Bank loyalty in times of crisis: why attachment alone is not enough in a war-torn economy? Open Access

Customer loyalty has long been recognized as a central construct in marketing studies, particularly in high-involvement and trust-based industries such as banking. Classic work distinguishes between attitudinal loyalty – reflecting psychological commitment and preference – and behavioral loyalty, which captures repeat patronage and usage intentions (Dick and Basu, 1994). In financial services, loyalty is particularly relevant because switching costs are high, trust is critical and customer retention is closely tied to financial system stability. Yet, despite its importance, the psychological mechanisms through which loyalty is formed and sustained remain contested, particularly under conditions of extreme uncertainty.

Recent research increasingly emphasizes emotional and psychological drivers of loyalty, moving beyond purely rational evaluations of price or service quality. In digital service environments, customer loyalty has been shown to depend not only on attachment to the brand but also on customers’ psychological engagement with service platforms (Levy, 2022). Psychological engagement in this study refers to an affective–experiential state emerging from meaningful interaction with the bank’s digital platform, rather than mere frequency of use or transaction volume. In fact, engagement is conceptualized as a multidimensional psychological state (Calder and Malthouse, 2016) that mediates the relationship between emotional attachment and loyalty. While this framework has received empirical support in stable economic contexts, it remains unclear whether these mechanisms operate similarly when consumers face prolonged crisis and existential threat.

This theoretical gap becomes particularly salient in wartime contexts, where uncertainty, perceived loss of control and emotional stress may fundamentally alter how consumers evaluate, engage with and remain loyal to service providers. Existing research on consumer behavior during crises has primarily focused on economic downturns or natural disasters (Campbell et al., 2020; Sheth, 2020), leaving wartime consumer decision-making – especially in banking – largely unexplored.

Crises often destroy financial systems; in Ukraine, however, they pose a puzzle: why do so many customers, despite the war, continue to be loyal to their banks? More than four years into Russia’s full-scale invasion, the Ukrainian banking sector has displayed remarkable resilience. Unlike in many conflict-affected countries – where financial institutions shut down or declared prolonged bank holidays, as recalled by Kyrylo hevchenko, former head of the National Bank of Ukraine – Ukrainian banks have remained fully operational, withstanding cyberattacks, sustaining daily transactions and emerging as one of the few profitable sectors in the national economy. In doing so, they have not only supported households and firms but also contributed significantly to tax revenues, becoming a stabilizing force in a time of profound uncertainty.

At the same time, Ukraine’s financial sector has undergone rapid digital transformation. Since the entry of domestic neobanks in 2017, and further accelerated by wartime constraints, digital banking has reshaped customer expectations through speed, usability and low-cost services – particularly among younger, urban and tech-savvy populations. Traditional banks now face increasing pressure to modernize, enhance their digital interfaces and improve customer experience. The competition has also intensified on the international front: global fintech firms such as British-based Revolut have strategically expanded into Ukraine, offering seamless user interfaces, zero-commission services and symbolic products such as the “Clear Sky” debit card in national colors. By early 2024, more than 700,000 Ukrainians abroad were using Revolut, sending over €1bn back to Ukraine – illustrating both the platform’s reach and its emotional resonance.

Despite a growing body of work on consumer behavior during economic crises and natural disasters (Campbell et al., 2020; Sheth, 2020), literature on civilian and consumer reactions to war remains scarce, particularly in the banking sector. Yurdagel and Baycur (2023) focused on consumer reaction to political tensions and wars, finding that conflicts lead consumers to change their purchasing behavior boycotting products from certain territories or brands that support specific territories. Bulyk and Havryliuk (2023) found that in times of war, Ukrainian consumers do not pay attention to the brand, but choose from what is available, with a growing attitude toward the use of digital platforms and online shopping. Few studies focused on consumers’ financial decision-making when everyday life is disrupted by conflict. Vranceanu et al. (2025) investigated how investment behavior of consumers in proximity of a war-affected country might be affected by the fear of war and uncertainty. However, there is still limited understanding of how consumers adjust their behaviors concerning financial services. This gap is particularly relevant for banking, where customers have to choose among different banking product providers and where trust and loyalty are critical for maintaining financial system stability and ensuring access to essential services in times of uncertainty.

In fact, in times of geopolitical instability, the choice of banking services may no longer be based solely on rational evaluations of cost or service quality, but increasingly shaped by emotional attachment with local financial institutions, as reinforced by consumer ethnocentrism – a psychological tendency to prefer local over foreign products based on moral, cultural or patriotic obligations (Zeugner-Roth et al., 2015).

Similarly, in the Ukrainian context, where banks are perceived not only as service providers but also as agents of national identity, psychological attachment may intensify affective loyalty to domestic banks, even in the presence of more efficient or stable alternatives (Zeugner-Roth et al., 2015). Conversely, some individuals may disidentify with domestic institutions due to dissatisfaction with public governance or service standards. Josiassen (2011) argued that consumer disidentification can drive preference for foreign providers perceived as more modern or competent, potentially weakening emotional bonds with local banks.

Against this theoretical and contextual backdrop, the aim of this study is to examine the psychological drivers of bank customer loyalty in a wartime economy. Specifically, building on Levy’s (2022) engagement model, developed and empirically validated in the context of digital banking services, this research investigates whether emotional attachment directly fosters loyalty, or whether its effect is fully mediated by customers’ psychological engagement with banks’ digital platforms, by posing the following research question: “What are the psychological drivers of customer attitudinal loyalty to a bank in a wartime economy?” By testing this model in the context of Ukraine, the study assesses whether loyalty mechanisms established in stable environments hold under conditions of prolonged conflict and uncertainty.

This research contributes to the literature on consumer behavior, banking and financial consumers behavior in crisis settings. It offers practical insights for Ukrainian financial institutions seeking to modernize, enhance emotional connections and retain customer loyalty, thus tackling financial system instability – interpreted as heightened risks of customer withdrawal, loss of trust in financial institutions and disruption of daily financial intermediation during periods of extreme uncertainty. More broadly, the findings may support policymakers in understanding the psychological dynamics that underpin financial resilience in postconflict societies.

The remainder of this article is organized as follows. Section 2 presents the conceptual model and develops the research hypotheses. Section 3 outlines the methodology and describes the sample. Section 4 presents the empirical results derived from the analysis. Section 5 provides a discussion of the main findings and Section 6 concludes the paper.

Customer loyalty in banking is increasingly recognized as a multidimensional construct that extends beyond transactional satisfaction or perceived service quality.

Following the seminal framework of Dick and Basu (1994), customer loyalty is defined as a bipartite concept: attitudinal loyalty and behavioral loyalty. Attitudinal loyalty reflects psychological, cognitive and emotional commitment; identification; and intention, of customers to a brand or service. Behavioral loyalty refers to repeated purchasing or usage patterns (Bowen and Chen, 2001; Dick and Basu, 1994; Jacoby and Kyner, 1973). While the former captures the deeper motivational aspects that explain why customers choose to remain with a provider and is particularly useful in predicting future intentions, the latter is often measured through observable indicators such as frequency of purchases or service use (Vesel and Zabkar, 2009). Importantly, loyalty must also be distinguished from mere usage preference or inertia. In banking, continued usage may be opportunistic, reflecting high switching costs, contractual constraint, or necessity rather than genuine loyalty. To address this distinction, scholars differentiate between active loyalty – characterized by positive word-of-mouth and willingness to recommend – and passive loyalty, which reflects continued patronage despite limited satisfaction or superior alternatives (Abbasi et al., 2011; Ahluwalia et al., 2000). In this study, loyalty is therefore approached as an active attitudinal construct capturing both psychological commitment and advocacy, rather than as simple service usage (i.e. behavioral loyalty). Consistent with this perspective, loyalty in banking is defined as customers’ intention to maintain their relationship with the bank and actively endorse it, even when alternative providers are available (Ladhari et al., 2011).

While traditional perspectives emphasized rational determinants such as efficiency, trust or switching costs (Fernandes Sampaio et al., 2019; Shankar and Jebarajakirthy, 2019), recent research highlights the centrality of psychological and emotional processes in sustaining long-term loyalty, particularly in digital service environments (Baumann et al., 2007; Levy, 2022), where loyalty increasingly depends on how customers psychologically relate to and engage with service platforms.

In today’s highly competitive and digitally driven financial landscape, loyal customers represent a stable source of revenue and a powerful driver of referrals – making loyalty not merely desirable, but strategically essential (Keisidou et al., 2013). Indeed, from a managerial perspective, loyal customers provide stable revenues, are less price-sensitive and contribute to profitability through long-term retention and advocacy. Research shows that even a modest increase in retention – around 5% – can translate into profit growth of 25–85% (Hallowell, 1996; Ladhari et al., 2011). Thus, loyalty represents both a psychological bond and an economic asset, making it a critical focus of inquiry in service industries such as banking.

From a theoretical standpoint, recent meta-analyses confirm that trust, commitment, empathy and responsiveness are among the strongest predictors of bank loyalty globally (Buhler et al., 2023). Building upon this, the Ukrainian context offers a compelling real-world laboratory. Despite being under attack, Ukraine’s banking system has maintained operations while accelerating digital service adoption. In such an environment, digital platforms serve not just functional needs but also become symbolic pillars of national resilience and trust.

This leads to the following guiding research question:

By addressing this question, the study aims to provide empirical insight into the psychological mechanisms that explain customer retention under crisis conditions and in the presence of increasingly competitive digital alternatives.

It is relevant to note that the present study is situated within the banking sector and draws on engagement and loyalty research developed across a range of service and digital contexts, whose findings are useful to identify general psychological mechanisms that are transferable to banking, due to their nature. For this reasons, Levy’s (2022) engagement model, developed and empirically validated in the context of digital banking services, serves as the primary theoretical anchor of the present study.

One of the key drivers of customer loyalty is emotional attachment (Levy and Hino, 2016; Thomson et al., 2005). Attachment theory (Bowlby, 1977) suggests that individuals form strong bonds with institutions that provide stability and security, which in banking translates to long-term relationships with trusted financial providers (Levy, 2022). Prior studies demonstrate that emotionally attached customers are less likely to switch banks, even when faced with alternative service providers (Moliner-Tena et al., 2019).

In war-affected regions, customer attachment may extend beyond brand experiences to include national solidarity. Research on consumer nationalism indicates that in times of crisis, individuals may prefer domestic brands to support local economic recovery or express collective solidarity (Balabanis and Diamantopoulos, 2016; Heslop et al., 2008). This suggests that Ukrainian bank customers may remain loyal not only due to positive banking experiences but also to broader identity-related considerations.

Thus, we propose:

Complementing attachment is psychological digital engagement – the degree to which customers actively interact, identify and immerse themselves with a bank’s digital platforms (Levy, 2022). Psychological digital engagement reflects a customer’s cognitive and emotional involvement with a brand’s platform or interface, often resulting from active interactions that are stimulating, rewarding or meaningful (Brodie et al., 2011; Calder and Malthouse, 2016). In digital banking, engagement may involve frequent use of the bank’s website or app, responsiveness to personalized offers or perceived control over one’s financial management experience (Islam et al., 2020).

Engagement not only strengthens brand–consumer relationships but also promotes behaviors that reinforce loyalty, such as advocacy, continued usage and resistance to competitor switching (Levy, 2022). Given the growing relevance of digital interfaces in banking – particularly in contexts like Ukraine, where service delivery has rapidly shifted online – engagement becomes a crucial determinant of whether customers remain loyal to a bank.

Following engagement theory, long-lasting customer relationships and profitable loyalty are achieved not merely through repeated usage, but through engaging customers with valuable and meaningful content (Pansari and Kumar, 2017). Consumer engagement is a multidimensional psychological construct that emerges from customers’ thoughts and feelings about their experience with a brand or service (Calder and Malthouse, 2016). Importantly, engagement is distinct from inertia or necessity-driven behavior as the frequency of interaction with a service does not, by itself, imply psychological engagement or attitudinal loyalty. Indeed, consistent with Levy (2022), the present study adopts a psychological–experiential perspective on the construct, focusing on the affective dimension of customer‘s emotional experiences with digital banking platforms during service delivery. This approach, building on prior works, conceptualizes engagement as a psychological state arising from interactive brand-related experiences, rather than as observable usage behavior (Brodie et al., 2011; Hollebeek et al., 2014). Therefore, this definition explicitly excludes engagement driven solely by necessity or inertia (e.g. safety constraints or lack of alternatives).

In line with this perspective, engagement is modeled as a multidimensional construct encompassing interaction, visual aesthetics, discovery, identity and civic orientation (Calder and Malthouse, 2016; Levy, 2022), reflecting how customers emotionally and cognitively relate to the digital banking environment rather than how frequently they use it.

This construct is critical in activating attitudinal loyalty in digital banking contexts (Brodie et al., 2011; Hinson et al., 2019). In fact, extant research consistently shows that psychological engagement with digital content leads to customer loyalty (Levy, 2022; van Asperen et al., 2018). In digital service contexts – and particularly in banking – engagement with digital interfaces has been shown to enhance positive brand experiences and, consequently, strengthen customer loyalty (Garzaro et al., 2020; Kosiba et al., 2018). Because it captures affective involvement rather than mere usage it is expected to foster active loyalty attitudes such as commitment and advocacy.

Thus, we propose the following hypothesis:

Emotional attachment has been identified as an antecedent to deeper cognitive and behavioral responses in consumer–brand relationships. Customers who feel emotionally connected to a brand are more likely to seek interactions with it, invest attention and time and experience a sense of pleasure in engaging with its service platforms (Hollebeek et al., 2014; Levy, 2022). In digital banking, emotional attachment can translate into a preference for interacting with the bank’s online services, as customers find reassurance, satisfaction or symbolic meaning in the brand. As such, emotional attachment not only drives loyalty directly but also increases psychological engagement by fostering sustained interaction and psychological involvement with the bank’s digital channels. Therefore:

However, emotional attachment does not necessarily translate immediately into active loyalty attitudes. In contexts characterized by uncertainty or crisis, attachment may remain symbolic or identity-based, without leading to advocacy or recommendation. Thus, psychological digital engagement is the key mechanism that transforms passive emotional attachment into active attitudinal loyalty, by reinforcing psychological connections through meaningful interactions (Levy, 2022) and acting as a mediator. While emotional attachment establishes the affective basis for maintaining a relationship with a bank, it does not by itself ensure advocacy or commitment – particularly under conditions of uncertainty. Instead, psychological engagement operationalizes that bond through active and meaningful participation with the brand’s digital service environment, thereby enabling loyalty attitudes.

In our context, given the growing relevance of digital interfaces in banking – particularly in warzones, where service delivery has rapidly shifted online – psychological engagement becomes a crucial mechanism through which emotional attachment is transformed into active loyalty, rather than a mere reflection of digital service usage. This psychological engagement then leads to attitudinal-advocacy loyalty outcomes, such as repeat usage and favorable word-of-mouth. The mediating role of engagement is also supported by broader literature on consumer–brand relationships, which suggests that engagement serves as the behavioral conduit through which emotional bonds translate into sustained commitment (Brodie et al., 2011; Vivek and Morgan, 2012).

Therefore:

Figure 1 shows the overall conceptual model and construct relationships.

This study extends Levy’s (2022) engagement model into a warzone context, testing whether emotional attachment continues to influence attitudinal loyalty directly or whether its effect is wholly mediated through psychological engagement with digital banking platforms. By doing so, it assesses how attitudinal loyalty mechanisms shift under extreme uncertainty.

Data were collected through a structured online questionnaire administered to a sample of 196 Ukrainian banking customers between October and December 2024.

Given the wartime context and the absence of a publicly accessible sampling frame of bank customers, a nonprobability purposive sampling approach was adopted. The survey was distributed through Ukrainian community networks using trusted local intermediaries, who shared the survey link via email and other digital and social media channels.

All participants involved in this study were informed about the purpose of the research and participated voluntarily. Before completing the survey, they were explicitly advised that their participation was based solely on informed consent and that they could withdraw at any time without consequence, as a matter of standard research ethics and common practice. Anonymity and confidentiality were fully guaranteed, and no identifying information was collected.

Table 1 offers the main descriptive statistics on the sample composition. The gender distribution is naturally skewed, with 68.9% identifying as female and 31.1% as male. This imbalance reflects the wartime context, where a significant portion of the male population has been mobilized for military service. As a result, women are increasingly responsible for managing household finances, making them particularly relevant respondents in assessing digital banking engagement and loyalty during this period. Regarding the banks chosen as respondents’ main financial service providers, the vast majority are headquartered in Ukraine (88.8%), while a smaller share rely on banks headquartered in foreign countries (4.7% Austria, 2% France, 2% Italy, 1% Slovakia, 1% Hungary, 0.5% USA) but with operating branches in Ukraine. Among these institutions, 76.5% are traditional banks, defined as banks operating a physical branch network in Ukraine while also offering digital banking services, whereas 23.5% are digital-only banks operating without a physical branch network in the country.

The survey was based on previously validated scales from the literature, and all responses were measured using a seven-point Likert scale ranging from 1 (“strongly disagree”) to 7 (“strongly agree”).

To empirically test the proposed conceptual model and analyze the relationship between the constructs, we applied a structural equation model (SEM) analysis following a two-step approach (Anderson and Gerbing, 1988). This method allows for the simultaneous examination of both the measurement and structural components of the model, ensuring robust assessment of the relationships between the latent constructs: emotional attachment, psychological engagement and customer loyalty. Consistent with previous studies (Anderson and Gerbing, 1988), a confirmatory factor analysis (CFA) was first conducted to assess the measurement model. The purpose of this step is to assess internal validity of items and reliability of constructs. Finally, in the second step, we tested the SEM to evaluate the hypothesized causal paths between constructs. All data analyses were performed using STATA SE 19.5.

Latent constructs are estimated by a multi-item scale of participants’ responses to the questionnaire. Bank customer loyalty is measured by three items adopted from Levy (2022) and Levy and Hino (2016), which assess customers’ intention to continue using the same bank, recommend it to others and express a preference for the bank over competitors. Emotional attachment was measured with four items adapted from Thomson et al. (2005) as in Levy (2022), reflecting affective commitment and personal connection with the domestic bank. Psychological engagement was assessed as a second-order construct through five latent dimensions adapted from Hollebeek et al. (2014), capturing cognitive and emotional involvement with the bank’s digital platforms and service environment: interaction, discovery, identity, civic orientation (adapted from Calder and Malthouse, 2016) and visual aesthetics (adapted from Moshagen and Meinald, 2010, as in Levy, 2022).

Before conducting the CFA, we assessed the suitability of the data set for factor analysis by evaluating the intercorrelations among observed variables. As recommended by Hahs-Vaughn (2016) and Watkins (2021), we examined the correlation matrix and found that several inter-item correlations exceeded 0.30, indicating sufficient shared variance among the items. This is a necessary condition for factor extraction and suggests that the items are meaningfully related.

To further assess factorability, we computed the determinant of the correlation matrix. The determinant was well above zero, indicating that the matrix is not singular and does not suffer from multicollinearity or perfect linear dependency. According to Thompson (2004) and Watkins (2021), a determinant close to zero would have suggested multicollinearity or a nonpositive definite matrix, which would compromise the validity of factor analytic procedures. We also conducted Bartlett’s Test of Sphericity to evaluate the null hypothesis that the correlation matrix is an identity matrix. The test was significant (p < 0.001), leading us to reject the null hypothesis and conclude that the observed variables exhibit substantial intercorrelations suitable for factor analysis. Finally, the Kaiser–Meyer–Olkin (KMO) Measure of Sampling Adequacy yielded a value of 0.916, which is considered “excellent” (Kaiser, 1974). Values above 0.90 provide strong evidence in favor of factor analysis. Taken together, these diagnostics confirm that the data are appropriate for latent structure analysis.

Three items were removed from the measurement model due to weak standardized loadings (i.e. below the commonly accepted threshold of 0.40 (Cutillo, 2019) or evidence of substantial cross-loadings on multiple factors. This step was taken to improve the overall construct validity of the model and to ensure a clearer factorial structure. The removal of these problematic items led to a more robust measurement model, with improved model fit and no further concerns related to discriminant or convergent validity. All remaining items loaded significantly and appropriately onto their respective latent constructs, supporting the internal consistency and reliability of the final scale. To assess the internal consistency and reliability of the measurement model, we computed Cronbach’s alpha for the full scale and for each latent construct individually. The Cronbach’s alpha for the entire scale, comprising all 23 observed items, was 0.94, indicating excellent internal consistency (Nunnally, 1994). For the individual constructs, all alpha values exceeded the commonly accepted threshold of 0.70, demonstrating acceptable to strong reliability across the scales (Hair et al., 2013). Specifically, the lowest alpha was observed for Engagement – Civic orientation at 0.81, while the highest was found for Psychological Engagement – Visual at 0.89, suggesting a high level of consistency among the engagement items related to visual interaction with the digital banking interface. From a validity perspective, the high internal consistency supports the construct validity of the scales, particularly in light of their theoretical basis and previous use in the literature (Brown, 2015). These results justify the use of these multi-item measures in the subsequent confirmatory factor analysis and structural equation modeling.

 Appendix reports item-level descriptive statistics (means and standard deviations) and corrected item–total correlations. All item–total correlations exceeded the recommended threshold of 0.30, indicating satisfactory internal consistency.

To assess the reliability and validity of the measurement model, we conducted a CFA using a two-step approach. First, we estimated a first-order measurement model including all latent constructs as distinct factors. The CFA results demonstrated no issues of convergent or discriminant validity. All items loaded significantly (p < 0.05) onto their respective constructs, with standardized factor loadings above the commonly accepted threshold of 0.5 (see Table 2). Postestimation fit indices indicated a good model fit: χ²(209) = 451.79, p < 0.001, with a χ²/df ratio below 3, root mean squared error of approximation (RMSEA) = 0.077 (p < 0.05), comparative fit index (CFI) = 0.920, Tucker–Lewis index (TLI) = 0.903 and standardized root mean squared residual (SRMR) = 0.064. These indices fall within acceptable ranges, suggesting a satisfactory representation of the measurement structure.

In addition to Cronbach’s alpha and average variance extracted (AVE), composite reliability (CR) was calculated for all constructs. All CR values exceeded the recommended threshold of 0.70, indicating satisfactory internal consistency (Hair et al., 2013). In the second step, a second-order CFA was estimated to model the higher-order latent construct of Engagement, composed of five first-order dimensions: Interaction, Visual, Discovery, Identity and Civic. Each of these dimensions loaded significantly (p < 0.05) and meaningfully onto the second-order construct, with standardized loadings as follows: Interaction = 0.65, Visual = 0.79, Discovery = 0.84, Identity = 0.79 and Civic = 0.79. No convergent or discriminant validity issues were detected for the second-order construct. The results support the theoretical structure of Engagement as a higher-order factor encompassing multiple dimensions of user involvement.

To test the hypothesized relationships among the latent constructs, a full SEM was estimated, based on the maximum likelihood estimation. This model integrates both the measurement and the structural components, allowing us to simultaneously assess the validity of the measurement model and the significance of the theoretical paths among constructs.

The structural results show a strong and statistically significant effect of Attachment on Engagement (β = 0.697, p < 0.001) and of Engagement on Loyalty (β = 0.637, p < 0.001), supporting the hypothesized mediation mechanism (H2 and H3). However, the direct effect of Attachment on Loyalty was not statistically significant (p = 0.454), and its 95% confidence interval included zero. Thus, H1 was rejected indicating that Engagement fully mediates the relationship between Attachment and Loyalty (H4 is supported). Indeed, the indirect effect (c-c′) from Attachment to Loyalty was statistically significant (β = 0.412, p < 0.001), as shown in Table 3. To formally confirm the mediating effect, we conducted a formal mediation analysis using the Baron and Kenny (1986) approach, adapted for use with SEM as proposed by Iacobucci et al. (2007). This procedure includes significance testing of indirect effects via multiple methods, including the Sobel’s test, Delta method and Monte Carlo simulation. All methods converged on the same conclusion: complete mediation. The indirect effect ranged from 0.443 to 0.444 (Delta = 0.444, Sobel = 0.444, Monte Carlo (with 5,000 simulations) = 0.443), with all 95% confidence intervals excluding zero, and z-values exceeding 5.5 across tests (p < 0.001).

These findings align with the theoretical proposition that customer engagement functions as a key mechanism through which attachment influences loyalty. Figure 2 presents the path model and its regression standardized coefficients along with their significance level.

Model fit was evaluated using several commonly accepted indices. The fit statistics indicate an acceptable model fit to the data: χ²(221) = 550.783, p < 0.001, with a chi-square to degrees of freedom ratio below 3. The RMSEA = 0.087 is slightly above the conventional threshold of 0.08 but still within the tolerable range for model adequacy in applied research. The CFI = 0.891 and TLI = 0.876, while slightly below the strict cutoff of 0.90, suggest marginally acceptable fit (Bentler, 1990; Hu and Bentler, 1999), particularly given the complexity of the model and the inclusion of a second-order factor. The SRMR = 0.082 is within acceptable limits, indicating a reasonable approximation of the observed covariance matrix.

Overall, the structural model supports the theoretical framework, with evidence for the mediating role of Psychological Engagement in the pathway from Emotional Attachment to Attitudinal Loyalty. This highlights the central role of customer engagement as a driver of relational outcomes in the banking context.

To ensure the stability of our findings, we conducted additional robustness analyses by estimating the full SEM on two restricted subsamples: customers using traditional banks only and customers whose main financial provider is a Ukrainian-headquartered bank.

For the traditional banks subsample (n = 150), the results confirm the centrality of psychological engagement. Engagement remains a strong and significant predictor of attitudinal loyalty (β = 0.615, p < 0.001), while the direct effect of emotional attachment on loyalty is not statistically significant (β = 0.161, p = 0.142). Mediation tests confirm that attachment influences loyalty only indirectly through engagement. The Sobel’s z-test is highly significant (z = 5.30, p < 0.001), and the indirect effect (approximately 0.44) is both statistically robust and substantively meaningful. Following the Baron and Kenny logic adjusted for SEM, the nonsignificant direct path combined with the strong indirect effect indicates full mediation: emotional attachment matters for attitudinal loyalty only to the extent that it fosters customer psychological engagement.

For the Ukrainian-headquartered banks subsample (n = 174), digital engagement again emerges as the key determinant of loyalty (β = 0.730, p < 0.001). In contrast, the direct effect of attachment not only loses significance but becomes essentially null (β = −0.011, p = 0.928). The mediation analysis shows an even stronger indirect pathway. The Sobel’s test is again highly significant (z = 5.66, p < 0.001), with an indirect effect of approximately 0.55. As in the previous subsample, all conditions for complete mediation are satisfied: attachment contributes to loyalty exclusively via its capacity to stimulate engagement.

Taken together, these robustness checks reinforce the main conclusion of the study: in a wartime economy, emotional attachment to banks alone is insufficient to secure customer attitudinal loyalty. Rather, loyalty is sustained through psychological engagement, especially with banks’ digital platforms, which provide continuity, reassurance and functional reliability during crisis. The fact that this pattern holds across both traditional banks and domestically headquartered banks underscores the generalizability of our findings within the Ukrainian context.

This study proposed and tested a conceptual framework to identify the drivers of bank customer attitudinal loyalty in a context of geopolitical instability, specifically in a country affected by ongoing war. In such conditions, banks face severe challenges in retaining their customer base (Piotrowska et al., 2025). Foreign financial institutions, which may offer lower transaction costs, more stable service infrastructures and reduced risk perceptions, might become more attractive and trustworthy alternatives. This intensifies the need for domestic banks to understand and strengthen the mechanisms that drive customer loyalty in these specific context where financial systems suffer from geopolitical uncertainty.

The results offer both theoretical and practical insights. First, the findings reveal that emotional attachment, conceptualized as affective commitment and personal connection with the bank, does not directly predict customer loyalty. Despite respondents expressing strong emotional bonds with their banks – especially in items such as feeling proud to be a customer, feeling highly regarded by the bank and perceiving alignment with the bank’s personality – this attachment did not significantly translate into loyalty attitudes, such as recommending the bank to others. Thus, H1 was rejected. This may reflect a context-specific disconnect: even emotionally committed customers may hesitate to fully endorse or rely on their financial institutions under conditions of heightened risk and uncertainty.

However, this relationship is mediated by psychological digital engagement, a second-order construct comprising five lower-order dimensions: interaction, visual appeal, discovery, identity and civic orientation. All five dimensions contributed significantly to the engagement construct, with visual aesthetics and discovery-related content having the highest coefficients. These findings suggest that digital experience quality plays a crucial role in activating emotional bonds and transforming them into behavioral loyalty, not only within commercial sectors during wartime (Bulyk and Havryliuk, 2023) but also in the banking sector.

Interestingly, interaction and civic orientation – while conceptually relevant – emerged as weaker (but significant) contributors to the engagement construct in this study. This suggests that, under wartime conditions, customers prioritize functionality, usability and identity-based aspects of digital banking over features related to two-way interaction or broader social value. In other words, engagement during crisis appears to be shaped more by immediacy, reliability and symbolic reassurance than by civic or environmental considerations (Casaló et al., 2008; Ladhari et al., 2011). Nonetheless, their weaker performance (compared to the other first-order engagement constructs) in this sample may reflect underdevelopment by banks or a temporary shift in consumer priorities. As stability returns, strengthening interactive functionalities and highlighting the social and environmental benefits of digital banking could become more effective strategies for deepening engagement and sustaining loyalty.

The removal of one loyalty item – related to customers’ intentions to use more services in the near future – also deserves mention. Its weak loading might suggest that, under the cloud of geopolitical uncertainty, customers may be reluctant to commit to future service uptake or might be more short sighted, even if they intend to maintain their current banking relationship. This nuance reflects how perceived instability may dampen proactive or future-oriented customer intentions, an important consideration for banks operating in volatile regions. Indeed, consumers might display higher levels of temporal discounting under uncertainty and be prone to the myopic bias (van der Wal et al., 2018).

Overall, the validated structural model shows that psychological engagement fully mediates the relationship between emotional attachment and attitudinal loyalty, indicating that emotional bonds with the bank translate into favorable evaluation and recommendation only when customers experience meaningful psychological engagement with the bank’s digital service platform (supporting hypotheses H2, H3 and H4). From a managerial perspective, this implies that banks cannot rely on emotional bonds alone but must actively create engaging digital environments to reinforce customer relationships. Making the attachment–loyalty path significant without mediation could be a long-term goal, but in the present circumstances, enhancing engagement remains the most viable strategic lever.

These results hold robust even in the subsamples restricted to traditional and Ukrainian banks. Contrary to expectations, attachment did not directly predict loyalty in contexts marked by high uncertainty. In other words, affective bonds with domestic institutions are not sufficient to ensure loyalty, but they require the activation mechanism of engagement – through functional, digital and identity-related interactions – to translate into behavioral intentions. This finding challenges the assumption that patriotic or affective attachment alone guarantees loyalty in times of difficulty (Poulsen, 2020), underscoring the primacy of engagement as a mediator. Therefore, the robustness analyses further reinforce the centrality of engagement in driving loyalty. While the full-sample model already indicated that attachment did not exert a direct effect on loyalty, this result became even clearer in the subsamples focusing on traditional banks and Ukrainian-headquartered banks, where attachment was strongly nonsignificant. This finding points to an important wartime dynamic: in contexts of crisis, loyalty appears governed less by sentiment and more by pragmatic considerations of safety, usability and reliability – in line with literature affirming that consumers change their behavior in times of uncertainty (Hasan et al., 2021). Attachment risks becoming “emotion without action” unless it is activated through functional and pragmatical engagement with the bank’s services and digital platforms. These results also suggest that in an environment marked by uncertainty and institutional fragility, customers may be more thoughtful when choosing a product or service with less attention to the brand (Bulyk and Havryliuk, 2023). Instead, engagement with digital interfaces provides a sense of control, immediacy and continuity – that usually tend to be lost in periods of crisis (Herzenstein et al., 2015) – which makes it a stronger predictor of loyalty. According to Thompson and Schlehofer (2008), consumers exhibit dispositional responses to crisis – such as control-seeking – which could influence how they cognitively and emotionally engage with financial services under uncertainty. For example, individuals with a control-based orientation may respond to geopolitical threats by actively engaging with digital banking tools as a way to reassert financial stability.

Moreover, attachment items such as “I am proud to be a customer of my bank” may partly reflect identity signaling or social desirability, which in times of war do not necessarily correspond to loyalty. This contrasts with findings in stable economies (e.g. Levy, 2022), where attachment exerts a direct effect on loyalty, and extends theory by showing that under conditions of extreme uncertainty, attachment only matters when it is mediated by engagement. In other words, in wartime economies, being proud of one’s bank is insufficient; loyalty emerges only when this pride is translated into meaningful and practical engagement with the bank’s services.

This study explored the psychological mechanisms underpinning customer attitudinal loyalty with their banks in a context of extreme geopolitical uncertainty – Ukraine during wartime – by proposing and testing a model centered on emotional attachment and psychological engagement with digital banking tools. The findings highlight that, while emotional attachment to one’s bank is present and meaningful, it does not directly predict loyalty. Instead, loyalty is driven by psychological digital engagement, which acts as a crucial mediating mechanism. Among the five dimensions of engagement, visual aesthetics and the discovery of relevant financial content emerged as the most influential, suggesting that banks’ digital interfaces and informational resources are critical levers for fostering loyalty in crisis conditions.

From a theoretical perspective, this study extends the literature on banking customer loyalty by explicitly demonstrating how Levy’s (2022) engagement model operates under conditions of extreme uncertainty. While Levy (2022) showed that psychological engagement mediates the relationship between attachment and loyalty in stable digital banking environments, our findings indicate that this mediating mechanism becomes essential – rather than complementary – in a wartime context, highlighting the contextual contingency of this relationship.

In peacetime settings, attachment may exert both direct and indirect effects on attitudinal loyalty; however, under prolonged geopolitical instability, attachment becomes an indirect driver whose influence is fully channeled through psychological engagement. In this sense, our study realizes Levy’s (2022) framework by empirically validating engagement as the central behavioral conduit through which emotional bonds are converted into active loyalty during crisis conditions. Engagement thus assumes a dual role: it is both a loyalty mechanism and a psychological coping response. This insight adds nuance to existing loyalty theory by demonstrating that the effectiveness of traditional affective drivers depends on contextual factors.

In line with recent calls to better understand consumer behavior under extreme conditions (Loxton et al., 2020), our findings offer evidence of how psychological mechanisms such as emotional attachment and digital engagement operate differently when customers face geopolitical instability and uncertainty, altering the functional role of affective constructs. Specifically, the absence of a direct effect of attachment on loyalty suggests that traditional affective drivers may be insufficient in maintaining customer relationships during crises unless supported by active digital engagement strategies implemented by banks through their online platforms and mobile applications. Such strategies may include enhancing the usability and visual appeal of digital interfaces, as well as emphasizing the bank’s social and environmental contributions through its digital services, providing informative and empowering financial content and facilitating meaningful interaction that strengthens customers’ sense of control.

This aligns with prior research showing that individuals under stress tend to reassess their behaviors and experience a reduced sense of control (Herzenstein et al., 2015), which is restored through digital engagement, due to a stress-induced reassessment process. In fact, under these conditions, individuals may reevaluate their service relationships and prioritize mechanisms that restore perceived control. In this context, emotional attachment alone may be insufficient to sustain loyalty; instead, customers require active psychological engagement with digital platforms to translate affective bonds into commitment and advocacy.

By showing that engagement becomes the key mediating factor in fostering loyalty, our study contributes to a growing but still underdeveloped stream of literature that explores how consumer loyalty mechanisms are reshaped by crisis contexts, especially in banking contexts. It also suggests that customer engagement with digital banking platforms during wartime may be shaped by control-seeking orientation, as proposed by Thompson and Schlehofer (2008). This psychological reaction may help explain variation in engagement levels in times of crisis, even among consumers with similar levels of emotional attachment to their bank, thus offering avenues for future research into segmentation and personalization in crisis-responsive digital banking strategies. Ukrainian banks, by maintaining reliable, responsive and meaningful digital interfaces, have effectively retained their customer base and reinforced resilience in the financial system – even when attachment as a driver was weaker.

Practically, the findings of this study offer several implications for banks operating in warzones or other environments characterized by extreme uncertainty. Most notably, the results underscore the critical role of psychological digital engagement in shaping customer loyalty when traditional trust anchors are weakened by geopolitical instability. Banks should prioritize the usability, functionality and pragmatical quality of their websites and mobile apps, with special attention to their interfaces and the provision of informative, empowering content that supports online customer decision-making and comparison among different options (Bulyk and Havryliuk, 2023). These elements were found to be the strongest drivers of engagement in this context among those proposed by Levy’s model (2022). Furthermore, while emotional attachment alone did not directly predict loyalty, it remains an important precursor to engagement. This suggests that banks should also work to reinforce customers’ sense of pride, recognition and alignment with the bank’s identity, thereby strengthening the emotional foundation on which engagement is built. Finally, lower-scoring dimensions such as interaction and civic orientation highlight areas for improvement: banks can enhance these aspects by enabling more responsive two-way communication and by promoting the social and environmental value of digital banking. Collectively, these strategies can help domestic banks retain customer trust and loyalty, even in the face of external threats and competition from international digital providers.

The findings also speak to broader concerns of financial resilience and inclusion during wartime. In conflict-affected contexts where physical banking services are frequently disrupted or potentially unsafe, digital platforms provide not only operational continuity but also a safer alternative for customers who may be reluctant or unable to visit a branch and more willing to spend more time online comparing different options (Bulyk and Havryliuk, 2023). Beyond their functional role, these platforms serve as symbols of institutional stability and trust, reinforcing the perception that the financial system remains accessible and dependable despite external shocks. By enabling secure, uninterrupted access to financial services, digital channels contribute to broader societal resilience and help maintain consumer confidence in times of crisis.

A particularly noteworthy finding of this study is that emotional attachment did not exert a significant direct effect on loyalty, especially in the subsamples restricted to traditional banks and Ukrainian banks. This result runs counter to prior research in stable economic environments, where attachment is typically a strong predictor of loyalty (Levy, 2022). In the Ukrainian wartime context, however, affective bonds with banks appear insufficient to sustain loyalty on their own. Instead, attachment must be translated into psychological engagement – through functional interaction with digital platforms, discovery of useful information, identity alignment and civic orientation – to influence customer loyalty. This suggests that while customers may feel proud of or connected to their banks, these sentiments remain largely symbolic unless reinforced by tangible, engaging experiences with banking services. In times of geopolitical instability, consumers may prioritize pragmatic factors such as convenience, security and digital accessibility (Bulyk and Havryliuk, 2023), thereby attenuating the direct influence of attachment. Theoretically, this finding extends existing models of bank loyalty by highlighting the contextual contingency of attachment’s role: what may be a strong direct driver of loyalty in peacetime becomes an indirect, engagement-dependent mechanism in wartime. In doing so, our study adds nuance to the understanding of consumer behavior in crisis settings (Campbell et al., 2020), showing that emotional attachment alone does not guarantee loyalty when functional reliability and digital engagement are perceived as more critical to survival and resilience.

Despite its contributions, this study is not without limitations. First, its context-specific focus on Ukraine during an ongoing armed conflict may be context-specific and restrict the generalizability of the findings. While the insights are particularly valuable for understanding consumer behavior in times of crisis, caution should be exercised in applying the results to different contexts. Nevertheless, the model and findings may still offer relevance to literature on the topic, which yet appears to be scarce and to other settings characterized by systemic disruptions, such as natural disasters, political unrest or economic crises. Second, although the constructs of emotional attachment and engagement capture essential psychological dimensions, customer relationships in the banking sector are multifaceted. Future research could benefit from incorporating additional financial, psychological or contextual variables, such as psychological traits of customers, to further expand the understanding on consumer loyalty both under peaceful conditions and under stress and uncertainty. In addition, although switching behavior is an important indicator of behavioral loyalty, the present study did not directly measure bank switching or changes in banking modality. In the wartime context, switching decisions may be constrained by institutional, contractual or safety-related factors, making attitudinal loyalty a particularly meaningful construct. Future research could examine whether emotional attachment and psychological engagement also predict actual switching behavior or migration toward foreign or fully digital banking providers, thus focusing on behavioral loyalty. Finally, the analysis was intentionally limited to the banking sector. On the one hand, this is a positive aspect, as there is a lack of literature on consumer behavior in times of crisis in the banking context. On the other hand, however, to assess the broader applicability of the model to any business (i.e. to brand loyalty), future research could extend the investigation to other service industries where customer loyalty is similarly shaped by digital engagement and perceived institutional reliability. However, these limitations, in turn, might also suggest future research directions.

The author gratefully acknowledge the invaluable support of Olha Panko and Kostiantyn Polishchuk who facilitated access to respondents and made this research possible.