Knowledge, Actionable Digital Skills, and Old-Age Anxiety: Evidence from Digital Financial Literacy Components Among Japanese Retail Investors

Abstract

Rapid digitalization has reshaped financial decision-making, and anxiety about later life is an important concern among middle-aged and older investors. Yet it remains unclear whether the traditional Big Three financial knowledge component captures the aspects of financial capability most closely associated with lower anxiety in digital financial environments. This study examines the association between old-age anxiety and digital financial literacy (DFL) components among digitally active Japanese retail investors aged 40–64. Using data from a large-scale survey of 94,695 investors, we estimate ordered probit models to examine overall DFL and its eight subdimensions. While overall DFL is negatively associated with anxiety about life after age 65, decomposing the index reveals substantial heterogeneity across components. The traditional Big Three financial knowledge component does not show a robust independent negative association with old-age anxiety once actionable and protective digital competencies are accounted for. In contrast, practical know-how, positive financial attitude, and self-protection are more consistently associated with lower anxiety. Supplementary heterogeneity analyses suggest that the positive conditional association between financial knowledge and anxiety is most visible among men aged 50–59, although these subgroup patterns should be interpreted cautiously. These findings do not imply that financial knowledge is unimportant. Rather, they suggest that Big Three financial knowledge alone may be an insufficient proxy for the dimensions of financial capability associated with lower self-reported old-age anxiety in digital financial environments. Given the cross-sectional design, the findings are interpreted as conditional associations rather than causal effects.

1. Introduction

A substantial body of research suggests that financial literacy is associated with improved retirement planning and reduced anxiety about later life. Early evidence from Japan and the United States indicates that individuals with higher levels of financial literacy—typically measured by the widely used “Big Three” questions on interest rates, inflation, and risk diversification—tend to report lower levels of anxiety about life after retirement (Kadoya & Khan, 2017; Kadoya et al., 2018). Similar findings have been documented in broader contexts, linking financial knowledge to retirement preparedness and financial well-being (Lusardi & Mitchell, 2007, 2014; Lusardi & Messy, 2023).

While previous research has documented a negative association between overall digital financial literacy (DFL) and anxiety about life after retirement, most empirical studies analyze financial literacy using either a composite index or relatively broad component groupings. Such approaches may mask important heterogeneity across the underlying dimensions of financial capability.

Building on earlier work that decomposed DFL into broader component groupings, the present study examines eight subdimensions of DFL. This decomposition is not intended to treat traditional financial knowledge as separate from DFL. Rather, financial knowledge is examined as one component within the broader DFL construct. The purpose is to assess whether the traditional Big Three financial knowledge component shows the same association with old-age anxiety as more actionable and protective DFL components, such as practical know-how and self-protection. Much of the financial literacy literature implicitly assumes that greater financial knowledge is associated with better financial preparedness and, by extension, lower anxiety about later life. However, whether this assumption holds in digital financial environments is not self-evident. Knowledge may increase awareness of financial risks, but lower anxiety may depend more directly on whether individuals possess actionable skills and protective capabilities to manage those risks. The rapid digitalization of financial services has increased both access to financial tools and exposure to complex information and risks. In this environment, the traditional Big Three financial knowledge component may not capture the full range of capabilities relevant to old-age anxiety, particularly actionable and protective competencies needed to navigate digital financial environments.

Recent research on digital financial literacy (DFL) suggests that financial capability is multidimensional, encompassing not only knowledge but also practical skills, decision-making abilities, and self-protection competencies (Lyons & Kass-Hanna, 2021; Lal et al., 2025). A previous study documented a negative association between overall DFL and old-age anxiety among Japanese investors (Amarsanaa et al., 2025). Yet, that aggregate relationship masked substantial heterogeneity across components.

In particular, preliminary evidence suggested that knowledge- and awareness-based dimensions were not uniformly associated with lower anxiety, whereas skill-based and protective components exhibited more consistent negative associations with anxiety (Amarsanaa et al., 2025). This raises an important unresolved question regarding the role of traditional financial literacy in increasingly digital financial environments:

Does the traditional Big Three financial knowledge component within DFL show the same negative association with old-age anxiety as actionable and protective digital financial competencies?

This study addresses this question using a large-scale survey of Japanese investors aged 40–64. Building on earlier work that examined broader component groupings of digital financial literacy, the present study refines that framework by examining eight distinct components.

From a behavioral perspective, greater knowledge may increase awareness of financial risks—such as longevity risk, inflation risk, and market volatility—without necessarily strengthening the individual’s perceived ability to manage these risks. In such cases, cognitive awareness may be associated with heightened anticipatory concern rather than lower anxiety. This possibility is consistent with research suggesting that information exposure can, under certain conditions, heighten concern in complex decision environments (Fernandes et al., 2014; Chhillar et al., 2025; Choung et al., 2023).

To address this question, we use a large-scale survey of 94,695 Japanese investors aged 40–64 to examine whether the traditional Big Three financial knowledge component within DFL shows the same association with old-age anxiety as more actionable and protective DFL components. Employing ordered probit models, we examine the association between old-age anxiety and the components of digital financial literacy, paying particular attention to financial knowledge as measured by the Big Three questions.

Rather than assuming that financial knowledge uniformly reduces anxiety, we assess whether the Big Three financial knowledge component exhibits the same negative association with old-age anxiety as digital knowledge, awareness, practical know-how, behavioral application, and self-protection skills when these components are jointly considered.

Importantly, given the cross-sectional nature of the data, the findings are interpreted as associations rather than causal effects.

This study contributes to the literature in three ways. First, it re-examines the role of the traditional Big Three financial knowledge component within a broader DFL framework. Rather than treating Big Three financial literacy as a sufficient proxy for financial capability, the study evaluates whether this knowledge-based component shows the same association with old-age anxiety as actionable and protective digital competencies. Second, it introduces an awareness–actionability perspective to interpret heterogeneity across DFL components. Knowledge- and awareness-based components may increase recognition of financial risks, whereas actionable and protective competencies may be more closely related to perceived ability to navigate digital financial environments. Third, using a large sample of digitally active Japanese retail investors aged 40–64, the study clarifies the population for which these associations are observed and avoids extrapolating the results to older adults or aging societies more generally.

2. Literature Review

2.1. Financial Literacy and Financial Well-Being

Financial literacy is widely recognized as a core determinant of individuals’ capacity to navigate increasingly complex financial environments and to attain higher levels of financial well-being. Lusardi and Mitchell (2007) provide seminal evidence that financial literacy is strongly associated with retirement planning and preparedness, showing that individuals who understand basic financial concepts are substantially more likely to calculate retirement needs and to hold retirement wealth, with marked differences across education, income, and demographic subgroups. Building on this, Lusardi and Messy (2023) emphasize that financial literacy encompasses not only knowledge of concepts and risks but also the skills and attitudes required to apply this knowledge to improve financial well-being and support participation in economic life. Recent evidence from the FILSA study also emphasizes that adult financial literacy is multidomain and heterogeneous across population subgroups, reinforcing the need to define the population of inference carefully when examining financial capability (Schuhen et al., 2022). Empirical studies generally confirm that higher financial literacy is associated with more prudent financial decisions, reduced susceptibility to behavioral biases, and greater comfort in using financial instruments, all of which translate into improved financial well-being. In country-specific evidence, Anghel and Pochea (2025) show that financial literacy is a key determinant of financial well-being in Romania, where widespread financial illiteracy and low inclusion contribute to unequal well-being and a potential poverty trap. Related work also links financial capability in later life to broader quality-of-life outcomes, including physical and mental health, often with gender-differentiated effects.

2.2. Digital Financial Literacy (DFL): Concepts, Determinants, and Measurement

Rapid digital transformation has broadened the scope of financial literacy to include digital competencies alongside traditional financial knowledge (Lyons & Kass-Hanna, 2021; Koskelainen et al., 2023). A bibliometric review of digital financial literacy further shows that the field has expanded rapidly across digital finance, financial inclusion, and technology-mediated financial behavior, while remaining conceptually diverse (Yadav & Banerji, 2023). While this literature emphasizes the multidimensional nature of digital financial literacy, it provides limited insight into whether traditional financial knowledge retains a component-level association with old-age anxiety once digital competencies are explicitly considered.

Empirical studies suggest that digital financial literacy is associated with improved financial well-being and confidence in managing digital financial tools (Choung et al., 2023). However, these studies typically focus on aggregate digital competencies rather than isolating the component-level association of traditional financial knowledge.

Prior research has operationalized digital financial literacy as a multidimensional construct combining financial and digital competencies (Lyons & Kass-Hanna, 2021). However, the psychological implications of individual components—particularly traditional financial knowledge—remain insufficiently examined.

2.3. DFL, Financial Behavior, and Psychological Outcomes

Recent behavioral studies highlight DFL’s influence on savings, stress, and subjective well-being. Chhillar et al. (2025) reported that DFL enhances financial well-being by reducing financial stress through improved financial behaviors among individuals employed in India. A growing body of research has linked digital financial competencies to subjective well-being and confidence in financial management (Choung et al., 2023). Yet, these analyses focus on aggregate measures and do not directly examine whether traditional financial knowledge retains an independent association with retirement-related anxiety once digital skill-based components are jointly considered. While these studies use the same 8-subdimension DFL construct as the current study and find positive effects on financial decision-making and well-being, their focus excludes older adults and psychological outcomes like old age anxiety.

Collectively, these studies signal that financial education interventions should integrate psychological and technological components to promote holistic well-being. While these studies consistently demonstrate that DFL is associated with financial well-being, financial stress, and confidence, they stop short of examining how digital financial competencies influence anticipatory psychological states, such as anxiety about later life. This gap motivates our empirical examination of whether the traditional Big Three financial knowledge component shows a different association with old-age anxiety from actionable and protective DFL components.

2.4. Gender Dimensions in Financial and Digital Knowledge

Gender remains a consistent differentiator in literacy outcomes. Classic research by Hsu (2016) showed how spousal learning incentives affect older women’s financial literacy development, while Cupák et al. (2018) provided cross-country evidence that much of the gender gap arises from differences in numeracy and financial experience. Previous studies also report gender differences in financial literacy across countries (Bucher-Koenen et al., 2017). More recent studies, such as Iwatsubo et al. (2025), reaffirm persistent gender gaps in numeracy, financial concept understanding, and retirement planning literacy. In the digital realm, Mishra et al. (2024) observed that women’s DFL significantly improves decision-making confidence in India. Sundarasen et al. (2023), through a bibliometric review, identified women’s financial literacy as an expanding frontier linking empowerment, inclusion, and digital access. Bajtelsmit (2024) further examined how gender differences influence retirement investment patterns, suggesting that tailored education and DFL programs can mitigate gender biases in risk preferences.

2.5. Awareness, Actionability, and Old-Age Anxiety

The relationship between DFL and old-age anxiety may depend on whether a given component primarily captures awareness of financial risks or actionable capability to manage them. Knowledge- and awareness-based components, including traditional financial knowledge, digital knowledge, and awareness of financial services or financial behaviors, may increase recognition of retirement-related uncertainty. Such recognition does not necessarily translate into lower anxiety if individuals do not simultaneously feel able to act on that knowledge.

By contrast, actionable and protective components, such as practical know-how and self-protection, may be more closely related to individuals’ perceived ability to navigate digital financial environments. These components may therefore show a more stable negative association with old-age anxiety. This framework does not imply a causal mechanism, because the present study uses cross-sectional data and does not directly measure perceived control or preparedness. Rather, it provides a conceptual lens for interpreting why different DFL components may exhibit different associations with old-age anxiety.

3. Data and Methods

3.1. Data

This study draws on data from the 2025 wave of the “Survey on Life and Money,” an online panel survey jointly administered by Rakuten Securities and the Kadoya Lab at Hiroshima University. The survey was conducted in January and February 2025 and targeted active account holders of security companies aged 18 and older who had logged in to their accounts at least once during the previous year. The survey collected detailed information on investment behavior, sociodemographic characteristics, economic conditions, and psychological preferences. To capture anxiety regarding life after retirement, the analytical sample was restricted to individuals aged 40–64, an age range in which concerns about post-retirement life are most salient (Weissman & Myers, 1980; Jorm, 2000; Kadoya et al., 2018). Key variables, including old age anxiety, educational attainment, financial literacy, and myopic view of the future, were extracted from earlier waves. We then excluded observations with missing data on main sociodemographic and economic variables, resulting in a final sample size of 94,695 observations. Because the survey targets active securities account holders who had logged in during the previous year, the population of inference is digitally active retail investors rather than the general population. Respondents in this sample are likely to be more financially engaged, more familiar with digital financial services, and wealthier than average adults. Accordingly, the results should be interpreted as evidence for digitally active investors approaching retirement age, not as estimates for older adults or aging societies in general.

3.2. Variables

3.2.1. Dependent Variable

The dependent variable in this study is old age anxiety. The following statement was used to define five levels of a respondent’s future anxiety: “I have anxieties about my life after 65 years old”. Responses were recorded on a five-point Likert scale ranging from 1 (“does not apply at all”) to 5 (“strongly applies to me”). Responses were recorded as an ordinal variable taking values from 1 (lowest anxiety) to 5 (highest anxiety). In addition, a binary indicator of old-age anxiety was constructed for descriptive statistics and robustness analyses. The dummy variable takes the value of 1 if the respondent selected “strongly agree” or “somewhat agree” and 0 otherwise.

3.2.2. Independent Variables

Our main independent variables are the DFL index and its eight subdimensions. These variables are used to examine whether overall DFL is associated with old-age anxiety and whether this association differs across conceptually distinct components of DFL. The eight subdimensions should be understood as an operational refinement of the broader multidimensional DFL framework rather than as a competing conceptual framework. Following Lal et al. (2025), the eight subdimensions allow us to distinguish traditional financial knowledge from basic digital knowledge, awareness-related dimensions, actionable digital financial skills, attitudinal and behavioral applications, and self-protection. This finer classification is necessary for the purpose of the present study because the central question is whether the traditional Big Three financial knowledge component shows the same association with old-age anxiety as more actionable and protective components of DFL. We follow the measurement approach of Lal et al. (2025), which operationalizes the broader DFL framework proposed by Lyons and Kass-Hanna (2021). In this approach, DFL is measured as a multidimensional capability covering knowledge, awareness, practical use, financial attitudes and behaviors, and self-protection. The index comprises eight subdimensions: (i) financial knowledge, (ii) digital knowledge, (iii) awareness of digital financial services (DFS), (iv) awareness of financial attitudes and behaviors (FAB), (v) practical know-how, (vi) positive financial attitude, (vii) positive financial behavior, and (viii) self-protection.

Financial knowledge, corresponding to traditional financial literacy, was measured by the “Big Three” questions on interest rates, inflation, and risk diversification. In the present analysis, financial knowledge is not treated as independent of DFL. It is treated as the traditional knowledge-based component within the DFL construct. The empirical question is therefore not whether DFL and financial knowledge have separate effects, but whether the financial knowledge component exhibits a different conditional association with old-age anxiety from other DFL components. The score was defined as the average number of correct responses, following the standard Big Three financial literacy approach (Lusardi & Mitchell, 2014). The remaining seven subdimensions were assessed by three Likert-scale questions each (1 = strongly disagree to 5 = strongly agree), with subdimension scores as the average of their items. The DFL subdimensions follow Lal et al. (2025), and the detailed survey items are available from the authors upon reasonable request. For the regression analyses, all subdimension scores were standardized using z-score normalization to facilitate comparability across coefficients. The DFL index used in the regressions was constructed by summing the standardized scores of the eight subdimensions. For descriptive statistics, however, the raw scores are reported before normalization.

Control variables include demographic characteristics (gender, age, marital status, children, education, employment status), socioeconomic variables (household income and assets), and psychological traits (risk aversion and myopic views of the future).

Table 1. Variable definitions.

Variable	Definition
Dependent Variable
Old age anxiety	Ordinal variable: 1 means “not at all” and 5 means “strongly agree” in response to the statement: “I feel anxious about my life after the age of 65”
Old age anxiety dummy	Binary variable: 1 = having anxiety about life after age 65 (selected strongly agree or somewhat true), 0 = otherwise (selected neither, not very applicable, or not at all)
Independent Variable
DFL index	Sum of the eight standardized DFL subdimension scores
Financial knowledge	Standardized score for Big Three financial knowledge
Digital knowledge	Standardized score for basic digital tool knowledge
Awareness of DFS	Standardized score for awareness of digital financial services
Awareness of FAB	Standardized score for awareness of positive financial attitudes and behaviors
Practical know-how	Standardized score for operational use of digital financial services
Positive financial attitude	Standardized score for positive financial attitudes in digital finance
Positive financial behavior	Standardized score for positive financial behaviors in digital finance
Self-protection	Standardized score for protection against digital financial scams and fraud
Other independent variable
Gender	Binary variable: 1 = male, 0 = female
Age	Continuous variable: respondents’ age
Age squared	Continuous variable: age squared
Age group	Ordinal variable: categorized by 3 groups of 40–49, 50–59, and 60–64
Marital status	Binary variable: 1 = married, 0 = otherwise
Child	Binary variable: 1 = have child(ren), 0 = otherwise
University degree	Binary variable: 1 = hold at least a bachelor’s degree, 0 = otherwise
Unemployed	Binary variable: 1 = do not have a job, 0 = otherwise
Household income	Continuous variable: the total annual income including tax for the household in 2024 (unit: JPY)
Log of Household income	Continuous variable: Natural log of the respondents’ household income
Household asset	Continuous variable: the total household financial assets
Log of Household asset	Continuous variable: Natural log of the respondents’ household financial assets
Risk aversion	Continuous variable: proxy measure of respondents’ risk attitudes based on the probability threshold at which individuals decide to carry an umbrella when going out, a commonly used survey-based approach to capture individual risk preferences.
Myopic view of the future	Binary variable: 1 = answering “somewhat agree” or “completely agree” with the idea that “the future is uncertain, so there is no point in thinking about it.” 0 = otherwise

provides definitions of all the variables.

3.3. Descriptive Statistics

Table 2. Summary statistics by gender.

Variable	Category	Mean (or %)	Std. Dev.	t-Test	Min	Max
Dependent variable
Old age anxiety dummy	Male	54.7%	0.498	t = 27.52 ***	0	1
	Female	64.3%	0.479
Old age anxiety	Male	3.446	1.227	t = 30.86 ***	1	5
	Female	3.710	1.160
Independent variable
DFL index a	Male	30.362	4.518	t = −10.69 ***	7	36
	Female	30.031	3.999
Financial knowledge b	Male	0.840	0.267	t = −52.65 ***	0	1
	Female	0.734	0.321
Digital knowledge b	Male	4.517	0.854	t = 0.67	1	5
	Female	4.521	0.801
Awareness of DFS b	Male	4.618	0.753	t = 1.55	1	5
	Female	4.626	0.675
Awareness of FAB b	Male	4.476	0.713	t = 16.09 ***	1	5
	Female	4.553	0.583
Practical know-how b	Male	4.147	0.900	t = −21.86 ***	1	5
	Female	4.006	0.937
Positive financial attitude b	Male	4.162	0.722	t = 8.91 ***	1	5
	Female	4.206	0.620
Positive financial behavior b	Male	3.874	0.857	t = −2.24 **	1	5
	Female	3.861	0.806
Self-protection b	Male	3.728	0.826	t = −34.85 ***	1	5
	Female	3.524	0.815
Other independent variable
Gender	All	69.9%	0.459	-	0	1
Age	Male	51.193	6.820	t = −28.64 ***	40	64
	Female	49.828	6.497
Age squared	Male	2667.189	705.944	t = −28.95 ***	1600	4096
	Female	2525.036	662.056
Marital status	Male	73.1%	0.444	t = −22.37 ***	0	1
	Female	65.9%	0.474
Child	Male	66.0%	0.474	t = 1.67 *	0	1
	Female	66.5%	0.472
University degree	Male	64.9%	0.477	t = −62.57 ***	0	1
	Female	43.5%	0.496
Unemployed	Male	5.2%	0.222	t = −13.34 ***	0	1
	Female	3.2%	0.176
Household income	Male	8,549,156	4,527,238	t = −28.70 ***	1,000,000	20,000,000
	Female	7,639,375	4,346,961
Log of Household income	Male	15.797	1	t = −29.03 ***	13.816	16.811
	Female	15.668	1
Household asset	Male	25,718,580	28,176,397	t = −23.84 ***	2,500,000	100,000,000
	Female	21,158,119	24,018,932
Log of Household asset	Male	16.456	1	t = −19.30 ***	14.732	18.421
	Female	16.303	1
Risk aversion	Male	0.545	0.248	t = −14.08 ***	0	1
	Female	0.521	0.220
Myopic view of the future	Male	13.3%	0.340	t = 13.02 ***	0	1
	Female	16.6%	0.372
Observations	Male	66,207
	Female	28,488

Note: ^a The DFL index reported in this table is the sum of the raw scores for the eight subdimensions before normalization. ^b Subdimension scores reported in this table are raw scores before normalization. Standardized variables are used in the regression analyses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

presents descriptive statistics of the model variables by gender. Although gender is not the primary focus of the study, the descriptive statistics are reported separately because gender differences in financial literacy, digital capability, and old-age anxiety are well documented in the literature and may be relevant for interpreting supplementary heterogeneity analyses. For the DFL subdimensions, the reported means represent the average score per item within each subdimension rather than the total score across items. Approximately 54.7% of men and 64.3% of women in the sample report anxiety about life after age 65; this difference is statistically significant (t = 27.5, p < 0.01). The mean level of old age anxiety is 3.4 (SD = 1.2) for men and 3.7 (SD = 1.2) for women on a five-point scale (with 5 indicating the highest level of anxiety). Because the sample consists of active investors, the observed levels of retirement-related anxiety may differ from those in the general population.

The mean DFL index score for men (30.4, SD = 4.5) is higher than that for women (30.0, SD = 4.0), and the difference is also statistically significant (t = −10.69, p < 0.01). Turning to the subdimensions, men exhibit significantly higher mean scores in financial knowledge (0.8, SD = 0.3), practical know-how (4.1, SD = 0.9), positive financial behavior (3.9, SD = 0.9), and self-protection (3.7, SD = 0.8) than women. In contrast, women have significantly higher mean scores in awareness of FAB (4.6, SD = 0.6) and positive financial attitude (4.2, SD = 0.7) than men. No statistically significant gender differences are observed for digital knowledge and awareness of DFS.

Regarding the control variables, the average age is 51.2 years among men and 49.8 years among women, respectively. Approximately 73.1% of men and 65.9% of women are married, while the proportion of respondents with children is 66.0% among men and 66.5% among women, respectively. In terms of education, 64.9% of men and 43.5% of women are university educated. 5.2% of men and 3.2% of women were unemployed. Mean annual household income and assets are higher among men, at 8,549,156 yen for income and 25,718,580 yen for assets. Regarding psychological factors, the average level of risk aversion is 0.54 for men and 0.52 for women, respectively. Approximately 13.3% of men report a myopic view of life, compared to 16.6% of women.

3.4. Methods

Given that the dependent variable is an ordinal discrete variable, we employed an ordered probit regression analysis to examine the association between old age anxiety and DFL, with a particular focus on its subdimensions. The baseline specification is given by using the following equation:

$$Y_i=f\left({DFL}_i, X_i{,\epsilon }_i\right)$$

(1)

where $Y_i$ is the $i$th respondent’s measure of the dependent variable, old age anxiety. ${DFL}_i$ is the DFL index or a vector of its subdimensions, and $X_i$ is a vector of individual demographic, socioeconomic, and psychological characteristics. ${\epsilon }_i$ is the error term. The main analysis is conducted using the pooled sample, controlling for gender, age, age squared, marital status, children, education, employment status, household income, household assets, risk aversion, and myopic view of the future. Gender- and age-specific analyses are treated as supplementary heterogeneity analyses rather than as the primary basis for the paper’s contribution. To avoid relying solely on comparisons of coefficients across separately estimated subgroup models, we additionally estimate pooled models with interaction terms for selected key DFL components.

The full specifications are presented in Equations (2) and (3):

$${Old age anxiety}_i= f({DFLIndex}_i,X_i,{\epsilon }_i)$$

(2)

$${Old age anxiety}_i= f({DFLComponents}_i,X_i,{\epsilon }_i)$$

(3)

where $DFLInde{x}_i$ denotes the standardized DFL index, and $DFLComponent{s}_i$ denotes a vector of the eight standardized DFL subdimensions: financial knowledge, digital knowledge, awareness of DFS, awareness of FAB, practical know-how, positive financial attitude, positive financial behavior, and self-protection. $X_i$ includes gender, age, age squared, marital status, children, university degree, employment status, household income, household assets, risk aversion, and myopic view of the future.

With respect to Equation (3), we estimated models in which each subdimension was included separately, as well as a model that includes all subdimensions simultaneously. This approach allows us to evaluate how the associations change with and without controlling for the other subdimensions. In particular, we examine whether the traditional Big Three financial knowledge component shows a conditional negative association with old-age anxiety once other digital-related, actionable, and protective components of DFL are taken into account.

To assess overlap among the DFL subdimensions, we calculated pairwise correlations and variance inflation factors (VIFs). The DFL subdimensions are conceptually related and empirically correlated, as expected for components of a broader DFL construct. The highest pairwise correlation among the DFL subdimensions was 0.744, observed between digital knowledge and awareness of DFS. The highest VIF among the DFL components was 3.05 for awareness of DFS, below conventional thresholds for severe multicollinearity. However, because several DFL components are conceptually and empirically related, coefficients from models including all components simultaneously should be interpreted as conditional associations within a correlated block of competencies. Therefore, sign changes across specifications are interpreted cautiously and are not treated as definitive evidence of substantive reversals.

4. Results

4.1. Main Results

Table 3. Pooled ordered probit estimation results.

Variables	Dependent Variable: Old Age Anxiety
	Model 1	Model 2-1	Model 2-2	Model 2-3	Model 2-4	Model 2-5	Model 2-6	Model 2-7	Model 2-8	Model 2-9
DFL index	−0.00531 ***
	(0.000646)
Financial knowledge		−0.000796								0.00647 *
		(0.00367)								(0.00377)
Digital knowledge			0.0184 ***							0.0361 ***
			(0.00352)							(0.00544)
Awareness of DFS				0.0115 ***						0.0233 ***
				(0.00352)						(0.00624)
Awareness of FAB					0.0282 ***					0.0941 ***
					(0.00365)					(0.00569)
Practical know-how						−0.0511 ***				−0.0359 ***
						(0.00367)				(0.00549)
Positive financial attitude							−0.0488 ***			−0.0778 ***
							(0.00382)			(0.00617)
Positive financial behavior								−0.0278 ***		0.0486 ***
								(0.00375)		(0.00516)
Self-protection									−0.108 ***	−0.136 ***
									(0.00384)	(0.00535)
Observations	94,695

Note: Robust standard errors are in parentheses. All models control for gender, age, age squared, marital status, children, university degree, employment status, household income, household assets, risk aversion, and myopic view of the future. *** p < 0.01, ** p < 0.05, * p < 0.10. The number of observations is 94,695 in all models. Shaded cells indicate the key coefficients discussed in the text. The shading is used only as a visual guide and does not represent any additional statistical classification.

reports the main pooled ordered probit estimates for the DFL index and its subdimensions. The purpose of this analysis is to examine whether overall DFL and its components are associated with old-age anxiety among digitally active Japanese retail investors aged 40–64. Model 1 includes the standardized DFL index. Models 2-1 through 2-8 include each standardized DFL subdimension separately. Model 2-9 includes all eight standardized subdimensions simultaneously.

The pooled estimates show that the overall DFL index is negatively associated with old-age anxiety. However, decomposing the index into its subdimensions reveals substantial heterogeneity across components. In the fully adjusted component model, the traditional Big Three financial knowledge component does not show a negative association with old-age anxiety. Instead, its conditional coefficient is weakly positive. Digital knowledge, awareness of DFS, awareness of FAB, and positive financial behavior also show positive conditional coefficients in the full component model.

By contrast, practical know-how, positive financial attitude, and self-protection are each negatively associated with old-age anxiety in the fully adjusted component model. These patterns suggest that the aggregate negative association between DFL and old-age anxiety is more closely reflected in actionable and protective components than in the traditional Big Three financial knowledge component.

Because the DFL subdimensions are correlated, the positive coefficients for knowledge- and awareness-related components should not be interpreted as evidence that these competencies increase anxiety. Rather, they indicate that, conditional on the other DFL components, these dimensions do not show the same negative association as actionable and protective components.

4.2. Supplementary Heterogeneity Analyses

Gender- and age-specific estimates are reported as supplementary heterogeneity analyses. These analyses suggest that the positive conditional association between financial knowledge and old-age anxiety is most visible among men aged 50–59, whereas the negative associations for practical know-how, positive financial attitude, and self-protection are more stable across subsamples. However, these subgroup patterns should be interpreted cautiously because differences across separately estimated models do not by themselves establish statistically significant subgroup differences.

To formally assess subgroup differences, we therefore estimated pooled interaction models. The gender interaction model shows that the interaction between financial knowledge and male is positive and statistically significant, suggesting that the association between financial knowledge and old-age anxiety differs by gender. In contrast, the interactions between male and practical know-how and between male and self-protection are not statistically significant. The age-group interaction model provides more limited evidence of age heterogeneity. The interaction between financial knowledge and the 50–59 age group is weakly positive, whereas the corresponding interaction for the 60–64 age group is not statistically significant. Interactions for self-protection are not statistically significant across age groups. Overall, the interaction results support a cautious interpretation of subgroup patterns, with stronger evidence for gender heterogeneity in the financial knowledge association than for broad age-group heterogeneity (

Table 4. Interaction test summary.

Variables	Dependent Variable: Old Age Anxiety
	Gender Interactions	Age-Group Interactions
Financial knowledge × Male	0.0293 ***	-
	(0.00751)
	p = 0.000
Financial knowledge × Age group (50–59)	-	0.0138 *
		(0.00774)
		p = 0.074
Financial knowledge × Age group (60–64)	-	−0.0166
		(0.0111)
		p = 0.135
Practical know-how × Male	−0.0104	-
	(0.00969)
	p = 0.284
Practical know-how × Age group (50–59)	-	0.0208 **
		(0.0100)
		p = 0.038
Practical know-how × Age group (60–64)	-	0.0144
		(0.0133)
		p = 0.280
Self-protection × Male	−0.00606	-
	(0.0101)
	p = 0.551
Self-protection × Age group (50–59)	-	−0.0161
		(0.0102)
		p = 0.115
Self-protection × Age group (60–64)	-	−0.0197
		(0.0144)
		p = 0.170
Observations	94,695

Note: Robust standard errors are in parentheses. The 40–49 age group is the reference category in the age-group interaction model. All models include the same controls as Table 3, except that age and age squared are replaced by age-group indicators in the age-group interaction model. *** p < 0.01, ** p < 0.05, * p < 0.10. Shaded cells indicate the key interaction terms discussed in the text. The shading is used only as a visual guide and does not represent any additional statistical classification.

).

4.3. Robustness Checks

We conducted several robustness checks to assess whether the main patterns depend on the ordered probit specification or on the ordinal coding of the dependent variable. First, we estimated ordered logit models using the same covariates. Second, we estimated OLS models treating the five-point old-age anxiety variable as a continuous outcome. Third, we estimated binary probit and logit models using the old-age anxiety dummy. The results are summarized in

Table 5. Robustness check summary.

Variables	Dependent Variable: Old Age Anxiety				Dependent Variable: Old Age Anxiety Dummy
	Ordered Logit Model		OLS		Binary Probit Model		Binary Logit Model
DFL index	−0.00763 ***		−0.00672 ***		−0.000197		−0.000698
	(0.00110)		(0.000674)		(0.000781)		(0.00130)
Financial knowledge		0.0130 **		0.00763 **		0.0281 ***		0.0455 ***
		(0.00636)		(0.00381)		(0.00469)		(0.00779)
Practical know-how		−0.0571 ***		−0.0386 ***		−0.0446 ***		−0.0732 ***
		(0.00937)		(0.00556)		(0.00658)		(0.0109)
Positive financial attitude		−0.140 ***		−0.0797 ***		−0.0728 ***		−0.122 ***
		(0.0106)		(0.00625)		(0.00714)		(0.0119)
Self-protection		−0.238 ***		−0.138 ***		−0.140 ***		−0.230 ***
		(0.00921)		(0.00537)		(0.00617)		(0.0102)
Observations	94,695

Note: Robust standard errors are in parentheses. The DFL index row is estimated using the DFL index specification, whereas the component rows are estimated using the full eight-component specification. To keep the table concise, only the key coefficients relevant to the main interpretation are reported. All models include the same controls as Table 3. *** p < 0.01, ** p < 0.05, * p < 0.10. Shaded cells indicate the key coefficients discussed in the text. The shading is used only as a visual guide and does not represent any additional statistical classification.

, which reports selected key coefficients from the full eight-component robustness specifications. These robustness checks should be interpreted as sensitivity analyses rather than as direct diagnostic tests of all assumptions underlying ordered-response models. We therefore use them to assess whether the main component-level patterns are sensitive to alternative outcome specifications, rather than treating estimator consistency as a substitute for the assumptions required by ordinal models.

The negative association between the overall DFL index and old-age anxiety is robust in the ordered logit and OLS specifications. However, the DFL index is not statistically significant in the binary probit and logit models using the dichotomized anxiety measure. This suggests that the aggregate DFL result is more clearly observed when the ordinal information in the dependent variable is retained.

The component-level patterns are more stable across specifications. Practical know-how, positive financial attitude, and self-protection consistently show negative associations with old-age anxiety across the ordered logit, OLS, binary probit, and binary logit models. In contrast, the traditional Big Three financial knowledge component does not show a negative association in these robustness checks. These results reinforce the main interpretation that actionable and protective components show more stable negative associations with old-age anxiety than the traditional financial knowledge component.

5. Discussion

5.1. Main Findings and Interpretation

This study examined whether the traditional Big Three financial knowledge component within DFL shows the same association with old-age anxiety as more actionable and protective digital financial competencies. The main finding is that overall DFL is negatively associated with old-age anxiety, but this aggregate association masks substantial heterogeneity across components. In particular, the Big Three financial knowledge component does not show a robust independent negative association with old-age anxiety once other DFL components are jointly considered. By contrast, practical know-how, positive financial attitude, and self-protection are more consistently associated with lower anxiety.

A key question raised by these findings is why traditional financial knowledge does not show the same negative association with old-age anxiety as actionable and protective DFL components. One of the central findings of this study is that traditional financial knowledge alone does not show the same robust negative association with old-age anxiety as actionable and protective DFL components. Supplementary heterogeneity analyses suggest that this positive conditional association is most visible among men aged 50–59, although the subgroup results should be interpreted cautiously.

In digital financial environments characterized by high information exposure and complexity, such heightened awareness may be associated with greater anticipatory concern rather than lower anxiety. A deeper understanding of financial risks may be associated with greater recognition of retirement-related uncertainty, particularly when not accompanied by practical competencies or confidence in navigating digital financial tools. This pattern is consistent with the observed finding that knowledge- and awareness-based components do not show the same negative association with self-reported old-age anxiety as more practical and self-protection-related DFL components.

One possible interpretation is an awareness–actionability gap. Knowledge- and awareness-based components may be associated with greater recognition of retirement-related risks, whereas actionable and protective components may be more closely related to the ability to manage those risks in digital financial environments. However, because the present study does not directly measure perceived control, preparedness, retirement confidence, or the channels through which financial and digital skills were acquired, this interpretation should be regarded as a conceptual explanation for the observed associations rather than as evidence of a tested psychological mechanism. The change in the coefficient of financial knowledge across model specifications may reflect statistical interdependence, shared variance, or suppression among correlated DFL components. In simpler specifications, financial knowledge may partly capture variance shared with broader digital competencies that are negatively associated with anxiety. Once these components are jointly included, the conditional coefficient for financial knowledge becomes positive or statistically insignificant. We therefore do not interpret this sign change as evidence of a substantive reversal, but as an indication that the Big Three component does not show the same stable negative association as actionable and protective components in the full component model. This pattern should be interpreted descriptively, given the cross-sectional design and the correlations among subdimensions.

5.2. Actionable and Protective Components

In contrast to traditional financial knowledge as measured by the Big Three questions, skill-based digital competencies are consistently associated with lower levels of old age anxiety across model specifications and demographic subsamples. Practical know-how, positive financial attitude, and self-protection exhibit stable negative associations, even when financial knowledge and other components are jointly included.

These dimensions reflect the capacity to act within digital financial environments—navigating platforms, evaluating service providers, and protecting against fraud—rather than merely understanding financial concepts. While conceptual knowledge may increase awareness of financial risks, actionable competencies may be more closely associated with perceived preparedness in complex financial settings (Mishra et al., 2024; Choung et al., 2023). One possible interpretation is that actionable digital competencies are more closely aligned with the kinds of capabilities that may help investors navigate digital financial environments, although the present data do not directly measure confidence, preparedness, or perceived control.

These associations, however, are not uniform across demographic groups. Supplementary heterogeneity analyses suggest some variation across demographic groups. In particular, the positive conditional association between financial knowledge and old-age anxiety appears most visible among men aged 50–59. However, these subgroup patterns should be interpreted cautiously, as the interaction tests provide stronger evidence for gender heterogeneity than for broad age-group heterogeneity.

While gender differences in financial literacy and digital capability are well documented (Mahdavi & Horton, 2014; Okamoto & Komamura, 2021; Lyu et al., 2025), the present results indicate that these differences may also be reflected in associations with old-age anxiety. However, gender is not treated here as a primary explanatory variable, but rather as an analytical lens through which variation in the associations between DFL components and anxiety can be observed.

5.3. Practical Significance and Relation to Previous Research

To illustrate the practical magnitude of the estimated associations, we computed marginal effects from the fully adjusted pooled ordered probit model including all DFL subdimensions simultaneously. Focusing on the probability of reporting the highest level of old-age anxiety, the estimates indicate that a one-standard-deviation increase in self-protection is associated with an approximately 3.9 percentage-point lower probability of reporting the highest anxiety category. A one-standard-deviation increase in positive financial attitude is associated with an approximately 2.2 percentage-point lower probability, and a one-standard-deviation increase in practical know-how is associated with an approximately 1.0 percentage-point lower probability. In contrast, a one-standard-deviation increase in financial knowledge is associated with an approximately 0.2 percentage-point higher probability of reporting the highest anxiety category, although this association is small and only weakly statistically significant.

These probability shifts are modest in magnitude, but they are consistent with the coefficient patterns reported in the main models. They also clarify that the negative associations are most pronounced for actionable and protective components, especially self-protection and positive financial attitude, rather than for the traditional Big Three financial knowledge component.

This study extends prior research showing that overall digital financial literacy is negatively associated with old age anxiety (Amarsanaa et al., 2025). Unlike the previous analysis, which examined aggregate DFL, the present study examines the component-level association of traditional financial knowledge within a multidimensional framework. By decomposing DFL into distinct subdimensions, the analysis shows that the negative aggregate association is more closely reflected in skill-based and protective components than in traditional financial knowledge itself.

In doing so, the study refines—rather than contradicts—the interpretation of earlier findings and contributes to a more differentiated understanding of how financial competencies relate to old-age anxiety among digitally active retail investors.

5.4. Implications, Limitations, and Future Research

The findings have implications for financial education, financial institutions, and policymakers. For financial education, the results suggest that programs focused only on conceptual knowledge may be insufficient in digital financial environments. Educational initiatives may need to combine traditional financial concepts with practical training in how to use digital financial services, evaluate service providers, and protect oneself from online fraud. For financial institutions, the findings suggest that customer support tools should not simply provide more information, but should help users translate information into concrete actions. For policymakers, the results highlight the importance of DFL frameworks that include practical and protective competencies, especially for digitally active middle-aged and older investors approaching retirement. Rather than replacing traditional financial education, these results indicate that knowledge-based approaches may need to be integrated with applied digital competencies to support both financial capability and anxiety-relevant preparedness in digital financial environments.

Several limitations should be acknowledged. First, the analysis relies on cross-sectional survey data, and the observed associations should therefore not be interpreted as causal effects. Second, the sample consists of digitally active retail investors who had logged in to their securities accounts during the previous year. These respondents are likely to be more financially engaged, more digitally exposed, and wealthier than the general population. Therefore, the findings should not be generalized to the general population, digitally excluded individuals, or older adults with limited financial market participation. Third, old-age anxiety is measured using a single survey item. Although this item captures a directly relevant concern about life after age 65, it does not separately measure perceived control, preparedness, retirement confidence, or subjective financial security. Fourth, the Big Three questions provide a widely used but narrow measure of traditional financial knowledge, and the summed standardized DFL index should be interpreted as an operational measure of selected DFL components rather than as a full psychometric assessment of financial capability. Fifth, the DFL subdimensions are correlated. Coefficients in the fully adjusted component model should therefore be interpreted as conditional associations within a correlated block of competencies, not as isolated effects of mutually independent traits. Sixth, the study does not observe objective retirement-income arrangements, such as pension wealth, annuity holdings, expected retirement benefits, or guaranteed income streams. Prior work suggests that guaranteed income can materially affect retirement portfolio choices and perceived retirement security (Waggle & Agrrawal, 2024). Therefore, the association between financial knowledge and old-age anxiety may partly reflect unobserved differences in awareness of retirement-income risks or differences in retirement-income structure. Finally, the study does not observe how financial and digital skills were acquired, such as through formal education, self-learning, hands-on experience, workplace exposure, or repeated use of digital financial services. Because practice-based learning may have different implications from purely informational exposure, the present data cannot determine whether the negative associations for practical know-how and self-protection reflect the content of these competencies, the way they were acquired, or both. Future research using longitudinal, experimental, or richer psychometric designs may clarify whether strengthening actionable and protective digital financial competencies can reduce old-age anxiety over time.

6. Conclusions

This study examined whether the traditional Big Three financial knowledge component within DFL shows the same association with old-age anxiety as actionable and protective digital financial competencies. Using a large-scale survey of digitally active Japanese retail investors aged 40–64, the analysis shows that overall DFL is negatively associated with old-age anxiety. However, component-level analyses indicate that this aggregate association is not reflected uniformly across all DFL dimensions.

The Big Three financial knowledge component does not exhibit the same robust negative association with old-age anxiety as actionable and protective components such as practical know-how, positive financial attitude, and self-protection. These findings do not imply that financial knowledge is unimportant. Rather, they suggest that Big Three financial knowledge alone may be an insufficient proxy for the dimensions of financial capability most closely associated with lower self-reported old-age anxiety in digital financial environments.

The results refine earlier evidence linking financial literacy to lower retirement-related anxiety by showing that the association may depend on which aspect of financial capability is being measured. Given the cross-sectional design, the single-item anxiety measure, and the active-investor sample, the findings should be interpreted as conditional associations rather than causal effects. Future research using longitudinal data, experimental interventions, richer measures of perceived control and retirement preparedness, and information on learning channels may further clarify how digital financial capability is associated with old-age anxiety over time.