What Determines Digital Financial Literacy? Evidence from a Large-Scale Investor Study in Japan

Abstract

The digitalization of financial systems has intensified risks such as cyber fraud, data breaches, and financial exclusion, particularly for individuals with low digital financial literacy (DFL). As digital finance becomes ubiquitous, DFL has emerged as a critical competency. However, the determinants of DFL remain insufficiently explored. This study aims to validate a comprehensive, theory-driven model that identifies the key sociodemographic, economic, and psychological factors that influence DFL acquisition among investors. Drawing on six established learning and behavioral theories—we analyze data from 158,169 active account holders in Japan through ordinary least squares regression. The results show that higher levels of DFL are associated with being male, younger or middle-aged, highly educated, and unemployed and having greater household income and assets. In contrast, being married, having children, holding a myopic view of the future, and high risk aversion are linked to lower DFL. Interaction effects show a stronger income–DFL association for males and a diminishing return for reduced education with age. Robustness checks using a probit model with a binary DFL measure confirmed the OLS results. These findings highlight digital inequalities and behavioral barriers that shape DFL acquisition. This study contributes a validated framework for identifying at-risk groups and supports future interventions to enhance inclusive digital financial capabilities in increasingly digital economies.

1. Introduction

In today’s financial landscape, digital ecosystems are rapidly reshaping how individuals manage money, credit, and investments. This transformation has introduced a new spectrum of risks that threaten both individual financial well-being and the broader stability of financial systems. Digital financial literacy (DFL) has therefore emerged as a critical competency for mitigating these risks and safely participating in the digital economy (Jose and Ghosh 2025; OECD 2024). The widespread adoption of smartphones has facilitated the integration of online banking, mobile financial platforms, AI-powered advisory tools, and electronic trading systems into everyday financial practices, promoting a parallel evolution in financial markets (Baker 2024; Morgan and Huang 2019; Zhang and Zhang 2015). Although these innovations offer unprecedented convenience, they also expose users to escalating risks, such as phishing attacks, identity theft, unauthorized access, and trading disruptions (Ahmed et al. 2024; Căciulescu et al. 2024; Chavleishvili et al. 2024; Kumiega et al. 2016; Serrano 2020).

In Japan, digital financial fraud has become an urgent issue. According to Bloomberg News (2025), The Japan Times (2025), and JPX (2025), recent cyberattacks and phishing schemes have resulted in over $700 million in losses. Similar vulnerabilities have been identified in other developed countries, including Canada and South Korea (OECD 2024). These incidents highlight the increasing vulnerability of digital financial systems and reinforce the urgent need for comprehensive digital competencies to ensure secure and informed financial participation. Although traditional financial literacy remains important, it is no longer sufficient to navigate the complexities of today’s digital financial systems (Koskelainen et al. 2023; Kovács and Terták 2024). DFL addresses this gap by integrating financial expertise with digital proficiency, enabling individuals to securely conduct digital transactions, critically evaluate financial technologies, and protect themselves from cybersecurity threats (Lyons and Kass-Hanna 2021; OECD 2024).

Globally, DFL has emerged as a critical concern across both developed and developing economies. OECD assessments indicate that, while countries such as the Netherlands, Canada, and South Korea report relatively high levels of digital financial proficiency, significant gaps persist among older adults, low-income populations, and women across regions (OECD 2024). Studies in the United States (Lyons and Kass-Hanna 2021) and the European Union (Soldatos and Kyriazis 2022; Koskelainen et al. 2023) likewise reveal disparities in DFL, often driven by variations in digital infrastructure, educational attainment, and trust in financial institutions. In emerging economies, structural constraints further complicate digital inclusion, as documented in research from India (Rajdev 2020; Kamble et al. 2024), Indonesia (Setiawan et al. 2022), and Nigeria (Adnan et al. 2023), where efforts to promote digital engagement are challenged by infrastructural and socioeconomic barriers. These findings are consistent with broader systematic reviews on the role of fintech in advancing financial inclusion, which caution against equating access with digital readiness (Dao Ha et al. 2025). Within this international landscape, Japan presents a paradox: despite its high technological development, it continues to face substantial DFL challenges among aging populations (Fukuda et al. 2022) and less digitally engaged segments (Nguyen et al. 2022b). These comparative patterns underscore the urgency of a multidimensional, theory-driven inquiry into the determinants of DFL in Japan.

Despite the growing recognition of the importance of DFL, the factors that influence its acquisition remain insufficiently explored. Although studies by Lyons and Kass-Hanna (2021) and the OECD (2024) have advanced the development of standardized DFL assessment tools, their contributions largely emphasize measurement rather than the underlying mechanisms shaping DFL acquisition. Moreover, much of the previous literature has focused on defining and measuring DFL within specific societal clusters. These studies often adopt a narrow perspective, lacking theoretical grounding and an analysis of complex mechanisms driving the acquisition of DFL (Jamnani and Jamnani 2024; Widaningsih and Firmialy 2024; Kautsar et al. 2024; Rajdev 2020; Azeez and Akhtar 2021; Ribeiro 2021; Mardhiyaturrositaningsih and Hakim 2023; Rekha and George 2022; Prasad et al. 2018; Adnan et al. 2023; Ravikumar et al. 2022; Setiawan et al. 2022).

In the context of technology-driven finance, DFL has emerged as a multidimensional construct that integrates financial knowledge, digital proficiency, cybersecurity awareness, and behavioral adaptability (Lyons and Kass-Hanna 2021; OECD 2024). It includes competencies such as understanding fundamental financial principles, conducting secure digital transactions, interpreting algorithmic recommendations, and demonstrating resilience in navigating a digitally saturated environment (Nguyen et al. 2024; Soldatos and Kyriazis 2022; Kumar et al. 2023; Mishra et al. 2024). Beyond traditional financial practices, such as budgeting, saving, and investing, contemporary DFL also requires familiarity with mobile applications, blockchain systems, and automated tools (Amnas et al. 2024; Liu et al. 2021; Subburayan et al. 2024), as well as the ability to manage digital credentials and authenticate securely (Ogunola et al. 2024). However, a substantial gap remains in understanding how these skills are acquired through a systematized, theory-driven lens. To address these gaps, the present study investigates the individual-level determinants of DFL in Japan. Specifically, it asks, what sociodemographic, economic, and psychological factors influence the acquisition of DFL among adult investors? By answering this question through a theory-driven empirical analysis, this study seeks to clarify the mechanisms through which individuals build DFL in the context of Japan’s evolving digital financial ecosystem.

To provide a more robust theoretical foundation for understanding the determinants of DFL, this study introduces a multidimensional conceptual framework that integrates socioeconomic, behavioral, and technological dimensions. Human capital theory (Becker 1962) conceptualizes DFL as a skill shaped by investments in education, employment, income, and financial assets, suggesting that individuals with greater socioeconomic resources are more likely to acquire digital financial competencies. The technology acceptance model (TAM) (Davis 1989) and the unified theory of acceptance and use of technology (UTAUT) (Venkatesh et al. 2016) highlight the influence of perceived usefulness, ease of use, social influence, and enabling infrastructure on motivating individuals to engage with digital financial tools. These models explain how interaction with financial technologies—here defined as fintech exposure, which refers to individuals’ usage frequency and access to digital financial platforms such as mobile banking apps, e-wallets, or robo-advisory services—fosters users’ confidence and facilitates adoption. Digital divide theory (van Dijk 2013) addresses demographic and structural barriers such as age, gender, marital status, and regional digital access that shape unequal opportunities for building DFL, especially among marginalized groups. From a behavioral perspective, social cognitive theory (Bandura 2023) emphasizes the importance of self-efficacy, observational learning, and digital confidence—defined as an individual’s perceived ability to navigate and engage with digital financial tools—in shaping individual engagement with digital finance. Complementing this, the theory of planned behavior (Ajzen 1991) highlights the role of psychological traits such as risk attitudes, future orientation, and behavioral intention in guiding technology-related decision-making. By explicitly linking each theoretical model to relevant socioeconomic or psychological variables, this integrated framework provides a comprehensive and empirically testable basis for understanding the diverse factors influencing DFL acquisition in the digital age.

Despite the growing importance of DFL, robust theoretical frameworks remain largely absent in much of the existing DFL research, including in studies by Rekha and George (2022), Widaningsih and Firmialy (2024), Rajdev (2020), and Kautsar et al. (2024). These studies often lack strong theoretical foundations and fail to sufficiently consider the structural and behavioral factors that influence DFL acquisition. The absence of such frameworks makes it difficult to conceptualize how socioeconomic or behavioral traits facilitate the development of digital financial competencies. For example, while the human capital theory (Becker 1962) conceptualizes knowledge, including DFL, as a skill shaped by education, employment, and experience, the existing research seldom incorporates these socioeconomic factors into analytical models. Notably, the studies by Kautsar et al. (2024) and Widaningsih and Firmialy (2024) overlook the role of human capital investment. Similarly, the digital divide theory (van Dijk 2013), which highlights structural barriers such as infrastructure limitations and demographic disparities that restrict digital access, is rarely applied in DFL research. Although some studies (e.g., Adnan et al. 2023; Kautsar et al. 2024; Prasad et al. 2018; Rajdev 2020; Setiawan et al. 2022) address disparities in financial literacy, systemic restrictions that disproportionately affect low-income individuals, older adults, and those with varying marital or household structures are often neglected. In addition, behavioral theories such as the social cognitive theory (Bandura 2023) and theory of planned behavior (Ajzen 1991), which emphasize the role of self-efficacy, social learning, and behavioral intent, are underutilized in the context of DFL. Although behavioral influences are occasionally acknowledged, few studies apply these theories systematically to explore how psychological dimensions such as risk attitudes, future orientation, and gender-specific financial behaviors affect digital financial decision-making.

To overcome these gaps, this study adopts a comprehensive conceptual and empirical framework for understanding DFL. We integrate well-established theories: the technology acceptance model (Davis and Granić 2024), the unified theory of acceptance and use of technology (Venkatesh et al. 2016), human capital theory (Becker 1962), digital divide theory (van Dijk 2013), social cognitive theory (Bandura 2023), and the theory of planned behavior (Ajzen 1991). This integrative model provides a structured foundation for identifying the sociodemographic, economic, and psychological determinants of DFL.

Our central hypothesis posits that individual characteristics, including gender, age, marital status, education, income, assets, employment status, parental status, risk attitudes, and future orientation, significantly influence DFL acquisition. These theories, in combination, offer a multidimensional lens for analyzing how cognitive, behavioral, structural, and technological factors shape engagement with DFL acquisition. For instance, they help to explain how individuals’ beliefs, motivations, access to digital resources, social influences, demographic characteristics, socioeconomic backgrounds, and perceived utility of digital tools shape their engagement with digital financial systems. Using a large-scale dataset of 158,169 active investors from Japan’s leading online securities platforms, we empirically validate these models and provide valuable insights into the acquisition of DFL in the digital era.

This study makes three contributions: First, it advances the theoretical understanding of DFL acquisition by systematically applying a multi-theoretic approach, offering explanatory depth beyond existing descriptive studies. Second, we propose an empirically testable model that identifies the pathways through which individuals acquire digital financial competencies. Third, it provides practical insights for policymakers and academic institutions seeking to design targeted interventions that enhance digital financial capabilities, promote safe participation in financial markets, and bridge the digital financial divide.

2. Data and Methods

2.1. Data

We employ a large-scale quantitative dataset from the 2025 wave of the “Survey on Life and Money,” which is an online survey conducted by Rakuten Securities and Hiroshima University. Data collection spanned two months: January and February 2025. It specifically targeted active account holders of securities companies aged 18 years and older who logged in to their websites at least once in the previous year. Beyond collecting quantitative data on demographic, economic, and psychological aspects, it included tailored questions to assess the respondents’ DFL. Since certain respondents had been part of the panel since 2022 or 2023, variables including educational attainment, financial literacy, and a shortsighted perspective on the future were consolidated from the 2022 and 2023 survey waves conducted between November and December of each year. After excluding missing values for all key variables of interest used in the analysis, the final dataset included 158,169 respondents, representing approximately 69.14% of the initial sample. We posit that the missing values are missing at random, and that their exclusion does not systematically affect our results. To verify that the exclusion of missing values did not introduce bias, we compared the means and standard deviations of key variables before and after simplification. The results showed no statistically significant differences, indicating that the missing values were likely missing at random and did not bias the results in any consistent or systematic direction. To preserve space, the complete results of the mean and standard deviation comparisons will be provided upon request.

Rakuten Securities was selected as the sampling frame due to its extensive market penetration and national relevance. As of March 2025, Rakuten Securities had approximately 12 million users, accounting for nearly 10% of Japan’s total population and about 40% of all online securities accounts nationwide, according to the Japan Securities Dealers Association (JSDA). These figures underscore the platform’s substantial presence in Japan’s digital investment landscape. While we recognize that the sample may overrepresent individuals who are more financially engaged, digitally literate, and economically secure, we believe it provides a meaningful and policy-relevant lens into the behaviors and capabilities of Japan’s active investor class. Nonetheless, we acknowledge that generalizability to the broader population including financially marginalized or digitally excluded groups may be limited. We therefore recommend that future research incorporate more representative or stratified sampling approaches to fully capture the diversity of DFL across Japan.

2.2. Variables

2.2.1. Dependent Variable

Item development and measurement of DFL: The dependent variable is DFL. Adapting Lyons and Kass-Hanna’s (2021) broad definition, we categorized DFL into five dimensions and eight sub-dimensions. The dimensions of basic knowledge and skills each include two sub-dimensions—financial literacy and digital literacy—which involve understanding fundamental financial concepts and the use of digital tools. Awareness encompasses two sub-dimensions: recognizing the availability of DFS and understanding positive financial attitudes and behaviors. Practical knowledge consists of a single sub-dimension that addresses the effective operational use of DFS applications. Decision-making encompasses two sub-dimensions that evaluate the application of positive financial attitudes and behaviors through DFS. Finally, self-protection comprises one sub-dimension that emphasizes safeguarding against online scams and fraud.

Each of the eight sub-dimensions consists of three questions that are carefully designed to measure the respective aspects of DFL. The questions assessing the basic financial knowledge sub-dimension were adapted from the widely recognized framework developed by Lusardi and Mitchell (2008) and complemented by Lyons and Kass-Hanna (2021) and the OECD (2024). This sub-dimension was measured using three multiple-choice questions in which respondents were required to select the correct answer for each question. These questions measure the respondents’ ability to understand the implications of interest rates, inflation, and risk diversification.

For all sub-dimensions except decision-making, specific questions were adapted from related research (Choung et al. 2023, 2025), with the overarching conceptual frameworks of Lyons and Kass-Hanna (2021) and the OECD (2024) being incorporated to ensure methodological rigor and consistency. Previous studies on DFL have largely overlooked decision-making as a dimension, which has created a gap that prevents a fully comprehensive measure of DFL. To address this limitation, this study follows the specific guidelines outlined by Lyons and Kass-Hanna (2021) to identify key indicators for measuring the two sub-dimensions of decision-making, while integrating excerpt questions from OECD (2024) reports to enhance the assessment’s precision and validity. Based on these sources, a well-validated set of six questions was developed, with three questions for each sub-dimension. Although the OECD-provided questions lacked meticulous detail in their original form, their adaptation within this study represents the first creation of a comprehensive set of questions that effectively measures both sub-dimensions of the decision-making dimension.

The remaining seven sub-dimensions were assessed using three Likert-scale questions. These questions required respondents to indicate their level of agreement or disagreement with the given statements, with the possible answers ranging from “strongly disagree” (1 point) to “strongly agree” (5 points). This systematic approach ensured that all sub-dimensions were measured comprehensively and consistently within the overarching DFL framework.

Table A1. (a) Descriptions of digital financial literacy questions (other than financial knowledge) using the Likert-scales items. (b). Descriptions of financial knowledge questions using Multiple choice options. Basic Financial knowledge *.

(a)
		1	2	3	4	5
		Strongly Disagree	Disagree	Neither Agree nor Disagree	Agree	Strongly Agree
Digital knowledge *
	I know how to turn on and off digital devices such as the mobile phone, computer, and tablet	1	2	3	4	5
	I know how to unlock digital devices via biometrics function (e.g., fingerprint, facial recognition)	1	2	3	4	5
	I know how to create a user account and manage it (e.g., change or reset password)	1	2	3	4	5
Awareness of DFS *
	I am aware that banking services are offered online or via mobile phone	1	2	3	4	5
	I am aware that payments can be made online or via mobile phone	1	2	3	4	5
	I understand the purpose and usage of online/mobile banking services	1	2	3	4	5
Awareness of positive financial attitudes and behaviors *
	I understand that excessive borrowing can lead to significant financial instability	1	2	3	4	5
	I am aware that long-term financial planning, including retirement savings and investment strategies, is crucial for achieving financial stability	1	2	3	4	5
	I understand that avoiding impulsive spending and making informed purchasing decisions are important for maintaining good financial health	1	2	3	4	5
Practical know-how of DFS *
	I know how to open an account on the DFS application or platform	1	2	3	4	5
	I know how to navigate the DFS application or platform	1	2	3	4	5
	I know how to make payment using mobile payment services	1	2	3	4	5
Positive financial attitudes *
	I believe in the effective management of my daily expenses, as well as in planning and saving for my long-term financial goals	1	2	3	4	5
	I understand the importance of using reliable and secure channels when sending money to family or friends	1	2	3	4	5
	I believe it is important to carefully evaluate my ability to manage additional debt before deciding to borrow money	1	2	3	4	5
Positive financial behaviors through DFS *
	I evaluate various digital financial services to determine the most suitable option for sending money internationally (remittances)	1	2	3	4	5
	I assess the security features and regulatory compliance of digital financial service providers before utilizing their services	1	2	3	4	5
	I select digital financial service providers with a proven track record of customer satisfaction and reliability	1	2	3	4	5
Self-protection from digital scams *
	I know how to avoid unnecessary fees for online/mobile financial transactions	1	2	3	4	5
	I have the ability to screen out deceptive saving schemes or borrowing services online or via mobile application	1	2	3	4	5
	I have the ability to detect and avoid voice phishing	1	2	3	4	5
(b)
Suppose you had $100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow? A. More than $102 B. Exactly $102 C. Less than $102 D. Do not know E. Refuse to answer
Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After 1 year, how much would you be able to buy with the money in this account? A. More than today B. Exactly the same C. Less than today D. Do not know E. Refuse to answer
Please tell me whether this statement is true or false. “Buying a single company’s stock usually provides a safer return than a stock mutual fund” A. True B. False C. Do not Know D. Refuse to answer

Note: * denotes the sub-dimensions being measured.

a and Table A1b in Appendix A.1 shows the specific questions used to assess each of the eight DFL sub-dimensions.

Reliability assessment using Cronbach’s alpha: Following the development of the DFL measurement scale, a reliability analysis was conducted, as shown in

Table 1. Cronbach’s alpha reliability scores for the digital financial literacy sub-dimensions.

Sub-Dimensions	Average Interitem Covariance	Scale Reliability Coefficient
Digital Literacy	0.614	0.811
Awareness of DFS	0.535	0.905
Awareness of FAB	0.416	0.828
Practical Know How	0.724	0.839
Positive Financial Attitude	0.348	0.779
Positive Financial Behavior	0.612	0.804
Self-Protection	0.583	0.801
DFL Index Less Financial Knowledge	0.383	0.9370

DFS: digital financial services; FAB: financial attitude and behavior.

, to evaluate the internal consistency of the investment. Cronbach’s alpha, a widely used statistical measure to evaluate scale reliability (Cronbach 1951), was used to examine the inter-correlations among the elements within each sub-dimension of the scale. Alpha values range from 0 to 1, and a higher Cronbach’s alpha value indicates stronger internal consistency, which is the extent to which items reliably measure the same underlying construct. In social science research, an alpha value of 0.70 or above is generally considered acceptable to ensure that the measurement instrument demonstrates sufficient reliability. Given the multidimensional nature of DFL, assessing internal consistency is essential to confirming the robustness and coherence of the measurement framework. Accordingly, reliability analysis was performed for each of the seven sub-dimensions, except for financial knowledge, which was treated as a continuous variable rather than a construct-based scale. The results of the reliability assessment were as follows.

The resulting coefficients (Cronbach’s alpha) for each sub-dimension, as well as for the overall DFL index, exceeded the widely accepted threshold of 0.7, confirming that there was satisfactory internal consistency across all constructs. These results support reliability of the instrument in comprehensively capturing the multifaceted nature of DFL. To further establish construct validity and examine the underlying dimensional structure of the instrument, an exploratory factor analysis (EFA) was conducted.

Prior to factor extraction, the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy was computed to assess the suitability of the dataset for factor analysis. The overall KMO value was 0.9454, which indicates excellent sampling adequacy and confirms that the correlations among valuables were sufficient to proceed with EFA. Additionally, individual KMO values for each variable are reported in

Table 2. Kaiser–Meyer–Olkin measure of sampling adequacy for individual items.

Variables	KMO
Digital literacy
q331	0.9690
q332	0.9648
q333	0.9636
Awareness of DFS
q334	0.9235
q335	0.9222
q336	0.9798
Awareness of positive financial attitude and behaviors
q337	0.9633
q338	0.9553
q339	0.9528
Practical know-how of DFS
q3310	0.9120
q3311	0.9002
q3312	0.9810
Positive financial attitudes
q3313	0.9664
q3314	0.9808
q3315	0.9623
Positive financial behaviors
q3316	0.9647
q3317	0.9081
q3318	0.9310
Self-protection from digital scams
q3319	0.9675
q3320	0.8876
q3321	0.8931
Overall	0.9454

to provide insight into whether specific items may be contributing weakly to the factor structure.

Given the results of the KMO test, exploratory factor analysis (EFA) was conducted using the principal components factor extraction method to assess the underlying structure of the DFL construct and to evaluate whether the questionnaire items effectively capture the intended dimensions of DFL. The analysis identified 21 factors and their eigenvalues and proportions of explained variance, as presented in

Table 3. Factor loadings from exploratory factor analysis.

Factor	Eigenvalue	Difference	Proportion	Cumulative
Factor 1	9.63050	7.32967	0.4586	0.4586
Factor 2	2.30083	1.05940	0.1096	0.5682
Factor 3	1.24143	0.35535	0.0591	0.6273
Factor 4	0.88608	0.06019	0.0422	0.6695
Factor 5	0.82589	0.19044	0.0393	0.7088
Factor 6	0.63545	0.07330	0.0303	0.7391
Factor 7	0.56214	0.00325	0.0268	0.7658
Factor 8	0.55889	0.03323	0.0266	0.7924
Factor 9	0.52567	0.05563	0.0250	0.8175
Factor 10	0.47004	0.03237	0.0224	0.8399
Factor 11	0.43767	0.01646	0.0208	0.8607
Factor 12	0.42121	0.02415	0.0201	0.8808
Factor 13	0.39706	0.05130	0.0189	0.8997
Factor 14	0.34576	0.00985	0.0165	0.9161
Factor 15	0.33591	0.02539	0.0160	0.9321
Factor 16	0.31052	0.01352	0.0148	0.9469
Factor 17	0.29700	0.03435	0.0141	0.9610
Factor 18	0.26265	0.01065	0.0125	0.9736
Factor 19	0.25200	0.05841	0.0120	0.9856
Factor 20	0.19359	0.08388	0.0092	0.9948
Factor 21	0.10972	.	0.0052	1.0000

.

The analysis revealed that the first three factors together accounted for approximately 63% of the total variance in the DFL construct. Factor 1 explained 45.86% of the variance, Factor 2 contributed 10.96%, and Factor 3 accounted for 5.91%. In accordance with the Kaiser criterion that recommends retaining factors with eigenvalues greater than 1, only the first three factors, with eigenvalues of 9.63050, 2.30083, and 1.24143, respectively, were retained.

Following factor extraction, the unrotated factor solution was examined to assess the underlying structure of the data. The results revealed a dominant first factor, with all items loading above the threshold of 0.50 (as shown in

Table 4. Factor loadings (pattern matrix) and unique variances.

Variable	Factor 1	Factor 2	Factor 3	Uniqueness
q331	0.5779	−0.4190	−0.0525	0.4877
q332	0.6578	−0.3432	−0.2102	0.4053
q333	0.7508	−0.3466	−0.2044	0.2743
q334	0.7849	−0.4297	−0.0448	0.1973
q335	0.7907	−0.4237	−0.0355	0.1940
q336	0.7810	−0.1622	−0.1601	0.3382
q337	0.6874	−0.3547	0.2912	0.3169
q338	0.7304	−0.1153	0.2999	0.3633
q339	0.6984	−0.0432	0.3439	0.3921
q3310	0.7186	0.1237	−0.3863	0.3191
q3311	0.6780	0.2874	−0.4115	0.2883
q3312	0.7261	−0.0483	−0.2318	0.4167
q3313	0.5479	0.3404	0.1543	0.5600
q3314	0.7281	0.1217	0.1215	0.4403
q3315	0.6908	−0.1920	0.3592	0.3569
q3316	0.6560	0.3110	0.2295	0.4203
q3317	0.5367	0.4825	0.2721	0.4051
q3318	0.5888	0.4058	0.2719	0.4146
q3319	0.6371	0.3925	−0.1187	0.4260
q3320	0.5987	0.5216	−0.2029	0.3284
q3321	0.5611	0.4096	−0.1871	0.4824

), which indicates a strong and coherent unidimensional construct. These findings reinforce the decision to retain Factor 1 as the basis for constructing a composite index that represents the core sub-dimensions of DFL. Accordingly, all 21 Likert-scale items were retained and used to compute the composite DFL Index. The financial literacy subdimension, assessed using 3 multiple-choice items, was excluded from the factor analysis due to its categorical format. Nonetheless, these items were retained in the final DFL index to ensure full representation of the construct.

Methodology for calculating the DFL Index: Following a reliability and validity assessment, which confirmed satisfactory internal consistency across the measurement instruments, we calculated the DFL Index using all 24 items. We adopted the methodology proposed by Lyons and Kass-Hanna (2021) and the OECD (2024) to calculate the DFL Index, which involved computing an average index for each sub-dimension and summing them to create a composite DFL score. The detailed calculation process is as follows.

First, because the financial knowledge index was based on multiple-choice questions, it was calculated as the number of correct responses (ranging from 0 to 3) averaged across the respondents. This approach aligns with previous studies (Kadoya and Khan 2020b; Katauke et al. 2023; Lal et al. 2022; Nguyen et al. 2022a; Watanapongvanich et al. 2022).

Second, because the remaining seven sub-dimensions were assessed using Likert-scale questions (1 = “strongly disagree” to 5 = “strongly agree”), the index for each sub-dimension was computed by summing the responses to all items within that sub-dimension and dividing by the total number of items.

Finally, the composite DFL score was calculated by summing the average scores for all eight sub-dimensions (range 7–36). Since the financial knowledge sub-dimension was measured on a different scale than the others, the composite index and sub-dimension scores were standardized using z-score normalization. This method rescales each index or any type of variable by centering the mean at 0 and setting the standard deviation to 1, effectively transforming it into a standard-normal distribution. Consequently, this method produces standardized scores that indicate how many standard deviations a value is above or below the mean, enabling researchers to examine and compare indices or variables with different units or value ranges (Lyons and Kass-Hanna 2021).

2.2.2. Independent Variable

The selection of independent variables was guided by six well-established behavioral and learning theories (Ajzen 1991; Bandura 2023; Becker 1962; Davis and Granić 2024; van Dijk 2013; Venkatesh et al. 2016). These frameworks provide a structured foundation for understanding the various factors that influence DFL acquisition. Accordingly, we categorized the independent variables into three main groups: sociodemographic (gender, age, educational attainment, marital status, having children, and unemployment), economic (household income and assets), and psychological (risk aversion and a myopic view of the future).

Table 5. Variables definitions.

Variables	Definition
Dependent variables
DFL Index	A continuous variable attained by summing the average scores of the eight sub-dimensions that measure digital knowledge, financial knowledge, awareness of digital financial services, awareness of positive financial attitudes and behaviors, practical knowledge of digital financial services, positive financial attitudes and behaviors, and self-protection against digital scams
DFL Median (for robustness test)	A binary variable equal to 1 if the respondent’s DFL score is equal to or above the 50th percentile (31 points and above), and 0 otherwise
Independent variables
Male	A binary variable equal to 1 if the respondent gender is male, and 0 otherwise
Age	Continuous variable: age of the respondents
Age Squared	Square of respondent age
University Degree	A binary variable equal to 1 if the respondent has a university degree at least, and 0 otherwise
Unemployed	A binary variable equal to 1 if the respondent is unemployed, and 0 otherwise
Married	A binary variable equal to 1 if the respondent is married, and 0 otherwise
Having a Child	A binary variable equal to 1 if the respondent has a child, and 0 otherwise
Household Income	A continuous variable indicating the respondent’s estimated annual income in Japanese yen for 2025
Log of Household Income	A natural log of the respondent’s estimated annual income in Japanese yen for 2025
Household Asset	A continuous variable indicating the respondent’s household financial asset balance in Japanese yen for 2025
Log of Household Assets	A natural log of the respondent’s household financial asset balance in Japanese yen for 2025
Myopic View of the Future	A binary variable equal to 1 if the respondent agrees with the statement “Since the future is uncertain, it is a waste to think about it,” and 0 if otherwise.
Risk Aversion	A continuous variable indicating the probability of the respondent will go out with an umbrella if it rains, reflecting the respondent’s risk aversion

provides detailed definitions and measurements of the study variables.

2.3. Descriptive Statistics

Table 6. Descriptive statistics for the variables.

Variable	Mean	Std. Dev.	Min	Max
Dependent Variables
** DFL Index	30.233	4.567	7	36
DFL Median	0.537	0.499	0	1
Independent Variables
Male	0.67	0.47	0	1
Age	46.34	12.26	18	90
Age Squared	2297.18	1177.61	324	8100
University Degree	0.64	0.48	0	1
Married	0.67	0.47	0	1
Having a Child	0.59	0.49	0	1
Unemployed	0.07	0.26	0	1
Household Income	JPY 7,694,153	JPY 4,284,923	JPY 1,000,000	JPY 20,000,000
Log of Household Income	15.69	0.62	13.82	16.81
Household Assets	JPY 21,574,242	JPY 25,514,145	JPY 2,500,000	JPY 100,000,000
Log of Household Assets	16.27	1.11	14.73	18.42
Myopic View of the Future	0.15	0.36	0	1
Risk Aversion	0.53	0.24	0	1
Observations	158,169

** Raw average scores before normalization.

presents the descriptive statistics for the variables used in this study. The average DFL score was 30.23, with approximately 53.7% of the participants scoring above the median. Regarding sociodemographic and economic characteristics, the sample had a mean age of 46 years, with males comprising approximately 67% of the population; moreover, 64% of participants held university degrees, 67% were married, and 59% had children. Only 7% of the respondents reported being unemployed. The average household income in 2025 is JPY 7,694,153 and the average household balance of financial assets is JPY 21,574,242. Risk aversion was notable among the sample: 53% identified as risk-averse, while 15% perceived the future as uncertain and believed it was a waste to plan for the future.

2.4. Methods

To empirically test the research question—what sociodemographic, economic, and psychological factors influence the acquisition of DFL among adult investors in Japan?—we adopt a theory-driven, quantitative modeling strategy using ordinary least squares (OLS) regression. Given that the primary dependent variable, “DFL Index,” is continuous in nature, the ordinary least squares (OLS) regression model was determined to be the most appropriate method for analyzing its relationship with key sociodemographic, economic, and psychological variables. This analytical approach is grounded in this study’s conceptual framework, in which six well-established theories inform the selection and categorization of explanatory variables. To reflect this theoretical structure, we apply a hierarchical linear regression approach, beginning with sociodemographic variables and progressively introducing economic and psychological constructs in subsequent model stages. This allows us to examine the unique and combined effects of each conceptual block on digital financial literacy outcomes. To investigate the relationship between DFL and the key variables among Japanese investors, the following equation was employed:

$$Y_i={\beta }_0+{\sum }_{j=1\dots p}{\beta }_j{X}_{ij}+ {\epsilon }_i,$$

(1)

where $Y_i$ is the measure of the dependent variable DFL Index, ${\beta }_0$ is the intercept of the model, $X_{ij}$ corresponds to the $j^{th}$ explanatory variable of the model (j = 1 to p), and ${\epsilon }_i$ is the error term. In addition to the main demographic, socioeconomic, and psychological variables, the model incorporates two theoretically motivated interaction terms: age × education and income × gender. These were selected based on the prior literature and the conceptual framework developed in the Introduction. The age × education interaction is informed by life-course learning theory and the cognitive aging literature (Salthouse 2004), which suggest that, while education typically enhances digital competencies, its influence may attenuate with age due to declining cognitive adaptability and generational differences in technology exposure. This interaction enables us to test whether the benefits conferred by formal education on DFL vary systematically across age cohorts. The income × gender interaction, on the other hand, addresses structural inequalities in financial and technological access. Drawing on digital divide and gender-equity research (e.g., Karim et al. 2023; World Economic Forum 2023), we hypothesize that men may derive greater benefit from rising income levels in terms of DFL acquisition than women, owing to persistent disparities in financial autonomy, digital engagement, and role expectations. These interaction terms allow us to explore how socioeconomic and demographic variables intersect to shape heterogeneous pathways of digital financial capability.

The full specifications of Equation (1) are represented in Equations (2)–(4), each of which occupies a distinct explanatory variable.

$$\begin{array}{c}\hfill DFL Index={\beta }_0+{\beta }_1{Male}_i+{\beta }_2{Age}_i+{\beta }_3{Age Squared}_i+{\beta }_4{University Degree}_i+{\beta }_5{Married}_i+{\beta }_6{Child}_i+\hfill \\ \hfill {\beta }_7{Unemployed}_i + {\epsilon }_i.\hfill \end{array}$$

(2)

$$\begin{array}{c}\hfill DFL Index={\beta }_0+{\beta }_1{Male}_i+{\beta }_2{Age}_i+{\beta }_3{Age Squared}_i+{\beta }_4{University Degree}_i+{\beta }_5{Married}_i+{\beta }_6{Child}_i+\hfill \\ \hfill {\beta }_7{Unemployed}_i+{\beta }_8{lHincome2024}_i+{\beta }_9{lHFasset2024}_i+ {\epsilon }_i.\hfill \end{array}$$

(3)

$$\begin{array}{c}\hfill DFL Index={\beta }_0+{\beta }_1{Male}_i+{\beta }_2{Age}_i+{\beta }_3{Age Squared}_i+{\beta }_4{University Degree}_i+{\beta }_5{Married}_i+{\beta }_6{Child}_i+\hfill \\ \hfill {\beta }_7{Unemployed}_i+{\beta }_8{lHincome2024}_i+{\beta }_9{lHFasset2024}_i+{\beta }_{10}Inter\_University Degree\&{Age}_i+\hfill \\ \hfill {\beta }_{11}Inter \_ Male\&lHincome{2024}_i+{\beta }_{12}{RiskAversion}_i+{\beta }_{13}{MyopicView}_i+{\epsilon }_i.\hfill \end{array}$$

(4)

To check for multicollinearity of the model, which could lead to spurious regression, we performed pairwise correlation and variance inflation factor (VIF) analyses. The results indicated low inter-variable correlations (r < 0.6), which is consistent with the results of Dormann et al. (2013), who suggest that correlations below this threshold generally pose minimal risk of multicollinearity in regression models, while the mean VIF value was 10.148. While this value may typically suggest moderate multicollinearity, it is considered acceptable in this context due to the inclusion of two theoretically essential interaction terms. It is well documented that interaction terms—especially when derived from correlated predictors—can inflate VIF values without necessarily undermining model validity. Therefore, the observed VIF levels do not raise substantial concerns about multicollinearity in our model (results available in

Table A2. Variance Inflation Factor (VIF) Test.

	VIF	1/VIF
Male	1.076	0.929
Log of Household income	3.613	0.277
Male × Log of Household income	3.052	0.328
University Degree	1.14	0.877
Age	57.162	0.017
University Degree × Age	3.186	0.314
Age Square	54.322	0.018
Married	1.774	0.564
Having a Child	1.705	0.587
Unemployed	1.428	0.7
Log of Household Asset	1.426	0.701
MyopicView	1.01	0.99
RiskAversion	1.028	0.973
Mean VIF	10.148	.

and

Table A3. Pairwise correlations test.

Variables	DFL Index	Male	Age	University Degree	Married	Having a Child	Unemployed	Household Income	Household Asset	Myopic View	Risk Aversion
DFL Index	1.000
Male	0.023 *	1.000
	(0.000)
Age	−0.024 *	0.179 *	1.000
	(0.000)	(0.000)
University Degree	0.106 *	0.137 *	−0.125 *	1.000
	(0.000)	(0.000)	(0.000)
Married	−0.006 *	0.080 *	0.209 *	0.044 *	1.000
	(0.026)	(0.000)	(0.000)	(0.000)
Having a Child	−0.034 *	0.046 *	0.324 *	−0.043 *	0.598 *	1.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Unemployed	−0.017 *	0.094 *	0.347 *	−0.008 *	−0.029 *	0.009 *	1.000
	(0.000)	(0.000)	(0.000)	(0.002)	(0.000)	(0.000)
Household Income	0.137 *	0.076 *	−0.040 *	0.185 *	0.400 *	0.246 *	−0.291 *	1.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Household Asset	0.171 *	0.104 *	0.331 *	0.175 *	0.143 *	0.102 *	0.150 *	0.350 *	1.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Myopic View	−0.012 *	−0.037 *	−0.063 *	−0.024 *	−0.031 *	−0.034 *	−0.019 *	−0.026 *	−0.056 *	1.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Risk Aversion	−0.014 *	0.054 *	0.103 *	0.071 *	−0.002 *	−0.007 *	0.051 *	0.011 *	0.100 *	−0.065 *	1.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.429)	(0.004)	(0.000)	(0.000)	(0.000)	(0.000)

* p < 0.1.

of the Appendix A.2).

We conducted robustness tests to validate the findings of the main model using an alternative measure of the DFL Index. For this purpose, we created a binary variable for the DFL Index, called “DFL Median,” assigning a value of 1 if the respondent’s raw DFL score was equal to or greater than the sample median score (31 points) and 0 otherwise. A similar approach was used by Choung et al. (2023, 2025) and Widyastuti et al. (2024a), which confirms the credibility of this type of measurement. Then, a probability regression estimate was employed to examine the relationship between this binary variable and key sociodemographic, economic, and psychological variables.

All empirical analyses, including the estimation of ordinary least squares (OLS) models and probit models for robustness checks, were performed using Stata version 18, and statistical significance was determined at a p-value threshold of 0.10.

3. Empirical Results

Table 7. Determinants of digital financial literacy acquisition: ordinary least squares estimation.

Variables	Model 1	Model 2	Model 3
		Dependent Variable: DFL Index
Male	0.0218 ***	0.0104 **	0.0138 ***
	(0.00529)	(0.00521)	(0.00523)
Age	0.0206 ***	0.00382 ***	0.00452 ***
	(0.00141)	(0.00141)	(0.00146)
Age Squared	−0.000217 ***	−7.94 × 10−5 ***	−7.57 × 10−5 ***
	(1.48 × 10−5)	(1.48 × 10−5)	(1.48 × 10−5)
University Degree	0.213 ***	0.115 ***	0.120 ***
	(0.00517)	(0.00525)	(0.00527)
Married	0.0182 ***	−0.0801 ***	−0.0772 ***
	(0.00645)	(0.00669)	(0.00671)
Having a Child	−0.0786 ***	−0.0666 ***	−0.0713 ***
	(0.00636)	(0.00627)	(0.00629)
Unemployed	−0.0125	0.0337 ***	0.0418 ***
	(0.0108)	(0.0111)	(0.0111)
Log of Household Income		0.148 ***	0.114 ***
		(0.00491)	(0.00724)
Log of Household Asset		0.130 ***	0.132 ***
		(0.00254)	(0.00255)
Inter_University Degree & Age			−0.00125 ***
			(0.000421)
Inter_Male & lHincome			0.0523 ***
			(0.00819)
Myopic View of the Future			−0.0172 ***
			(0.00662)
Risk Aversion			−0.125 ***
			(0.0101)
Constant	−0.509 ***	−4.358 ***	−1.832 ***
	(0.0324)	(0.0719)	(0.0539)
Observations	158,169	158,169	158,169
R-squared	0.014	0.048	0.049

Standard errors in parentheses. *** p < 0.01, ** p < 0.05.

presents the results of the hierarchical OLS models that were used to examine the factors that shape DFL acquisition. Model 1 included sociodemographic variables, Model 2 included economic factors, and Model 3 incorporated interaction variables and psychological characteristics.

Among the sociodemographic variables, being male most strongly promotes the acquisition of DFL, with the results being consistent and significant across all models at the 5% significance level. Age positively influences the DFL association. However, the significant negative coefficient of age squared indicates a nonlinear relationship, where individuals develop DFL as they age (younger-to-middle age), but beyond a certain point, their ability to acquire new competencies begins to decline. Individuals with university degrees consistently demonstrate significantly greater levels of DFL acquisition across all models at the 1% significance level. Marital status shifts notably across models. While being married initially promotes DFL acquisition in Model 1, it becomes a constraining factor after controlling for economic and psychological factors in Models 2 and 3. Similarly, having children hinders the ability of individuals to acquire DFL at the 1% significance level across all models. Unemployment does not show a significant relationship with DFL acquisition in Model 1, but after economic variables are introduced in the subsequent models, it seems to enhance DFL acquisition.

Regarding economic variables, both household income and household assets strongly increase DFL acquisition, which underscores the importance of financial stability in supporting individuals’ ability to develop DFL. In the final model, the interaction between age and education reveals a diminishing return effect, where the positive influence of education on digital financial literacy becomes weaker at higher ages. Moreover, the positive coefficient for the interaction between being male and household income indicates that the relationship between income and digital financial literacy is more pronounced among male respondents. Beyond demographic and economic factors, psychological variables play a significant role, as a myopic view of the future and risk aversion prevent individuals from effectively acquiring DFL.

To assess the robustness of our findings, we re-estimated the main models by employing a probit regression, using the “DFL Median,” which is binary in nature. The results from the probit specification across all models, as shown in

Table 8. Probit regression results of digital financial literacy acquisition with the binary DFL median as the dependent variable.

Variables	Model 1	Model 2	Model 3
	Dependent Variable: DFL Median
Male	0.118 ***	0.108 ***	0.110 ***
	(0.00695)	(0.00703)	(0.00705)
Age	0.0126 ***	−0.00864 ***	−0.00717 ***
	(0.00185)	(0.00190)	(0.00198)
Age Squared	−0.000166 ***	7.29 × 10−7	−2.50 × 10−6
	(1.94 × 10−5)	(2.00 × 10−5)	(2.00 × 10−5)
University Degree	0.248 ***	0.125 ***	0.126 ***
	(0.00678)	(0.00706)	(0.00709)
Married	0.0389 ***	−0.0815 ***	−0.0775 ***
	(0.00850)	(0.00909)	(0.00912)
Having a Child	−0.0954 ***	−0.0792 ***	−0.0825 ***
	(0.00838)	(0.00849)	(0.00851)
Unemployed	−0.0258 *	0.0181	0.0280 *
	(0.0142)	(0.0150)	(0.0151)
Log of Household Income		0.171 ***	0.134 ***
		(0.00667)	(0.00981)
Log of Household Asset		0.184 ***	0.184 ***
		(0.00346)	(0.00347)
Inter_University Degree & Age			−0.0416 ***
			(0.00895)
Inter_Male & lHincome			−0.0229 *
			(0.0137)
Myopic View of the Future			−0.00180 ***
			(0.000568)
Risk Aversion			0.0552 ***
			(0.0111)
Constant	−0.312 ***	−5.228 ***	−2.935 ***
	(0.0426)	(0.0985)	(0.0730)
Observations	158,169	158,169	158,169

Robust standard errors in parentheses. *** p < 0.01, * p < 0.1.

, are consistent with those obtained from the OLS regression models presented earlier. Specifically, the direction, magnitude, and statistical significance of several key explanatory variables remain largely unchanged. This consistency across models enhances confidence in the reliability and stability of our results and suggests that the relationships are insensitive to the estimation technique used.

4. Discussion

Using a comprehensive theory-driven approach, we investigate the determinants of DFL acquisition among active investors in Japan. Drawing on well-established theoretical frameworks and using a large-scale dataset comprising 158,169 respondents, this analysis offers a pioneering and extensive empirical examination of DFL acquisition. These results strongly align with the theories underpinning this study’s theoretical framework, indicating that individual differences in sociodemographic, economic, and psychological factors significantly influence the acquisition of DFL.

Gender emerged as a strong predictor of DFL, with men exhibiting higher levels of DFL acquisition. This disparity is consistent with both the social cognitive theory (Bandura 2023), which emphasizes learning through observation, modeling, and reinforcement in social contexts, and the digital divide theory (van Dijk 2013), which highlights unequal access to digital skills and resources based on structural and cultural factors. In the Japanese context, traditional gender norms often position men as the primary financial decision-makers, encouraging their greater involvement in activities such as investment management and risk assessment (Kadoya and Khan 2020a). These norms contribute to an unequal distribution of financial responsibilities within households, with men more frequently being tasked with managing their household’s finances (Okamoto and Komamura 2021). Consequently, men gain more hands-on experience of financial concepts and digital financial platforms, which fosters the development of DFL through experiential learning and increased digital engagement (Yoshino et al. 2020). This result is consistent with DFL studies by Choung et al. (2023) and Setiawan et al. (2022), who reported similar gender disparities in South Korea and Indonesia. Conversely, Kass-Hanna et al. (2022) observed narrowing gender gaps in DFL in some South Asian and Sub-Saharan African countries, indicating that Japan may lag in this respect.

Age has a nonlinear effect on DFL, a finding that aligns with life-cycle learning theories (Salthouse 2004) and the digital divide theory (van Dijk 2013). Younger individuals in Japan benefit from early exposure to technology and ICT education, which strengthens their digital competencies (Moreno and Castillo 2022). Middle-aged individuals develop DFL through accumulated experience and apply financial knowledge to informed decision-making (Okamoto and Komamura 2021). However, aging can reduce digital engagement and cognitive adaptability, limiting the ability to adopt new financial technologies and the overall development of financial literacy (Kadoya and Khan 2020b; Salthouse 2004). This nonlinear trend is echoed in OECD (2018), which found that, while middle-aged adults often exhibit strong digital proficiency, older populations remain at a disadvantage, especially in areas like online banking, e-government services, and cybersecurity awareness.

Higher education strengthens DFL by equipping individuals with the cognitive and technical skills to navigate financial systems (Becker 1962; OECD 2021a). In Japan, universities are increasingly integrating fintech education and fostering analytical thinking and responsible financial decision-making (Yasukawa and Yanagihara 2021). Research shows that higher education increases engagement with digital financial tools such as e-payments and online banking, thus enhancing financial competency (Yoshino et al. 2020). Government-led initiatives, including the Financial Literacy Guidelines from Japan’s Financial Service Agency (2024) and Nomura Research Institute (2024), further support education-driven fintech adoption and fraud prevention. By improving financial literacy, education facilitates the effective use of financial technologies, which reinforces its role in DFL development (Yoshino et al. 2020). This finding aligns with Rajdev (2020) and Adnan et al. (2023), who emphasize the central role of educational attainment in enhancing DFL across various countries. However, the diminishing educational effect we observe at older ages is less frequently addressed in prior DFL research and may reflect age-related cognitive and digital exposure gaps.

However, marital status and parenting create barriers to DFL by limiting time, financial resources, and personal autonomy in households. In Japan, financial responsibilities are often divided, with primary earner-led decision-making reducing individual engagement with digital financial tools (Bandura 2023; Ikeda 2019). Traditional gender roles further constrain financial educational opportunities, particularly for women (Kuroda 2019). Parenting responsibilities, which are largely placed on women (Saito 2024), divert attention from financial learning, restricting access to digital platforms and financial education (Tamura 2024). These factors collectively hinder the development of DFL and financial independence. Our results on marital and parental status are supported by Lee (2023), who found that caregiving roles reduce digital engagement in South Korea. However, they contrast with findings by Bäckström et al. (2022), who found that expecting parents or those with children increasingly use digital tools to navigate both financial and emotional transitions, which indicates that engagement may vary across the parental journey (caregiving norms) and contextual differences in digital ecosystem maturity.

Unemployment fosters DFL through necessity-driven engagement with online job-seeking tools. Japan’s high digital penetration (Yoshino et al. 2020) and availability of job search platforms (Kaji 2022) have helped unemployed individuals gain employment opportunities and financial awareness. Exposure to financial education initiatives supports practical applications such as managing unemployment benefits and loans (Furusawa 2014). Additionally, DFL equips job seekers with the ability to assess financial risks, including fraudulent schemes such as “Yami Baito” scams (Hire Planner 2024). By engaging with digital financial services, individuals develop financial awareness and strengthen their ability to make informed decisions regarding job searching and financial management. This finding differs from studies such as that by Williams et al. (2023), who linked unemployment status to lower DFL and financial inclusion, and argue that unemployed individuals often lack the financial resources and motivation to engage with digital financial tools. Our results suggest a unique adaptation in Japan, where digital necessity among the unemployed may drive financial engagement.

Economic resources, such as household income and assets, boost DFL by providing access to essential tools and opportunities (Becker 1962; Venkatesh et al. 2016). Financially stable individuals in Japan invest more in education, which affords them the cognitive skills needed to understand financial concepts and form responsible behaviors (Kadoya and Khan 2020b). It also facilitates access to technology, such as smartphones and internet services, boosting digital competency and engagement with digital financial services (Nomura Research Institute 2018). Typically, wealthier households seek and benefit from professional financial advice (Lal et al. 2023), which enhances their expertise in using DFS tools and risk management strategies. Ultimately, economic resources strengthen DFL by enabling informed financial decision-making and digital engagement. These findings mirror those of Yang et al. (2023) and Widyastuti et al. (2024b), who report that income and asset levels are major enablers of DFL due to improved access to fintech tools and resources.

Among the interaction terms, this study found that the interaction between age and having a university degree is negative, which indicates that the positive effect of a university degree on DFL decreases with age. This can be explained through the lens of age-related cognitive decline (Lenehan et al. 2015; Murman 2015), which demonstrates that cognitive ability reduces as people age. Such age-driven skill losses could potentially weaken education’s influence on DFL among aging investors. Generational disparities in digital exposure add a second layer: younger individuals, often digital natives, are inherently more familiar with online financial systems, whereas older cohorts may lack both access to and adaptability in such environments. Regarding the interaction term between gender (Male) and household income, we observed a positive and significant coefficient, which underscores that the effect of income on improving DFL is more pronounced for males compared to the female cohort. This disparity likely stems from structural inequalities—men often gain greater access to financial tools and platforms as income rises, while women may face barriers like reduced financial autonomy and limited exposure to financial education. Karim et al. (2023) highlight that financial empowerment, such as entrepreneurship, is less responsive for women even when they have equivalent resources. The Global Gender Gap Report (World Economic Forum 2023) further underscores persistent gender-based gaps in income and economic participation, which may moderate the income effect on women’s DFL.

Beyond the interaction terms included in this study, future research should explore additional theoretically grounded interactions to future elucidate the multifaced dimensions of DFL. For instance, a marital status × income interaction could assess whether intra-household financial decision-making dynamics vary by income level, particularly in contexts where financial responsibilities are disproportionately borne by one partner. Similarly, an education × employment status interaction may reveal whether the practical application of formal education through active labor force participation enhances digital financial engagement. A risk aversion × age interaction could also be insightful, potentially capturing how conservative financial attitudes, when combined with age-related technological skepticism, may jointly suppress DFL acquisition. These proposed interactions are theoretically anchored in human capital theory, the theory of planned behavior, and the digital divide literature, and they represent promising avenues for future empirical inquiry. Investigating such interactions can contribute to a nuanced and intersectional understanding of how digital financial capabilities are shaped across diverse social, economic, and behavioral dimensions.

Risk aversion hinders DFL by fostering a preference for certainty and limiting engagement with digital financial tools and protective strategies (Bandura 2023). In Japan, skepticism toward new technologies and reliance on traditional systems has slowed digital adaptation, as seen in the gradual shift away from “hanko” (The Japan Times 2024). This resistance restricts financial awareness and the adoption of e-payment services (Yoshino et al. 2020). Additionally, the reluctance to embrace change stifles proactive financial behavior and hands-on experimentation (Tanaka 2025), preventing individuals from strengthening responsible financial practices and effectively navigating digital financial services. The mixed effects of a myopic view of the future on DFL reflect the tension between short-term pragmatism and long-term planning. A study of Japan’s fintech sector highlights that the country has traditionally favored cash transactions (One Step Beyond 2025), a preference that inadvertently hinders awareness and DFS competencies. Moreover, a focus on short-term outcomes likely hinders the acquisition of financial knowledge and positive financial attitudes, which require foresight and strategic planning, which is consistent with the findings of Kadoya and Khan (2020b) and Lal et al. (2024).

These findings contribute to theory by deepening our understanding of digital financial literacy (DFL) through the lens of six conceptual frameworks. First, consistent with human capital theory (Becker 1962), individuals with higher education, income, and asset ownership exhibit greater DFL, which reinforces the notion that socioeconomic resources enhance individuals’ capacity for skill acquisition. Notably, the diminishing impact of education at older ages suggests that human capital accumulation may be moderated by life-cycle constraints, extending the theory’s applicability. Second, the results support both the technology acceptance model (TAM) (Davis 1989) and the unified theory of acceptance and use of technology (UTAUT) (Venkatesh et al. 2016), highlighting how employment status, fintech exposure, and income facilitate the adoption of digital tools. The findings show that perceived utility and the use of financial services, driven by employment needs or income-based access, enhance DFL, aligning with these models’ assumptions regarding behavioral intent and enabling conditions. Third, this study affirms the relevance of digital divide theory (van Dijk 2013), as gender, age and parenting roles emerge as critical sources of inequality in DFL acquisition. The observed interaction between gender and income further illustrates how structural disadvantages compound digital exclusion, which reinforces the theory’s emphasis on both access and skill gaps. Fourth, social cognitive theory (Bandura 2023) is extended through evidence that social roles, digital self-efficacy, and experiential learning, particularly among unemployed or parenting individuals, significantly shape digital engagement. These findings underscore the importance of observational leaning and perceived competence in navigating digital financial environments. Fifth, our findings validate the theory of planned behavior (Ajzen 1991), as psychological traits such as risk aversion and future orientation significantly predict DFL. Risk-averse individuals are less likely to engage with digital finance, while those with stronger future orientation exhibit more proactive learning and participation. These psychological traits influence attitudes and behavioral intentions, as positioned by the theory. Taken together, the empirical findings confirm, refine, and extend these six theories, demonstrating that DFL acquisition is shaped by a constellation of structural, behavioral, and technological factors. This integrated, multidimensional perspective contributes to a richer understanding of digital financial inclusion in advanced economies.

Broader Implications of DFL

Although this study adopts a micro-level behavioral analysis, it is important to briefly consider the broader macroeconomic relevance of DFL. Previous research has shown that well-functioning financial systems, robust regulatory environments, and inclusive access to financial services can contribute significantly to economic growth and stability, particularly in lower- and middle-income countries (Creane et al. 2004; Serres et al. 2006). While our analysis does not directly examine these macro-level dynamics, our findings suggest that individual DFL—by promoting safer, informed, and inclusive digital participation—may serve as a complementary mechanism that enhances the effectiveness of financial regulation (Ofosu-Mensah Ababio et al. 2024), facilitate foreign investment, and build resilience during financial crises (OECD 2021b; Klapper et al. 2012; Banna and Alam 2021). These insights are especially salient in the post-global crisis context, where financial inclusion and digital competencies have become central to economic recovery and policy design (Guo and Naseer 2025). Future research could usefully explore the extent to which individual-level DFL aggregates to influence broader patterns of financial stability, market efficiency, and economic development.

5. Conclusions

This pioneering study presents one of the most comprehensive empirical investigations of the determinants of DFL acquisition among adult investors in Japan. The findings of this study highlight critical implications for policymakers and academic institutions seeking to promote DFL in Japan, especially in terms of addressing barriers such as marital status, parenting, myopic views of the future, and risk aversion. Policymakers should prioritize initiatives that enhance access to education and digital tools for lower-income households, married individuals, and parents using gamified mobile apps with bite-sized DFL lessons to fit their constrained schedules. AI-driven platforms can support risk-averse individuals with low-risk simulations of digital financial tools, countering hesitancy and fostering informed decision-making. To address myopic views of the future, financial planning applications that promote long-term behavior should be incentivized.

In light of Japan’s aging population, inclusive fintech training tailored to older adults is essential. Community-based programs delivered through local centers and libraries, along with age-friendly digital interfaces, can help alleviate technological apprehension and promote digital financial engagement among seniors. Strategic partnerships with healthcare and pension institutions may further facilitate the integration of DFL into aging-related services. In parallel, regulatory frameworks must evolve in step with digital transformation of financial services. Policymakers should collaborate with fintech firms, cybersecurity experts, and consumer protection agencies to develop balanced, transparent, and adaptive regulations that address algorithmic transparency, data privacy, and fraud prevention. Initiatives such as the implementation of standardized digital safety labels and simplified risk disclosures could empower consumers to make informed decisions when using financial technologies. Finally, DFL education efforts should be demographically stratified. Short-form, app-based DFL modules may better suit time-constrained groups such as working parents and young professionals, whereas long-form community-based instruction may be more appropriate for retirees and digitally underserved populations. Public broadcasting platforms and national financial education campaigns can play a vital role in promoting financial inclusion and demystifying fintech tools.

Additionally, inclusive fintech training offered online and in community centers can bridge the digital divide between older adults and women. Academic institutions should embed DFL in university curricula, leverage virtual reality case studies to equip students to navigate the digital economy, and create supportive learning environments for parents and workers to mitigate family related constraints. Public–private partnerships can further scale-up these technologies to ensure affordability and inclusivity, empower safe financial participation, and foster resilient financial ecosystems.

Despite its contributions, this study has several limitations that should be considered when interpreting the results. First, the data are drawn exclusively from active account holders of Rakuten Securities, a population likely to be more financially engaged, digitally literate, and economically secure than the general public. While this sample offers valuable insights into Japan’s digitally active investor segment, it limits the generalizability of our findings to populations that are financially marginalized or digitally excluded. Given that Rakuten Securities clients typically have regular internet access, investment experience, and familiarity with financial technology, the estimated relationships, particularly those concerning income, education, or employment status, may differ in both magnitude and direction for underserved groups. For instance, individuals with lower income or education who do not use online financial services may exhibit even lower digital financial literacy (DFL) than our estimates suggest. As such, these findings should be interpreted with caution when they are applied to the general population, especially to those facing structural barriers to financial and digital inclusion. Future research should consider using nationally representative or stratified samples to validate and extend these results to broader population groups. Second, the use of cross-sectional data restricts our ability to draw causal inferences or assess how DFL evolves over time. Although our theoretical framework provides strong justification for the observed associations, the findings remain correlational in nature. Future research using longitudinal or panel data could provide deeper insights into how DFL changes across individuals’ life course or in response to external shocks such as economic downturns or digital innovation. Third, this study relies on OLS estimation without formally correcting for potential endogeneity in the model. Due to the absence of a clearly excludable instrument within the dataset and the lack of a designated treatment or focal explanatory variable, instrumental variable estimation was not pursued. As a result, the findings should be interpreted as associative rather than strictly causal.