The Association of Financial Knowledge, Attitude, and Behavior with Investment Loss Tolerance: Evidence from Japan

Abstract

Investment loss tolerance refers to an investor’s willingness to hold financial instruments after experiencing value declines and is considered essential to long-term investment success. Financial literacy, comprising financial knowledge, attitude, and behavior, has been widely identified as a key factor in promoting rational financial decisions. A recent study by Homma et al. suggests that the three components can help prevent panic selling during market crises, such as the COVID-19 pandemic. However, that study relies on binary behavioral indicators within crisis-specific contexts, limiting the generalizability of their findings. To address these gaps, the present study quantitatively measures investment loss tolerance using a generalized hypothetical loss scenario and investigates the associations of financial literacy components. Using a large-scale dataset of 161,223 active investors from one of Japan’s largest online securities firms, we conducted ordered probit and probit regression analyses while controlling for demographic, socioeconomic, and psychological factors. The results reveal that financial knowledge, attitude, and behavior all have statistically significant positive effects on investment loss tolerance. These findings indicate that financial literacy enhances investors’ capacity to withstand losses and discourages premature asset liquidation, even outside crisis-specific contexts. The evidence supports policies aimed at improving financial literacy to foster more resilient investor behavior and promote long-term financial well-being.

1. Introduction

Investment loss tolerance refers to the extent to which investors are willing to continue holding financial instruments after they have incurred losses. Investors with low investment loss tolerance tend to react sensitively to short-term losses and often sell off their holdings prematurely (Benartzi and Thaler 1995; Odean 1998; Gneezy and Potters 1997), thereby missing potential gains during subsequent market recoveries. In contrast, investors with high investment loss tolerance are more likely to retain their instruments despite experiencing some losses, thereby positioning themselves to benefit from substantial returns when the market rebounds (Huang et al. 2021). This capacity to withstand short-term volatility and avoid frequent trading is widely considered a core principle of long-term asset management (Bihary et al. 2020). While stock prices can be highly volatile over short periods such as months or a year, they generally exhibit a tendency to grow steadily over longer periods, such as several years or decades (Siegel 2022). This pattern is even more pronounced in the case of mutual funds in general. Therefore, having a high investment loss tolerance is considered a rational approach to long-term investing.

Theoretical perspectives in behavioral finance help explain why many investors still struggle to sustain this tolerance. According to Benartzi and Thaler’s (1995) framework of myopic loss aversion, individuals are prone to overweight short-term losses relative to long-term gains, leading to premature liquidation. Similarly, the disposition effect (Odean 1998) highlights investors’ tendency to sell losing assets too quickly while holding onto winners for too long. Both biases underscore the importance of mechanisms such as enhanced financial literacy that can mitigate these cognitive and emotional pitfalls and foster more resilient investment behavior.

Many previous studies have shown that financial literacy promotes rational decision making (Homma et al. 2025; Bawalle et al. 2025; Bucher-Koenen and Ziegelmeyer 2014; Ahmad and Shah 2022; Nieddu and Pandolfi 2021; Takahashi et al. 2022; Lusardi and Mitchell 2014; Yamori and Ueyama 2022; Khan et al. 2021; Kristanto and Gusaptono 2020). Extending this evidence beyond investment behavior, Ilan and Mugerman (2025) found that financially literate households make superior mortgage choices, underscoring the broad and cross-domain relevance of financial literacy in shaping sound financial outcomes. Bucher-Koenen and Ziegelmeyer (2014) found that individuals with high financial literacy were less likely to engage in panic selling during financial crises and more likely to retain devalued assets. Ahmad and Shah (2022) suggested that financial literacy improves the quality of investment decisions and performance. Similarly, Nieddu and Pandolfi (2021) showed that enhancing financial literacy helps individuals recognize and mitigate heuristic biases, thereby enabling more informed investment decisions. Bayar et al. (2020) also found that higher financial literacy is associated with higher risk tolerance among investors.

This study examines financial literacy by categorizing it into three components: financial knowledge, financial attitude, and financial behavior. According to the OECD, financial literacy comprises the awareness, knowledge, skills, attitudes, and behaviors necessary to make sound financial decisions and ultimately achieve financial well-being. The OECD framework highlights three core elements: financial knowledge, financial behavior, and financial attitudes (Atkinson and Messy 2012). Previous research shows that these three components, when considered together, have significantly greater predictive power for financial outcomes than a single financial literacy measure focusing solely on financial knowledge (Singh and Mittal 2023; Vieira et al. 2019; Bhushan and Medury 2014). Moreover, studies indicate that financial attitude and financial behavior play important mediating roles in the relationship between financial knowledge and investment decisions (Fessler et al. 2020; Nano 2015). Evidence also suggests that having financial knowledge or earning a higher income alone does not guarantee sound financial behavior (Sam et al. 2022), highlighting the need to incorporate psychological and behavioral dimensions. This comprehensive approach is consistent with behavioral finance theory, which recognizes that financial decisions are shaped by cognitive, emotional, and behavioral factors. Research further demonstrates that financial attitudes can exert a stronger influence than financial knowledge in shaping financial management practices (Kartini et al. 2020). Homma et al. (2025) found that all three components significantly reduce the likelihood of panic selling, reinforcing their relevance to rational decision making. Therefore, this study adopts these components as valid and distinct indicators of rational investment behavior. In this context, financial knowledge refers to the ability to understand basic financial concepts and facts; financial attitude reflects psychological tendencies and values toward money, such as a preference for long-term saving; and financial behavior captures the extent to which individuals effectively apply their financial knowledge in real-world situations, such as investment management and financial planning (Homma et al. 2025).

Financial knowledge, behavior, and attitude together constitute a multi-dimensional framework that more effectively explains investment loss tolerance than traditional, knowledge-based measures alone. While conventional financial literacy emphasizes the cognitive understanding of financial concepts, research indicates that loss tolerance is shaped by three interconnected dimensions: financial knowledge (both objective and subjective), behavioral patterns influenced by psychological biases, and underlying attitudes toward risk and financial decision making (Ahmed et al. 2021; Sobaih and Elshaer 2023). Financial knowledge provides the foundation for risk assessment, but behavioral factors, such as overconfidence, loss aversion, and anchoring, can account for up to 91% of the variance in risk-taking behavior, highlighting that central role of psychological processes (Mangala and Sharma 2014; Çatak and Arslan 2023). Moreover, financial attitudes act as crucial mediators between knowledge and behavior, with subjective financial knowledge (perceived competence) often influencing risk tolerance differently from objective knowledge, particularly in moderating demographic effects on investment choices (Ahmed et al. 2021). This multi-dimensional perspective is essential because investors’ decision making under uncertainty involves complex cognitive and emotional processes that extend beyond technical knowledge, including mental accounting, future orientation, and emotional responses to potential losses (Reddy and Mahapatra 2017). Accordingly, understanding investment loss tolerance requires analyzing not only what investors know but also how they process risk, their behavioral tendencies under stress, and their attitudes toward uncertainty, factors that collectively determine whether theoretical financial knowledge translates into effective real-world investment decisions.

While the literature on investor behavior is extensive, most prior studies have focused on risk tolerance, the willingness to accept uncertain returns, rather than investment loss tolerance, which reflects the capacity to endure realized or unrealized loss without premature liquidation (Grable and Rabbani 2023; Ahmed et al. 2021; Akdeniz and Turan 2021). Moreover, these studies have typically emphasized financial knowledge as the primary explanatory factor, often neglecting the roles of financial attitudes and behaviors (Grable and Rabbani 2023; Ahmed et al. 2021). Akdeniz and Turan (2021), Corter (2011), and Jain and Kesari (2020) demonstrate that cognitive and behavioral biases, such as overconfidence, loss aversion, and mental accounting, significantly explain variation in risk-tolerance. These findings indirectly highlight the potential of financial knowledge, behavior, and attitude to mitigate such bias (Shivangi and Chaudhury 2025; Adil et al. 2022). The study most closely related to ours, Homma et al.’s (2025), examined how financial knowledge, attitude, and behavior influenced panic selling during market downturns. However, their analysis relied on a binary crisis-specific indicator, whether investors sold assets during the COVID-19 pandemic, which neither captures the degree of loss that investors are willing to tolerate nor generalizes to non-crisis conditions. Our study addresses these limitations by introducing a continuous measure of investment loss tolerance and examining its relationship with financial literacy, financial knowledge, financial attitude, and financial behavior in a generalized, non-crisis market context. In doing so, it extends the literature beyond risk tolerance, provides a more nuanced understanding of investor resilience, and contributes empirical evidence on how cognitive, psychological, and behavioral dimensions of financial literacy jointly shape responses to portfolio losses.

This study examines financial literacy by categorizing it into three components, financial knowledge, financial attitude, and financial behavior, and investigates how each independently and collectively relates to investors’ tolerance for losses in a generalized, non-crisis investment context. To guide the analysis, we pose the following research questions: (1) Does financial knowledge increase investment loss tolerance? (2) Do financial attitudes contribute to higher investment loss tolerance? (3) Does financial behavior strengthen investment loss tolerance? Drawing on a large dataset of retail investors from one of Japan’s largest online securities companies, we hypothesize that each of these components is positively associated with investment loss tolerance. Building on prior studies showing that financial literacy mitigates irrational investment behavior, we argue that improvements in financial knowledge, along with supportive attitudes and behaviors, enhance investors’ capacity to remain committed to their holdings during market downturns, thereby reducing premature liquidation.

This study contributes to the behavioral finance literature in three key ways. First, we quantified investors’ tolerance for investment losses by constructing a generalized loss tolerance index based on realistic market scenarios. Second, this study provides, for the first time, empirical evidence that financial knowledge, attitudes, and behavior positively relate to investment loss tolerance. Third, we offer practical and policy-related implications aimed at enhancing the investment decision making of individual investors.

2. Data and Methods

2.1. Data

This study utilized data from the 2025 wave of the “Survey on Life and Money,” an online survey collaboratively administered by Rakuten Securities and the Kadoya Lab of Hiroshima University. Data collection occurred in January and February 2025, focusing on account holders of Rakuten Securities aged 18 and older who had logged into their account at least once in the previous year. Both the dependent and most of the independent variables used in this study are drawn from the 2025 wave. However, because some part of the dataset follows panel respondents across multiple years, several key variables, specifically financial knowledge, financial behavior, and financial attitude, were partially measured in the 2022 and 2023 waves, as these constructs are relatively stable over time.

2.2. Variables

2.2.1. Dependent Variable

We define investment loss tolerance as the extent of financial loss an individual is willing to endure while maintaining possession of an investment asset. This study specifically examines the investment loss tolerance in mutual funds. Investment loss tolerance was derived through a hypothetical scenario approach, a method commonly employed in both academic research and financial advisory practices to explore individual financial behaviors and investment decision-making processes (Grable and Lytton 2003; FinaMetrica 1998). The survey question utilized in this study is as follows.

Q1. Suppose you invest JPY 1 million in an investment trust and experience a loss. Up to what level of loss would you be willing to continue holding the investment? (Just one)

• JPY 990,000 (JPY 10,000 loss or 1% loss);
• JPY 900,000 (JPY 100,000 loss or 10% loss);
• JPY 800,000 (JPY 200,000 loss or 20% loss);
• JPY 700,000 (JPY 300,000 loss or 30% loss);
• JPY 600,000 or less (JPY 400,000 loss or more, or 40% loss or more).

Based on the responses, a discrete dependent variable was established for the primary estimate, categorized into the following ranges: 1%, 10%, 20%, 30%, and 40% or more. To enhance the robustness of the results, a binary variable was also utilized as an alternative dependent variable. A value of 1 was assigned to respondents capable of enduring a loss of 30% or greater, and a value of 0 was assigned otherwise. Given that a 20% decline in stock prices from a recent peak is commonly interpreted as an indicative of a bear market (Chen 2024), the 30% threshold, representing a loss exceeding this conventional benchmark, was used to identify respondents as possessing higher tolerance for investment losses.

2.2.2. Independent Variable

Our main independent variables are the three indicators of financial literacy: financial knowledge, financial attitude, and financial behavior. To evaluate these dimensions, we used survey questions adopted from the methodology by Kadoya and Khan (2020). The coding process for each variable is elaborated in detail as follows.

First, financial knowledge was measured using three questions on interest rate calculations, aiming to assess the understanding of compound interest, inflation, and risk assessment (see Appendix A). Each correct response was awarded one point, while incorrect answers received zero points. The financial knowledge score was calculated as the average of these three item scores.

Second, financial attitude was measured using two statements related to expenditure and the preference for immediate gratification (see Appendix A). Responses were recorded on a five-point Likert scale, ranging from 1 (“Completely applicable”) to 5 (“Completely not applicable”). Respondents who selected option 4 (Disagree) or 5 (Strongly Disagree) were assigned one point, while all other response were scored as zero points. The final financial attitude score was calculated as the average of the two statements, with higher values indicating a more favorable financial attitude.

Third, financial behavior was assessed using four questions related to purchasing decisions, bill payments, financial planning, and money management. Each item was rated on a five-point Likert scale ranging from 1 (“Completely applicable”) to 5 (“Not at all applicable”). Respondents who selected 1 (Completely applicable) or 2 (Somewhat applicable) were assigned a score of one point, while all other responses were scored as zero (see Appendix A). The final financial behavior score was computed as the average of the individual scores from these four questions.

To isolate the independent effect of financial knowledge, attitudes, and behaviors on investment loss tolerance, a set of socio-demographic control variables was included in the estimation model. These variables included gender, age, marital status, number of children, years of education, employment status, household income, household assets, risk aversion, and myopic view of the future. For some variables within the panel dataset, specifically financial knowledge, financial attitude, financial behavior, years of education, and myopic view of the future, data were obtained from the 2022 or 2023 data waves.

The dependent variable (investment loss tolerance) was measured in the 2025 wave, whereas the main independent variables (financial knowledge, financial behavior, and financial attitude) were partly measured from the 2022 and 2023 waves. Since this is partly a panel survey, the panel respondents were followed across waves. These literacy measures are considered relatively stable constructions that do not fluctuate substantially over short periods of time, which supports their use even when measured in an earlier wave.

Table 1. Definitions of variables.

Variable	Definition
Dependent Variables
Investment loss tolerance	Discrete variable: How much loss respondents can withstand if they invest JPY 1 million in an investment trust (1% loss/10% loss/20% loss/30% loss/40% or more loss)
Investment loss tolerance dummy	Binary variable: 1 = respondents can withstand a loss of 30% or more if they invest JPY 1 million in an investment trust, 0 = otherwise
Independent Variables
Financial knowledge	Discrete variable: average score of three financial knowledge questions
Financial attitude	Discrete variable: average score of two financial attitude questions
Financial behavior	Discrete variable: average score of four financial behavior questions
Gender	Binary variable: 1 = male, 0 = female
Age	Continuous variable: respondents’ age
Age squared	Continuous variable: age squared
Marital status	Binary variable: 1 = having a spouse, 0 = otherwise
Number of children	Continuous variable: the number of children
Education year	Continuous variable: years of education
Having a job	Binary variable: 1 = having a job, 0 = otherwise
Household income	Continuous variable: the total annual income including tax for the household in 2024 (unit: JPY)
Household asset	Continuous variable: the total household financial assets
Risk aversion	Continuous variable: respondents’ risk aversion (the answer to the following question: when you usually go out with an umbrella, what is the probability of rain?)
Myopic view of the future	Discrete variable: 1 = completely opposite, 2 = somewhat opposite, 3 = cannot say, 4 = somewhat agree, 5 = completely agree with the idea that “the future is uncertain, so there is no point in thinking about it.”

presents the detailed definitions of the dependent, independent, and control variables.

2.3. Descriptive Statistics

The descriptive statistics presented in

Table 2. Descriptive statistics.

Variable	Mean	Std. Dev.	Min	Max
Dependent Variable
Investment loss tolerance	0.245	0.128	0.010	0.400
Investment loss tolerance dummy	0.471	0.499	0	1
Independent Variable
Financial knowledge	0.791	0.301	0	1
Financial attitude	0.623	0.381	0	1
Financial behavior	0.761	0.292	0	1
Gender	0.670	0.470	0	1
Age	46.38	12.28	18	90
Age squared	2302	1181	324	8100
Marital status	0.670	0.470	0	1
Number of children	1.134	1.115	0	12
Education year	15.08	2.088	9	21
Having a job	0.896	0.305	0	1
Household income	7,690,000	4,287,000	1,000,000	20,000,000
Log of household income	15.69	0.622	13.82	16.81
Household assets	21,590,000	25,540,000	2,500,000	100,000,000
Log of household assets	16.27	1.111	14.73	18.42
Risk aversion	0.535	0.238	0	1
Myopic view of the future	2.428	0.966	1	5
Observations	161,223

indicate that on average, respondents were able to tolerate an investment loss of 24.5% when investing JPY 1 million in an investment trust. Furthermore, 47.1% of the respondents reported the ability to endure losses of 30% or greater. The mean scores and standard deviations (SDs) for financial knowledge, attitude, and behavior among the respondents were recorded as 0.791 (0.301), 0.623 (0.381), and 0.761 (0.292), respectively. Regarding demographic characteristics, 67.0% of the respondents were male, with a mean age of 46. Additionally, 67.0% were married, and the respondents reported having more than one child on average. Concerning socioeconomic status, the respondents had an average educational attainment of 15 years, and 89.6% engaged in some form of employment. The average household income and assets were JPY 7.7 million and JPY 21.6 million, respectively. Regarding psychological characteristics, the average level of risk aversion was measured at 0.54, while the average degree of myopic view was 2.4.

2.4. Methods

This study conducted regression analyses to examine the relationships among financial knowledge, financial attitude, financial behavior, and investment loss tolerance. Given that the primary dependent variable was ordinal in nature, we employed an ordered probit model for the main analysis. To assess the robustness of the findings, we performed an additional probit regression utilizing a binary dependent variable.

These analytical approaches are particularly appropriate for the present investigation, as they are widely used in the field of financial behavior research. For instance, Sung and Hanna (1996) applied ordered probit analysis to examine the demographic and socioeconomic determinants of household risk tolerance. Similarly, Fang et al. (2021) implemented both ordered and binary probit models to investigate the influence of wealth accumulation on household risk tolerance in the Chinese context. These prior studies provide empirical support for the use of ordered and binary probit models in analyzing financial decision making and behavior.

The estimation models are delineated by Equations (1) and (2), corresponding to the ordered and binary probit analyses.

$$Y_{1i}=f\left(F{K}_i, F{A}_i, F{B}_{i,}{X}_i, {\epsilon }_i\right)$$

(1)

$$Y_{2i}=f\left(F{K}_i, F{A}_i, F{B}_{i,}{X}_i, {\epsilon }_i\right)$$

(2)

where $Y_{1i}$ and $Y_{2i}$denote the levels of investment loss tolerance exhibited by the $i$th respondent. The variables FK, FA, and FB represent financial knowledge, financial attitude, and financial behavior, respectively. X represents a vector encompassing individual demographic, socioeconomic, and psychological variables. $\epsilon$ is the error term. The comprehensive specifications for Equations (1) and (2) are as follows:

$$\begin{array}{l}{Investmentlosstolerance}_i\\ ={\beta }_0+{\beta }_1{Financialknowledge}_i\\ +{\beta }_2{Financialattitude}_i\\ +{\beta }_3{Financialbehavior}_i+{\beta }_4{Gender}_i\\ +{\beta }_5{Age}_i+{\beta }_6{Agesquared}_i\\ +{\beta }_{7}Marialstatu{s}_i+{\beta }_{8}Numberofchildre{n}_i\\ +{\beta }_{9}Educationyea{r}_i+{\beta }_{10}Havingajo{b}_i\\ +{\beta }_{11}Logofhouseholdincom{e}_i\\ +{\beta }_{12}Logofhouseholdasset{s}_i\\ +{\beta }_{13}Riskaversio{n}_i\\ +{\beta }_{14}Myopicviewofthefutur{e}_i+{\epsilon }_i\end{array}$$

(3)

$$\begin{array}{l}Investmentlosstolerancedumm{y}_i={\beta }_0+{\beta }_1\\ Financialknowledg{e}_i\\ +{\beta }_{2}Financialattitud{e}_i\\ +{\beta }_{3}Financialbehavio{r}_i+{\beta }_{4}Gende{r}_i\\ +{\beta }_{5}Ag{e}_i+{\beta }_{6}Agesquare{d}_i\\ +{\beta }_{7}Maritalstatu{s}_i+{\beta }_{8}Numberofchildre{n}_i\\ +{\beta }_{9}Educationyea{r}_i+{\beta }_{10}Havingajo{b}_i\\ +{\beta }_{11}Logofhouseholdincom{e}_i\\ +{\beta }_{12}Logofhouseholdasset{s}_i\\ +{\beta }_{13}Rislaversio{n}_i\\ +{\beta }_{14}Myopicviewofthefutur{e}_i+{\epsilon }_i\end{array}$$

(4)

We assessed multicollinearity by examining both correlation coefficients and variance inflation factors (VIFs), as substantial interdependence among independent variables can lead to biased or unstable estimation results. The correlation coefficients among the independent variables were all below 0.3, and the corresponding VIF values were consistently below 2. These results indicate that multicollinearity is not a concern in any of the estimated models.

3. Estimation Results

3.1. Regression Results

Table 3. Results of the ordered probit regression analysis.

Variable	Dependent Variable: Investment Loss Tolerance
	Model 1	Model 2	Model 3	Model 4
Financial knowledge	0.6539 ***			0.6221 ***
	(0.0098)			(0.0099)
Financial attitude		0.1746 ***		0.0881 ***
		(0.0075)		(0.0079)
Financial behavior			0.2690 ***	0.1450 ***
			(0.0097)	(0.0103)
Gender	0.3123 ***	0.3819 ***	0.3753 ***	0.3220 ***
	(0.0061)	(0.0060)	(0.0060)	(0.0061)
Age	0.0233 ***	0.0273 ***	0.0293 ***	0.0241 ***
	(0.0016)	(0.0016)	(0.0016)	(0.0016)
Age squared	−0.0004 ***	−0.0004 ***	−0.0004 ***	−0.0004 ***
	(0.0000)	(0.0000)	(0.0000)	(0.0000)
Marital status	−0.0936 ***	−0.1081 ***	−0.1076 ***	−0.0973 ***
	(0.0073)	(0.0073)	(0.0073)	(0.0073)
Number of children	−0.0135 ***	−0.0205 ***	−0.0188 ***	−0.0154 ***
	(0.0030)	(0.0030)	(0.0030)	(0.0030)
Education year	−0.0201 ***	−0.0085 ***	−0.0095 ***	−0.0203 ***
	(0.0014)	(0.0014)	(0.0014)	(0.0014)
Having a job	−0.0147	−0.0255 **	−0.0238 **	−0.0168
	(0.0102)	(0.0102)	(0.0102)	(0.0103)
Log of household income	−0.0004	0.0186 ***	0.0198 ***	0.0034
	(0.0057)	(0.0057)	(0.0057)	(0.0057)
Log of household assets	0.2230 ***	0.2403 ***	0.2357 ***	0.2119 ***
	(0.0030)	(0.0030)	(0.0030)	(0.0031)
Risk aversion	−0.1226 ***	−0.1273 ***	−0.1310 ***	−0.1246 ***
	(0.0119)	(0.0119)	(0.0119)	(0.0119)
Myopic view of the future	−0.0230 ***	−0.0256 ***	−0.0275 ***	−0.0076 **
	(0.0029)	(0.0030)	(0.0029)	(0.0030)
/cut1	2.2394 ***	2.7284 ***	2.8005 ***	2.3104 ***
	(0.0826)	(0.0823)	(0.0825)	(0.0827)
/cut2	3.4245 ***	3.8890 ***	3.9618 ***	3.4971 ***
	(0.0827)	(0.0824)	(0.0826)	(0.0828)
/cut3	4.1002 ***	4.5548 ***	4.6282 ***	4.1739 ***
	(0.0828)	(0.0826)	(0.0827)	(0.0829)
/cut4	4.5428 ***	4.9919 ***	5.0658 ***	4.6175 ***
	(0.0829)	(0.0826)	(0.0828)	(0.0830)
Observations	161,223	161,223	161,223	161,223
Pseudo R-squared	0.0421	0.0332	0.0337	0.0431
Log likelihood	−229,796	−231,919	−231,799	−229,557

Robust standard errors in parentheses. *** p <0.01 and ** p < 0.05.

presents the regression results examining the relationship between investment loss tolerance and the three dimensions of financial literacy: financial knowledge, financial attitude, and financial behavior. Models 1 to 3 report the individual effects of financial knowledge, attitude, and behavior, respectively, while Model 4 incorporates these three variables simultaneously.

In the ordered probit models presented, the cut points (/cut1 to/cut4) represent the thresholds separating categories of investment loss tolerance. All are statistically significant, indicating well-differentiated response levels. The pseudo R-squared values range from 0.0332 to 0.0431, suggesting that the models explain a moderate but typical proportion of the variance for cross-sectional survey data. Among the models, Model 4 demonstrates the best fit, with the highest pseudo R-squared (0.0431) and the least negative log likelihood (−229,557), indicating improved explanatory power when financial knowledge, attitude, and behavior are jointly included.

The ordered probit regression results in Table 3 indicate that financial knowledge, financial attitude, and financial behavior are each significantly and positively associated with loss tolerance. Specifically Model 1 shows that financial knowledge is positively associated with loss tolerance, suggesting that investors with a higher level of financial knowledge are more capable of enduring losses in investment. Model 2 reveals a positive relationship between financial attitude and investment loss tolerance, indicating that investors with a healthier financial attitude are better equipped to cope with potential investment losses. Similarly, Model 3 finds that financial behavior is positively linked to investment loss tolerance, implying that investors with better financial behavior are more resilient to investment loss.

In Model 4, where three financial literacy components are included simultaneously, their effects remain statistically significant. The marginal effects, reported in

Table 4. Estimated marginal effects on loss tolerance.

	Investment Loss Tolerance	Investment Loss Tolerance Dummy
	Marginal Effects	Marginal Effects
Financial knowledge	0.0246	0.2195
Financial attitude	0.0035	0.0365
Financial behavior	0.0057	0.0648
Gender	0.0127	0.1082
Age	0.0010	0.0114
Age squared	0.0000	−0.0002
Marital status	−0.0038	−0.0376
Number of children	−0.0006	−0.0069
Education year	−0.0008	−0.0093
Having a job	−0.0007	−0.0015
Household income	0.0001	−0.0081
Log of household income	0.0084	0.0834
Household assets	−0.0049	−0.0445
Log of household assets	−0.0003	−0.0038
Risk aversion	0.0246	0.2195
Myopic view of the future	0.0035	0.0365

, indicate that a one-unit increase in financial knowledge raises the probability of tolerating a 30% investment loss by 2.46 percentage points, from an average predicted probability of 15.74% to 18.20%. Analogously, a one-unit increase in financial attitude and financial behavior raises this probability by 0.35 and 0.57 percentage points, that is, from 15.74% to 16.09% and 16.31%, respectively. These findings suggest that each dimension independently contributes to variations in loss tolerance, even when controlling for covariates, with financial knowledge exerting the most economically meaningful effect, and behavior and attitude showing smaller but directionally consistent effects.

With respect to the control variables, being a male, age, and the log of household assets are consistently positively associated with investment loss tolerance in all models. In Models 2 and 3, the log of household income has a significantly positive impact on investment loss tolerance. Conversely, age squared, marital status (i.e., having a spouse), number of children, education year, risk aversion, and myopic view are negatively associated with investment loss tolerance. Additionally, employment status exhibits a significantly negative impact on investment loss tolerance in Models 2 and 3. However, the marginal effects of the control variables indicate that gender and risk aversion have the greatest economic significance.

3.2. Robustness Check

To verify the robustness of the primary findings, we conducted a probit regression analysis using a binary proxy dependent variable for investment loss tolerance (i.e., investment loss tolerance dummy), as shown in

Table 5. Results of the probit regression analysis.

Variable	Dependent Variable: Investment Loss Tolerance Dummy
	Model 1	Model 2	Model 3	Model 4
Financial knowledge	0.6316 ***			0.5940 ***
	(0.0118)			(0.0119)
Financial attitude		0.1893 ***		0.0989 ***
		(0.0088)		(0.0094)
Financial behavior			0.2986 ***	0.1755 ***
			(0.0115)	(0.0123)
Gender	0.2821 ***	0.3498 ***	0.3426 ***	0.2928 ***
	(0.0072)	(0.0071)	(0.0071)	(0.0072)
Age	0.0297 ***	0.0336 ***	0.0358 ***	0.0308 ***
	(0.0019)	(0.0019)	(0.0019)	(0.0019)
Age squared	−0.0005 ***	−0.0005 ***	−0.0005***	−0.0005 ***
	(0.0000)	(0.0000)	(0.0000)	(0.0000)
Marital status	−0.0973 ***	−0.1112 ***	−0.1110 ***	−0.1018 ***
	(0.0086)	(0.0086)	(0.0086)	(0.0086)
Number of children	−0.0165 ***	−0.0234 ***	−0.0217 ***	−0.0186 ***
	(0.0036)	(0.0035)	(0.0035)	(0.0036)
Education year	−0.0247 ***	−0.0140 ***	−0.0151 ***	−0.0250 ***
	(0.0017)	(0.0016)	(0.0016)	(0.0017)
Having a job	−0.0013	−0.0117	−0.0100	−0.0041
	(0.0124)	(0.0123)	(0.0123)	(0.0124)
Log of household income	−0.0264 ***	−0.0084	−0.0068	−0.0218 ***
	(0.0067)	(0.0067)	(0.0067)	(0.0067)
Log of household assets	0.2385 ***	0.2542 ***	0.2489 ***	0.2257 ***
	(0.0036)	(0.0036)	(0.0036)	(0.0036)
Risk aversion	−0.1175 ***	−0.1227 ***	−0.1269 ***	−0.1205 ***
	(0.0137)	(0.0136)	(0.0137)	(0.0137)
Myopic view of the future	−0.0279 ***	−0.0283 ***	−0.0299 ***	−0.0102 ***
	(0.0034)	(0.0035)	(0.0034)	(0.0035)
Constant	−3.9457 ***	−4.3704 ***	−4.4555 ***	−4.0374 ***
	(0.0967)	(0.0960)	(0.0962)	(0.0970)
Observations	161,223	161,223	161,223	161,223
Pseudo R-squared	0.0645	0.0530	0.0540	0.0665
Log likelihood	−104,282	−105,567	−105,456	−104,056

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

. This approach enables us to assess whether the results remain consistent under different distributional assumptions. The results indicate that financial knowledge, financial attitude, and financial behavior remain positively associated with investment loss tolerance at a 1% significance level, consistent with the primary findings. When using the binary proxy for loss tolerance, the marginal effects (Table 4) show that a one-unit increase in financial knowledge raises the probability of tolerating a loss of 30% or more by 21.95 percentage points, from 47.09% to 69.04%. The corresponding impacts of financial attitude (+3.65 percentage points, to 50.74%) and financial behavior (+6.48 percentage points, to 53.57%) remain positive but comparatively smaller. Taken together, these patterns reinforce the primary finding that financial knowledge plays a substantially stronger role than attitude or behavior in shaping loss tolerance.

Regarding the control variables, being a male, age, and the log of household assets are positively associated with the investment loss tolerance dummy in all models. In contrast, age squared, marital status (having a spouse), number of children, education year, risk aversion, and a myopic view of the future are negatively associated with the investment loss tolerance dummy. Additionally, in Models 1 and 4, the log of household income has a significantly significant negative impact on investment loss tolerance.

4. Discussion and Conclusions

This study set out to examine how the three dimensions of financial literacy, financial knowledge, financial attitude, and financial behavior shape investment loss tolerance among retail investors in Japan. Consistent with our hypothesis, the results demonstrate that each dimension of financial literacy exerts a significant and positive effect on investment loss tolerance. The findings hold across both ordered probit and probit regression models, reinforcing the robustness of the evidence. In other words, investors with higher financial knowledge, healthier financial attitudes, and more responsible financial behaviors are more resilient when facing potential or realized investment losses.

The empirical evidence strongly supports our hypothesis that financial literacy enhances investor resilience in the face of losses. By integrating knowledge, attitude, and behavior, this study shows that financial literacy provides a multi-dimensional buffer against panic-driven liquidation. This aligns with behavioral finance theory, which posits that financial decisions are shaped not solely by cognitive knowledge but also by attitudes and behavioral tendencies under stress (Tandon 2024; Kahneman and Tversky 1979). The robustness checks further confirm that these effects are not model-specific, underscoring the validity of the conclusions.

The findings resonate with the theoretical arguments outlined in the Introduction Section. Financial knowledge equips investors with a sound understanding of market dynamics and risk–return tradeoffs, enabling them to recognize that short-term losses do not necessarily undermine long-term profitability (Lusardi and Mitchell 2014; Ahmad and Shah 2022). At the same time, financial attitudes, such as future orientation and confidence in long-term planning, mitigate emotional responses to losses (Chu et al. 2014). Finally, prudent financial behaviors such as diversification and consistent monitoring help investors implement disciplined strategies that reduce the likelihood of premature liquidation (Munizu et al. 2024). Together, these three elements provide a comprehensive framework for explaining why some investors can endure temporary losses, while others cannot.

Our study extends and generalizes previous findings in important ways. Homma et al. (2025) demonstrated that financial knowledge, attitudes, and behaviors reduced panic selling during the COVID-19 crisis, highlighting their protective role under extraordinary conditions. However, their operationalization of investor behavior was binary, i.e., whether investors sold during a crisis or not, thus overlooking the variation in the degree of losses that investors were willing to withstand. By contrast, our study introduces a continuous measure of investment loss tolerance in a non-crisis context, thereby capturing more nuanced investor behavior and broadening the generalizability of results.

Moreover, much of the earlier literature focused on risk tolerance as the outcome of interest and often relied solely on financial knowledge as the explanatory factor. For example, Grable and Rabbani (2023) and Ahmed et al. (2021) found that individuals with greater financial knowledge display higher risk tolerance. Similarly, Reddy and Mahapatra (2017) emphasized that knowledge enhances the ability to accept uncertainty in financial markets. However, these studies typically treated financial literacy as a one-dimensional construct and neglected the roles of attitudes and behaviors. Recent studies (Jain and Kesari 2020; Tiwari 2024) have begun to suggest that psychological and behavioral aspects matter, but evidence remains fragmented. By systematically incorporating all three dimensions of financial literacy, our findings demonstrate that risk-related outcomes such as investment loss tolerance are better explained within a multi-dimensional framework.

Beyond financial literacy, the control variables also yielded insights consistent with the prior literature. Male investors, older individuals, and those with greater household assets consistently exhibited higher investment loss tolerance, echoing patterns reported in risk tolerance research (Alber and Gamal 2019; Maritz and Oberholzer 2019). The positive association with age, though nonlinear due to the negative effect of age squared, suggests that tolerance initially rises with experience but may diminish in later life stages when risk capacity declines. Household income showed mixed effects, being positively associated in some models but being negative in robustness checks, suggesting that wealth accumulation rather than income flow may be the most stable determinant of resilience. In contrast, having a spouse, more children, higher education years, greater risk aversion, and a myopic outlook were negatively associated with tolerance, indicating that family responsibilities, conservative attitudes, or short-termism constrain investors’ willingness to endure losses. These patterns highlight that demographic, psychological, and socioeconomic contexts interact with financial literacy in shaping investment decisions.

Our findings offer several practical implications for policymakers, regulators, and financial institutions that go beyond the generic call for “more education.” This study demonstrates that not only financial knowledge but also financial attitudes and behaviors significantly shape investors’ tolerance for losses, highlighting that traditional educational efforts focused solely on conveying financial facts are insufficient. Instead, effective programs must also target attitudes (e.g., fostering patience and long-term orientation) and behaviors (e.g., disciplined saving and portfolio management), which our results show independently strengthen investor resilience. For policymakers, this means designing financial literacy curricula that explicitly integrate psychological and behavioral training alongside knowledge dissemination; for example, classroom simulations, gamified investment exercises, and scenario-based learning could prepare individuals to handle volatility without resorting to premature liquidation. Financial institutions, in turn, should move beyond generic advisory services by offering interventions tailored to clients’ attitudes and behavioral tendencies—such as nudges to discourage impulsive trading, tools to encourage goal-based investing, or counseling to reinforce long-term perspectives—since these can be more effective than standard product education alone. By implementing these multi-dimensional approaches, both policymakers and financial institutions can strengthen investor resilience at scale, which in turn promotes greater financial stability in retail markets. This contribution is particularly relevant in the Japanese context, where household participation in securities markets remains relatively low and investor confidence is often shaken by downturns.

While our findings highlight the positive associations of financial knowledge, behavior, and attitude with investment loss tolerance, several limitations should be acknowledged. First, investment loss tolerance is measured based on subjective responses derived through generalized hypothetical scenarios, which may not fully capture actual investor behavior. However, this approach is valuable because it avoids the need to control for individual-specific circumstances and external environments, which would be necessary when using behavioral indicators based on real market activity. Second, this study remains concerned about the accuracy of self-reported responses to questions about financial behavior and attitudes, as such responses may be subject to social desirability bias. Third, the use of cross-sectional data from customers of a single online securities firm may limit the generalizability of our findings, and the possibility of endogeneity cannot be fully ruled out given the observational nature of the dataset. Finally, financial literacy variables (partially from 2022/2023) and risk tolerance (from 2025) were measured in different waves. While literacy is typically stable over time, the possibility of minor measurement error or concerns about reverse causality cannot be entirely excluded. Although these limitations may not significantly alter the conclusions, future research should incorporate both cross-sectional and longitudinal data from more diverse populations to enhance the external validity of the results.