of 36 respondents, 80.56% of which were male (19.44% female) and approximately 61.11% were married (38.89% single). Two-thirds of the respondents were 18–35 years old and 33.33% were 36 or over. All respondents held their full driver’s license and over 86% had at least two years of driving experience.

The pretest results show that the Cronbach’s alphas (CAs) for the modeled constructs exceeded the 0.7 threshold, confirming their reliability (Hair et al., 2010) . Based on the results, minor amendments were made to some of the items’ wording. Overall, the pretest results demonstrated that the survey questions were clear, and easily understood by, the respondents. The findings also showed that the required average time to complete the survey was 10 min, revealing an acceptable respondent burden (Frazer and Lawley, 2000) .

The questionnaire commenced with an introduction outlining the study’s objectives, while also assuring the respondents of the survey’s anonymity, confidentiality and its purely voluntary nature. The introduction also noted that there are no right/wrong answers in the survey. The questionnaire’s second part started with screening questions to ensure the participants were at least 18 years old and held their full driver’s license, followed by one of two scenarios describing different SDC driving situations (see Appendix 1). In each scenario, participants also responded to the items measuring the modeled constructs (see Section 3.4). To administer the survey, the measurement items were randomly ordered across the participants. In the questionnaire’s final section, we requested the respondents’ demographic information (e.g. gender/age).

A total of 368 valid questionnaires (i.e. 186 limited self-driving scenario and 182 full selfdriving scenario) was collected using purposive and snowball sampling (Hair et al., 2017, 2020) . Of these, 77% were male (23% female). Given the non-probability nature of our sampling method, our objective is not to generalize to the entire population (e.g. of Facebook users), but to examine key differences across potential SDC users. In terms of age, 20.65% of the respondents were 18–25 years old, 49.46% were 26–35 years old and 29.89% were 36 years or over. In terms of marital status, 59.51% were single (40.49% married). Moreover, over 76% has held their full driver’s license for at least two years.

We deployed established multi-item scales (see Appendix 2) to gauge the modeled constructs. The items were measured on five-point Likert scales ranging from strongly disagree (1) to strongly agree (5). To measure PBC, Collier and Sherrell’s (2010) four-item scales were adapted to suit our research context (J€orling et al., 2019) . A sample item reads: “I would feel in control using this car.” Second, perceived safety was gauged by using Xu et al.’s (2018) three-item instrument. A sample item reads: “I would feel safe using this car.” Third, CE was measured as a higher-order construct comprising three dimensions: cognitive processing (i.e. cognitive engagement, three items), affection (i.e. emotional engagement, four items) and activation (i.e. behavioral engagement, three items; Hollebeek et al., 2014) . A sample item reads: “I would feel very positive when I use this car.” Finally, intent to purchase an SDC was measured by deploying Zeithaml et al.’s (1996 ) two-item scale. A sample item reads: “If this car is available, I would intend to purchase it in the future.”

The data were analyzed by using partial least-squares structural equation modeling (PLSSEM). For the lower-order constructs (i.e. PBC, perceived safety, engagement and SDC purchase intent), the measurement model’s construct reliability and validity were first checked, yielding the internal consistency, convergent validity and discriminant validity results shown in Table 1 (Hair et al., 2017) . Overall, the findings reveal that all factor loadings exceeded the value of 0.7, exception for PS3 (0.58), which was, however, retained owing to its significant loading (i.e. p < 0.001), suggesting the reliability of the measurement items.

Moreover, CAs and composite reliability (CR) were greater than 0.7, and the average variance extracted (AVE) exceeded 0.5, indicating convergent validity (Hair et al., 2020) . Moreover, discriminant validity was evaluated by examining any cross loadings and by Note(s): Values in italic represent the factor loadings of the measurement items; ***p < 0.001 (based on t (4,999), a two-tailed test) deploying the Fornell–Larcker test. As shown in Table 1, the item factor loadings exceeded their correlation with the other constructs and the square root of the AVE for each construct exceeded its correlation with each of the other constructs (see Table 2), confirming discriminant validity of the modeled constructs.

We next verified the measurement model for the higher-order CE construct. As shown in Table 3, the variance inflation factor (VIF) values remained under the critical threshold of 5 and all indicator weights, with the exception of COG (i.e. cognitive processing), which was, however, retained owing to its significant factor loading (i.e. p < 0.001), and loadings were significant, validating the higher-order engagement construct. 1. Perceived behavioral control 2. Perceived safety 3. Cognitive processing 4. Affection 5. Activation 6. Intent to purchase an SDC

1 0.83 0.68*** 0.47*** 0.66*** 0.71*** 0.71***

2 0.81 0.52*** 0.73*** 0.72*** 0.74*** 3 4 5

6 0.82 0.74*** 0.66*** 0.66*** 0.89 0.87*** 0.85*** 0.89 0.86***

Table 4 shows PBC’s positive effect on intent to purchase an SDC and engagement, thus supporting H1. CE, in turn, positively impacts consumers’ intent to purchase an SDC. To measure engagement’s mediating effect, a bootstrapping procedure with 5,000 resamples was employed (Hair et al., 2017) . The results show that the direct/indirect effects of PBC on consumers’ intent to purchase an SDC were significant. In addition, the 95% confidence intervals did not contain the value of 0 (see Table 4), suggesting engagement’s partial mediating effect in the association of PBC and intent to purchase an SDC, thus partially supporting H2.

Next, we evaluated the moderating effect of perceived safety in the association of PBC and CE. When testing a moderating effect, the moderator’s (i.e. perceived safety’s) direct effect on the endogenous construct (i.e. engagement) needs to be included. The results in Table 4 show that perceived safety has a direct, positive impact on engagement and a negative moderating effect on the positive relationship of PBC and engagement, supporting H3. That is, the positive effect of PBC on engagement is weaker for consumers who perceive SDCs to be safer (vs less safe), as shown in Figure 3. Specifically, while the green line in Figure 3 represents the respondents’ high perceived SDC safety (0.33), the red line represents their low perceived safety (0.47).

Construct

Indicator

Weight

Loading

VIF Direct effect H1: c’ 5 perceived behavioral control → intent to purchase an SDC a 5 perceived behavioral control → consumer engagement b 5 consumer engagement → intent to purchase an SDC Perceived SDC safety → consumer engagement Mediating effect H2: ab 5 perceived behavioral control → consumer engagement → intent to purchase an SDC Moderating effect H3: a1 5 perceived behavioral control*perceived safety → consumer engagement VIF 2.04 Consumer engagement 0.64 0.49 Intent to purchase an SDC 0.79 0.67 Model fit 0.07 Note(s): VIF: variance inflation factor; NA: not applicable; SRMR: standardized root mean square residual; ***p < 0.001; *p < 0.05 (based on t (4,999) and a one-tailed test)

Furthermore, we evaluated the VIFs, effect sizes (f2), coefficient of determination (R2), predictive relevance (Q2) and standardized root mean square residual (SRMR). As shown in Table 4, the VIFs remained below the value of 5, indicating that collinearity is not an issue in the data (Hair et al., 2017) . To assess the effect sizes f2 for the structural model relationships, the results show that PBC exerts a medium effect (0.25) on CE and a small effect (0.06) on the respondents’ intent to purchase an SDC, while engagement also exerts a large effect (1.45) on their intent to purchase an SDC (Cohen, 1988) . In terms of the moderation effect, the interaction term’s f2 effect size displayed a value of 0.01, indicating a medium effect size (Kenny, 2018) . According to Figure 2, engagement’s R2 value is 0.64, and 0.79 for intent to purchase an SDC. The Q2 values of engagement (0.49) and intent to purchase an SDC (0.67) exceeded the value of 0, confirming the model’s predictive relevance. Finally, at an SRMR of 0.07 (i.e. remaining under the 0.08 cutoff), the model was found to offer a good fit to the data (Henseler et al., 2016) .

Drawing on the TPB (Ajzen, 1991) , this study developed and tested a theoretical model that explores the effect of PBC on SDC engagement and its ensuing effect on SDC purchase intent. Moreover, we examined the potentially moderating role of consumer-perceived safety in the association of PBC and engagement, as outlined. The following main implications for SDC marketing researchers arise from our findings.

First, by examining the effect of PBC on SDC purchase intent, the results provide novel insight into SDCs’ capacity to radically transform users’ driving experience. That is, while consumers are typically found to value having an extensive level of control over their traditional driving experience, this is substantially reduced when driving an SDC, creating an interesting tension that merits further scrutiny (Baccarella et al., 2020; Schneble and Shaw, 2021) . The findings suggest that the higher PBC, the higher a consumer’s intent to purchase an SDC (Gkartzonikas et al., 2022; Jing et al., 2019) , suggesting that SDC manufacturers should boost (vs reduce) consumers’ perceived behavioral SDC-related control (e.g. by allowing users to take over control over their SDC when needed; Herrmann et al., 2018) .

Relatedly, while the TPB suggests that consumers’ behavioral intent is directly affected by their PBC (Ajzen, 1991; Hollebeek et al., 2022a) , as the present SDC-based findings corroborate, consumers’ level of SDC engagement is also found to mediate this association.

That is, PBC exhibits both a direct and indirect effect on the user’s intent to purchase an SDC (i.e. as mediated via engagement), revealing more nuanced insight into these dynamics (Rather et al., 2022; Villagra et al., 2021) . This result makes sense, as the higher one’s perceived control when driving an SDC, the greater one’s expected (e.g. cognitive/behavioral) resource investment in interacting with the SDC, thus boosting the individual’s SDC engagement (Kumar et al., 2019; Hollebeek et al., 2019) . This finding, therefore, counters Papagiannidis et al. (2014) , who suggest a non-significant association of control and engagement.

Moreover, this finding extends the work of Hollebeek et al. (2022b), who examine the role of social influence-based control in impacting stakeholders’ (e.g. consumers’) engagement. While these authors, like the present findings, suggest a key effect of control on engagement, they focus on interpersonal control (vs user-perceived control over a product), thus adding important further SDC-based insight. Specifically, by suggesting engagement as a key mediator in the association of PBC and consumers’ SDC purchase intent, this finding advances scholarly understanding that future researchers are encouraged to validate across contexts. For example, to what extent does consumer-perceived control affect their engagement across (e.g. self-driving technology-based) contexts? In which cases is the direct (vs mediated) effect stronger? Additional factors associated with social influence and/ or the consumer’s psychological state may also be examined to derive further understanding of SDC-related engagement and purchase intention (e.g. Jiang et al., 2020; Naqvi et al., 2020, 2021b; Yushi et al., 2018) .

Second, we examined the potentially moderating effect of consumer-perceived SDC safety in the association of their PBC and engagement (see Figure 1). Consistent with H3, this finding shows that perceived safety moderates the positive effect of PBC on engagement. Specifically, the interaction effects shown in Figure 3 reveal that the effect of PBC on engagement is stronger for consumers who perceive SDCs as less safe (vs safer), revealing pertinent insight. That is, the role of perceived control in generating engagement is more important for those consumers who perceive SDCs to be less safe (vs safer). This finding is plausible, because individuals who view SDCs to be less safe will consider their perceived behavioral SDCrelated control as a major factor in their developing SDC engagement, thus extending existing research identifying the paramount role of perceived vehicle safety to the SDC context (Nunes et al., 2018) . Put differently, if a consumer’s perceived SDC safety is high (vs low), his/her perceived behavioral SDC-related control may not necessarily raise his/her SDC engagement and vice versa.

Finally, this study is among the first to empirically examine consumers’ engagement with SDCs prior to, or at the point of, the latter’s market introduction (i.e. well before SDCs become mainstream or enter the growth phase of the PLC). Therefore, the reported findings are important for SDC theoreticians (Fu et al., 2010) . More broadly, the findings show that even prior to a product’s market introduction, consumers may already engage with it, reflecting their pre-usage, advance or anticipation engagement.

This study also generates pertinent managerial implications for SDC manufacturers and marketers. First, the results demonstrate the importance to design products that feature high (vs low) SDC-related PBC, which is – in turn – expected to raise their SDC engagement. To do so, SDC marketers may wish to educate their consumers regarding how to take control over their SDC (e.g. in cases of emergency), including through the provision of detailed SDC manuals and consumer support and training facilities (Tyagi and Aswathy, 2021) .

Second, we advise marketers to favorably shape consumers’ SDC safety perceptions, which can, for instance, be done by emphasizing specific SDCs as safer alternatives to conventional cars (Anderson et al., 2016; Herrmann et al., 2018) . This suggestion is important, given our finding that consumer-perceived SDC safety boosts their SDC engagement, as shown in Table 4. That is, the safer consumers perceive an SDC to be, the greater their SDC engagement is. In other words, those consumers who perceive SDCs as safer (vs less safe) are likely to display higher SDC engagement, in turn favorably impacting their intent to purchase an SDC, thus offering important benefits to marketers. For example, engaged consumers have been shown to spend more, offer greater positive word of mouth and to be willing to pay a price premium (Brodie et al., 2011; Pansari and Kumar, 2017) , revealing their strategic importance for SDC manufacturers. We, thus, recommend that SDC marketing efforts to focus on nurturing consumers’ SDC engagement as key metric to support their SDC adoption.

Third, the findings have implications for policy makers, including those tasked with raising SDC accessibility to the public. For example, legislators may wish to put forward new legal bills or requirements that command SDC users to fully understand and be able to skillfully operate, their SDC (e.g. through SDC-related driver testing or examination that differs from traditional driver’s licensing requirements; Nunes et al., 2018) . Relatedly, as SDC uptake rises, updates or upgrades to roading infrastructure are also required (e.g. by adding more 5G communication networks; Herrmann et al., 2018) , thus offering further policy implications.

Despite its contributions, this study also incurs limitations that offer further research opportunities. First, this study conducted a scenario-based experiment to explore the proposed TPB-informed model in the SDC context. While the findings are important for SDC researchers and marketers, future researchers may wish to extend our results, including by conducting quasi- or field experiments in real driving situations. For example, further research may invite their participants to drive an actual (e.g. Tesla, Pony.ai or Waymo) SDC and gauge their responses to the modeled constructs.

Second, the data were collected from potential (vs actual) SDC users, given SDCs’ limited availability to date, as outlined. However, despite SDCs being at the (pre-)introduction stage of the PLC, the category is expected to rapidly develop in the coming years (Anderson et al., 2016; Nielsen and Haustein, 2018) . Future research may, therefore, replicate the proposed model in or across different new product or service categories (e.g. smart devices/automated personal assistants, etc.).

Third, we deployed cross-sectional data to investigate the proposed research issues. Therefore, further researchers may adopt longitudinal research designs to track the proposed dynamics over time, thus yielding further insight. Relatedly, researchers could adapt the proposed model (e.g. by adding constructs including consumer trust, perceived risk or those that explicate consumers’ sentiment toward different SDC brands, such as brand love).

Fourth, while we explored the potentially moderating role of consumer-perceived safety in the association of PBC and engagement, other moderating factors may be applied (e.g. perceived privacy, security or coping appraisal). Likewise, while the present data were collected during the pandemic, pandemic-related issues were not directly explored. However, the pandemic context may heighten or alter consumers’ SDC-related perceptions. For example, a rising demand for driverless taxis may be observed that reduces the risk of contracting COVID-19 (Dogerlioglu-Demir et al., 2022) . Consequently, future examination of the proposed model with direct reference to relevant pandemic-related issues (e.g. social distancing) may yield important further insight (Kapser et al., 2021) .

Fifth, while we adopted the TPB to guide our analyses, scholars may also use other theoretical perspectives to inform their analyses. For example, given SDCs’ nature as a radical innovation, technology adoption theories may be used, including the technology acceptance model (Davis, 1989) , the technology readiness index (Parasuraman, 2000) or the stimulus– organism–response model (Naqvi et al., 2021a) that have been adopted in related research (e.g. Naqvi et al., 2019, 2021b) . In addition, the theory of reasoned action may be used to explore consumer attitudes and subjective norms in their SDC adoption and usage process (Ajzen and Fishbein, 1980) .

Finally, though the deployed dependent variable was consumers’ intent to purchase an SDC, further researchers may include other potential outcome variables, including product/ brand-related word of mouth, referrals or actual purchase behavior, thus offering further insight.