Self-Reported Investment Performance and Loneliness Among Individual Investors in Japan

Abstract

This study examines the association between self-reported recent investment performance and loneliness among 158,080 individual investors in Japan using the 2025 Survey on Life and Money. Loneliness was measured using three UCLA loneliness items, and investment performance was classified into five self-reported categories. We estimated probit models with break-even performance as the reference category, adjusting for demographic, socioeconomic, financial, and other observed characteristics, and conducted robustness and Wald tests. In the fully adjusted model, slight and considerable profits were associated with 2.23- and 7.34-percentage-point lower probabilities of loneliness, respectively, relative to break-even performance. These negative associations remained statistically significant under alternative definitions of loneliness. In contrast, results for reported losses were less consistent, with considerable losses positively associated with loneliness under some alternative measures. Wald tests rejected coefficient symmetry between considerable profits and considerable losses, but not between slight profits and slight losses. Joint tests did not find statistically significant evidence of differences in investment-performance coefficients by gender or age group. These findings highlight an association between perceived investment outcomes and loneliness, but the cross-sectional design and self-reported measures preclude causal conclusions.

1. Introduction

Financial circumstances have been associated with loneliness, particularly in research on older adults. Drost et al. (2024) found that loneliness was associated with subjective financial strain, whereas changes in household income and net assets showed no comparable relationship. Similarly, Bierman et al. (2024), using longitudinal data, documented an association between financial strain and greater loneliness in later life. These findings suggest that the subjective experience of financial circumstances may be relevant to loneliness in ways that are not fully captured by objective economic resources alone. In Japan, loneliness and social isolation have also become explicit policy concerns. The Act on the Advancement of Measures to Address Loneliness and Isolation, which came into effect in April 2024, established a national framework for addressing these issues, while the government’s subsequent priority plan outlined measures to strengthen coordinated policy responses (Cabinet Office of Japan, 2025). This policy context underscores the importance of understanding loneliness across different social and economic domains, including the financial experiences of individual investors.

This distinction between objective resources and subjective financial experience is also evident in the broader financial well-being literature. Netemeyer et al. (2018) conceptualized perceived financial well-being in terms of current money-management stress and expected future financial security and showed that these subjective evaluations were associated with overall well-being even after accounting for objective financial measures and other life domains. Thus, how individuals perceive and evaluate their financial circumstances may contain information relevant to well-being beyond conventional indicators such as income and wealth. Against this background, investors’ evaluations of their recent investment performance can be viewed as one specific dimension of subjective financial experience.

Related evidence on stock-market fluctuations and wealth shocks further highlights the potential relevance of investment outcomes to psychological well-being. McInerney et al. (2013) found that wealth losses associated with the 2008 stock-market crash were accompanied by increased self-reported depression and antidepressant use among older adults in the United States, although the study found no corresponding increase in clinically validated measures of depressive symptoms. Engelberg and Parsons (2016) documented an inverse relationship between daily stock returns and hospital admissions, particularly for psychological conditions. Schwandt (2018) further showed that stock-market-induced wealth shocks affected physical and mental health outcomes among retirees. Although these studies focus on mental or physical health rather than loneliness, they suggest that investment-related financial experiences may have implications beyond conventional measures of economic resources. However, they do not establish whether investors’ own subjective evaluations of recent performance are associated with loneliness.

Research on individual investors provides further evidence that investment experiences are related to subjective well-being and satisfaction. Merkle et al. (2015), using a panel survey of brokerage clients, showed that experienced investor happiness was associated with relative investment performance and success in active trading. They also found that the determinants of experienced happiness did not fully correspond to those of anticipated happiness. Related experimental evidence indicates that social comparison can affect how investors evaluate their own outcomes. Andraszewicz et al. (2023) found that upward social comparison increased risk-taking and trading activity while reducing investors’ satisfaction with their own performance. Together, these studies show that investors’ evaluations of investment performance are associated with happiness and satisfaction and can also be influenced by relative performance and social comparison.

Investing also has an important social dimension. Hong et al. (2004) showed that households characterized by greater social interaction were more likely to participate in the stock market. Andraszewicz et al. (2023) further demonstrated that information about other investors’ performance can affect both trading behavior and satisfaction with one’s own performance. However, social interaction, social comparison, and loneliness are distinct constructs and should not be treated as interchangeable. Prior research has linked investment experiences to investor happiness and satisfaction, while a separate study has documented associations between financial strain and loneliness. However, relatively little is known about whether individual investors’ own recent investment performance is associated with loneliness. This question lies at the intersection of research on subjective financial experiences, investor well-being, and loneliness.

Taken together, these strands of literature provide a conceptual basis for examining the relationship between perceived investment performance and loneliness. Favorable evaluations of investment outcomes may coincide with greater perceived financial security, reduced financial strain, or greater confidence, all of which could be associated with lower loneliness. Accordingly, a more favorable assessment of recent investment performance may be associated with a lower likelihood of reporting loneliness. However, this relationship need not be symmetric across gains and losses. Investors may evaluate their outcomes relative to personal expectations, reference points, or the performance of other investors, rather than solely in relation to zero returns. Moreover, loneliness itself and unobserved psychological characteristics may influence investment decisions or subjective assessments of performance. We therefore examine profit and loss categories separately, using break-even performance as the reference, and assess whether the observed associations exhibit asymmetry. These mechanisms and the direction of the relationship cannot be identified from the present cross-sectional data.

The present study examines this question using the 2025 Survey on Life and Money, with an analytical sample of 158,080 individual investors in Japan. We examine the association between a UCLA-based binary measure of loneliness and respondents’ self-reported investment performance over the preceding year using probit models with a broad set of demographic, socioeconomic, financial, and other individual characteristics. Given the cross-sectional nature of the data and the self-reported measure of investment performance, the estimates are interpreted as conditional associations rather than causal effects of objectively measured investment returns.

The main results indicate that respondents reporting slight or considerable profits are significantly less likely to be classified as lonely than those reporting break-even performance. These negative associations persist across alternative operationalizations of loneliness. In contrast, the associations for reported losses are less consistent across specifications, with considerable losses showing a positive association under some alternative measures. Formal tests provide evidence of asymmetry between considerable profits and considerable losses, while the corresponding evidence for slight profits and losses is less conclusive. This study contributes to the literature in three respects. First, it connects research on investors’ subjective financial experiences with the literature on loneliness and well-being. Second, rather than focusing only on income, wealth, or general financial strain, it examines investors’ own recent perceived investment outcomes as a distinct financial experience. Third, it identifies a robust negative association between reported investment profits and loneliness while documenting important differences in the patterns observed across profit and loss categories. The mechanisms underlying these associations remain unresolved and require further investigation.

2. Materials and Methods

2.1. Data

This study used data from the 2025 wave of the Survey on Life and Money, an online survey jointly conducted by Rakuten Securities, a major online brokerage firm in Japan, and the Kadoya Lab at Hiroshima University. Data collection was carried out between January and February 2025. The target population consists of individuals aged 18–90 years residing in Japan who had logged into their Rakuten Securities account at least once during the preceding year. The survey collected information on respondents’ demographic, socioeconomic, and psychological characteristics. It also included questions related to loneliness and investment performance. The same 2025 survey and self-reported investment performance measure were previously used by Lal et al. (2026) to examine horizon-specific inflation expectations. The present study focuses on loneliness as a distinct outcome. The 2025 survey included 230,976 respondents, of whom 230,154 provided valid responses to all three loneliness items. Regarding investment performance, 9738 respondents selected the “I do not know” category, and a total of 10,069 respondents had no usable response to this question. After excluding observations with unavailable investment-performance information and missing values in variables required for estimation, the final analytical sample comprised 158,080 respondents. Some respondents had also participated in the 2022 or 2023 survey waves. Educational attainment was obtained from respondents’ earliest available survey responses and merged with the 2025 data.

2.2. Variables

The dependent variable is a binary measure of loneliness derived from the three-item UCLA loneliness scale developed by Hughes et al. (2004). The reliability and validity of a Japanese-language adaptation of this scale have been examined by Igarashi (2019). Respondents were asked: (1) “How often do you feel that you lack companionship?” (2) “How often do you feel left out?” and (3) “How often do you feel isolated from others?” Each item was measured using three response categories: “hardly ever or never” (1), “some of the time” (2), and “often” (3).

Following the binary operationalization adopted in our previous research (Kuramoto et al., 2024), respondents were classified as lonely (loneliness = 1) if they answered “some of the time” or “often” to at least one item, and as not lonely (loneliness = 0) otherwise. This indicator captures the occurrence of any reported loneliness rather than persistent or severe loneliness. The internal consistency of the three items was acceptable in the analytical sample (Cronbach’s alpha = 0.7534; N = 158,080).

To examine the sensitivity of the findings to this relatively inclusive definition, we also constructed three alternative measures: (1) the summed three-item loneliness score, ranging from 3 to 9; (2) a binary indicator equal to one for a summed score of 6 or higher; and (3) a binary indicator equal to one when at least two of the three items were answered “some of the time” or “often.”

The main explanatory variable is self-reported investment performance over the preceding year. Respondents were asked to evaluate their overall investment gains and losses during this period, including outcomes from investments that had been sold. The six response categories are described in English as follows: (1) considerable profit overall, (2) slight profit overall, (3) approximately break-even, (4) slight loss overall, (5) considerable loss overall, and (6) I do not know.

Responses 1–5 were coded as five mutually exclusive performance categories, with break-even serving as the reference category in the regression models. Respondents who selected “I do not know” were excluded from the main analysis. Because this measure reflects respondents’ own assessments, it may capture subjective evaluations and recall as well as actual investment outcomes. Accordingly, the analysis concerns self-reported investment performance rather than objectively verified portfolio returns.

We included the following observed covariates: gender, age, marital status, presence of children, years of education, employment status, urban residency, annual household income, household financial assets, risk aversion, and a myopic view of the future.

Household income and financial assets were reported in categorical ranges. For categories with defined lower and upper bounds, we assigned the midpoint of each interval. For the lowest and highest open-ended categories, we assigned the corresponding category boundary values. Specifically, annual household income below JPY 1 million was assigned JPY 1 million, while income of JPY 20 million or more was assigned JPY 20 million. Similarly, financial assets below JPY 2.5 million were assigned JPY 2.5 million, while assets of JPY 100 million or more were assigned JPY 100 million. Both variables were entered in logarithmic form.

Risk aversion was proxied by respondents’ reported threshold probability of rain at which they would normally carry an umbrella when going out. The responses ranged from 0% to 100% in increments of 10 percentage points. The measure was reverse-coded and rescaled to range from 0 to 1, with higher values indicating greater risk aversion. This measure is treated as a proxy rather than a direct measure of financial risk preferences.

The myopic view of the future was measured using agreement with the statement, “The future is uncertain, so there is no point in thinking about it,” on a five-point scale, with higher scores indicating greater agreement. Educational attainment was obtained from each respondent’s earliest available survey response. The other covariates were obtained from the 2025 survey, except that earlier responses from previous survey waves could be used for the myopia measure when contemporaneous information was unavailable. Table 1 presents the detailed definitions of all variables used in this study.

Table 1. Variable Definitions.

Variable	Definition
Dependent Variable
Loneliness	Binary variable: 1 = participants have feelings of loneliness some of the time or often, 0 = otherwise
Independent Variable
Investment Performance Considerable Loss	Binary variable: 1 = experiencing a considerable loss in respondent’s investment, 0 = otherwise
Investment Performance Slight Loss	Binary variable: 1 = experiencing a slight loss in respondent’s investment, 0 = otherwise
Investment Performance Break-Even (Base)	Binary variable: 1 = breaking even on respondent’s investment, 0 = otherwise
Investment Performance Slight Profit	Binary variable: 1 = achieving a slight profit in respondent’s investment, 0 = otherwise
Investment Performance Considerable Profit	Binary variable: 1 = achieving a considerable profit in respondent’s investment, 0 = otherwise
Gender	Binary variable: 1 = male, 0 = female
Age	Continuous variable: respondents’ age
Marital Status	Binary variable: 1 = having a spouse, 0 = otherwise
Having Children	Binary variable: 1 = having at least one child, 0 = otherwise
Years of Education	Continuous variable: years of education
Unemployed	Binary variable: 1 = not having a job or being a housewife or househusband, 0 = otherwise
Urban Residency	Binary variable: 1 = living in Tokyo special wards or government-designated city areas, 0 = otherwise
Household Income	Continuous variable: the total annual gross household income in 2024 (unit: JPY)
Log of Household Income	Continuous variable: log (Household Income)
Household Assets	Continuous variable: the total household financial assets
Log of Household Assets	Continuous variable: log (Household Assets)
Risk Aversion	Continuous variable: measuring risk aversion based on the probability of rain at which the respondent would carry an umbrella, reverse-coded so that higher values indicate greater risk aversion.
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.”

2.3. Descriptive Statistics

Table 2 shows the descriptive statistics for the variables used in the estimation. Approximately 73.5% of respondents were classified as lonely based on the binary measure used in this study. Among respondents, 3.8% reported a considerable loss, 5.1% a slight loss, 9.0% break-even performance, 63.1% a slight profit, and 19.0% a considerable profit. Men accounted for 67.5% of the sample, and the mean age was 46.42 years. Approximately 67.3% were married, 59.3% had at least one child, and the mean number of years of education was 15.10. About 10.4% were not currently employed, and 15.7% lived in Tokyo special wards or government-designated cities. Mean annual household income and household assets were approximately JPY 7.74 million and JPY 21.90 million, respectively. The mean risk-aversion score was 0.535, and the mean score for the myopic view of the future was 2.422.

Table 2. Descriptive Statistics.

Variable	Mean	Std. Dev.	Min	Max
Dependent variable
Loneliness	0.735	0.441	0	1
Independent variable
Investment Performance Considerable Loss	0.0379	0.191	0	1
Investment Performance Slight Loss	0.0509	0.220	0	1
Investment Performance Break-Even	0.0904	0.287	0	1
Investment Performance Slight Profit	0.631	0.483	0	1
Investment Performance Considerable Profit	0.190	0.392	0	1
Gender	0.675	0.468	0	1
Age	46.42	12.29	18	90
Marital Status	0.673	0.469	0	1
Having Children	0.593	0.491	0	1
Years of Education	15.10	2.083	9	21
Unemployed	0.104	0.305	0	1
Urban Residency	0.157	0.364	0	1
Household Income	7,736,000	4,294,000	1,000,000	20,000,000
Log of Household Income	15.69	0.619	13.82	16.81
Household Assets	21,900,000	25,680,000	2,500,000	100,000,000
Log of Household Assets	16.29	1.109	14.73	18.42
Risk aversion	0.535	0.238	0	1
Myopic View of the Future	2.422	0.966	1	5
Observations	158,080

We also examined the prevalence of loneliness by gender and age. The proportion classified as lonely was 74.07% among men and 72.46% among women. By age, the corresponding proportions were 74.85% among respondents younger than 30 years, 76.32% among those aged 30–44 years, 74.55% among those aged 45–59 years, and 64.24% among those aged 60 years or older. Table 3 presents these distributions.

Table 3. Distribution of Loneliness by Gender and Age.

Loneliness Dummy	Gender		Age
	Female	Male	Under 30	30–44	45–59	60 or Older
0	14,136	27,682	3412	13,674	15,536	9196
	27.54%	25.93%	25.15%	23.68%	25.45%	35.76%
1	37,194	79,068	10,156	44,076	45,513	16,517
	72.46%	74.07%	74.85%	76.32%	74.55%	64.24%
Total	51,330	106,750	13,568	57,750	61,049	25,713
	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%

Note: Each cell reports the number of respondents, followed by the percentage within the corresponding gender or age group. Loneliness = 1 indicates respondents classified as lonely under the original binary definition, and Loneliness = 0 indicates otherwise. Pearson chi-square tests indicate statistically significant associations between loneliness and both gender (χ²(1) = 46.06, p < 0.001) and age group (χ²(3) = 1417.92, p < 0.001).

2.4. Methods

Because loneliness is measured as a binary variable, we estimated probit regression models of the following form:

$$\mathrm{Pr}\left({\mathit{Loneliness}}_i=1\mid I{P}_i,X_i\right)=\mathsf{\Phi }\left(\alpha +{\beta }^{\mathrm{T}}I{P}_i+{\gamma }^{\mathrm{T}}{X}_i\right)$$

where Loneliness_i indicates whether respondent i is classified as lonely, IP_i is a vector of indicators for the respondent’s self-reported investment performance over the preceding year, X_i is a vector of demographic, socioeconomic, financial, and psychological characteristics, and $\mathsf{\Phi }\left(\cdot \right)$ denotes the standard normal cumulative distribution function. Break-even performance serves as the reference category.

Model 1 includes only the investment-performance indicators. Model 2 adds gender, age, marital status, presence of children, years of education, employment status, and urban residency. Model 3 further includes the logarithms of household income and household assets. Model 4 additionally includes risk aversion and a myopic view of the future. Robust standard errors are reported for all models.

To assess the robustness of the main findings, we estimated alternative specifications using the summed loneliness score as the dependent variable in an OLS model and two alternative binary loneliness indicators in probit models. Robust standard errors were used throughout these analyses.

For the fully adjusted probit model, we calculated average marginal effects for each investment-performance category relative to break-even performance, together with 95% confidence intervals. We also obtained adjusted predicted probabilities for the five performance categories. To examine asymmetry, Wald tests were conducted for the null hypotheses that the coefficients for slight loss and slight profit sum to zero and that the coefficients for considerable loss and considerable profit sum to zero. A joint Wald test was also performed to assess whether both loss-category coefficients were zero.

For the heterogeneity analysis, we estimated the fully adjusted specification separately by gender and by four age groups: younger than 30 years, 30–44 years, 45–59 years, and 60 years or older. In addition, pooled probit models incorporating interactions between investment-performance categories and gender or age groups were estimated. Joint Wald tests were used to assess whether the interaction coefficients were collectively equal to zero.

All pairwise correlations were below 0.65, and all variance inflation factors were below 3.10, indicating no substantial multicollinearity among the explanatory variables.

All statistical analyses were conducted using Stata 16 (Stata Corp LLC, College Station, TX, USA).

3. Results

3.1. Main Results

Table 4 reports the probit estimates, with break-even investment performance serving as the reference category. Across all four specifications, the coefficients for slight profit and considerable profit remain negative and statistically significant. In the fully adjusted model (Model 4), the coefficients are −0.0721 and −0.2263, respectively (both p < 0.01). By contrast, neither slight loss nor considerable loss is statistically significantly associated with loneliness in this specification. The corresponding coefficients are 0.0254 and 0.0087, respectively.

Table 4. Regression Results Using the Probit Model.

Variable	Dependent Variable: Loneliness
	Model 1	Model 2	Model 3	Model 4
Investment Performance Break-Even (Base)
Investment Performance Considerable Loss	0.0008	0.0082	0.0072	0.0087
	(0.0213)	(0.0215)	(0.0215)	(0.0215)
Investment Performance Slight Loss	0.0127	0.0228	0.0247	0.0254
	(0.0193)	(0.0195)	(0.0195)	(0.0195)
Investment Performance Slight Profit	−0.0861 ***	−0.0851 ***	−0.0748 ***	−0.0721 ***
	(0.0123)	(0.0125)	(0.0126)	(0.0126)
Investment Performance Considerable Profit	−0.2440 ***	−0.2492 ***	−0.2305 ***	−0.2263 ***
	(0.0138)	(0.0140)	(0.0143)	(0.0143)
Gender		0.1247 ***	0.1268 ***	0.1287 ***
		(0.0076)	(0.0076)	(0.0077)
Age		−0.0077 ***	−0.0083 ***	−0.0085 ***
		(0.0003)	(0.0004)	(0.0004)
Marital Status		−0.2129 ***	−0.1551 ***	−0.1545 ***
		(0.0094)	(0.0100)	(0.0100)
Having Children		−0.1666 ***	−0.1566 ***	−0.1541 ***
		(0.0091)	(0.0091)	(0.0091)
Years of Education		−0.0203 ***	−0.0155 ***	−0.0159 ***
		(0.0017)	(0.0018)	(0.0018)
Unemployed		0.1094 ***	0.0461 ***	0.0463 ***
		(0.0119)	(0.0125)	(0.0125)
Urban Residency		−0.0562 ***	−0.0395 ***	−0.0420 ***
		(0.0094)	(0.0095)	(0.0095)
Log of Household Income			−0.1234 ***	−0.1223 ***
			(0.0073)	(0.0073)
Log of Household Assets			0.0125 ***	0.0125 ***
			(0.0038)	(0.0038)
Risk aversion				0.1110 ***
				(0.0148)
Myopic View of the Future				0.0241 ***
				(0.0037)
Constant	0.7313 ***	1.5622 ***	3.2012 ***	3.0772 ***
	(0.0116)	(0.0323)	(0.1040)	(0.1053)
Observations	158,080	158,080	158,080	158,080
Pseudo-R-squared	0.00291	0.0223	0.0240	0.0246
Log likelihood	−91,065	−89,291	−89,135	−89,086

Note: Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

These findings indicate that, under the original binary definition of loneliness, negative associations are observed for the profit categories but not for the loss categories relative to break-even performance. However, differences in statistical significance alone do not establish asymmetry between gains and losses. Formal tests of asymmetry and robustness analyses using alternative loneliness measures are therefore presented below. The pseudo-R-squared of Model 4 is 0.0246, indicating a modest improvement in likelihood-based model fit relative to the intercept-only model.

In the unadjusted descriptive analysis, the proportions of respondents classified as lonely were 76.79% for considerable loss, 77.16% for slight loss, 76.77% for break-even performance, 74.06% for slight profit, and 68.70% for considerable profit. These descriptive patterns indicate lower loneliness prevalence among respondents reporting profits, particularly considerable profits. However, they do not account for differences in respondents’ observed characteristics.

3.2. Heterogeneity Analysis

We next examined heterogeneity in the association between self-reported investment performance and loneliness by gender and age. Table 5 presents the subgroup-specific probit estimates. The negative coefficients for considerable profit were statistically significant across all gender and age subgroups. The coefficients for slight profit were also negative across all subgroups, although their statistical significance varied. Among respondents aged 30–44 years, considerable loss was negatively associated with loneliness at the 10% significance level, whereas slight loss was positively associated at the 5% level.

Table 5. Regression Results by Gender and Age.

Variable	Dependent Variable: Loneliness
	Gender		Age
	Male	Female	Under 30	30–44	45–59	60 or Older
Investment Performance Considerable Loss	0.0097	0.0077	0.1084	−0.0783 *	0.0179	0.0445
	(0.0244)	(0.0465)	(0.1179)	(0.0412)	(0.0343)	(0.0419)
Investment Performance Slight Loss	0.0200	0.0405	0.0719	0.0835 **	−0.0043	0.0184
	(0.0227)	(0.0385)	(0.0800)	(0.0376)	(0.0310)	(0.0392)
Investment Performance Slight Profit	−0.0669 ***	−0.0769 ***	−0.0655	−0.0927 ***	−0.0682 ***	−0.0292
	(0.0148)	(0.0240)	(0.0450)	(0.0230)	(0.0201)	(0.0273)
Investment Performance Considerable Profit	−0.2119 ***	−0.2450 ***	−0.2429 ***	−0.2528 ***	−0.2195 ***	−0.1744 ***
	(0.0169)	(0.0272)	(0.0507)	(0.0255)	(0.0230)	(0.0323)
Gender	−	−	0.0083	0.0642 ***	0.2241 ***	0.3024 ***
			(0.0241)	(0.0124)	(0.0125)	(0.0219)
Age	−0.0060 ***	−0.0124 ***	0.0403 ***	0.0089 ***	−0.0152 ***	−0.0317 ***
	(0.0004)	(0.0006)	(0.0058)	(0.0014)	(0.0014)	(0.0019)
Marital Status	−0.2319 ***	−0.0634 ***	−0.1457 ***	−0.1967 ***	−0.1152 ***	−0.1885 ***
	(0.0128)	(0.0163)	(0.0331)	(0.0180)	(0.0161)	(0.0242)
Having Children	−0.1215 ***	−0.1869 ***	−0.1147 ***	−0.1817 ***	−0.1831 ***	−0.1461 ***
	(0.0116)	(0.0150)	(0.0414)	(0.0155)	(0.0146)	(0.0238)
Years of Education	−0.0129 ***	−0.0231 ***	−0.0302 ***	−0.0066 **	−0.0083 ***	−0.0161 ***
	(0.0021)	(0.0034)	(0.0072)	(0.0031)	(0.0028)	(0.0042)
Unemployed	−0.0866 ***	0.1703 ***	0.3830 ***	0.2800 ***	0.1782 ***	0.1548 ***
	(0.0172)	(0.0188)	(0.0759)	(0.0312)	(0.0255)	(0.0200)
Urban Residency	−0.0382 ***	−0.0413 **	−0.0699 **	−0.0517 ***	−0.0315 **	−0.0023
	(0.0118)	(0.0162)	(0.0305)	(0.0158)	(0.0156)	(0.0239)
Log of Household Income	−0.1408 ***	−0.1221 ***	−0.2168 ***	−0.2066 ***	−0.1969 ***	−0.1237 ***
	(0.0091)	(0.0125)	(0.0269)	(0.0151)	(0.0125)	(0.0145)
Log of Household Assets	0.0153 ***	0.0127 *	0.0727 ***	0.0416 ***	0.0112 *	−0.0306 ***
	(0.0046)	(0.0069)	(0.0169)	(0.0070)	(0.0060)	(0.0085)
Risk aversion	0.1173 ***	0.0933 ***	0.1303 ***	0.1211 ***	0.1310 ***	0.1655 ***
	(0.0175)	(0.0279)	(0.0503)	(0.0255)	(0.0236)	(0.0364)
Myopic View of the Future	0.0395 ***	−0.0077	0.0261 **	0.0165 ***	0.0076	0.0579 ***
	(0.0044)	(0.0066)	(0.0121)	(0.0061)	(0.0060)	(0.0091)
Constant	3.2839 ***	3.3844 ***	2.3203 ***	3.2883 ***	4.5488 ***	4.8976 ***
	(0.1318)	(0.1820)	(0.3944)	(0.2093)	(0.1875)	(0.2559)
Observations	106,750	51,330	13,568	57,750	61,049	25,713
Pseudo-R-squared	0.0261	0.0254	0.0193	0.0286	0.0285	0.0316
Log likelihood	−59,502	−29,443	−7504	−30,704	−33,640	−16,237

Note: Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

To formally assess whether the associations differed across subgroups, we estimated pooled probit models including interactions between investment-performance categories and gender or age groups. Joint Wald tests did not reject the null hypothesis that the interaction coefficients were collectively zero for either gender (χ²(4) = 1.38, p = 0.8483) or age groups (χ²(12) = 16.88, p = 0.1542). Thus, despite differences in the statistical significance of individual subgroup coefficients, the interaction tests did not provide statistically significant evidence of heterogeneity in the investment-performance coefficients across gender or age groups.

3.3. Robustness Checks and Marginal Effects

Table 6 presents robustness checks using alternative operationalizations of loneliness. The negative associations for slight profit and considerable profit remain statistically significant across all three alternative specifications: OLS using the summed three-item loneliness score, probit using a summed score of 6 or higher, and probit using a binary indicator requiring positive responses to at least two loneliness items.

Table 6. Robustness Checks Using Alternative Loneliness Measures.

Variables	OLS: Summed Score	Probit: Score ≥ 6	Probit: ≥2 Items
Investment Performance Considerable Loss	0.1477 ***	0.0501 **	0.0195
	(0.0266)	(0.0196)	(0.0195)
Investment Performance Slight Loss	0.0221	0.0064	0.0030
	(0.0229)	(0.0178)	(0.0177)
Investment Performance Slight Profit	−0.1844 ***	−0.1395 ***	−0.1292 ***
	(0.0148)	(0.0115)	(0.0115)
Investment Performance Considerable Profit	−0.3839 ***	−0.3140 ***	−0.3093 ***
	(0.0168)	(0.0135)	(0.0133)
Gender	0.1678 ***	0.1075 ***	0.0709 ***
	(0.0089)	(0.0073)	(0.0071)
Age	−0.0121 ***	−0.0079 ***	−0.0086 ***
	(0.0004)	(0.0003)	(0.0003)
Marital Status	−0.3299 ***	−0.2207 ***	−0.2182 ***
	(0.0115)	(0.0091)	(0.0090)
Having Children	−0.2541 ***	−0.1469 ***	−0.1368 ***
	(0.0106)	(0.0086)	(0.0084)
Years of Education	−0.0276 ***	−0.0186 ***	−0.0174 ***
	(0.0021)	(0.0017)	(0.0016)
Unemployed	0.1081 ***	0.0303 **	−0.0022
	(0.0152)	(0.0121)	(0.0117)
Urban Residency	−0.0450 ***	−0.0314 ***	−0.0391 ***
	(0.0112)	(0.0091)	(0.0089)
Log of Household Income	−0.1879 ***	−0.1042 ***	−0.0928 ***
	(0.0084)	(0.0067)	(0.0066)
Log of Household Assets	0.0136 ***	−0.0008	−0.0037
	(0.0045)	(0.0036)	(0.0035)
Risk aversion	0.1314 ***	0.1304 ***	0.1299 ***
	(0.0175)	(0.0139)	(0.0136)
Myopic View of the Future	0.0291 ***	0.0319 ***	0.0301 ***
	(0.0043)	(0.0034)	(0.0033)
Constant	8.8892 ***	2.1020 ***	2.3406 ***
	(0.1230)	(0.0974)	(0.0958)
Observations	158,080	158,080	158,080
R-squared	0.0591
Adjusted R-squared	0.0590
Log likelihood	−297,329	−100,498	−106,348
F-value	620
Pseudo-R-squared		0.0309	0.0292

Note: Robust standard errors in parentheses: *** p < 0.01, ** p < 0.05, * p < 0.1.

The results for the loss categories are less consistent. Considerable loss is positively and significantly associated with loneliness when the summed score is used in OLS and when the binary outcome is defined using a threshold of 6 or higher. However, this association is not statistically significant when at least two loneliness items must be endorsed. Slight loss is not statistically significant in any of the three alternative specifications. These findings suggest that the negative association for reported profits is relatively robust to the operationalization of loneliness, whereas the association for reported losses is more sensitive to the outcome definition.

Table 7 reports average marginal effects and adjusted predicted probabilities from the fully adjusted probit specification. Relative to break-even performance, slight profit is associated with a 2.23-percentage-point lower probability of being classified as lonely (95% CI: −2.98 to −1.48 percentage points), while considerable profit is associated with a 7.34-percentage-point lower probability (95% CI: −8.22 to −6.45 percentage points). Neither loss category has a statistically significant average marginal effect under the original binary loneliness definition.

Table 7. Average Marginal Effects and Adjusted Predicted Probabilities.

AME	dy/dx	Delta-Method Std. Err.	z	p-value	[95% Conf. Interval]
Investment Performance Considerable Loss	0.002604	0.0064757	0.40	0.688	−0.0100881	0.015296
Investment Performance Slight Loss	0.0075961	0.005819	1.31	0.192	−0.0038091	0.0190012
Investment Performance Slight Profit	−0.0222946	0.0038191	−5.84	0.000	−0.0297798	−0.0148094
Investment Performance Considerable Profit	−0.0733631	0.0045264	−16.21	0.000	−0.0822347	−0.0644915
Predictive Margins	Margin	Delta-Method Std. Err.	z	p-value	[95% Conf. Interval]
Investment Performance Considerable Loss	0.7656064	0.0054585	140.26	0.000	0.7549079	0.7763048
Investment Performance Slight Loss	0.7705985	0.0046512	165.68	0.000	0.7614822	0.7797148
Investment Performance Break-Even	0.7630024	0.003564	214.09	0.000	0.7560172	0.7699877
Investment Performance Slight Profit	0.7407079	0.0013744	538.93	0.000	0.7380141	0.7434017
Investment Performance Considerable Profit	0.6896394	0.0026975	255.66	0.000	0.6843523	0.6949264

Formal Wald tests were also conducted to examine gain–loss symmetry. The null hypothesis that the coefficients for considerable loss and considerable profit sum to zero was rejected (χ²(1) = 50.25, p < 0.001), whereas the corresponding hypothesis for slight loss and slight profit was not rejected (χ²(1) = 2.68, p = 0.1013). A joint test of the two loss coefficients also failed to reject the null hypothesis that both coefficients were zero (χ²(2) = 1.70, p = 0.4283). These results provide evidence against symmetry between the considerable-profit and considerable-loss coefficients, but the evidence is less conclusive for the slight-profit and slight-loss categories. The tests concern coefficients in the original probit specification and do not establish causal relationships.

4. Discussion

This study examined the association between self-reported recent investment performance and loneliness among individual investors in Japan. Respondents reporting slight or considerable profits were significantly less likely to be classified as lonely than those reporting break-even performance. These negative associations remained statistically significant across alternative operationalizations of loneliness. In the fully adjusted probit model, slight and considerable profits were associated with 2.23- and 7.34-percentage-point lower probabilities of loneliness, respectively. In contrast, the results for losses were less consistent across specifications, with considerable losses showing a positive association under some alternative loneliness measures. Formal Wald tests rejected gain–loss symmetry for the considerable-profit and considerable-loss coefficients, but not for the slight-profit and slight-loss coefficients. Furthermore, joint interaction tests did not provide statistically significant evidence that the investment-performance coefficients differed by gender or age group. These findings should be interpreted as conditional associations rather than causal effects.

The negative association observed for the profit categories complements previous evidence linking investment experiences to investors’ subjective well-being and satisfaction. Merkle et al. (2015) showed that experienced investor happiness was associated with relative performance and success in active trading, suggesting that investors’ subjective experiences are related to how they evaluate their investment outcomes. The present study complements this literature by examining loneliness, which is conceptually distinct from happiness and satisfaction, in relation to investors’ own recent perceived performance. Importantly, the negative association with reported profits remained statistically significant after adjusting for household income, household financial assets, and other observed characteristics. However, these adjustments do not eliminate the possibility of unobserved confounding, and household financial assets may themselves partly reflect recent investment outcomes. Nevertheless, the findings suggest that investors’ subjective financial experiences may be relevant to research on loneliness and well-being alongside conventional measures of economic circumstances.

The differences observed across profit and loss categories deserve particular attention. In the present sample, approximately 82% of respondents reported positive investment performance, whereas only about 9% reported break-even performance. Consequently, break-even outcomes may not necessarily have been perceived as neutral by respondents. In a setting where most surveyed investors reported profits, break-even performance could represent a relatively unfavorable outcome when compared with the experiences of other investors. The observed pattern may therefore partly reflect a distinction between reporting profits and not reporting profits, rather than an asymmetric response around an objectively meaningful zero-return reference point. Because the survey does not identify respondents’ individual reference points or their perceptions of other investors’ performance, this interpretation remains speculative. Moreover, the distribution of reported performance limits the generalizability of the findings to periods in which investment losses are more widespread.

The observed pattern also does not provide a straightforward application of conventional loss-aversion arguments. Merkle (2020) found that loss aversion was substantially stronger for anticipated investment outcomes than for experienced returns and that investors’ evaluations of experienced gains and losses were much closer to loss neutrality. Similarly, Grosshans and Zeisberger (2018) showed experimentally that investor satisfaction depends not only on the final return but also on the path through which that return is achieved. Investors were most satisfied when an asset initially declined and subsequently recovered and least satisfied when the price path followed the reverse pattern, regardless of whether the final return was positive or negative. A related asymmetry was documented by Kaustia and Knüpfer (2012), who found that favorable recent returns experienced by local peers increased stock-market entry, whereas unfavorable peer returns did not produce a corresponding reduction. Although these studies examine different outcomes and mechanisms, they illustrate that psychological and behavioral responses to investment-related gains and losses need not be symmetric. The present analysis does not directly measure reference points, return paths, or social comparisons and therefore cannot determine which, if any, of these mechanisms accounts for the observed associations.

Several psychological mechanisms could potentially contribute to the negative association between reported profits and loneliness. Favorable investment outcomes may coincide with greater financial satisfaction, perceived financial security, confidence, or more favorable evaluations of one’s financial circumstances. Previous research has linked perceived financial well-being to broader well-being (Netemeyer et al., 2018) and investment performance to investor happiness and satisfaction (Merkle et al., 2015). However, the present analysis does not directly test these mechanisms. An important alternative explanation is that respondents’ mood, optimism, or other psychological characteristics may influence both their assessments of investment performance and their responses to loneliness questions. Because both measures were self-reported in the same survey, common method variance cannot be excluded. Reverse causality also remains possible, as loneliness may influence investment decisions or the subjective evaluation of investment outcomes. Accordingly, the observed associations should not be interpreted as evidence of a specific psychological mechanism or causal direction.

The findings should also be considered in light of several limitations. First, the analysis uses cross-sectional observations from the 2025 survey and does not establish temporal ordering or causality. Although some respondents participated in earlier survey waves, the present analysis does not exploit repeated observations to examine within-person changes or the temporal relationship between loneliness and investment performance. Second, investment performance is based on respondents’ own assessments rather than objectively verified brokerage returns. These assessments may reflect recall, subjective reference points, and psychological characteristics in addition to actual investment outcomes. Third, the models do not directly account for health status, personality traits, or detailed characteristics of investment activity, which may be associated with both reported performance and loneliness. Fourth, the sample consists of respondents drawn from the customer base of a single Japanese brokerage and may not represent the general population or investors in other settings. In addition, excluding respondents who could not assess their investment performance or had missing covariate information may introduce selection bias. Fifth, the original binary loneliness measure uses a relatively inclusive threshold and does not distinguish occasional from more persistent loneliness. Although the alternative measures support the negative associations for reported profits, the findings for losses vary across specifications. Finally, the available data do not allow us to identify the psychological and social mechanisms underlying these associations. Future research using repeated measures of loneliness, objective investment outcomes, and additional psychological and financial information would help clarify these issues.

5. Conclusions

This study examined the association between self-reported recent investment performance and loneliness among 158,080 individual investors in Japan. Respondents reporting slight or considerable profits were significantly less likely to be classified as lonely than those reporting break-even performance, and these negative associations remained robust across alternative loneliness measures. In contrast, the associations for reported losses were less consistent across specifications. Formal tests provided evidence against symmetry between considerable profits and considerable losses, but not between slight profits and slight losses.

These findings contribute to behavioral research on financial experiences and well-being by linking investors’ own recent perceived investment outcomes to loneliness, an outcome that has received relatively little attention in research on investment performance. The findings also demonstrate that the negative association between reported profits and loneliness remains after adjustment for conventional financial characteristics such as household income and assets. However, the results do not establish that investment gains reduce loneliness or identify the mechanisms underlying the observed associations. The cross-sectional design, self-reported measures, and potential for reverse causality and unobserved confounding limit causal interpretation.

Future research combining longitudinal measures of loneliness with objectively measured investment outcomes and relevant psychological and financial characteristics would help clarify the temporal direction and mechanisms of these relationships. Such research could also examine whether the associations differ across market conditions and investors’ subjective reference points.