Social Stratification, Self-Image Congruence, and Mobile Banking in Colombian Cities

Abstract

This study examines the antecedents of the intention to use m-payment applications for lower-income urban users. Using the unique socioeconomic stratification system of Colombia, this is the first study to integrate facilitating conditions and self-image congruence constructs with the technology acceptance model for this consumer group. In elucidating the perceptions of these m-payment users, we give relevance to their personal characteristics and lifestyles and articulate how the self-concept is influenced by sources of information. Perceived usefulness and perceived ease of use were the main direct antecedents, respectively, while risk had less influence. Moreover, facilitating conditions and congruence were significant in explaining intention to use and served as antecedents of resistance for using m-payment apps. The managerial implications are that marketing messaging and application design should thematically integrate representations, images, and expressions shared among these consumers to increase self-image congruence and enhance perceived utility and ease of use of m-payment apps.

Keywords:

• Self-image congruence
• socio-economic stratification system
• mobile banking
• Latin America
• technology acceptance

Introduction

Mobile banking has been conceived as a solution for bringing vulnerable populations closer to economic inclusion (Kansal 2016). The use of this technology has been enabled by high penetration rates of mobile devices and internet access and the development of cost efficient, technology-enabled products and services (Hussain et al. 2019). Yet full adoption of digital payment technologies has not yet been reached for lower income consumers. Worldwide, there were 950 million mobile payment users by the end of 2019, but the number is expected to grow to 1.3 billion by 2023 (de Best 2020). For Colombia, there is an equally interesting uptake of mobile banking occurring. According to the BBVA (2015), mobile banking in Colombia had reached 2.4 million inhabitants, while mobile banking average a 17% growth rate from 2016 to 2019 (Ríos 2021). By mid-2019, there were 4.12 million total electronic savings accounts (CAE), with only 2.15 million active, while the Savings Accounts with Simplified Procedure (CATS) – which have the dual functional of mobile banking and in-store mobile payments (m-payments) – numbered 2.95 million with 1.75 million active accounts.¹ One year later, the CAE accounts increased slightly to 4.16 million, but a higher proportion of them were active (2.39 million). There was a more dramatic increase for CATS, which rose from 2.95 to 7.29 million accounts with 5.45 million active accounts and Bancolombia a la Mano, Daviplata, and Movii as leading financial technologies offering this service in Colombia, particularly targeting lower-income consumers (Palau 2020). The growth of more than 210% for CATS goes hand-in-hand with the government’s increased delivery of subsidies to vulnerable populations, such as Families in Action, and specific subsidies to mitigate the financial effects of the COVID-19 pandemic. This explosive growth was triggered by expectations of receiving government subsidies, making digital payments, limitations on mobility, restrictions on face-to-face financial services – due to social distancing measures implemented during the pandemic – and the growth efforts of banks and Fintech companies. In fact, the 2020 appointment of the new Information and Communications Technology Minister came with the expressed mandate of focusing on telemedicine, telework and mobile banking, the latter of which to facilitate that “financial services can reach Colombia deeper…and have the possibility of marketing through technology” (Correa 2020). However, these factors may not be sufficient to explain the specific determinants of acceptance and use of m-payments. Furthermore, consumers with lower income would likely have less experience with banking and m-payment services and would be more accustomed to using cash derived often through informal work (Jagtap 2019; Hasan et al. 2020), making their circumstances more precarious during the pandemic.

Despite the touted importance of mobile banking for facilitating financial inclusion, research on this consumer group is scarce for urban Latin American markets; which, in 2018, contained 85% of populations – and expected to increase by 2050.² The high degree of urbanisation implies that mobile banking technology design and marketing should be targeted better toward lower income consumers in urban settings (Rahman et al. 2017). Colombia offers an interesting context for examining the intention to use m-payments for poorer urban dwellers because it is the only country in the world to use a socioeconomic stratification system (SES) in urban areas to spatially socio-economically categorise its cities based on the characteristics of the houses, the immediate built environment, and the urban context (Giménez-Santana et al., 2018). The SES was developed in the 1980s and legalised in 1994 under Law 142, article 101.1. This approach was employed in lieu of using data on household income due to the hazards of collecting reliable income data, particularly in poorer areas where the informal economy is active. Using a scale of 1-6, the government subsidises the utility bills and provides other public services for the lower socio-economic areas (strata 1-3) and charges a tax premium on residents living in strata 4-6 (Chica-Olm et al., 2020; Guevara and Shields 2019).

Strata 1-3 were designated as neighborhoods with the poorer conditions, characterised as “low-low”, “low” and “medium-low”, and strata 4-6 were the wealthiest. Despite its intended purpose as a socio-spatial identifier based on the “physical characteristics of dwellings and their immediate environments” (Guevara and Shields 2019, 229), the SES became a de facto socio-economic identifier that characterised standards of living (Bogliacino et al., 2018), with an implicit assumption that neighborhood residents share similar values and interests and have similar economic resources (Chica-Olmo et al., 2020). Furthermore, if there are infrastructure improvements even adjacent to the area, the stratum classification would rise.

Uribe-Mallarino (2008, 147) described the SES as “structured structures that function as structuring structures.” As such, a particular self-awareness is both created and reinforced – in that every urban resident becomes conscious of their socio-economic classification and its associated narratives or stereotypes (Bogliacino et al., 2018). This would inevitably shape the self-image. Stated differently, the SES has a direct role in shaping one’s perception of their own “abilities, limitations, appearance, and characteristics” (Graeff 1996). The SES would also shape the self-concept, which Onkvisit and Shaw (1987, 14, italics added) likened to an individual learning about “capabilities that he cannot achieve as well as skills he is capable of attaining”, thus informing the opportunity set and directing goal prioritization. The potential influence of the SES on technology product adoption then is an interesting backdrop against which to study mobile banking usage in Colombia, particularly through the theoretical lenses of technology acceptance and self-image congruence.

Onkvisit and Shaw (1987, 14) theorized on how the self-self-concept was formed – such that as a person evaluates the environment, they are also evaluating themselves in relation to “others in a socially determined frame of reference.” This results in an interplay between what people think of themselves and what they believe others think of them. The self-concept is also learned through continuous interaction with others and the environment; it is stable and consistent, purposeful, and unique and can result in either positive or negative evaluation outcomes.

Self-image has been theorized as an influence on purchase decisions (Graeff 1996), whereby an object’s image becomes an important criterion for product adoption, shaping attitudes and beliefs about the good, particularly as more relevant information is gathered (Kleijnen, de Ruyter, and Andreassen 2005). Recent studies have examined some determinants for the use of technologies including consumer lifestyles and ecosystem-related variables (Chouk and Mani 2019) or product novelty (Karjaluoto et al. 2019); however, these studies do not explicitly examine specific types of consumers who, in this case, are designated as lower-class through a government-mandated system. This is important because the designation informs the socially determined frame of reference and shapes the self-concept that, in turn, influences product adoption decisions. Therefore, we present a model to explain how urban consumers in the lower SES interact with m-payment technologies through the following research question:

 

Research Question: Is there an association between social stratification, facilitating conditions, and self-image congruence with respect to m-payment acceptance and usage for urban customers in the lower socioeconomic strata?

Our objective is to identify the main antecedents of the intention to use m-payment applications for lower-income consumers in emerging market cities. We achieve this by combining the Technology Acceptance Model (TAM) (Davis, Bagozzi, and Warshaw 1989) with the relevant individual constructs of congruence (Antón, Camarero, and Rodríguez 2013; Karjaluoto et al. 2019; Kressmann et al. 2006), risk (Lee 2009), and resistance to innovation (Laukkanen 2016; Heidenreich and Kraemer 2015; Ram and Sheth 1989). We posit that the resistance arising from psychological and functional barriers reflects a disposition to evaluate the value proposition of mobile banking in determining adoption and use. We also argue that self-image congruence is negatively associated with the psychological barriers of resistance and that facilitating conditions can help explain functional barriers. By incorporating resistance and risk as barriers for adoption behavior into our model we take into account possible constructs that limit acceptance and use of m-payment by lower income users. We also incorporate the positive determinants of congruence, facilitating conditions, perceived ease of use, and perceived usefulness. We found that the total effect analysis for each construct was significant, but perceived ease of use and perceived usefulness were the main antecedents for the intention to use (Davis, Bagozzi, and Warshaw 1989; Antón, Camarero, and Rodríguez 2013; Karjaluoto, et al. 2019; Kressmann et al. 2006). The addition of these constructs along with facilitating conditions will hopefully shed light on how the ease of access, exposure to advertising media and mobile banking in the urban environment might influence acceptance and use behavior among lower-income consumers (Mathur et al. 2018; Gupta and Srivastav 2016).

Theoretical framework

Academic research has addressed acceptance and use of mobile banking from various perspectives, especially mobile applications for using digital money in transactions. A systematic literature review from 2018 to 2021 on this subject area revealed that there were 119 articles published in academic journals listed on the SCOPUS database,³ with 2019 and 2020 as the years with the most prolific number of publications (45%). Of these, about 60% (64 of the 119) were based on technology acceptance models, such as the Technology Acceptance Model (TAM), the Theory of Planned Behavior (TPB), the Unified Theory of Technology Acceptance and Use (UTAUT) for Unified Theory of Acceptance and Use of Technology) (Abdul Aziz et al. 2020; Nur and Panggabean 2021; Malek et al. 2017; Flavian et al. 2020; Hee et al. 2020; Liébana-Cabanillas et al. 2018; Patil et al. 2020; Wong et al. 2021) – all of which were derived from the Theory of Reasoned Action (TRA) (Ajzen and Fishbein 1973; Fishbein and Ajzen 1975). Another 15% analyzed the use of applications for mobile payment from a perspective of various enablers or barriers (Cheng et al. 2019; Zhang, Luximon, and Song 2019; Li et al. 2020; Pal et al. 2020; Rootman and Krüger 2020; Chang and Yeh 2020). In addition, nearly all were quantitative in nature, with the exception of Moghavvemi et al. (2021), who used in-depth interviews to understand acceptance of mobile payments of merchants. The overall trend here shows a growing subject interest in line with the development of the mobile payments industry previously evidenced. However, there is still much to be learned about m-payments as a promising field of study, particularly with growth of this technology expected both in transactions through mobile payments and in academic research.

Given the state of the field, this present study contributes to the technology acceptance model by including self-image congruence and facilitating conditions as relevant antecedents to understand the implications of acceptance and use of mobile payment applications, particularly when this technology is developed for lower income users.

Technology acceptance model

The technology acceptance model (TAM) is a critical model for understanding the intention to use mobile banking. It was derived from the Theory of Reasoned Action (Ajzen and Fishbein 1973) and quickly became the main antecedent of use behavior (Davis 1989). Awa, Ojiabo, and Emecheta (2015, 79) pointed out that TAM provides the basis for “unveiling the impacts of external variables on adoption decisions”, which is relevant for this study given the context of the SES. Indeed, the economic, utilitarian, and attitudinal aspects of this model are manifested in the constructs of ease of use – the perception that technology performance behavior is devoid of effort (Davis 1989; Venkatesh and Davis 2000) – and perceived usefulness, that is, the perceived benefits of using a given technology compared to previous technologies (Venkatesh et al. 2003). As such, TAM established that perceived ease of use (PEOU) influences perceived usefulness (PU) and the intention to use a technology and that perceived usefulness influences the intention to use (IU) (Venkatesh and Davis 2000; Venkatesh and Bala 2008). TAM has been used in explaining mobile banking acceptance and value co-creation (Mostafa 2020) the field of Marketing and Technology Management (Luarn and Lin 2005; Muñoz-Leiva, Climent-Climent, and Liébana-Cabanillas 2017), and in understanding the use of Near Field Communication (NCF) payments – money transfers between two devices at close range – (Bailey et al. 2017; Ooi and Tan 2016; Khalilzadeh, Ozturk, and Bilgihan 2017; de Kerviler et al. 2016) and for QR code payments (Liébana-Cabanillas et al. 2018; de Luna et al. 2019). These studies evidence the positive influence that perceived ease of use and perceived usefulness have on intention to use, as well as the positive influence of perceived ease of use on perceived usefulness.

This relationship was also evidenced for mobile banking (Bhatiasevi 2016; Farah, Hasni, and Abbas 2018; Giovanis et al. 2019; Mutahar et al., 2018). As such, we hypothesize the following:

H1: PEOU positively influences PU

H2: PEOU positively influences IU

H3: PU positively influences IU

It is common to integrate additional constructs into TAM to understand acceptance of a technology within a particular context (Awa, Ojiabo, and Emecheta 2015). Venkatesh and Bala (2008) proposed to include variables related to individual influences that would affect ease of use and perceived usefulness. Hence, the behavior of lower-income users within the context of the socioeconomic stratification system in Colombia could be approached from aspects associated with personality such as the perceived risk of this type of application (Parasuraman and Colby 2014), the congruence with self-image (Kleijnen, de Ruyter, and Andreassen 2005; Kang, Hong, and Lee 2009) and innovation resistance (van Klyton et al. 2021).

Perceived risk (PR)

PR is the uncertainty that users may feel toward a technology because of potential negative consequences or a less than expected outcome (Featherman and Pavlou 2003). Following Pavlou’s (2003) argument, perceived risk can be segmented into environmental and behavioral uncertainty, associated with technology infrastructure and risks associated with the expectation of performance of actors involved in the transaction, respectively. He further argued that from the point of information retrieval to product purchase, both environmental and behavioral risk constitute a general expectation regarding the behavior of the seller and the platform, and the likelihood of protecting the user’s information constituting a unidimensional construct (Pavlou 2003). For mobile payments, perceived risk could be understood as a user’s subjective expectation of financial losses, privacy violations, lack of performance or personal inconveniences – created by the uncertainties of using the service (Forsythe and Shi 2003; Pavlou 2003; Yang et al. 2015). Furthermore, Nepomuceno, Laroche, and Richard (2014) expressed that the intangibility of services increases the perception of risk because they are difficult to evaluate. In emerging market countries, the mobile payment context also includes the lack of experience with mobile banking and financial products representing two types of intangibilities: the absence of cash and the digital delivery of financial services (van Klyton et al. 2021). For lower-income consumers, this would theoretically heighten perceived risk.

These consumers are more likely to engage in informal transactions and thus perceive greater risk in using m-payments. In fact, van Klyton et al. (2021) found that some of these consumers believed that if they lost the mobile phone, they would also lose the digital money. In the context of informal economies, there might be greater hesitation or distrust in the act of giving the information needed to make a mobile transaction or that the perception of losing the money would have a greater impact on them economically – both could result in resistance to mobile banking technologies.

Risk has been shown to negatively influence intention to use technologies, specifically for internet banking (Alalwan, Dwivedi, and Rana 2017). In addition, Jeon, Sung, and Kim (2020) found that perceived risk does not significantly influence acceptance intention for self-service kiosks, in part, because innovativeness had a moderating effect on this relationship. Furthermore, PR is an antecedent for mobile banking acceptance intention to use both directly (Lee 2009; Muñoz-Leiva, Climent-Climent, and Liébana-Cabanillas 2017; Natarajan et al. 2017) and indirectly through the mediation of PU and PEOU (Martins, Oliveira, and Popovič 2014; Mutahar et al. 2018). Although these studies do not explicitly focus on m-payment solutions for low-income and underbanked users, they afford us to propose the following hypotheses:

H4: Risk negatively influences PEOU

H5: Risk negatively influences PU

H6: Risk negatively influences IU

Resistance to the m-banking application

Resistance to innovation occurs when an innovation is perceived to disrupt the user’s norms, state of satisfaction or beliefs (Ram and Sheth 1989; Heidenreich and Spieth 2013; van Klyton et al. 2021). While both passive and active innovation resistance exist, this present study focuses on active innovation resistance (AIR) – the deliberate form of resistance that results from functional and psychological barriers that emerge from unfavorable product evaluations producing a negative attitudinal outcome (Talke and Heidenreich 2014; Joachim, Spieth, and Heidenreich 2018). Yu and Chantatub (2016) examined these barriers for East Asian consumers and found that only the tradition barrier was not a significant determinant of resistance to mobile banking. Several studies examine AIR using survey data, including Mani and Chouk (2016) who found that, for smart products, perceived novelty negatively influenced consumer resistance, whereas perceived uselessness and intrusiveness (which itself is positively influenced by privacy concerns) positively influenced resistance. They also found that self-efficacy – one’s perception about their own ability to use a technology – negatively influenced resistance. This finding is interesting with respect to lower-income consumers because it emphasizes how these users reflect on their own abilities as they process information about the technology – a semblance of congruence.

Kaur et al. (2020) found that usage, risk, and value barriers impeded the intention to use mobile payment solutions (mps), such that users were likely likely to recommend mps to others when only usage and value barriers were present. Risk was significant only on intention to use, but not for intention to recommend. This suggests that even if people resist using mobile payments for themselves because of risk, they might still recommend it to others. Furthermore, their study showed that psychological barriers (i.e., tradition and image) were not significant predictors of intention to use or of intention to recommend.

The innovation resistance theory (IRT) posits that psychological and functional barriers (i.e., risk) influence resistance (Ram and Sheth 1989). As stated above, AIR involves evaluating a technology, its features, attributes, and proposed benefits. This process allows for the emergence of risk as a perception. Other studies offer meaningful contributions to understanding IRT. For example, Hong (2020) found that PEOU and PU had a negative influence on innovation resistance, which negatively influenced the intention to adopt. Rammile and Nel (2012) evidenced the influence of different types of resistance on the use of mobile phone banking (RUMB) on PU and PEOU. Mohammadi (2015) also found a significant negative influence of resistance on PEOU and PU for mobile banking in Iran, which was explained by youth and education. Similarly, Raza, Umer, and Shah (2017) found that for mobile banking usage in Pakistan, resistance negatively influenced PEOR, but was positively associated with perceived usefulness, which illustrates that resistance does not always impair an individual’s perception of the benefits of mobile banking. Lastly, although Arif et al. (2016) had already integrated resistance with the technology acceptance, they established a direct relationship to Attitude. Because Attitude was not measured in this present study, our model is more aligned with the approach of Rammile and Nel (2012) and Mohammadi (2015). Accordingly, we put forth the following hypotheses:

H7: Resistance negatively influences PEOU

H8: Resistance negatively influences PU

On the relationship between resistance and risk

In alignment with IRT, a product or service cannot be considered risky if it has not previously been known to some degree. Therefore, we posit that resistance can precede risk. While this may seem contradictory at face value, we theorize that the degree of perceived alignment between the product and the potential user’s image of themselves (i.e., congruence) has an indirect effect on risk by virtue of its influence on resistance, such that the greater the disconnect a user perceives between the image of the innovation and their self-image, the more likely the user would begin to perceive using the technology as a risky endeavor. In this way, self-image congruence (discussed below) then would help to mitigate psychological barriers. The effect of congruence on resistance could also distort perceptions of usefulness and ease of use for the innovation. However, these variables would not be rooted in uncertainty and unpredictability (i.e., risk) as IRT suggests, rather, they would emanate from self-image incongruence, which in this context of Colombia could be partly shaped by the stratification system. Hence, we hypothesize the following:

H9: Resistance positively influences Risk

Facilitating conditions

In developing the UTAUT model, Venkatesh et al. (2003, 453) defined facilitating conditions (FC) as the extent to which an individual “believes that an organizational and technical infrastructure exists to support the use of the system.” This depiction captures the technological and environmental factors that may affect the intention to use a technology. FC has been traditionally considered as an antecedent of PEOU. Karahanna and Straub (1999) argued that support and accessibility are psychological origins of PEOU. FC influences technology use because of the strong effect that organizational environments exert on behavioral control (see also Ajzen 1991), such that favorable FC will yield a higher likelihood of intention to use, particularly when moderated by age. FC has been included in studies to explain facilitators and barriers of mobile banking adoption (Chemingui and Lallouna 2013; Pal et al. 2020).

Baishya and Samalia (2020) used an extended UTAUT model to examine smartphone adoption among lower-income consumers in India and found that FC was significant in predicting use behavior, particularly for technology-skilled younger people. However, their combining urban and rural data could be problematic because the two segments have different demonstrated needs, self-image, and perceptions of facilitating conditions. In contrast, Hussain et al. (2019) focused exclusively on urban lower-income consumers in Bangladesh and found that FC had a positive and significant effect on behavior intention for m-payment adoption. Pipitwanichakarn and Wongtada (2019) found that urban street vendors in Thailand (who would constitute a lower socioeconomic status) felt pressured to adopt technology for m-payments because their relatively wealthier customers, used these systems. Although their study was theorized through TAM and thus did not utilize FC, their findings underscore the importance of technological and social aspects that influence behavioral intention.

In this present study, we theorize that the social stratification plays a role in shaping users’ perception of facilitating conditions. As stated earlier, the SES measures the characteristics of the neighborhood, local infrastructure, as well the condition of the house itself (Guevara and Shields 2019), which coincides with a positive correlation between internet speeds and quality of neighborhood,⁴ potentially influencing an individual’s belief that the “organisational and technical infrastructure exists to support the use of the system” (Venkatesh et al. 2003, 453). In fact, Giménez-Santana, Caplan, and Drawve (2018) argued that the stratum is front of mind for urban Colombians and is an early questions asked when considering a house purchase.

Given that FC favors the evaluation of the characteristics of a technology, we propose that FC has an influence on functional barriers associated with resistance, such that:

H10: FC positively influences PEOU

H11: FC negatively influences Resistance

Moreover, as hypothesized above, we argue that congruence and facilitating conditions may be relevant for explaining the formation of resistance to the use of m-payment applications for low-income users. As we show in the next section, congruence can indirectly affect risk through the formation of psychological barriers that are expressions of resistance. In like manner, the formation of functional barriers could be explained by the existence of facilitating conditions, which influences risk.

Self-image congruence

Self-image congruence is a determinant in product evaluation (Graeff 1996) and a key psychological driver of adoption of technologies (Antón, Camarero, and Rodríguez 2013). It pertains to the psychological comparison that individuals make between themselves (i.e., the self-image) and a product-user image (Sirgy et al. 1997). Higher levels of self-image congruence are theorized to foster particular behaviors, such as purchase or usage of a product (Johar and Sirgy 1991; Antón, Camarero, and Rodríguez 2013) or adoption of a service innovation (Kleijnen, de Ruyter, and Andreassen 2005). Self-image informs the ability to self-designate, which also plays a role in technology adoption (Rogers 1995). Furthermore, both actual and ideal self-image can positively influence attitudes and intention to use (Kleijnen, de Ruyter, and Andreassen 2005). The self-concept underlies self-congruence and forms as individuals process information around them including digital infrastructure and the neighborhood quality in Colombian cities. The positive correlation between the digital infrastructure and neighborhood quality is captured within the SES code, which structures and reinforces particular information about the identities of urban Colombians (Uribe-Mallarino 2008). We know from the work of Onkvisit and Shaw (1987) that self-image is informed by the information a person integrates and processes. For example, Farhat and Khan (2012) showed that an increase in the information intake regarding information and communication technologies (ICTs) is positively related to self-image congruence, leading to greater intention to use. This has implications for how certain users integrate technology knowledge for trying new innovations.

In urban settings, the formation of self-image for consumers living in lower SES designated areas is salient in differentiating acceptance and technology use behavior because these customers would be subjected to multiple influences in their surroundings.⁵ The shared values and interests of the residents within a given stratum constitutes a collectivism, affirming the critical role that social influence plays in technology adoption (Hussain et al. 2018). This is line with the findings of Hussain et al. (2019) – discussed in the previous section – who found that behavior intention for m-payment adoption often required support from others when difficulties arose in using m-payment. We contend that these consumers are exposed to higher-level goods – technology, education for their children, and access to financial services – as they move through and experience both the dynamic nature of cities and variegated socio-economic strata, potentially catalyzing a desire to consume such goods. Hence, the consumer lifestyle factors of urban settings would increase self-congruence with technology goods (Chouk and Mani 2019; Wu et al. 2020) and alter the “consumer’s usage situation” (Kleijnen, de Ruyter, and Andreassen 2005). This aligns with Gupta and Srivastav (2016) findings that aspirational consumption for lower-income consumers is driven by personal growth, emotional intimacy, and community service.

As pointed out by Escalas and Bettman (2005), reference groups are key for self-brand formation, where shared beliefs and other similarities with the people around a person become a vital source of information for brand meaning. Consumers who utilize reference groups begin to construct their self-identities for a given brand. The context of urban consumers living in lower SES areas complicates this line of thinking because these users would theoretical have multiple points of reference – other lower-income consumers within their stratum (the actual one), the stratum classification itself and their exposure to other SES strata within the city. Guevara and Shields (2019, 231) argued that the SES classification shapes the identities of neighborhoods and extends “identity as a form of social representation” beyond the statistical measurement archived within DANE (the Colombian National Statistical Office). It reinforces a socioeconomic positionality that is widely perceived in the country to be tied to income levels. Furthermore, in an urban context, with all strata represented, these consumers would witness high rates of mobile technology adoption and usage by wealthier city dwellers (Dahana, Kobayashi, and Ebisuya 2018; Pipitwanichakarn and Wongtada 2019), which has implications for who constitutes a reference group within the self-image congruence model for lower income users. Every m-payment transaction activity would provide new information for lower-income consumers, shaping the self-concept (Quester et al. 2000) and linking it with the “symbolic value” of that service (Grubb and Grathwohl 1967). The interplay between the real self and ideal self is in part informed by the SES and both dictate “specific behavior patterns” (Onkvisit and Shaw 1987, 14). As such, the perception of facilitating conditions and self-congruence for m-payment systems, would be affected by the context of the SES.

In sum, when congruence is present and there is a personal connection between a product and the potential consumer, a purchase is likely to occur (Johar and Sirgy 1991). Furthermore, there is both a direct influence of congruence on the IU for mobile applications (Wu et al. 2020), and the potential for congruence to lead to a sustained intention to use (Kang, Hong, and Lee 2009). Recent work also supports the notion that congruence reduces resistance to smart services (Chouk and Mani 2016. Quester et al. (2000) found that the actual- and ideal-self images play a part in both functional and status-related product evaluation, such that congruence has a direct effect on resistance from an active resistance perspective (Heidenreich and Spieth 2013; Joachim, Spieth, and Heidenreich 2018). We therefore propose the following hypothesis:

H12: Congruence negatively influences resistance

H13: Congruence positively influences IU

It is worth noting that a key difference between image as a barrier in the innovation resistance model and the image of self-congruence is that the latter entails a more explicit contextualization of the innovation’s image based on the self-concept, which in this case is influenced by the stratification label. Hence, self-congruence is an inherently more internal evaluation process (and thus more subjective) than the image barrier of innovation resistance theory.

Methodology

Data and sample

Our study included a sample of 397 lower income urban users from Colombia who are classified as “low-low”, “low” and “medium-low” in the SES. In August of 2019, questionnaires were administered through a Colombia-based marketing firm to current mobile bank users and potential users that were in their opt-in database. We used non-probability sampling in four Colombian cities based on a convenience sample of email addresses (Fielding et al. 2008) delivered through a mass email service provider with an opt-in database. Hulland et al. (2018) argue that convinienceconvenience sampling can be an appropiateappropriate mean for testing the “veracity of proposed theoretical effects.”. Isolating this group of consumers in Colombia was achievable through the SES.⁶ The 1^st, 2^nd, and 3^rd strata were selected from the provider’s database. The surveys obtained were further depurated by income, extracting only respondents who earned USD550 or less. Table 1 describes the sample’s characteristics, including gender, age, occupation, education level, and income. Our target respondents were registered in the 1st, 2nd, and 3rd strata, constituting lower-income consumers. As shown, 19.5% of the sample earns below the minimum wage and 52.6% between one and two standard deviations above the minimum wage.

Table 1. Demographic profile of respondents.

Gender:	Male: 47%; Female 53%	Non high school graduate: 8%
Occupation	Employees: 59%	High school graduate: 21%
	Informal employees: 16%	2 year degree: 34%
	Unemployed: 14%	Bachelor’s degree: 36%
	Retired: 1%	Master or PhD: 1%
	Student: 3%	USD$275 (minimum wage) or less: 29.5%
	Entrepreneur: 5%	USD$276-USD$550: 52.6%
n = 397		USD$551-800: 18%

Instrument development

We measured the constructs proposed in the model by seven scales of measurement. The surveys employed a Likert scale from 1 to 5 (with 5 meaning best) and contained a total of 26 items. The scales of perceived ease of use and perceived usefulness were adapted from Venkatesh et al. (2012). Facilitating conditions was adapted from Venkatesh et al. (2012) and Arvidsson (2014). The congruence scale was adapted from Escalas and Bettman (2005, 2003). Resistance to mobile banking was adapted from Yu and Chantatub (2016) and Risk was adapted from Lee (2009), and from Featherman and Pavlou (2003) to include financial, technical and security aspects. The Risk construct contained questions that elicited both positive and negative responses. Therefore, to ensure internal consistency of this measurement we inverted the Likert-scaled responses such that all items reflected as 1 for the highest risk and 5 for the least risky. Each item used in each construct is detailed in Appendix A. A pilot of 30 surveys was implemented for assuring face validity. Once the tests were completed the collection of surveys continued.

Data analysis and results

Outer model assessment

The study utilized partial least squares (PLS) techniques and SMARTPLS software, which provides two levels of assessment, the measurement model (outer model) and the structural model assessment (inner model) presented below. The thresholds used are presented in Table 2. Our final model ascertained the construct’s internal consistency with a Cronbach’s alpha and composite reliability between 0.7 and 0.9 (Nunnally and Bernstein 1994; Hair et al. 2017) and convergent validity and average variance extracted of at least 0.50 (Hair et al. 2017), which indicates that more than 50% of the construct variance can be explained by its indicators. Furthermore, factor loadings obtained high values and were significant, with exception of two items (FC1, FC4) from the Facilitating Conditions scale that were close to the threshold. However, the relevance of these items with respect to lower-income markets made them worthy of retention, therefore we applied a 0.6 threshold as discussed in Bagozzi and Yi (1988). Every scale, including Facilitating Conditions, obtained factor loading averages above 0.7, completing the confirmation of reliability and convergent validity, as shown in Table 3.

Table 2. Thresholds for PLS-SEM.

Validation stages	Indicators	Minimum criteria
Measurement instrument for reflective constructs	Internal consistency: Cronbach’s Alpha (CA).	≥ 0.70 (Nunnally and Bernstein 1994)
	Reliability: Compound reliability (CR).	≥ 0.70 (Fornell and Larcker 1981)
	Convergent validity: Average variance extracted (AVE).	≥ 0.5 (Fornell and Larcker 1981)
	Convergent validity: Size of the loads and significance.	≥ 0.6 (Bagozzi and Yi 1988) and p < 0.01
	Discriminant validity: Square roots of extracted variance (AVE)	≥ Values below the diagonal (Chin 1998; Fornell and Larcker 1981).
	Discriminant validity: Heterotrait-Monotrait Ratio	≤ 0.9 (Hair, et al. 2017)
Structural model	Coefficient of determination (value ).	≤0.75 and > 0.50 relevant, ≤ 0.50 and ≥ 0.26, moderate, and < 0.25 weak (Hair, et al. 2014)
	Significance of structural relationships	p < 0.0010
	Predictive relevance	> 0 (Stone 1974; Geisser 1974; Tenenhaus et al. 2005).

Source: Hair et al. (2017).

Table 3. Reliability and convergent validity.

	Factor loadings	p value	Cronbach’s alpha	Composite reliability	Average variance extracted
FC1	0.671	0.000	0.734	0.827	0.547	FC1
FC2	0.782	0.000				FC2
FC3	0.837	0.000				FC3
FC4	0.652	0.000				FC4
CON1	0.845	0.000	0.875	0.911	0.720	CON1
CON2	0.903	0.000				CON2
CON3	0.805	0.000				CON3
CON4	0.839	0.000				CON4
R1	0.838	0.000	0.865	0.906	0.707	R1
R2	0.860	0.000				R2
R4	0.858	0.000				R4
R5	0.805	0.000				R5
EOU1	0.884	0.000	0.902	0.932	0.774	EOU1
EOU2	0.892	0.000				EOU2
EOU3	0.859	0.000				EOU3
EOU4	0.882	0.000				EOU4
RES1	0.841	0.000	0.840	0.904	0.758	RES1
RES2	0.872	0.000				RES2
RES3	0.898	0.000				RES3
PU1	0.858	0.000	0.893	0.926	0.758	PU1
PU2	0.881	0.000				PU2
PU3	0.894	0.000				PU3
PU4	0.848	0.000				PU4
IU1	0.907	0.000	0.922	0.950	0.865	IU1
IU2	0.957	0.000				IU2
IU3	0.924	0.000				IU3

For assessing discriminant validity, the Fornell and Larcker (1981) criterion was performed, which showed that the square root of each average variance extracted (AVE) was larger than the correlation between constructs, implying that each construct is sufficiently distinct from each other (highlighted in bold in Table 4). Supporting the Fornell and Larcker criterion, we obtained Heterotrait-Monotrait (HTMT) ratios below 0.8, confirming the existence of discriminant validity, as shown in Table 4. Simply put, a low value of HTMT means that each construct is better explained by its own items than the items of other constructs (Henseler, Ringle, and Sarstedt 2015).

Table 4. Discriminant validity: Fornell and Larcker criterion and HTMT (Heterotrait-Monotrait Ratio).

	Congruence	Facilitating conditions	Intention to use	Perceived ease of use	Perceived utility	Resistance	Risk
Congruence	0.849	0.659	0.558	0.464	0.726	0.337	0.153
Facilitating Conditions	0.557	0.739	0.650	0.737	0.799	0.398	0.158
Intention to Use	0.531	0.569	0.930	0.600	0.715	0.575	0.239
Perceived Ease of Use	0.444	0.651	0.550	0.880	0.736	0.404	0.240
Perceived Usefulness	0.663	0.652	0.653	0.663	0.870	0.547	0.148
Resistance	–0.319	–0.326	–0.509	–0.353	–0.476	0.871	0.387
Risk	–0.152	–0.141	–0.225	–0.220	–0.137	0.350	0.841

Note: The bold italic values illustrate is the Fornell and Larcker criterion value.

Determination coefficients (R²) were calculated for each dependent construct of the model, obtaining weak results for Risk and Resistance which is consistent with the number of relations that influence these constructs (Table 5). We found that the R² for Intention to Use and Perceived Ease of Use were moderately relevant, while the R² for Perceived Usefulness was relevant above 0.5 (Hair et al. 2014). In addition, the Stone-Geisser blindfolding test was executed with an omission distance of 7 for assuring the predictive relevance of the model (Geisser 1974; Franke and Sarstedt 2019). This test calculates the Q², which should be above zero (Hair et al. 2017). For each predicted construct, the Q² obtained adequate values, completing our assessment of the predictive validity of the model.

Table 5. Predictive validity.

Dependent constructs	R²	Q²
IU	0.478	0.397
PEU	0.453	0.343
PU	0.514	0.383
Resistance	0.133	0.095
Risk	0.122	0.073

Table 6. Hypothesis testing results.

Hypotheses	Original sample	T statistics	P values	Result
Perceived ease of use -> Perceived utility	0.576	12.2997452	0.000	Supported
Perceived ease of use -> Intention to Use	0.184	3.26542497	0.001	Supported
Perceived usefulness -> Intention to use	0.410	6.42480462	0.000	Supported
Risk -> Perceived ease of use	–0.091	1.85179488	0.064	Not supported
Risk -> Perceived utility	0.097	2.35463494	0.019	Supported
Risk -> Intention to use	−0.103	2.26997335	0.023	Supported
Resistance -> Perceived ease of use	−0.127	2.9070285	0.004	Supported
Resistance -> Perceived utility	−0.307	6.62874646	0.000	Supported
Resistance -> Risk	0.350	7.63574841	0.000	Supported
Facilitating conditions -> Perceived ease of use	0.597	15.0894985	0.000	Supported
Facilitating conditions -> resistance	−0.215	3.39573695	0.001	Supported
Congruence -> Intention to use	0.162	3.13579682	0.002	Supported
Congruence -> Resistance	−0.199	3.18606535	0.001	Supported

Inner model assessment

The inner model establishes the relations between constructs by their path coefficients and evaluates the significance of those estimations from the PLS algorithm. For each path estimation, a significance test was executed using the bootstrapping technique, creating 5000 subsamples using the original data obtained from the sample to estimate a better standard deviation and T test. Thirteen path coefficients were found to be significant at the 5% level, with only the influence of Risk over Perceived Ease of Use found to be insignificant (p = 0.056) (Table 6).

Total effect analysis (Table 7) was conducted to isolate not only the direct effects represented by the hypotheses tested but also the contribution of the indirect effects represented in the whole model (Streukens and Leroi-Werelds 2016). Our analysis focused on explaining the Intention to Use as the most relevant dependent variable of the model. All total effects were found to be significant (p < 0.05), affording comparisons of the magnitude of these effects. We can affirm that Perceived Ease of Use (TE = 0.419), Perceived Usefulness (TE = 0.410), Risk (TE = 0.295) and Congruence (TE = 0.205) were the most important positive antecedents of the Intention to Use m-payment apps. Additionally, we found that Resistance (TE=-0.215) and Risk (TE=-0.102) were significant and negative antecedents of the Intention to Use.

Table 7. Total effect analysis.

	Total effect	T statistics	P values
Congruence -> Intention to use	0.205	3.802	0.000
Facilitating conditions -> Intention to use	0.296	7.337	0.000
Perceived ease of use -> Intention to use	0.419	8.479	0.000
Perceived utility -> Intention to use	0.410	6.425	0.000
Resistance -> Intention to use	–0.215	6.011	0.000
Risk -> Intention to use	–0.102	2.077	0.038

Post-hoc analysis was conducted to identify differences between users and potential users. The database was divided into 279 actual users and 118 potential users and Multigroup Analysis (MGA) was applied. After verifying adequate conditions of reliability and convergent validity, differences were found in three hypotheses: Congruence over Resistance, Facilitating Conditions over Resistance, and Resistance over Perceived Utility. For each one, the parametric test result was significant, which was different to the other hypotheses. Table 8 presents the results for the three significant hypotheses.

Table 8. Parametric test for multigroup analysis: Only three significant hypotheses.

	Path coefficients-diff (Actual users – potential users)	t-Value (|Actual users vs potential users|)	p-Value (Actual users vs potential users)
Congruence -> Resistance	0.429	3.51	0.000
Facilitating Conditions -> Resistance	−0.389	2.890	0.002
Resistance -> Perceived Utility	0.252	2.535	0.006

Table 9 presents the significance of each path coefficient for the three selected hypotheses and the differences in the path coefficients estimated. The negative influence of congruence was significant for potential users but not for actual users. The negative influence of facilitating conditions over resistance was significant for actual users but not significant for potential users. Furthermore, the negative influence of resistance over perceived utility was significant for both groups evaluated but it was bigger for potential users.

Table 9. Path coefficients after bootstrapping results.

Path coefficient	Path coefficients potential users	Path coefficients original actual users	t-value potential users	t-value actual users	p-value potential users	p-value actual users
Congruence -> Resistance	–0.427	0.002	4.640	0.031	0.000	0.488
Facilitating Conditions -> Resistance	0.088	–0.302	0.740	4.329	0.230	0.000
Resistance -> Perceived Utility	–0.448	–0.196	5.885	3.628	0.000	0.000

Discussion and conclusion

As predicted, a multigroup post-hoc analysis revealed notable differences between potential and actual users. First, resistance to use mobile payment apps is negatively influenced by Self-Image Congruence for potential users and by the perception of existing Facilitating Conditions for actual users. Thus, we argue that resistance may manifest in different ways, serving to reconfigure the context of m-payments for lower income urban people. While potential users seem to look for a fit between their identities and the use of m-payment apps, actual users seem to overlook this particularity, possibly because of awareness of the application or the experience using the app. Psychological and functional barriers arise for actual users when they lack resources and opportunities for using the app correctly, though these barriers were not relevant for potential users. Moreover, even with these different sources of influence, resistance is still relevant to explain the perception of utility but seems to be of greater importance for potential users who have not yet used m-payment apps. It would be expected that resistance may be stronger for inexperienced users, particularly the underbanked or have little knowledge of mobile services. The differences found are relevant for m-payment acceptance in this context, especially when there is a risk of discontinuance of these applications, which have already been observed in the literature (Huang et al. 2020; van Klyton et al. 2021).

This study yields clear implications for theory. We found that perceived usefulness was the main direct construct that influences the intention to use m-payment apps, followed by Congruence. This implies that lower income consumers would need to perceive benefits and utilities from the app’s services. The importance of congruence in our model was underscored by the country’s unique stratification system, which contributes an interesting angle to our understanding of self-congruence for m-payment technologies. For although there are a few studies that examine the relationship between these variables (Karjaluoto et al. 2019; Antón, Camarero, and Rodríguez 2013; Klabi 2020; Reddy and Ahmad 2020), neither the self-concept nor self-image congruity have been examined to take into account how the “surrogate spatialisation of economic divisions” imposed by the socioeconomic stratification system (Guevara and Shields 2019, 223) affect the influence among the relationship of these variables. Our findings showed that the respondents felt that the use of this technology is congruent to their identity, particularly in relation to the presence of digital payments in the cities in which they live.

Risk was the least important, but significant, direct antecedent of the intention to use. Risk is relevant because it runs counter to the expectation that lower income users would become followers instead of possible innovators (Rogers 1995). When total effects were analyzed, perceived ease of use took on more importance because of indirect effects. While perceived usefulness manifested the strongest total effect, ease of use was second. This confirms that TAM beliefs remain relevant and should be included in explaining m-payment use behavior. Facilitating conditions is also a relevant antecedent of the intention to use when total effect analysis is done. Its importance is similar to Congruence and very relevant for the urban context where access to networks, smartphones, and stores that use m-payments may be available but not necessarily affordable for lower-income users.

Similarly, Risk is the least important factor for intention to use, particularly for total effect analysis. This is important for understanding the personal characteristics of the users in this study, especially when considering the greater influence of resistance to use mobile banking. While the perception of not needing new ways to make payments in stores (i.e., Resistance) affects intention to use, we can further nuance the impact by looking at the moderating variables of PU, PEOU, and Risk, such that the effect of Resistance on IU is stronger when moderated by PU than by PEOU or Risk.

Therefore, we conclude that TAM beliefs are not sufficient for understanding acceptance and use behavior of urban lower income users; rather, their personal characteristics should be taken into consideration. This stance is afforded by incorporating Congruence, Resistance, and Risk into the model to understand how these users express their identities, evaluate alternatives, and take into account security, financial and technical concerns associated with adoption.

From a managerial perspective, we affirm that urban lower income users still need the regular elements for accepting m-payment apps such as ease of use, perceived utility, and facilitating conditions. However, they also want to develop an identity as buyers in a context in which technology plays as relevant a role as buying behavior. Therefore, we recommend that marketing communications better visualize opinion leaders in using the technology, that is, role models who, for this consumer group, reflect an aspirational lifestyle that is accessible to them. This would complement the belief that people have higher levels of trust between themselves or other people rather than with institutions (i.e., the bank) (Lappeman, Ransome, and Louw 2019; Barki and Parente 2006). Furthermore, we affirm that PEOU and PU are relevant antecedents of intention to use, as shown in other studies.

In our case, it is important to understand how PEOU and PU are constituted and the relevant constructs that influence them. Congruence and resistance were shown to be direct and indirect antecedents of perceived utility, respectively, and facilitating conditions and resistance are important antecedents of PEOU. Therefore, to stimulate PU, we recommend that marketing communications should thematically integrate representations, images, and expressions that are shared among residents of strata 1, 2, and 3 (and in fact, integrating the designation itself into marketing strategies), which would influence the self-concept, and thus increase self-image congruence for these consumers. This would allow prospective users to feel that their lives fit with the use of mobile payment alternatives and with the rational arguments for eliminating functional barriers. This would enhance the perception of the utility and ease of use of m-payment apps over cash or other payment methods. At the same time, application design should be adjusted to include language and symbols that are accessible for users and should be simplified to further enhance user experience. The communications should assure potential users of the existence and benefits of the ecosystem with supportive resources, networks and opportunities for using m-payment technology. The data of this study was collected before the COVID-19 pandemic, however, the value proposition that m-payment apps reflect is in line with the benefits of reducing the physical circulation of money or objects (i.e., credit cards, bills).

Limitations and future directions

This study is based on a non-probabilistic sample in a single country. This represents a limitation for the generalizability of the results to emerging market countries, which would need a broader study and a random sample that allows statistical inference. Future areas of research could include differentiating the influence of different reference groups, urban lower-income consumers from wealthier urban residents, in acceptance and use behavior of m-payment apps. This would contribute to the self-image congruence literature in examining how role models are used for these groups, namely, the impact of marketing communications on the use decision. Finally, we argue for examining government initiatives that drive the massification of mobile banking for financial inclusion and mitigating health and safety issues in the context of shaping self-image congruence and resistance.

Disclosure statement

No potential conflict of interest was reported by the authors.

Correction Statement

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

Notes

1 https://bancadelasoportunidades.gov.co/index.php/es/cuentas-de-ahorro; accessed: January 20, 2021.

2 Source: https://www.weforum.org/agenda/2018/06/latin-america-cities-urbanization-infrastructure-failing-robert-muggah/; accessed: October 5, 2020.

3 The search equation applied was (TITLE (("mobile payment") ("acceptance" OR "use" OR "adoption")) AND KEY (("mobile payment"))) AND (LIMIT-TO (DOCTYPE, "ar")) AND KEY (("mobile payment"))), without a specific range of years.

4 https://colombiatic.mintic.gov.co/679/articles-161478_archivo_pdf.pdf; accessed January 20, 2021.

5 The allocation of strata in cities is variegated, ranging from 1-6, while all rural communities are classified as stratum 1.

6 Estratificación Socioeconómica - Metodología; accessed December 15, 2020.

Appendix A.

Measurement scales items and sources

Table

Code	Constructs and items	Sources
	Perceived ease of use
PEOU1	Learning to use this mobile application for paying in stores is easy for me	Venkatesh, et al. (2012)
PEOU2	Interacting with this mobile payment app for purchasing in stores is clear and understandable for me
PEOU3	Using this m-payment application does not imply any efforts
PEOU4	Developing skills to purchase face-to-face with this m-payment app is simple for me
	Perceived usefulness
PU1	In my daily life is useful to pay with this mobile application	Venkatesh, et al. (2012)
PU2	Using this m-payment app for purchasing in stores let me get what I want
PU3	Using this m-payment application to pay in stores is more practical than by other means
PU4	Using this m-payment application to pay in stores is faster than by other means
	Self-image congruence
C1	Making purchases through this m-payment application reflects who I am	Escalas and Bettman (2003)
C2	I identify myself with buying in stores through this mobile application
C3	Using the mobile application to buy things is a way of communicating to others who I am
C4	Buying in stores through this mobile payment app works well for me, it is compatible with me
	Facilitating conditions
FC1	There are many shops, stores, places where you could pay through this mobile payment app	Venkatesh, et al. (2012); Arvidsson (2014)
FC2	I have the necessary resources to pay for my purchases in stores through this mobile application
FC3	I have the necessary knowledge to pay for my purchases in stores through this mobile application
FC4	When I have doubts or difficulties in using this mobile application, I can find others to help me
	Perceived risk
R1	I am worried about making mistakes when entering the information to use this mobile application to pay in stores	Lee (2009); Featherman and Pavlou (2003)
R2	I am afraid of making mistakes when paying ins stores with this mobile application and that the bank will not give back my money
R3	It makes me afraid to think that this mobile application does not work properly
R4	I feel totally safe providing personal information through this m-payment application
R5	I am concerned about using this mobile payment app to purchase in stores, because I think that other people could access my account
	Resistance to use mobile banking
RUMB1	I do not consider it necessary to use this mobile application to make payments	Yu et al. (2005)
RUMB2	I consider that the other existing payment options offered by banks are sufficient to cover my needs
RUMB3	I consider that this mobile application is not necessary because I already have other options that serve to pay
	Intention to use m-payment
IU1	I intend to use this type of application to pay in stores digitally through my mobile phone	Taylor and Todd (1995)
IU2	I will definitely use this type of application to pay in stores
IU3	In the next few weeks, I see myself buying in stores through this type of mobile application