Adoption and use of digital financial services: A meta analysis of barriers and facilitators

The study of digital financial services is not recent; however, since COVID-19, new attention has been given to solutions that avoid face-to-face interactions. Accordingly, the number of studies on the topic of digital financial services has increased, covering new topics, such as digital wallets. Thus, this study presents a weight and meta-analysis that synthesises previous literature on digital financial services use. Using 121 articles, four analyses were performed for all financial solutions, but also per main service, such as digital banking, digital management and payment services, and digital wallets. Results show the relevance of technology adoption factors such as perceived usefulness, ease of use, but also security and trust dimensions for individuals to use these solutions. Additionally, results from the moderation analysis show that more money-oriented cultures tend to use these services more when perceiving them as easy to use. The findings of this study support practitioners and future research.


Introduction
Nowadays, the presence of mobile and/or digital solutions in individuals' lives is unquestionable (Baptista & Oliveira, 2016).Especially since the occurrence of the Covid-19 pandemic, people have given preference to digital solutions that require less human contact (Al-Qudah et al., 2022).One of the main affected areas was finance/payment services.These types of services are increasingly shifting to a digital environment, providing an easier way to manage payments, control finances, and proceed with payments, among other functionalities (Kaur et al., 2020).These services bring advantages not only for their users but also to banks, financial institutions, and almost any organisation that operates with digital payments (Chawla & Joshi, 2019).In this work, digital financial services are studied overall and per main area.In fact, these services might be divided into three main fields: digital banking, digital management and payment services, and digital wallets.
Digital banking is defined as a financial service made through mobile or digital means, allowing to perform transactions, trading, visualisation of history, and sometimes permitting advisory services and cross-selling products (Baptista & Oliveira, 2015).On the other hand, digital management and payment services are not implicitly connected to a specific bank since FinTechs usually offer them.These solutions allow users to invest their money, provide investment proposals, define financial goals, perform payments, and present visualisations of history and performance (Gerlach & Lutz, 2021).Finally, digital wallets are defined as mobile applications that work as a wallet in a way that allows performing payments and storing information such as credit cards, passport details, several payment accounts, passwords, etc. (George & Sunny, 2022).
The number of studies on these services has increased greatly in the last few years, mainly covering the three areas mentioned above.These results reflect the maturity of digital financial services, especially digital banking.Given that the literature on the topic is somewhat scattered and the research process has become complex, there is a need to provide a comprehensive and synthesised picture of the progress in the field of digital financial services by evaluating and integrating the findings described in the literature, overall and per service, especially due to the emergence of more recent digital financial services, like digital wallets.Therefore, it is now relevant to understand what literature has investigated on antecedents of use, over the years, and understand how these factors might change across services.Hence, this study intends to answer the following research questions: (1) What are the main motivators and barriers influencing individuals' intention to use digital financial services?; (2) How do these factors vary across different digital financial services?; (3) How do these factors vary across different sociodemographic contexts?
To answer the above questions (1) and ( 2), a weight and metaanalysis will be performed for all digital financial services and an analysis per main service: digital banking, digital management and payment services, and digital wallets.Moreover, a moderation analysis will be developed in order to compare results across different sociodemographic contexts, providing an answer to the research question (3).A meta and weight analysis is considered a stronger method to integrate quantitative results than the usual literature reviews (Schmidt & Hunter, 2016), being a robust method that provides a concise understanding of a topic, uncovering relationships, disparities in results, and gaps (Urbach et al., 2009).Thus, one of the main reasons to use meta-analysis is its capacity to compare effect sizes across different effects.Besides the meta-analysis, a weight analysis is also performed, allowing the identification of the most frequently used constructs (Jeyaraj et al., 2006).The combination of a meta and weight analysis provides one of the most comprehensive analysis available (Baptista & Oliveira, 2016).Overall, these techniques together allow us to summarise the different studies, identify the most important variables, and quantify the moderating effects.Therefore, we believe these are strong methods that allow to answer the proposed research questions.
The article makes the following contributions.First, to the best of our knowledge, this is the first time that a weight and meta-analysis has been performed for this whole set of digital financial services, especially regarding digital management and payment services and digital wallets, providing an overall perspective of the status of digital financial services, but also a discriminated view of each major service.Secondly, it will contribute to a better understanding of the most-used variables and theories, facilitating theory development in the fields of information systems and financial research.Thus, researchers can make use of these findings to help in the selection of factors to analyse the use of digital financial services.Finally, creating a meta and weight analysis per service will allow an understanding of how digital financial services may differ between them in terms of main motivators and barriers.
The paper is structured as follows.The next section presents the research methodology with a description of the selected studies and the merging process.Section 3 presents the results, including some descriptive statistics, the meta and weight analysis results, and the moderator's analysis.Section 4 introduces the discussion of results and their implications.Section 5 presents the limitations and future research directions, followed by the conclusion in Section 6.

Literature review
There is a broad set of digital financial services, according to what they allow to perform, which organisations are behind the service, and which functionalities are available, among others.Three primary services can be identified: digital banking, digital management and payment services, and digital wallets.Digital banking is the most common digital financial service.It is usually associated with a banking organisation and allows banking activities to be performed either electronically or online (Al-Dmour et al., 2019).These banking services are available at any time, allowing tasks to be executed anytime and anywhere, avoiding queues, and reducing operating costs (Inder et al., 2022).Among the functions available, digital banking enables the performance of various transactions, trading, and visualisation of history and sometimes permits access to advisory services and cross-selling products (Baptista & Oliveira, 2015).Overall, these services can be boosted with the implementation of emerging technologies, such as adaptative business intelligence (Arjun et al., 2021).However, although this might present benefits, the technology acceptance depends on a variety of factors, especially security and convenience ones.On the other hand, although digital management and payment services perform many similar activities to those of digital banking, they are usually offered by FinTechs (Mainardes, Costa, & Nossa, 2023) or e-commerce organisations like Amazon (Jena, 2022).These services allow users to invest money, provide investment proposals, define financial goals, make payments, and present visualisations of history and performance (Gerlach & Lutz, 2021).In fact, due to the high amounts of data collected and available, several advanced models of deep learning or reinforcement learning are being developed to support decision making (Singh et al., 2022).Finally, digital wallets are the newest form of digital financial service.These are defined as mobile applications that substitute real wallets, allowing users to make payments without the participation of financial intermediaries (Tran Le Na & Hien, 2021) and store information, such as credit cards, passport details, several payment accounts, passwords, etc. (George & Sunny, 2022).In fact, digital wallets are increasingly being investigated, and not only for financial purposes.For example, blockchain based wallets for health information are also being studied (Maher et al., 2023;Mittal, Gupta, Chaturvedi, Chansarkar, & Gupta, 2021).Overall, the maturity of the digital financial services is clear, however, there are clear opportunities due to the digital advancements, like advanced methods and blockchain.Given this, it is important to understand what literature has investigated on this topic.

Methods
A meta-analysis allows one to summarise the quantitative findings from prior research on similar topics.Thus, one of the main reasons to use the meta-analysis is its capacity to compare effect sizes across different effects (Schmidt & Hunter, 2016).Although having the limitation of only evaluating quantitative studies, the number of quantitative articles on the topic of digital finance services is strongly increasing, being therefore especially relevant nowadays to provide this analysis.Thus, despite some variable criticism (Borenstein et al., 2009), some researchers consider that a meta-analysis is more robust when compared to descriptive literature reviews, creating a consolidated view of the research on a topic (Schmidt & Hunter, 2016).Given this, we believe that meta-analysis is appropriate for our study.Meta-analyses allow researchers to compute a pooled estimate of the explanatory variables used to explain the target ones, combining different effect sizes from various studies.This meta-analysis followed the standard meta-analytical calculation procedures (Sampaio et al., 2017).This method requires the selection of quantitative studies that present the following metrics: sample size, and correlation coefficients.These measures are needed to compute the pooled estimate of the relationships studied.After the collection of studies, it is necessary to merge all constructs whose definition is the same but might have different names.This process will also allow us to understand the most used dependent variable, as well as the most tested relationships.Having this, it will then be possible to compute the meta-analysis.
Additionally, a weight analysis was conducted, as it enables estimating an independent variable's importance, indicating its predictive power to the target variable (Jeyaraj et al., 2006).The combination of a meta and weight analysis provides one of the most comprehensive analysis available (Baptista & Oliveira, 2016).After data collection, the weight is calculated by dividing the number of times the independent variable is significant by the total number of relationships between the independent variable and a dependent one.Each dependent variable will therefore be classified according to its weight.The next sub-section presents the steps described above to conduct the meta and weight analysis, following the structure of previous meta-analysis investigations (Naranjo Zolotov et al., 2018).

Selection of studies
This study followed a meta-analytical study approach (Schmidt & C. Neves et al.Hunter, 2016).In particular, we applied a bivariate meta-analysis to test direct effects and meta-regressions to analyse the possible moderators.We detail each step of our meta-analysis below.Studies were collected during the month of November 2022 from the most well-known databases, such as Scopus, Science Direct and Web of Science (Webster & Watson, 2002).Queries were built on those databases using the fields available, the logical operators (AND/OR) and the most relevant keywords.First, to guarantee that quantitative articles were found, keywords like questionnaire, survey, regression, or structural equation modelling were used.Then, to define the context of the studies, keywords like digital banking, digital payment, digital wallet, and digital assistant were used.Other variations of the keywords were also searched, such as mobile/online banking, electronic wallet, and e-wallet.Finally, to capture the specific adoption and use behaviour, keywords like intention to adopt, intention to use, adoption, and use were also added.From the queries, 3510 initial articles were found.A rigorous analysis was made on these, based on a set of criteria: (1) studies published under a peer-review process; (2) studies on the scope of the analysis (excluding specific contexts, such as clinical studies, medicine, tourism, e-learning, etc.); (3) quantitative studies that show correlation values and sample sizes; (4) independent datasets; (5) studies using datasets already included were removed to avoid study bias by using the same sample (Wood, 2008).These criteria was established to make sure only quantitative peer-reviewed studies on the topic were selected, following the recommendations of previous recent meta-analysis (Franque et al., 2021;Neves et al., 2022).After this screening process, 121 articles met all criteria, being able to be used in the meta and weight analysis.Several pieces of information were recorded from each article, namely the publication year, source, independent variable, dependent variable, the correlation coefficient of each relationship, method, type of technology, sample size, country, and journal.Fig. 1 describes the selection process.

Variables merger and relationships
At the onset of the analyses, the main information to be extracted from the articles is the relationships between dependent and independent variables.Nevertheless, the occurrence of variables with different names but the same meaning is recurrent.For example, performance expectancy and perceived usefulness have been identified as holding the same meaning.Therefore, before starting the analysis, it is crucial to merge variables into one single name when representing the same concept.This process was conducted for both dependent and independent variables.The coding scheme and data extraction procedures followed previous research in meta-analyses (Rosario et al., 2016).After deciding the coding classification criteria, two research assistants coded the effect sizes.A third judge resolved the disagreements.After this process, 384 relationships were identified; however, only 30 could be used, given that only relationships that occurred three or more times can be used to perform the meta-analysis (Baptista & Oliveira, 2016).From the selected relationships, the main dependent variables identified were intention to use, use behaviour, intention to continue using, attitude and satisfaction.The variables codification is presented in Appendix A.

Moderation analysis
When testing for the significance of independent variables, it is also possible to evaluate moderation effects.This is especially relevant to understand how the factors that influence individuals intention to use digital financial services vary across different socio-demographic contexts.Therefore, several variables were collected as possible moderators of relationships.Table 1 presents the collected variables to be tested as moderators, as well as a short description.Some socio-economic variables were selected, such as the human development index (HDI) and the global innovation index (GII).HDI is a measure that summarises the standard of living, life expectancy and education and measures the overall development of a country.GII comprises 81 indicators and measures the economic innovation level of a country.Overall, it is expected that countries with higher levels of HDI and GII will present higher levels of use of digital financial solutions, given the solid digital component of these solutions.Also, more developed economies tend to invest more in the development of these solutions in order to facilitate payments and money investments (Hassan & Wood, 2020).Moreover, Hofstede's cultural indicators were collected (Hofstede & Minkov, 2010), given the substantial impact cultures might have on technology adoption and use (Tam & Oliveira, 2017).Hence, uncertainty avoidance, masculinity, individualism, long-term orientation, power distance and indulgence were collected.

Results
This section is composed of a descriptive analysis of the articles found and analysed.This includes ana analysis of the countries where studies have been conducted, the type of digital service, and the journals in which the topic of digital financial services have been published.Then, the results of the meta and weight analysis are presented, respectively.On this sub-section, four meta and weight analysis results are presented on overall digital financial services, and then per service, respectivelydigital banking, digital payment assistants, and digital wallets.Finally, the moderation analysis results are described.This last analysis was only conducted for all digital financial services, and not individually per service, given that moderation analysis can only be significant when working with more than twenty relationships (Geyskens et al., 2009).

Descriptive analytics
From the 121 datasets, samples covered 37 countries.Fig. 2 presents the cumulative sample size per country, showing the top three countries as India (22% of the respondents), Brazil (6% of the respondents) and China (5% of the respondents).Regarding the research years, Fig. 3 summarises the distribution of studies over the years, categorised by type of digital financial solution.The selected studies range from 2009 to 2022, presenting an increased tendency over the years since the last three years represent 51% of the studies.Moreover, the increasing interest in digital wallet research is noticeable, given its strong presence in recent years.
Regarding the source of the articles, Table 2 summarises the number of articles per journal.As described, articles mainly belong to the finance, management, and information systems fields, especially to the following journals: international journal of bank marketing (25%), journal of retailing and consumer services (6%), and international journal of information management (5%).Finally, in terms of the main areas of digital financial services, the majority of articles are about digital banking (69 studies), followed by digital management and payment services (34 studies), and lastly, digital wallets (18 studies).In fact, digital wallets are being increasingly studied, showing great prominence in recent studies in the digital financial arena.

Weight and meta-analysis results
Following the recommendations from recent meta-analyses, we only included studies that reported the correlation effects in our analysis, reducing potential conversion biases (Roth et al., 2018).Collecting a set of quantitative measures is required to calculate this, namely the statistically significant and non-significant relationships, the effect size, and the sample size (Cook, 1991).The meta-analytical effect sizes were corrected by sample size (Hunter & Schmidt, 2004).The random-effects model was used instead of the fixed-effects one, given the high variance in terms of effect sizes and the heterogeneity between studies due to different samples (Borenstein et al., 2010).This method considers within and between variance between studies, while the fixed effect model only considers the variation within studies.Given this, the random effect is a much more realistic method and widely used on meta-analysis (Franque et al., 2021;Neves et al., 2022).We also evaluated the publication bias, which refers to the possible tendency for studies with significant and positive results to get published instead of the studies with no statistically significant or negative results.This bias can influence the meta-analysis results.Therefore, Cochran's Q, I 2 heterogeneity measure, failsafe number and Egger's test were evaluated.In Cochran's Q, the heterogeneity is proven by the significance level.The I 2 presents heterogeneity in a range from 0% to 100%.We also assessed the failsafe number (FNS) (Rosenthal & Rubin, 1991).The failsafe number comprises the number of non-significant or unpublished studies that would be necessary to refute the findings of each relationship tested.Finally, the effect size was also checked for evidence of publication bias using Egger's test (Egger et al., 1997), where we evaluated whether the data distribution was a representative sample rather than asymmetric.Additionally, we performed meta-regression in order to analyse possible moderators in the main relationships tested.The analyses were conducted using packages of R (version 4.0.2).
In order to provide a comprehensive view of digital financial  services, an overall model was estimated, including all services, as well as three individual ones for each solution: digital banking, digital management and payment services, and digital wallets.Starting with the main model, Table 3 presents the meta-analysis results.For the main model, four target variables were identified: intention to use, intention to continue to use, attitude, and perceived usefulness.The first columns present the number of times the relationship was examined (O), the cumulative sample size (N), the correlation found in the studies corrected by sample size (r), and the confidence interval (lower and upper bounds) of 95% confidence level.Regarding this last aspect, one can consider the relationship as not statistically significant only when the interval includes the value zero.Otherwise, they are considered statistically significant.Finally, regarding heterogeneity, the Q Cochran measure and the scale-free index of heterogeneity (I2) are presented, indicating whether the data refute the homogeneity hypotheses and the percentage of variance in the dataset resulting from heterogeneity (Higgins & Thompson, 2002), respectively.This structure is present in   Botella, 2006).Therefore, we can confirm that the random effect model is adequate, and a moderation analysis can be investigated.Additionally, to provide an accurate assessment of the asymmetry, the Egger regression resulted to be non-significant, concluding no evidence of publication bias (Naranjo Zolotov et al., 2018).
The results show that out of the 30 relationships, 25 are statistically significant, and five are not statistically significant.Perceived risk to intention to use (r=-0.08),cost to intention to use (r=-0.12),privacy concerns to intention to use (r=0.01),perceived risk to attitude (r=-0.39)and privacy concerns to attitude (r=0.10) are not significant relationships.On the other hand, regarding intention to use digital financial solutions, the top five most significant impacts are from compatibility (r=0.57),attitude (r=0.55),price value (r=0.38),perceived usefulness (r=0.37), and convenience (r=0.36).Regarding intention to continue to use digital financial services, the greatest impacts are satisfaction (r=0.48),trust (r=0.26),perceived usefulness (r=0.25), and perceived ease of use (r=0.14).The most significant impacts on attitude are from perceived usefulness (r=0.39),compatibility (r=0.37),benefits (r=0.36) and subjective norms (r=0.35).Finally, perceived ease of use is statistically significant to perceived usefulness (r=0.55).
After analysing the meta-analysis results, it is relevant to understand the findings of the weight analysis.Weight analysis enables estimating an independent variable's importance, indicating its predictive power to the target variable (Jeyaraj et al., 2006).The weight is calculated by dividing the number of times the independent variable is significant by the total number of relationships between the independent variable and a dependent one.When the weight is 1, all relationships analysed were statistically significant, whereas a weight of 0 indicates no relationship was found to be statistically significant.Moreover, if the variable is tested five or more times, it is "well-utilised".If tested fewer than five times but with a weight equal to 1, the variable is a "promising predictor" (PP).On the other hand, if a variable has a weight higher or equal to 0.8 and was examined five or more times, it is considered a "best predictor" (BP) (Jeyaraj et al., 2006).By combining both a meta and weight analysis, it is possible to get a more comprehensive and complementary view of the significance and strength of the explanatory variables into the target ones.This way, it is possible to identify the strongest predictors.
Of all relationships, 20 were considered best predictors, and two were considered promising predictors.The best predictors of intention to use are perceived usefulness, perceived ease of use, subjective norms, trust, attitude, hedonic motivations, price value, habit, personal innovativeness, convenience, and compatibility.Awareness is considered a promising predictor.Satisfaction, perceived usefulness, trust, and perceived ease of use are best predictors of intention to continue to use.Also, perceived usefulness, perceived ease of use, and subjective norms are best predictors of attitude.Compatibility is a promising indicator.Finally, perceived ease of use is a best predictor of perceived usefulness.
The next step in the analysis was to evaluate the results per digital financial service.The first analysis was conducted for digital banking services, the most common digital financial service.The results are presented in Table 4. Regarding heterogeneity, the results show high heterogeneity levels (around 90%).Regarding the coefficients' significance, the meta-analysis results show that 22 out of 28 relationships are statistically significant.Therefore, the top five variables with the greatest significant impacts on intention to use digital banking are attitude (r=0.58),perceived usefulness (r=0.41),price value (r=0.39),trust (r=0.32),and hedonic motivations (r=0.28).Regarding attitude, the main predictors are perceived usefulness (r=0.39),subjective norms (r=0.39),compatibility (r=0.37), and perceived ease of use (r=0.26).The most significant variables impacting satisfaction are perceived usefulness (r=0.37) and information quality (r=0.21).Regarding intention to continue using digital banking, the significant variables with the greatest impacts are satisfaction (r=0.40),perceived usefulness (r=0.29), and trust (r=0.29).The most influencing factor for digital banking adoption are intention to use (r=0.56),perceived usefulness (r=0.41), and perceived ease of use (0.36).Finally, the results indicate that intention to use positively influences the use behaviour of digital banking (r=0.33 and perceived ease of use impacts perceived usefulness (r=0.58).
Regarding the weight analysis, the best predictors of intention to use digital banking are perceived usefulness, perceived ease of use, attitude, trust, price value, convenience, and personal innovativeness.Perceived usefulness and perceived ease of use are also best predictors of attitude.On the other hand, compatibility and subjective norms are promising indicators of attitude.The promising predictors of satisfaction are information quality and perceived usefulness.Satisfaction is the best predictor of intention to continue to use digital banking, whereas perceived usefulness and trust are promising indicators.Regarding the adoption of digital banking, the best predictor is perceived usefulness, while perceived ease of use and intention to use are promising predictors.Intention to use is a best predictor of use behaviour, and perceived ease of use is a promising indicator of perceived usefulness.
The second analysis was conducted for digital management and payment services only.The results are presented in Table 5.First, the results regarding heterogeneity are very similar to the overall model, showing high heterogeneity levels (around 90%).The results show that five out of 19 relationships are not statistically significant.Therefore, the top five variables with the greatest significant impacts on intention to use digital management and payment services are attitude (r=0.55),habit (r=0.37),perceived usefulness (r=0.30),trust (r=0.27),and perceived ease of use (r=0.26).Regarding intention to continue to use digital management and payment services, the main predictors are satisfaction (r=0.65),trust (r=0.21),perceived usefulness (r=0.20),perceived risk (r=-0.15),and perceived ease of use (r=0.15).The most significant variables impacting attitude are perceived usefulness (r=0.40) and perceived ease of use (r=0.32).Finally, perceived ease of use positively affects perceived usefulness (r=0.52).
Regarding the weight analysis, the best predictors of intention to use digital management and payment services are perceived usefulness, trust, perceived ease of use, and subjective norms.Habit and attitude are promising indicators.Regarding intention to continue to use, satisfaction, perceived usefulness, trust, and perceived ease of use are promising predictors.Perceived ease of use is a best predictor of attitude, while perceived usefulness is a promising predictor.Finally, perceived ease of use is a best predictor of perceived usefulness of digital management and payment services.
Finally, a third analysis was conducted for digital wallets, for which results are presented in Table 6.The results regarding heterogeneity show that three out of the eight relationships did not reject the homogeneity hypothesis.The remaining ones present high heterogeneity

levels (around 90%
).The results show that one out of eight relationships is not statistically significant.Therefore, the top five variables with the greatest significant impacts on intention to use digital wallets are perceived usefulness (r=0.38),attitude (r=0.36),trust (r=0.26),subjective norms (r=0.24), and perceived security (r=0.20).
Regarding the weight analysis, perceived usefulness, perceived ease of use, trust, and facilitating conditions are best predictors of intention to use a digital wallet.On the other hand, subjective norms, perceived security, and attitude are promising indicators of intention to use.

Moderation results
Along with the meta and weight analyses, a moderation analysis was conducted in order to understand changes in effect size.The results are presented in Appendix B for the relationship between perceived usefulness and intention to use digital financial services (Table B.1), perceived ease of use and intention to use digital financial services (Table B.2), subjective norms and intention to use (Table B.3), and the relationship between perceived usefulness and intention to use digital banking (Table B.4).For this type of analysis, only the relationships that were analysed several times (>20) and present high heterogeneity (Geyskens et al., 2009) can be selected.From all relationships, significant moderators were only found between perceived ease of use and intention to use digital financial services.Therefore, IDH (z=3.47),power distance (z=3.35),masculinity (z=2.33),uncertainty avoidance (z=2.20), and indulgence (z=-2.77)are statistically significant moderators.

Discussion
Digital financial services have been studied over the years, passing   through digital banking, digital management and payment services, not necessarily connected to banks, and the most recent servicedigital wallets.The current study indicated that plenty of relationships have been studied.Therefore, from the literature review, 384 relationships were analysed based on 121 quantitative articles.The meta and weight analyses showed the statistically significant and most used constructs to explain a set of target variables.
The results show that the most important predictors of intention to use financial services overall are dimensions from technology acceptance models, such as attitude, perceived usefulness, perceived ease of use, compatibility, and price value.These constructs are widely used in theories like the technology acceptance model (TAM), theory of planned behaviour (TPB), and unified theory of acceptance and use of technology (UTAUT).Nevertheless, previous studies point out that not only technology factors should be considered.Pondering money as a sensitive issue, many articles include security and trust dimensions (Merhi et al., 2019).Especially in these technologies, it is still important to highlight how organisations protect users' data and continually build trust in the service.Trust is also one of the best predictors of intention to continue using digital financial services.Finally, it is also salient to note that individuals who feel more open to innovation or achieve some enjoyment while using these types of solutions are more willing to use them (Rahman et al., 2020), given the great impact of personal innovativeness and hedonic motivations.Fig. 4 presents the meta-analysis results from the overall model of digital financial services, where the continuous arrows represent the best predictors and the non-continuous ones are the promising predictors.
Additionally, moderators were found in the relationship between perceived ease of use and intention to use digital financial technologies, namely on the variables IDH, power distance, masculinity, uncertainty avoidance, and indulgence (with a negative effect).The findings show that the relationship between perceived ease of use and intention to use digital financial services is stronger in countries with a higher human development index.The result is not surprising since cultures with higher income or education usually create individuals more willing to try innovative or digital solutions, like digital financial ones.Moreover, regarding cultural factors, in cultures with high power distance, high masculinity, high uncertainty avoidance and low indulgence, the relationship between perceived ease of use to intention to use is also stronger.Power distance refers to the relationship between a superior and a subordinate (Mora, 2013).Cultures with high power distance tend to accept an inequality distribution of power and accept and not question authority.Also, cultures with higher levels of masculinity tend to clearly have distinct gender roles or unequal pay, for example, where success and money are drivers of society (Hofstede & Minkov, 2010).
Regarding uncertainty avoidance, societies with high uncertainty avoidance tend to make calculated and safe decisions that can lead to expected outcomes, avoiding risky choices (Hofstede & Minkov, 2010).Finally, indulgence refers to the tendency of people in a society to fulfil their desires.Therefore, societies with low indulgence tend to suppress gratification, being more pessimistic and inflexible and characterised by a more pessimistic approach (Hofstede & Minkov, 2010).Overall, the results suggest that more rigid and more money-oriented societies tend to be the ones where the effort to use impacts the use of digital financial services the most.In these cultures, the intention to use these services is higher when the effort to use them is perceived as very low.

Results per main digital financial service
When comparing the three main different digital financial services, the results show that digital banking is the most mature topic, not only because it is the one with more collected articles but also due to the target variables, not only focusing on intention to use, but also on many post-adoption stages (Baptista & Oliveira, 2016).Fig. 5 presents the results of the meta-analysis for digital banking.Overall, the results are very similar to the overall model, showing the relevance of not only technological but also security factors to use digital banking.
Regarding digital management and payment services, the main difference from the other services is the presence of habit as a best predictor of intention to use, as seen in Fig. 6.Habit is defined as the extent to which individuals perform behaviours due to learning or routine (Venkatesh et al., 2012).This finding suggests that people with a certain familiarity with this service will present greater intention to use them.In fact, this type of digital service tends to require more knowledge since they are frequently used to invest money and define investment positions (Gerlach & Lutz, 2021).Therefore, only people that feel more comfortable with these services will be willing to use them.
On the other hand, the least mature topic is digital wallets, as can be seen in Fig. 7.The Meta-analysis results are somewhat similar to the overall model, showing both technological and security dimensions as primary motivators to use.However, it is relevant to notice the facilitating conditions factor is not present in the other two models.Facilitating conditions measure the degree to which individuals believe they have the necessary resources to use a technology or that organisational and technical support exists when needed (Venkatesh et al., 2012).This finding suggests that individuals are more willing to use mobile wallets when feeling that support for their use exists, both from an infrastructural point of view or the provider organisations, proving the relevant need of companies to pass a supportive and trustworthy image to their customers (Jaiswal et al., 2022).

Theoretical implications
Concerning implications for theory, this work provides a comprehensive picture of digital financial services, overall and per main service area.The meta-analysis allowed to have a cumulated influence of each explanatory variable on the dependent variable, as well as an evaluation of the significance, while the weight analysis provided a weight.Therefore, it was possible to obtain a model of the best predictors of digital financial services use.Based on this, we conclude that the main factors used are the dimensions from technological acceptance theories, like TAM, TPB and UTAUTeven on more recent investigations (e.g.Alnemer, 2022; Khan, 2022) -, adding dimensions related to trust and security, indicating that other constructs related to these security dimensions should be continually tested, as they are proven to be strong predictors, as found in previous relevant works (Oliveira et al., 2016;Talwar et al., 2020).Therefore, the resulting models provide both a synthesis and a support basis for future research.Moreover, it contributes to a better understanding of the tendencies on the digital financial services field, proving the relevance of new services, like digital wallets (Chawla & Joshi, 2021), that can still further be investigated, since the number of used predictors is low, compared to the other services that are more mature.This shows that there is still room to deep investigate these digital wallets (Mohd Thas Thaker et al., 2022), where the provided synthesis model can be a starting point.For example, recent research is starting to analyse how digital wallets can affect the financial stability and awareness of the user, or if it impulses more purchases (Lee et al., 2022).Additionally, from the weight analysis, we can see that although variables like facilitating conditions, perceived risk, and benefits have been used continuously, their weight is low, meaning that they have been found as continuously not significant.From these results, researchers can make use of these findings to help in the selection of factors to analyse the use of digital financial services.For example, promising indicators can be further explored, while the non-significant and with low weight variables can be disregarded.Overall, the same theories have been applied to different digital financial services, however, other theories could also be tested, making use of the socio-technical perspective of IS, for example, to study antecedents of use, but also extending to it to the study of outcomes (Sarker et al., 2019), as well as understanding the use of different intelligent systems, like Internet of Things, virtual reality (Arjun et al., 2021), cloud-based blockchain to also improve financial security (Amponsah et al., 2022).
C. Neves et al.For example, understand how the inclusion of artificial intelligence can help in performing financial investments through these digital services (Akbarighatar et al., 2023).

Practical implications
Regarding practical implications, the results provide relevant suggestions to the practitioner side in order to increase the use of digital financial services.The results suggest that besides the standard technological dimensions, it is important that organisations show how data is protected and treated, creating a trustworthy environment.The security dimensions play a major role in these digital services; as shown, they are present in all types of digital financial services.Another important finding is that individuals who are more familiar or have more support to use digital financial services are more willing to use them.Therefore, it is important to engage users through knowledgeable interaction, like demonstrations, and overall increase the digital literacy of individuals.

Limitations and future research
The study is not without limitations.The first one is related to the articles used to perform the analysis.Although 121 articles are considered a significant number of articles, some of the existing studies had to be discarded due to the inexistence of correlation coefficients and/or sample sizes.This way, only quantitative articles were selected.Moreover, when analysing digital wallets per se, only 18 articles were available; however, we find it reasonable given the topic's novelty.Also, most articles analysed only one country, for which we recommend future studies with more country comparisons, given the strong prospect of cultural values to influence technological behaviours.Finally, most studies have focused on technological and security dimensions, which leaves the question if other theories or constructs could also be relevant.For example, given the sensitiveness of data, theories that investigate privacy issues, like privacy calculus theory, or the protection motivation theory, that comprises the individuals' perceptions about the risk's preventive behaviours.Further studies could analyse and/or validate this, resorting to a mixed methods design, for example, to better understand consumer behaviour regarding digital financial services.

Conclusion
The study of digital financial services started many years ago.However, digital financial solutions have gained renewed attention, primarily as preventive COVID-19 solutions have been implemented to avoid face-to-face interactions and facilitate the performance of tasks anywhere and anytime.Therefore, given the vast research on the topic, we applied a weight and meta-analysis based on 121 articles, resulting in 384 identified relationships.The study was conducted for all digital financial services, but also individually for the three main identified areas: digital bankingthe most common, digital management and payment services, and digital walletsthe most recent service.As a result, this study presents a comprehensive view of the best and most promising predictors of the behavioural intention to use digital financial services, emphasising the importance of technological, but also security and trust dimensions for individuals to use these solutions.
Moreover, the moderation analysis results show that more moneyoriented cultures tend to use these services more when perceiving them as easy to use.Overall, this study not only illustrates the state of the art of digital financial services but also serves as support for future research, opening the door to new theories that might affect users' behaviours that can go beyond technological and security factors.The study also strongly supports the practitioner side, providing several recommendations for digital financial services use strategies.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 4 .
Fig. 4. Model resulting from the meta-analysis of digital financial services Note: Numerical values represent the average beta.Continuous arrows are "best predictors".

Fig. 5 .
Fig. 5. Model resulting from meta-analysis of digital banking services Note: Numerical values represent the average beta.Continuous arrows are "best predictors".

Fig. 6 .
Fig. 6.Model resulting from meta-analysis of digital management and payment services Note: Numerical values represent the average beta.Continuous arrows are "best predictors".

Fig. 7 .
Fig. 7. Model resulting from meta-analysis of digital wallet Note: Numerical values represent the average beta.Continuous arrows are "best predictors".

Table 1
Moderators description.
(Nations, 2023)tion SourceHuman development index (HDI)Index that comprises three indicators of a country: standard of living, life expectancy, and years of schooling(Nations, 2023)

Table 2
Selected studies per journal.

Table 2
(continued ) all the meta-analysis results.It can be observed that only one relationship did not reject the homogeneity hypothesis (p<0.01)(subjectivenorms to intention to use).Moreover, most relationships present a high heterogeneity level, above 90%, meaning that the variability among effect sizes is not caused by sampling error, but true heterogeneity between studies(Huedo-Medina, Sánchez-Meca, Marín-Martínez, &

Table 3
Meta and weight analysis results for all digital financial services.
Note: (O) number of observations taken from the analysis of the studies; (N) number of accumulated samples of the assessed studies; r =correlation found in the studies correct by sample size; CI (95%)=confidence interval; Q= test of heterogeneity at the individual; I 2 =scale-free index of heterogeneity; (*) = p < 0.01; ns= not significant; NC = not calculated; Egger's intercept = Asymmetry test.

Table 4
Meta and weight analysis results for digital banking services.

Table 5
Meta and weight analysis results for digital management and payment services.

Table 6
Meta and weight analysis results of digital wallet.

Table B .1
Moderation results between perceived usefulness and intention to use digital financial services.

Table B . 2
Moderation results between perceived ease of use and intention to use digital financial services.Moderation results between subjective norms and intention to use digital financial services.

Table B . 4
Moderation results between perceived usefulness and intention to use digital banking.