Payment method preferences and digital financial adoption among Ecuadorian consumers: a cross-sectional comparison of 2025 and 2026 survey samples

1 Introduction

The transformation of payment methods constitutes one of the most relevant changes within contemporary financial digitalization. Everyday transactions, traditionally dominated by cash, are now developing in a more diverse ecosystem where debit cards, credit cards, electronic transfers, banking applications, mobile payments, digital wallets, and person-to-person payment systems coexist (Antonio et al., 2024; Ibrahim et al., 2024; Khando et al., 2022; Teker et al., 2022; Yang et al., 2021). This transition does not imply the immediate disappearance of cash, but rather a progressive reorganization of payment preferences, determined by technological availability, banking access, perceived security, consumer trust, and household economic conditions. According to the World Bank, in developing economies the proportion of adults who made or received digital payments increased from 35% in 2014 to 57% in 2021, evidencing the progress of financial inclusion through electronic channels (Demirgüç-Kunt et al., 2022).

The growth of digital payments is also observed in global financial infrastructure. The Bank for International Settlements has indicated that the use of cashless payment methods continued to grow in 2024, although cash maintains a relevant presence in different consumption contexts (Bank for International Settlements, 2026). This coexistence between traditional and digital instruments shows that the choice of a payment method does not respond solely to technological innovation, but also to established habits, transaction costs, access to financial services, risk perception, and the consumer's prior experience (Annisaa et al., 2023; Ghouse et al., 2024; Lu and Wung, 2020; Pal et al., 2020). Consequently, the analysis of payment preferences requires considering both the transformations of the financial system and the practical decisions that individuals make in their daily activities.

In Latin America and the Caribbean, the expansion of digital payments has been driven by mobile phone penetration, e-commerce, banking innovation, and behavioral changes resulting from the COVID-19 pandemic (Astarloa et al., 2021; Frisancho and Parrado, 2021; Vlaicu, 2025). The Inter-American Development Bank maintains that the pandemic accelerated the adoption of digital payments due to the increase in consumers who opted for these methods as a result of physical restrictions and the perception of greater security compared to cash use (Banco Interamericano de Desarrollo, 2022). However, this adoption has not been homogeneous, as gaps associated with income, age, educational level, trust in financial institutions, internet access, and familiarity with digital tools persist (Ba and Dieng, 2025; Panetta et al., 2025; Petríková and Kočišová, 2024). Therefore, the study of payment methods in Latin American economies requires understanding not only the growth of electronic operations but also the social and economic conditions that influence their use.

In the Ecuadorian case, the digitalization of payments has shown significant growth in recent years. The Central Bank of Ecuador reported that in 2023, electronic payment methods processed a value equivalent to 1.9 times the Gross Domestic Product, reaching the highest level of utilization in the last decade (Banco Central del Ecuador, 2024a). Likewise, the institution reported that the number of operations with electronic payment methods tripled between 2019 and 2023, reflecting greater incorporation of transfers, cards, and digital channels into the national financial dynamic (Banco Central del Ecuador, 2024b). These precedents show that the country is undergoing a transition process toward more digitized payment methods, but the evidence indicates that consumer preferences change more slowly and unevenly across groups.

From the consumer's perspective, payment methods are not neutral instruments, but decisions linked to spending organization, security, ease of use, household economy, and consumption experience. Cash may be preferred by those seeking greater immediate budget control or maintaining traditional purchasing habits (Hernández et al., 2014). The debit card may be associated with banking access, direct fund availability, and operational ease. Credit cards, in turn, may be related to financing, payment deferral, promotional benefits, or debt capacity (Liu and Dewitte, 2021). Similarly, the adoption of banking applications, online payments, P2P payments, and mobile wallets depends on technological trust, perceived usefulness, and perceived security in transactions (Chawla and Joshi, 2019).

In this context, the present study is justified by the need to generate recent empirical evidence on payment method preferences and digital financial adoption among Ecuadorian consumers. The research project from which this article derives examines payment preferences in relation to consumption habits, sociodemographic factors, and household economic conditions. The availability of two independent survey waves collected in consecutive years provides an opportunity to assess the short-interval stability or variation of aggregate payment patterns during a period of continued expansion of electronic payment infrastructure in Ecuador. The comparison is not intended to establish longitudinal change in the same individuals, nor is it linked to a single exogenous policy shock; rather, it evaluates whether the composition of reported preferences and digital-use indicators differed across the 2025 and 2026 survey samples. The contribution of the study lies in identifying persistent and varying patterns of cash, debit-card, credit-card, and digital-tool use, as well as factors associated with payment-method choice in the Ecuadorian context.

2 Methodology

2.1 Research design

The study followed a quantitative, non-experimental, repeated cross-sectional comparative design with a descriptive-explanatory scope. Two independent groups of respondents were surveyed in 2025 and 2026; the same individuals were not followed over time. Accordingly, the analysis compares distributions and associations between two survey samples and does not estimate longitudinal trajectories, individual-level transitions, or causal effects of time. The term 'survey year’ is therefore used as a grouping variable rather than as evidence of temporal evolution.

2.2 Population and sample

The target population consisted of economically active Ecuadorian consumers who use different payment methods in their daily consumption activities. The analytic sample comprised valid responses from two independent survey databases: 767 observations from 2025 to 438 from 2026. Because the two waves differed in sample size and may also differ in sociodemographic composition, estimates are interpreted primarily as findings for the surveyed samples rather than as nationally representative population parameters.

After integrating both databases, a consolidated dataset was created and a 'survey year' variable was added to identify the collection wave. Records with incomplete information required for the analyses or internally inconsistent responses were excluded during data cleaning. The final consolidated sample contained 1,205 valid records. The unequal sample sizes (767 vs. 438) were retained rather than artificially balanced because each wave represents an independent data collection. This imbalance is explicitly considered when interpreting comparisons between the two samples.

The survey was administered online using Google Forms. The questionnaire link was distributed electronically via email to potential participants during the 2025 and 2026 data-collection periods. Participation was voluntary, and respondents completed the questionnaire electronically. Eligible participants were Ecuadorian consumers who were economically active and reported using payment methods in their daily consumption activities. Incomplete questionnaires and records containing inconsistent information were excluded from the final analysis.

To assess the comparability of the two independent samples, gender, age, income level, and employment status were compared between the 2025 and 2026 waves using chi-square tests and standardized mean differences. Absolute SMD values below 0.10 were considered indicative of adequate balance, while values above 0.20 were interpreted as evidence of relevant compositional imbalance.

2.3 Data collection instrument

Information was collected through a structured consumer survey containing items on payment-method preference, familiarity with online payments, banking-application use, knowledge and use of person-to-person (P2P) payments, perceived security of payment methods, willingness to use mobile wallets, and sociodemographic characteristics. All variables analyzed in this study were measured as single-item observed indicators rather than as multi-item latent constructs. Therefore, internal consistency measures such as Cronbach's alpha or composite reliability were not applicable. The questionnaire was designed to capture specific payment behaviors, preferences, and perceptions rather than psychometric dimensions.

The main dimensions considered were:

• a) Payment method preferences: cash, debit card, and credit card.
• b) Digital financial adoption: use of banking applications, familiarity with online payments, use of P2P payments, and willingness toward mobile wallets.
• c) Perceived security: perception of security in online and in-person payments.
• d) Sociodemographic characteristics: gender, age, city, employment status, and income.
• e) Consumption context: type of establishment where certain payment methods are used, such as supermarkets, gas stations, restaurants, cafes, discount stores, and department stores.

2.4 Study variables

The main dependent variable was payment method preference, classified into three categories: cash, debit card, and credit card.

As independent variables, the survey year, gender, age, income level, employment status, city, familiarity with online payments, frequency of banking app use, knowledge and use of P2P payments, perceived security, willingness to use mobile wallets, and type of device used for transactions were considered.

An exploratory digital financial adoption classification was also derived from observed indicators related to mobile-banking use, online-payment familiarity, P2P use, and willingness toward mobile wallets. This classification was used for descriptive profiling and should not be interpreted as a validated latent scale unless psychometric evidence for a multi-item construct is provided.

2.5 Data analysis procedure

The analysis was carried out in several phases. First, the 2025 and 2026 databases were cleaned, reviewing missing values, inconsistent responses, duplicates, and homogeneity in variable names. Subsequently, both bases were integrated into a single data matrix, incorporating the “year” variable to distinguish the collection period.

In a second phase, a descriptive analysis was developed using absolute frequencies, percentages, measures of central tendency, and distribution of the main sociodemographic and financial variables. This analysis allowed characterization of the participants and description of the most used payment methods in each year.

In a third phase, the 2025 and 2026 survey samples were compared with respect to preferences for cash, debit cards, and credit cards, as well as banking-app use, familiarity with online payments, P2P use, and willingness to use mobile wallets. Chi-square tests of independence were applied to categorical variables. Because multiple inferential tests were conducted, Holm's sequential correction was used as a sensitivity procedure to control the family-wise error rate. The adjusted results did not alter the substantive conclusions obtained from the unadjusted analyses.

Subsequently, a multinomial logistic regression model was estimated to identify factors associated with payment-method preference, using cash as the reference category. The model estimated the relative odds of preferring debit or credit cards rather than cash according to sociodemographic, economic, and digital indicators. Overall model fit was evaluated by comparing the complete model with the intercept-only model using a likelihood-ratio test. McFadden, Cox–Snell, and Nagelkerke pseudo-R² measures were calculated. Residual deviance, the Akaike Information Criterion (AIC), and the Bayesian Information Criterion (BIC) were also reported. Multicollinearity was assessed using variance inflation factors adjusted for the degrees of freedom of categorical predictors. Predictive performance was evaluated through a stratified 70/30 training-test split. Accuracy, Cohen's kappa, sensitivity, specificity, precision, recall, F1-score, and balanced accuracy were calculated.

Complementarily, a decision-tree classifier was used to explore combinations of sociodemographic and digital variables associated with payment-method preference. The dataset was divided using the same stratified 70% training and 30% holdout test procedure. The tree was fitted using the training sample and evaluated using the test sample. Its performance was assessed using the confusion matrix, accuracy, Cohen's kappa, sensitivity, specificity, precision, recall, F1-score, and balanced accuracy. Given its modest out-of-sample performance, the tree was interpreted as an exploratory classifier rather than as a highly accurate predictive model.

2.6 Proposed statistical techniques

The statistical techniques used in this study were selected according to the specific objectives of the analysis, as summarized in Table 1. Frequencies, percentages, and descriptive tables were used to characterize the sample. Cross-tabulations and chi-square tests were applied to compare the 2025 and 2026 survey samples, while chi-square tests and measures of association were used to examine relationships between variables. Multinomial logistic regression was employed to explain payment-method preference, and a decision tree was used to identify consumer profiles. In addition, a digital financial adoption classification was applied to evaluate patterns of digital adoption among respondents. The correspondence between each analytical objective and the statistical technique applied is presented in Table 1.

Table 1. Statistical analysis plan according to the study objectives.

2.7 Software used

Data processing and analysis were performed using R software, utilizing statistical packages oriented toward cleaning, transformation, descriptive analysis, modeling, and data visualization. Among the packages used are tidyverse, dplyr, ggplot2, nnet, caret, rpart, factoextra, and cluster.

2.8 Ethical considerations

The research was developed for academic and scientific purposes, guaranteeing the confidentiality of the information provided by participants. The data were analyzed in an aggregated manner, without individually identifying the respondents.

All behavioral and perceptual variables were self-reported. Consequently, estimates may be affected by recall bias, social-desirability bias, common-method variance, and over-reporting or under-reporting of digital-payment use. Self-reported familiarity or frequency should not be interpreted as equivalent to objectively verified transaction records. In addition, because the sampling frame and probability of selection must be explicitly documented, generalization beyond the surveyed participants should be made cautiously.

3 Results

3.1 Sample characterization

The consolidated database consisted of 1,205 respondents from two independent data-collection waves: 767 participants in 2025 and 438 in 2026. In the pooled sample, women represented 62.57%; the most frequent age groups were 25–34 years (46.72%) and 35–44 years (27.63%) (Table 2). The most common income ranges were USD 400–750 (28.46%) and USD 751–1,000 (26.97%), and 67.30% reported full- or part-time employment.

Table 2. Demographic comparability of the 2025 and 2026 survey samples.

To examine the comparability of the two independent survey samples, gender, age, income level, and employment status were compared using chi-square tests and standardized mean differences. Relevant compositional differences were observed for gender (SMD = 0.529, p < 0.001), income level (SMD = 0.436, p < 0.001), employment status (SMD = 0.322, p < 0.001), and age (SMD = 0.228, p = 0.010). Therefore, the two samples were not demographically equivalent. These differences were addressed analytically by including survey year and the principal sociodemographic variables as covariates in the multinomial regression model. Consequently, unadjusted comparisons between waves were interpreted cautiously.

3.2 General payment method preference

The debit card was the preferred payment method for the largest proportion of consumers, with 562 responses (46.64%). In second place was cash, with 362 cases (30.04%), followed by the credit card, with 281 cases (23.32%). These results show that although bank-based methods predominate in the sample, cash maintains a relevant participation in payment preferences.

3.3 Comparison of payment method preferences between 2025 and 2026

The debit card was the most frequently selected payment method in both samples. Its share was 48.37% in the 2025 sample and 43.61% in the 2026 sample. Cash was 29.86% and 30.37%, respectively, while credit-card preference was 21.77% and 26.03%. These values describe differences between independent samples and should not be interpreted as within-person change or longitudinal evolution. The comparative distribution is presented in Table 3 and visualized in Figure 1.

Table 3. Payment method preference by year.

The chi-square test did not show a statistically significant association between survey year and payment-method preference, χ²(2) = 3.54, p = 0.170. Thus, the observed descriptive differences between the two independent samples do not provide statistical evidence of a change in the overall distribution of payment preferences. This conclusion also remains unchanged under Holm correction for multiple testing.

Figure 1 summarizes the descriptive distribution of payment preferences in the two independent survey samples. The debit card has the largest share in both samples, cash is similar across samples, and the credit-card share is higher in the 2026 sample; however, the overall between-sample difference is not statistically significant.

3.4 Comparison of digital financial adoption indicators between 2025 and 2026

Three indicators linked to digital financial adoption were analyzed: familiarity with online payments, use of P2P payments, and frequency of banking app use. Table 4 summarizes the comparative results and the corresponding chi-square tests.

Table 4. Digital financial adoption indicators by year.

Familiarity with online payments was high in both samples, and frequent banking-app use exceeded half of respondents in the 2026 sample. P2P use remained comparatively limited, with a substantial proportion reporting that they had never heard of P2P payments. None of the three indicators differed significantly by survey year, and Holm correction did not change these conclusions.

3.5 Association between payment preference and sociodemographic variables

Chi-square tests were applied and Cramer's V was calculated to evaluate the association between payment method preference and sociodemographic variables. The results are presented in Table 5.

Table 5. Association between payment preference and sociodemographic variables.

Payment method preference was significantly associated with gender, income level, and employment status. In the case of gender, cash showed a higher relative proportion of women among those who preferred it, while the credit card showed a more balanced distribution between men and women. Income level showed a stronger association: among those who preferred cash, there was a higher concentration in the ranges of 400–750 dollars and less than 400 dollars, while the credit card presented greater relative presence in middle-income ranges. Age did not show a statistically significant association with payment preference.

3.6 Association between payment preference and digital financial adoption

The relationship between payment preference and digital adoption indicators is summarized inConsumers who preferred credit and debit cards showed greater familiarity with online payments than those who preferred cash. Additionally, lack of knowledge of P2P payments was more frequent among those who indicated cash as the preferred method. The strongest association corresponded to the method perceived as safest for online purchases. In this case, a clear correspondence was observed between the preferred method and the method considered safest: those who preferred credit cards tended to identify it as the safest instrument, while those who preferred debit cards also recognized it as the safest option for online purchases (Table 6).

Table 6. Association between payment preference and digital indicators.

As a sensitivity check for multiplicity, Holm's sequential adjustment was applied to the family of reported chi-square tests. The inferential pattern was unchanged: non-significant year comparisons remained non-significant, and the reported associations with p < 0.001 or p = 0.002 remained below the adjusted family-wise threshold.

3.7 Multinomial logistic regression model

A multinomial logistic regression model was estimated to examine the factors associated with payment-method preference, using cash as the reference category. The model compared the odds of preferring a credit card or debit card rather than cash as a function of sociodemographic, economic, and digital financial indicators.

The complete model provided a significantly better fit than the intercept-only model, likelihood-ratio χ²(66) = 460.12, p < 0.001. The residual deviance was 2086.07, with a log-likelihood of −1043.04, an Akaike Information Criterion (AIC) of 2222.07, and a Bayesian Information Criterion (BIC) of 2568.48. The model produced a McFadden pseudo-R² of 0.181, a Cox–Snell pseudo-R² of 0.317, and a Nagelkerke pseudo-R² of 0.361, indicating moderate explanatory capacity.

Survey year was not a statistically significant predictor of the preference for either credit cards or debit cards over cash. Therefore, after accounting for the sociodemographic, economic, and digital indicators included in the model, membership in the 2025 or 2026 survey sample was not significantly associated with payment-method preference. This finding should be interpreted as an adjusted cross-sectional association and not as evidence of longitudinal stability or individual-level change.

Table 7 shows that perceived security in online purchases and willingness to use mobile wallets were significantly associated with preference for bank-based payment methods. Respondents who considered the credit card the safest method for online purchases had 4.45 times the odds of preferring a credit card rather than cash. Familiarity with online payments was associated with 1.67 times the odds of preferring a debit card rather than cash. In contrast, not using a banking application and reporting low willingness to use mobile wallets were associated with lower odds of preferring debit cards over cash.

Table 7. Significant predictors of the multinomial model.

The classification performance of the model was moderate in the complete analytical sample. The overall accuracy was 58.34%, with a Cohen's kappa of 0.341. Classification performance was highest for debit-card preference, with a sensitivity of 0.671, precision of 0.616, and F1-score of 0.642. For cash preference, the sensitivity was 0.478 and the F1-score was 0.526, whereas for credit-card preference the sensitivity was 0.544 and the F1-score was 0.529.

To evaluate out-of-sample predictive performance, the data were divided into a stratified 70% training sample and a 30% test sample. In the holdout test sample, the model achieved an accuracy of 51.39% and a Cohen's kappa of 0.222. This accuracy was modestly higher than the no-information rate of 46.67%, p = 0.041. Test-set classification performance was highest for debit-card preference, with a sensitivity of 0.631 and an F1-score of 0.591. Performance was lower for cash preference, with an F1-score of 0.466, and for credit-card preference, with an F1-score of 0.400. These findings indicate that the model had moderate explanatory capacity but limited out-of-sample predictive performance.

The multicollinearity assessment indicated low adjusted variance inflation values for survey year and familiarity with online payments, and moderate values for gender, banking-application use, and P2P use. However, higher adjusted VIF values were observed for age, income level, employment status, perceived online security, and willingness to use mobile wallets. Therefore, individual coefficients involving these predictors were interpreted cautiously. Overall, the results should be understood as evidence of statistical association within the fitted model rather than as causal effects or evidence of strong predictive accuracy.

3.8 Decision tree for classifying payment method preference

A decision tree was estimated to explore classification of payment-method preference using sociodemographic and digital variables. The first split was based on the payment method perceived as safest for online purchases, consistent with the bivariate and multinomial results. However, the tree should be interpreted as exploratory until its out-of-sample performance is documented.

Figure 2 shows that the root node corresponds to the debit card as the predominant category in the total sample. From this node, the tree divides consumers according to the method they consider safest for online purchases. The main classification rules are summarized in Table 8.

Table 8. Main rules of the decision tree.

Table 8 confirms that the perception of security in digital environments clearly organizes payment preferences. When consumers consider the credit card to be the safest method for online purchases, the model classifies the predominant profile as a credit card user. In contrast, when security is associated with PayPal or debit cards, willingness to use mobile wallets makes it possible to distinguish between consumers oriented toward cash and consumers oriented toward debit cards.

To evaluate the out-of-sample performance of the decision tree, the dataset was divided using a stratified 70% training sample and a 30% holdout test sample. The tree was fitted using the training data and subsequently evaluated on the test set. In the training sample, the model achieved an accuracy of 52.78% and a Cohen's kappa of 0.295. In the test sample, accuracy was 51.94%, with a Cohen's kappa of 0.278. The test accuracy was significantly higher than the no-information rate of 46.67% (p = 0.025), although the improvement was modest. Class-specific performance showed the highest sensitivity for credit-card preference (0.750), followed by debit-card preference (0.476) and cash preference (0.407). Precision was 0.485 for credit cards, 0.635 for debit cards, and 0.423 for cash, while the corresponding F1-scores were 0.589, 0.544, and 0.415, respectively. The complete class-specific performance metrics are presented in Table 9.

Table 9. Performance of the decision tree in the test sample.

The decision tree was trained on 70% of the sample and evaluated on a stratified 30% holdout test set. Overall test accuracy was 51.94%, Cohen's kappa was 0.278, and the no-information rate was 46.67%.

As shown in Table 9, the decision tree performed best in identifying credit-card preference, which presented the highest sensitivity and balanced accuracy. However, overall predictive performance remained moderate. Therefore, the tree was interpreted as a validated exploratory classifier rather than as a highly accurate predictive model.

3.9 Cluster analysis and consumer profiles

To identify consumer profiles, a k-means cluster analysis was applied to categorical variables transformed into binary variables. First, the optimal number of clusters was explored using the elbow method and the silhouette coefficient.

Figure 3 shows a progressive decrease in the within-cluster sum of squares as the number of groups increases. Although no completely marked inflection point is observed, the reduction tends to stabilize after the first clusters; therefore, this criterion was used in an exploratory manner.

Figure 4 presents the average silhouette coefficient for alternative cluster solutions. The highest value was obtained for k = 2, indicating that a two-cluster solution had the strongest internal separation according to this criterion. A three-cluster solution was nevertheless retained solely for exploratory interpretability because it distinguished traditional, transitional, and banked profiles. This choice is not presented as the statistically optimal clustering solution and should be interpreted cautiously.

The three-cluster solution distributed participants into 244 consumers in cluster 1, 390 in cluster 2, and 571 in cluster 3. The characterization is presented in Table 10.

Table 10. Distribution and characterization of the clusters.

Cluster 1 concentrates the most traditional profile, since almost half of its members prefer cash and 81.97% report income below USD 400. Cluster 2 shows greater diversification of payment methods, with a balanced presence of cash and credit, although debit remains predominant. Cluster 3 represents the most banked profile, with a dominant preference for debit cards and a greater concentration in the income ranges of USD 400–750 and USD 751–1,000.

Figure 5 shows a partial overlap between groups, especially between clusters 2 and 3. This indicates that the identified profiles do not constitute completely separate segments, but rather gradual patterns of financial behavior. Therefore, the clusters should be interpreted as exploratory profiles of consumption and financial adoption, rather than as rigid or definitive categories.

3.10 Synthesis of the results

Overall, the results show that debit cards were the most frequently preferred payment method in the pooled survey data, followed by cash and credit cards. No statistically significant differences were identified between the independent 2025 and 2026 samples in the overall distribution of payment preferences or in the main digital-adoption indicators. The bivariate analyses, multinomial model, and decision tree nevertheless indicate that payment preference is associated with economic, occupational, and digital factors, particularly familiarity with online payments, perceived security, and willingness to use mobile wallets. These associations should be interpreted as cross-sectional rather than causal.

The evidence suggests that the transition toward digital payment methods does not appear as a direct replacement of cash, but rather as a coexistence of instruments in which consumers combine traditional methods, cards, and digital tools according to their level of trust, income, and technological familiarity.

4 Discussion

The findings suggest that payment behavior among the surveyed Ecuadorian consumers is characterized by coexistence, segmentation, and differentiated adoption rather than a simple replacement of cash by digital instruments. Debit cards were the most frequently preferred method in the pooled sample, while cash retained a substantial share. The 2026 sample showed a higher descriptive share of credit-card preference and a lower share of debit-card preference than the 2025 sample, but the overall difference was not statistically significant. Because the two surveys involved different participants, these contrasts cannot be interpreted as longitudinal evolution or within-person change.

This finding contributes to the debate on the transition toward economies less dependent on cash by showing that the growth of electronic payment methods does not imply the automatic disappearance of traditional methods. In line with previous studies indicating the persistence of cash due to budget control, consumption habits, and trust, the results show that cash continues to play a relevant role for certain consumer segments, especially those with lower income levels or lower integration with digital tools (Hernandez et al., 2017). Therefore, this study helps nuance the discourse on financial digitalization by showing that the adoption of electronic payment methods and the persistence of cash are not mutually exclusive phenomena, but simultaneous dimensions of everyday financial behavior.

A central result is that survey year was not a statistically significant predictor of payment preference. This finding should not be interpreted as evidence that preferences were unchanged in the Ecuadorian population over time; rather, within these two independent survey samples, year membership did not explain a statistically detectable difference after accounting for the variables included in the analysis. The evidence therefore directs attention toward income, digital familiarity, banking-app use, perceived security, and willingness to use mobile wallets as correlates of payment-method preference.

The association between income level and payment preference confirms that the choice of payment methods is linked to the economic structure of households. Consumers with income below USD 400 were less likely to prefer credit and debit cards over cash. This result suggests that functional banking inclusion does not depend solely on access to financial instruments, but also on having economic conditions that allow them to be used regularly, safely, and conveniently. At this point, the study engages with the literature on digital financial inclusion, which warns that technological access does not always translate into effective use, especially when economic inequalities, trust gaps, or limitations in financial experience persist (Ba and Dieng, 2025; Petríková and Kočišová, 2024).

Familiarity with online payments also emerged as a relevant factor. Consumers familiar with digital payments were more likely to prefer debit cards over cash. This finding supports the idea that the debit card functions as a bridge instrument between traditional banking and digital financial adoption. Unlike the credit card, which may be associated with financing, debt, or promotional benefits, the debit card is linked to the immediate availability of funds, spending control, and operational ease. Thus, its predominance in the sample can be interpreted as a form of moderate digital adoption, in which consumers accept bank-based channels but maintain a cautious relationship with credit instruments or more advanced technologies such as mobile wallets and P2P payments.

Another important result is the limited expansion of P2P payments. Although some consumers reported having used these systems, a considerable proportion stated that they had not heard of them. This shows that digital financial adoption does not advance uniformly across all its dimensions. Consumers may be familiar with online payments and use banking applications, but still be unaware of or not use more specific mechanisms such as person-to-person payments. This finding broadens the understanding of digital adoption by showing that it should not be treated as a homogeneous category. Conceptually, the study suggests distinguishing between basic digital familiarity, mobile banking use, and adoption of advanced digital solutions. This distinction is relevant for future research, as it avoids assuming that the use of a banking application automatically equals full adoption of the fintech ecosystem.

Perceived security was the factor with the strongest explanatory power in the study. In the bivariate analyses, the multinomial model, and the decision tree, the payment method considered safest for online purchases clearly organized payment preference. Consumers who considered the credit card the safest method were more likely to prefer it over cash, while those who identified the debit card as safer tended to prefer it. This result confirms that perceived security does not operate only as a general attitude toward digital payments, but as a specific criterion guiding the choice of payment instrument. Consequently, trust in the payment method becomes a central mechanism for explaining digital financial adoption.

This finding has relevant theoretical implications. The literature on digital payment adoption often highlights variables such as perceived usefulness, ease of use, trust, risk, and technological compatibility (Chawla and Joshi, 2019; Pal et al., 2020; Yang et al., 2021). The present study contributes to this field by showing that, in the Ecuadorian context, perceived security not only predicts adoption intention but also differentiates concrete preferences among cash, debit, and credit. In addition, the decision tree confirms that security operates as a primary segmentation variable. This means that consumers do not choose only based on availability or habit, but also on a comparative evaluation of which instrument they consider most trustworthy in digital environments.

Willingness to use mobile wallets also played an important role. Categories associated with low willingness or rejection of mobile wallet use reduced the probability of preferring cards over cash. This result suggests that openness toward mobile technologies does not necessarily imply current use, but it does reveal a favorable attitude toward the digitalization of payments. In other words, willingness to use mobile wallets may function as an indicator of digital readiness. Consumers who are not willing to load a card into a phone or an application tend to remain closer to traditional payment methods, while those who show openness to these mechanisms are positioned closer to banked or digitally active profiles.

The cluster analysis complements this interpretation by identifying differentiated consumer profiles. The first group, characterized by greater dependence on cash and low income, represents a traditional profile that may face economic or trust-related barriers to adopting digital payment methods. The second group, with more diversified preferences and a strong presence in 2026, reflects a transitional stage in which cash, credit, and debit coexist. The third group, mainly oriented toward debit cards, represents a banked profile with relatively greater income stability. Although the clusters should be interpreted with caution due to the influence of the survey year on their composition, they provide exploratory evidence of consumer heterogeneity and make it possible to move beyond an aggregate reading of payment behavior.

From a practical perspective, the results have implications for financial institutions, businesses, and policymakers responsible for financial inclusion. For financial institutions, the study shows that promoting digital payment methods requires strengthening trust and perceived security, not only expanding the technological offer. Adoption campaigns should differentiate between consumers who already use cards, consumers who maintain a preference for cash, and consumers who are unfamiliar with mechanisms such as P2P payments. For businesses, the findings suggest that maintaining flexible payment infrastructure remains necessary, since the coexistence of cash, debit, and credit continues to be part of consumer behavior. For public policy, the results show that digital financial inclusion must address economic and educational barriers, especially among lower-income segments.

The main contribution of this study is to provide recent empirical evidence on payment preferences and their associated economic, sociodemographic, and digital factors in an understudied Latin American setting. Rather than demonstrating longitudinal evolution, the analysis compares two independent survey waves and combines descriptive, inferential, and exploratory predictive approaches to examine heterogeneity in payment behavior.

Furthermore, the study contributes methodologically by combining multinomial logistic regression, decision tree analysis, and cluster analysis. This strategy allows the phenomenon to be observed at three levels: statistical association, probabilistic explanation, and profile segmentation. The convergence among these approaches strengthens the interpretive validity of the findings, especially by showing that perceived security, digital familiarity, and willingness to use mobile wallets repeatedly appear as relevant dimensions. This analytical triangulation constitutes a strength of the study and offers a replicable methodological route for analyzing payment preferences in other emerging contexts.

Several limitations must be considered. First, the 2025 and 2026 databases are independent cross-sectional waves; the same individuals were not followed, so the study cannot establish longitudinal evolution or individual-level changes. Second, sample sizes differed substantially (767 vs. 438), and demographic equivalence between waves must be explicitly evaluated before attributing descriptive differences to survey year. Third, the sampling frame, recruitment process, survey distribution mode, response rate, and inclusion/exclusion criteria must be fully documented to allow readers to assess external validity and generalizability. Fourth, all behavioral and perceptual measures are self-reported and may be affected by recall bias, social-desirability bias, common-method bias, and over-reporting of digital-payment use; reported use cannot be equated with objective transaction records. Fifth, indicators such as perceived security, mobile-wallet willingness, and online-payment familiarity are treated as observed survey variables; if any were measured with multi-item scales, their psychometric properties and measurement invariance across waves should be reported. Sixth, the multinomial and decision-tree analyses require fuller diagnostic and validation evidence before strong predictive claims are made. Finally, the cluster analysis is exploratory: the silhouette criterion favored k = 2, whereas k = 3 was retained for interpretive purposes, and cluster composition was strongly related to survey year.

Despite these limitations, the study provides useful cross-sectional evidence on payment-method preferences and digital financial adoption among the surveyed Ecuadorian consumers. The findings indicate that financial digitalization should not be evaluated only through aggregate growth in electronic transactions, but also through consumers' trust, economic conditions, familiarity with digital tools, and the coexistence of multiple payment instruments. Broader population-level inference will require probability-based or well-documented sampling, comparable survey waves, and ideally longitudinal or repeated cross-sectional designs with appropriate weighting and measurement controls.

5 Conclusions

This study compared payment-method preferences and indicators of digital financial adoption among Ecuadorian consumers surveyed in two independent cross-sectional waves in 2025 and 2026. Based on 1,205 valid responses, debit cards were the most frequently preferred payment method in the pooled sample, followed by cash and credit cards. These findings point to coexistence between traditional and bank-based instruments among the surveyed participants rather than to a complete replacement of cash.

The 2026 sample displayed a lower descriptive share of debit-card preference and a higher share of credit-card preference than the 2025 sample; however, the overall difference was not statistically significant. Familiarity with online payments, P2P use, and banking-app use also showed no statistically significant between-sample differences. Because different participants were surveyed in each year and the samples were unequal in size, these results should not be interpreted as longitudinal evolution or as proof of population-level temporal change.

The association analyses and multinomial model indicate that income, familiarity with online payments, banking-app use, perceived security, and willingness to use mobile wallets are associated with preference for bank-based payment methods over cash. Perceived security was also the main splitting variable in the exploratory decision tree. These findings support the relevance of trust and digital readiness, but the predictive interpretation of the models should remain cautious until the requested multicollinearity, fit, and validation diagnostics are fully reported.

The exploratory cluster analysis identified profiles characterized by greater cash dependence, diversified payment preferences, and stronger debit-card orientation. However, because the silhouette criterion favored two clusters and the three-cluster solution was retained for interpretive purposes, these profiles should be viewed as exploratory rather than definitive market segments. Future research should use better-documented and comparable sampling procedures, objective transaction data where possible, and longitudinal or appropriately weighted repeated cross-sectional designs.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Víctor Hugo Fernández-Bedoya, Universidad Nacional Mayor de San Marcos, Peru

Reviewed by: Ghalia Nassreddine, Rafik Hariri University, Lebanon

Ira Gjika, Mediterranean University of Albania, Albania