九州大学学術情報リポジトリ

Kyushu University Institutional Repository

An empirical study on how financial literacy

contributes to preparation for retirement
Yeh, Tsung-ming
School of Economics, Kyushu University

http://hdl.handle.net/2324/4485345

出版情報：Journal of pension economics and finance. First View, pp.1-23, 2020-10-08. Cambridge
University Press
バージョン：
権利関係：© The Author(s), 2020.
 An empirical study on how financial literacy contributes to
preparation for retirement
Tsung-ming Yeh

Professor, School of Economics, Kyushu University

Email: [email protected]

Cite this article: Yeh T (2020). An empirical study on how financial literacy
contributes to preparation for retirement. Journal of Pension Economics and
Finance pp. 1–23. https://doi.org/10.1017/S1474747220000281

Abstract: This study provides empirical evidence on the mechanisms through which
financial literacy may be associated with saving for retirement, in the three phases of the
decision-making process—information perception, information search and evaluation,
and decision-making and implementation. The results indicate that financial literacy has
significantly positive effects on one’s awareness of post-retirement financial needs,
comparing alternatives when purchasing financial products, displaying fewer present-
time, and planning for and setting aside funds for retirement. Financial literacy not only
directly contributes to planning for future, but also indirectly via a reduction in
behavioral biases.

Keywords: Financial literacy; Saving for retirement; Loss aversion; Present time bias
Acknowledgement: The author is grateful for comments and suggestions by the
referees and the editor, as well as participants at the 26th Conference on the Theories
and Practices of Securities and Financial Markets, the 31st Australasian Finance and
Banking Conference, the Asian Growth Research Institute, the Economic Engineering
Research Workshop at Kyushu University, and the 2019 International Conference of
Taiwan Finance Association. The author also acknowledges financial support from
JSPS KAKENHI (JP17K03807); Nomura School of Advanced Management; and
Education and Research Center for Mathematical and Data Science, Kyushu University.

1
 An empirical study on how financial literacy contributes to preparation for

retirement

The existing literature has in general demonstrated that a lack of financial

literacy is found to be also associated with poor financial decisions or lower financial

wellbeing; individuals with low financial literacy are less likely to invest in the stock

market (van Rooij et al., 2011), to save for post-retirement (Lusardi and Mitchell, 2007),

to accumulate wealth (Lusardi and Mitchell, 2008), are more likely to take out high-cost

mortgages (Moore, 2003), to have debt problems (Lusardi and Tufano, 2009), and to have

incurred a loss from sub-prime mortgages during the 2008 financial crisis (Gerardi et al.,

2010). However, the mechanisms by which financial literacy leads to the reported

outcomes have not been fully explored—Why or how do financially literate people

prepare for future such as retirement? Planning for retirement is an important long-term

financial decision that matters a great deal for life post-retirement when one is usually no

longer earning. Under-saving for retirement will result in a dramatic drop in income,

consumption and life quality. The mechanism in which financial literacy is associated

with preparing for retirement is a research topic of both theoretical and practical

importance.

This study explores the possible channels through which financial literacy can

play relevantly positive roles in the three phases of a financial decision-making process

in a financially dominated life situation, using the taxonomy suggested by Daxhammer

and Facsar (2012)—information perception, information search and evaluation, and

decision-making and implementation. This study hypothesizes the phase-by-phase roles

of financial literacy, that is, an increase in one’s awareness of post-retirement living needs,

2
the ability to compare alternative financial products, and the ability to prepare for

retirement.

Existing literature also suggests that the reason people are not saving enough is

primarily behavioral biases or heuristics, such as self-control and procrastination (Thaler

and Shefrin, 1981), status quo bias (Samuelson and Zeckhauser, 1988) and peer effect

(Duflo and Saex, 2002). It is plausible that individuals succumb to behavioral biases

because they lack financial literacy. For instance, if people do not understand their

financial choices or cannot grasp general concepts, they can easily make mistakes and fall

back on simple heuristics (Agnew, 2011). This study explores the indirect or mediation

effects of financial literacy—financial literacy may contribute via a reduction in

behavioral biases.

The empirical tests provide evidence as to the effect of financial literacy on the

financial behaviors or biases related to saving for retirement, by utilizing a dataset

generated by a survey conducted by Japan’s central bank (Bank of Japan), conducted

during 2015 on 25,000 Japanese respondents for a total of 51 questions. The richness of

data and information enables more rigorous empirical tests which shed new light on the

relationship between financial literacy and behaviors. The survey questions range across

a wide range of financial literacy questions, financial behaviors including retirement

saving and investment in financial products, as well as behavioral tendencies or heuristics.

The survey comprised 30 financial literacy questions ranging from inflation, interest,

spending, risks, pension, and insurance. As more than general financial literacy is required

for the decision of saving for future, I use the respondent’s answers to these 30 questions

to compile a comprehensive financial index using factor analysis. The primary analyses

are based on non-student respondents aged 20–60.

3
 The empirical results can be summarized as follows. First, similar to studies

conducted in the United States and some European countries, Japanese respondents are

not uniformly distributed in their understanding of financial information. For instance,

43% of respondents fail to understand the relationship between inflation and purchasing

power, and 45% fail to understand compound interest rates correctly. Second, financial

literacy is associated with comparison of alternatives when investing in financial products

such as stocks, mutual funds, mortgages, or life insurances. Such prudent behaviors help

respondents make choices that are less costly or better fit one’s needs or investment goals,

achieving the saving/investing goals more efficiently and effectively. Third, lack of

financial literacy predisposed an individual to biases relevant to financial decisions.

Respondents with lower financial literacy manifest a greater degree of present time or

loss aversion bias. Such biases may have influenced respondents to the effect that they

refrain from saving at all or fail to follow through with rational investment strategies.

Lastly, financial literacy helps one be better able to perceive post-retirement life style,

and to implement a saving/investment plan, after controlling for the effects of other

possible factors. This is conducive to a higher chance of successfully securing funds

necessary for post-retirement life. Additional tests also find that financial literacy not only

contributes directly to saving for retirement, but also indirectly a reduction of behavioral

biases, although the indirect effect is only modest compared with direct effect.

This study makes several contributions. While previous studies have

documented the positive effects of financial literacy on financial decisions, the

mechanisms in which financial literacy plays a role have not been fully explored. The

empirical results of this establish that financial literacy is associated with those behaviors

or biases in the different phases of saving decision for retirement—it increases one’s

4
awareness of post-retirement living needs and the ability to compare alternative financial

options, and reduces behavioral biases, which eventually directly and indirectly enhance

one’s ability to save for retirement. The results complement the findings of previous

studies that people with higher financial literacy are more likely to invest in the stock

market and accumulate greater net wealth (Behrman et al., 2012; Gustman et al., 2012;

Jappelli and Padula, 2013; Lusardi and Mitchell, 2008; van Rooij et al., 2012). This

connection can be explained by the results of this study that more financially literate

people are less subject to biases, allowing them to generate greater wealth from

investments. Previous research finds that more financially literate people tend to choose

mutual funds or take loans with lower costs or fees (Disney and Gathergood, 2013;

Klapper et al., 2013; Moore, 2003), which can also be attributed to making use of financial

knowledge to search for information and compare alternative products.

In addition, while most previous studies use only three or four questions

primarily related to inflation and interest to measure financial literacy, this study develops

a comprehensive measure, compared against most studies using only three or four simple

questions about interest and inflation to measure one’s level of financial literacy. An

important decision such as saving for retirement required more than general financial

literacy. Both theoretical and empirical results illustrate that multiple dimensions of

financial knowledge are relevant in assisting or guiding one’s prudent financial behaviors.

Some common knowledge of spending, financial products, markets, and the associated

risks are necessary for an informed decision-making process, at least in the decision to

save for the future.

Furthermore, the results in this study can have implications for the question of

whether financial education can be effective in promoting prudent financial behaviors.

5
There is little consensus in the literature on the efficacy of financial education (Hastings,

et al., 2013), even though positive effects of financial literacy have been reported. For

instance, Clark and d’Ambrosio (2008) show that while individuals attending employer-

sponsored seminars have intentions to improve savings behavior after attendance, they

do not necessarily follow through. Choi et al. (2002) find similar tendencies. The results

in this study can provide supporting evidence for financial education aiming to increase

financial knowledge regarding interest, inflation, risk-and-return, spending, insurance,

and pensions. Greater financial literacy via focused education may have the potential to

mitigate biases and prompt prudent financial decision-making, increasing one’s financial

welfare. Such questions await further research in the future.

Literature review and hypotheses

Literature review

Behrman et al. (2012) define financial literacy as “the ability to process

economic information and make informed decisions about household finances.” In

practice, questionnaires are used to quantify an individual’s financial literacy. In most

studies, an individual is asked to answer three or four questions regarding inflation and

interest rates (Choi et al., 2010; Disney and Gathergood, 2013; Gathergood et al., 2017;

Gaudecker, 2015; Jappelli and Padula, 2013; Klapper et al., 2013; Lusardi and Mitchell,

2008). Exceptions are Behrman et al. (2012), which asked 12 questions, and van Rooij et

al. (2012), which asked a total of 16 questions. The justification for using only a limited

number of questions to measure financial literacy is that knowledge pertaining to interest

rates can be a good predictor of one’s financial literacy. While some studies simply use

the percentage of correctly answered questions to proxy one’s financial literacy, some

studies, e.g., Gaudecker (2015), Klapper et al. (2013), van Rooij et al. (2012), use factor
6
analysis or principal component analysis to compute a composite index to be suited to

empirical tests. The current study uses an index based on a larger set of questions related

to financial knowledge in a variety of dimensions.

Previous studies have documented systematic variations in financial literacy that

can be explained by demographic and cognitive factors. In general, those with lower

financial literacy are more likely to be elderly, female, unemployed, less healthy

individuals or households (Agarwal et al., 2009; Calvet et al., 2009; Jappelli and Padula,

2013). Higher cognitive ability, manifested in higher mathematical performance during

the teens or higher education attainment, is also associated with higher financial literacy

(Gathergood and Weber, 2017; Jappelli, 2010; Jappelli and Padula, 2013). In addition,

people with home-ownership, or higher income, higher property value, or less consumer

debt are found to be more financially literate (Disney and Gathergood, 2013; Gathergood

and Weber, 2017). In this current study, the empirical studies account for these

demographic and socioeconomic factors as control variables in explaining an individual’s

financial literacy.

One of the most important questions in the literature involves the implications

of financial literacy. Previous studies conducted in different countries have consistently

reported positive effects of financial literacy on prudent financial behaviors. For instance,

more financially literate people are more likely to avoid the disposition effect (Dhar and

Zhu, 2006), plan for retirement (Lusardi and Mitchell, 2008), shun mutual funds with

expensive fees (Choi et al., 2010), use loans or other financing sources with lower interest

cost (Disney and Gathergood, 2013; Klapper et al., 2013; Moore, 2003), are less likely to

experience debt problems (Lusardi and Tufano, 2009), and diversify stock portfolios

(Gaudecker, 2015). Gaudecker (2015) also concludes that financial literacy may

7
supplement professional advice. These results are also consistent with the finding that

people who are more financially literate accumulate greater wealth (Behrman et al., 2012;

Gustman et al., 2012; Jappelli and Padula, 2013; Lusardi and Mitchell, 2008; van Rooij

et al., 2012). However, the mechanism through which financial literacy contributes to

financial behaviors or outcomes is not clear. This current study aims to fill the gap by

advancing some possible channels through which financial literacy can contribute to an

individual’s saving behavior for the future.

Hypotheses of the effects of financial literacy

This study focuses on an individual’s preparation for post-retirement life, which

is a substantive decision in one’s financial life and has been the subject of previous

research. How can financial literacy have an impact on the process of saving for

retirement? Daxhammer and Facsar (2012) suggest that there are three phases of a

decision-making process in a financially significant life situation (Loerwald and

Stemmann, 2016). Financial literacy, “the ability to process economic information and

make informed decisions about household finances” (Behrman et al., 2012), can play an

important role in the process, resulting in decisions with varying degrees of quality.

During the first phase of information perception, investors create a picture of

their environment to reduce the uncertainty in the process. The intensity of information

perception influences to what extent existing information can be used, or what additional

information from external sources is required and acquired via active searching. In the

context of saving/investing for retirement, investors need to be able to imagine their post-

retirement life, estimating the amount of living expenses needed to maintain a certain

desired living standard. Financial literacy, including knowledge of spending behaviors,

8
can help investors grasp their post-retirement living expenditures. Financial literacy can

also increase efficiency and reduce costs associated with information searching and

acquisition. Based on the aforementioned, the following hypothesis can be formulated:

H1: People with greater financial literacy are more able to be aware of their post-

retirement needs and know their estimated living expenditures for post-retirement life.

In the second phase of information processing and evaluation, investors consider

the relevant information to prepare for the decision. In order to make an informed decision,

investors need to compare alternative options in terms of return and cost. Decisions in

these matters can be considered substantial enough to merit one’s time and effort being

spent on information search and comparison. Therefore, a broad range of knowledge is

necessary when investors are making financial decisions in purchasing stocks or funds,

taking out a mortgage or buying long-term insurance policies. Comparing alternatives can

be beneficial by saving costs such as fees or finding a more suitable product to fit one’s

needs.

H2: People with greater financial literacy are more likely to compare alternatives when

making financial investments.

The final phase of decision making and its implementation then completes the

process of decision making. In the context of saving for retirement, this is the stage in

which an individual formulates and implements the plan to meet the financial needs post-

retirement. Financial knowledge related to markets, risk, investment products, and so on

helps one in implementing his or her saving plan.

However, it has been reported that attitude toward money influences behaviors

9
regarding saving for life post-retirement (MacFarland et al., 2004). Thaler and Shefrin

(1981) suggest that under-saving for retirement can be blamed on limited self-control and

procrastination. Self-control bias occurs when investors are not always consistently or

persistently pursuing a particular investment or saving goal. Self-control bias is related to

present time bias. People often act as if they are using a hyperbolic discount function

(Laibson, 1997), hence preferring spending in the present and may not make sufficient

saving for retirement and are thus more likely to suffer post retirement or due to an

unexpected situation. There is evidence to show that present bias induces higher levels of

credit card debt and lower savings (Laibson, 1997; Meier and Sprenger, 2010).

Another widely reported behavioral bias that may be relevant for saving for

retirement is loss aversion bias. Considering the context of stock investment, loss aversion

arises when losses are felt more than comparable gains. Based on insights into prospect

theory (Kahneman and Tversky, 1979), loss-aversive investors with stock gains behave

in a risk-averse manner and are concerned about losing their gains, but their risk attitude

changes as soon as they make losses. They tend to hold onto falling shares and the

respective loss, while making profits by prematurely selling their rising shares. Such

investors are vulnerable to increased risk-seeking in situations where losses have already

been incurred, possibly leading to greater losses (Loerwald and Stemmann, 2016). Loss

aversion bias can also lead to status quo bias, referring to the tendency to do nothing, and

thus influence one’s under-preparing for retirement (Samuelson and Zeckhauser, 1988).

In light of the research suggesting such biases as one reason behind poor savings,

this current study explores the possibility that financial literacy can reduce such bias. For

example, financially literate individuals can anticipate their future life from the

perspective of a life cycle and comprehend the importance of disciplined savings and

10
investment behavior; thus, they are less subject to present-time bias. If financially literate

investors are more capable of recognizing the vulnerability of prioritizing losses over

gains in financial decisions (less subject to loss aversion bias), they are more likely to

succeed in their preparation for post-retirement life. In fact, a laboratory experiment by

Agnew and Szykman (2005) suggest that lack of financial literacy may make individuals

susceptible to biases.

In summary, it can be hypothesized that financial literacy may directly contribute

to saving for retirement and that it may also contribute indirectly by reducing behavioral

biases, which has been suggested as a reason for the phenomenon of poor savings.

H3: Individuals with greater financial literacy are more likely to prepare for post-

retirement life (direct effect).

H4: People with greater financial literacy are less likely to display behavioral biases and

are therefore more likely to prepare for post-retirement life (indirect effect).

Sample and data

The empirical study investigates the hypotheses by tapping into the datasets

obtained from a national survey, conducted by Japan’s central bank (Bank of Japan) in

2015, on 25,000 Japanese regarding their financial literacy and financial behaviors. For

the purpose of this study, the empirical tests exclude respondents who are under the age

of 20 or those who are students, as they are less likely to be required to make significant

independent financial decisions.

Table 1 summarizes the socioeconomic characteristics for the sample. The

respondents are evenly distributed between genders. The average age of respondents is

aged 50, with 11.6% in their 20s, 19.2% in their 30s%, 18.0% in their 40s, 17.6% in their

11
50s, 20.4% in their 60s, and 13.3% in their 70s. In terms of occupation, the largest group

is respondents employed by companies (34%), followed by housewife or househusband

(22%), unemployed (16%), employed part-time (14%), self-employed (7%), civil servant

(4%) and others. With regard to final educational attainment, 50% of respondents had

graduated from 2-year or 4-year colleges, and 43% had graduated from a senior high

school or vocational school. The remaining 7% are those with only a degree of junior high

school (mandatory education) or graduate school.

Table 1

The survey also obtained information on the respondent’s annual household

income and financial wealth (including deposits, stocks, and financial assets). Thirty

seven percent reported annual income between 2.5 and 5 million yen, followed by 22%

between 5 and 7. 5 million yen, 20% below 2.5 million yen, and 13% between 7. 5 and

10 million yen. Those earning an income higher than 10 million yen account for only 8%

of the sample. On the other hand, household financial wealth is relatively more evenly

distributed. The proportion of each wealth category ranges from 7% to 22%. The lowest

wealth (zero) and highest wealth category (>20 million yen) are both 18%.

Financial Behaviors and biases

The remaining items in Table 1 show the financial behaviors to be investigated

in this study. This first set of variables pertain to whether the respondent is aware of the

amount necessary for the present and post-retirement. While 88% and 73% of respondents

are aware of their current one-month incomes and expenses, respectively, only a half of

the respondents indicated awareness of their post-retirement living expenditures. Among

the respondents who are aware of their post-retirement living estimates, 36% indicated

that they had a saving plan and only 27% have secured the necessary funds.

12
 The second set of variables is about whether one searches for information to

compare with alternative products when buying investment products such as stocks,

mutual funds, or foreign exchange products, or when taking out a loan with an amount

greater than one’s three-month living expenses. Only 63% of respondents make

comparisons when purchasing investment products, and 54% when taking out loans.

The third set of variables is related to behavioral biases. For present-time bias,

the survey question (Q1_10) asks “How much do you agree or disagree that the following

statement applies to you personally? If I had the choice of (1) receiving 100,000 yen now

or (2) receiving 110,000 yen in 1 year, I would choose (1), provided that I can definitely

receive the money.” The original question asks the respondent to choose from a scale of

1 (agree) to 5 (disagree). I recoded this variable by reversing the order of the scale so that

a higher value now indicates greater present-time bias. The average and median are 3.21

and 3, respectively. On average, Japanese respondents display no considerable biases.

For loss aversion bias, Q6 asks “Suppose that, if you invested 100,000 yen, you

would either get a capital gain of 20,000 yen or a capital loss of 10,000 yen at 50%

probability. What would you do? Choose (1) I would invest, or (2) I would not invest.”

The variable was recoded into a dummy, with the value of 1 if the respondent chose (2)

to indicate loss aversion bias, which is the tendency to place greater weights on losses

rather than on an equivalent amount of gains. On average, 79% of respondents display

loss-aversion bias.

Financial literacy index

Survey questions relating to financial literacy can be divided into six categories,

with each associated with knowledge relating to “inflation,” “interest,” “pension,”

“insurance,” “spending,” and “risk.” The original questions are presented in the appendix

13
at the end of this paper. Table 2 summarizes the performance of the respondents in terms

of their financial literacy for respondents over the age of 20. Table 2 reports the percentage

of respondents correctly answering the questions, ranging from 24% to 81%. The question

with the lowest correct answer rate, 24%, concerns the relationship between bond price

and interest rate. Questions about compound interest calculation also have low correct

answer rates, around 41% to 54%.

Table 2

An index of the financial literacy level is constructed as follows. First, for each

of the six categories, I tabulated the number of questions that a respondent correctly

answered, which is then standardized (with an average of zero and a variance of one) to

obtain a score that represents the respondent’s financial literacy relating to that category.

Take, for example, the category of interest rate knowledge, which includes eight questions.

The number of questions that a respondent correctly answers can range from zero to eight.

The mean and standard deviation for all 25,000 respondents are 3.89 and 2.5, respectively.

Thus, a respondent who correctly answers four questions obtains a standardized score of

0.042=(4−3.89)/2.5, which serves as a measure of this respondent’s financial literacy

regarding the interest rate. Similarly, I calculated the standardized scores for the

respondents’ financial literacy regarding pension, inflation, insurance, spending, and risk,

respectively.

Next, the financial literacy scores of the six categories are then combined by a

factor analysis using the iterative principal factor method, following van Rooij et al.

(2011; 2012). The advantage of factor analysis is that it accounts for correlation among

the questions (as well as the respondent’s performances) in different categories. The factor

analysis computes a composite index that is used as a measure of one’s financial literacy

14
for subsequent analyses. In Table 3, Panel A reports the financial literacy scores for each

of the six categories, as well as the composite index, for the 23,714 non-student adults

(over the age of 20). Also reported in Panel B are their pairwise correlation coefficients.

The indices are all positively and statistically correlated with one another at the 1% level.

Table 3

Empirical tests and results

Empirical tests are conducted to test the hypotheses regarding the relationship

between financial literacy and the aforementioned outcome variables. The main part of

the analysis uses the sample set excluding students, underaged respondents, and those

above sixty years, as the primary interest of this research is one’s preparation for

retirement. However, additional robustness tests will utilize other sample sets, for

example, those who are not full-time employees of companies or government agencies

(and may be eligible for employer-provided retirement pensions).

When explaining the causal effect of financial literacy on the individual's

outcome variable, the possibility of unobservable factors exists. Some measures are taken

to address the issue of endogeneity of financial literacy to the extent possible within the

data sets. First, the estimation accounts for control variables used and reported in previous

studies that may also affect the outcome variables, such as demographic characteristics.

Second, the instrument for financial literacy, which is not correlated to the outcome

variable, are used in the analyses. Commonly used instruments include education

attainments (Klapper et al., 2013; van Rooij et al., 2012), the mathematical ability during

the teens (Gathergood and Weber, 2017; Jappelli and Padula, 2013), the experience of

family members (Behrman et al., 2012: van Rooij et al., 2011), or the number of

15
universities or newspaper circulating in the neighborhood (Klapper et al., 2013).

Due to data constraints, the empirical study uses as the IV an indicator variable

for those who responded in the survey that their “parents or guardians teach them how to

manage finances.” The understanding of financial education during childhood is

significant for financial literacy. Those who have received financial education at home

may better understand the importance of finance and might be more motivated to acquire

financial knowledge as they grow up. Tang and Peter (2015) report that financial

education and parents' financial experience exert a positive impact on young adults'

financial knowledge. Moreno-Herrero et al. (2018) also assert that students' financial

literacy is associated with discussions on money matters with parents. Meanwhile,

financial education at home is an exogenous experience, since it is provided to the

respondents (the children) exogenously. However, it is also influenced by a responder’s

demographic background such as parents' wealth and education, which may affect their

children’s education and income. Therefore, by controlling for a family's socioeconomic

factors, financial education at home can be a plausible instrument for financial literacy.

IV regressions in the empirical analyses will control for the respondents’ demographic

characteristics such as education, occupation, and income level.

In this study, I present both results with and without using the instrumental

variable estimating method. Furthermore, the propensity score matching method is also

employed to match those with high financial literacy with comparable peers with low

financial literacy, based on a set of relevant covariates. Statistical tests are performed to

test whether there are differences in the financial outcomes between the high- and low-

literacy groups.

16
 The factors determining financial literacy level

Table 4 summarizes the financial literacy composite index stratified by the

respondent’s socioeconomic characteristics. Female respondents have statistically lower

financial literacy than males, consistent with previous findings based on different

countries. Financial literacy also increases with age, a result somewhat different from

previous studies which show older people are less financially literate (Agarwal et al.,

2009; Calvet et al., 2009; Jappelli and Padula, 2013). The cohort with the highest literacy

is those in their 60s, followed by those in their 70s, 50s, 40s, 30s, and 20s. This tendency

is similar when comparing the sub-indices. One possible explanation is that the literacy

index used in this study is based on a wider range of financial knowledge, which may

increase with age and life experiences.

Table 4

In terms of occupation, civil servants and unemployed respondents report higher

financial literacy index of 0.29 and 0.24, respectively. Civil servants probably perform

better because they possess a wider range of knowledge gained from their civil servant

examination preparations. Approximately three quarters of the unemployed respondents

are in their 60s and above, who, as previously indicated, are also more literate than cohorts

of other ages. Education also matters in that respondents with higher education

attainments are more financially literate. The last two columns show that financial literacy

is higher for those with greater household financial wealth or annual income. In general,

analysis of variance indicates significant differences in financial literacy among groups

defined by the aforementioned demographic and socioeconomic factors.

Table 5 presents the results of regression of a respondent’s composite financial

literacy index on his or her socioeconomic characteristics, including age, gender,

17
household annual income (or financial asset wealth), occupations, education attainments,

and areas of residence (broadly divided into nine areas). 1 Coefficients are estimated

based on robust standard errors. The results can also serve as references for the exogeneity

tests for financial literacy to be performed in the subsequent analyses.

Table 5

The first column shows that financial literacy is significantly associated with

one’s socioeconomic characteristics, consistent with univariate results. Female

respondents have significantly lower financial literacy. Financial literacy also increases

with age. Civil servants and the unemployed group are more financially literate compared

to the reference group of company employees, while other categories are either

significantly less financially literate (e.g. part-timers) or do not differ from company

employees. Furthermore, people with higher education degrees or greater household

income (or financial asset wealth) have significantly higher financial literacy than those

in the lowest income (wealth) cohort.

In column 2, the estimation is performed using respondents without experiences

of investing in financial assets such as stocks, funds, or foreign exchanges. The exclusion

is meant to account for possibility that respondents with investment experiences may

acquire more financial knowledge. The results remain similar to the first two columns.

1
These 9 areas are Tohoku (including prefectures of Aomori, Iwate, Miyagi, Akita,
Yamagata, Fukushima, and Hokkaido), Kanto (Ibaragi, Tochigi, Gunma, Saitama, Chiba,
Kanagawa, and Yamanashi), Tokyo, Chubu (Niigata, Toyama, Ishii, Fukui, Nagano, Gifu,
Shizuoka, Aichi, and Mie), Keihan (Kyoto and Osaka), Kinki (Shiga, Hyogo, Nara, and
Wakayama), Chugoku (Tottori, Shimane, Okayama, Hiroshima, and Yamaguchi),
Shikoku (Tokushima, Kagawa, Ehime, and Kochi), and Kyushu (Fukuoka, Saga,
Nagasaki, Kumamoto, Oita, Miyazaki, Kagoshima, and Okinawa).

18
Similarly, column 3 is estimated on respondents without mortgages in their household, as

those with a mortgage may also develop financial literacy, particularly interest-related

knowledge. The results remain similar. To compare with previous studies that focus on

household heads, Column 4 only includes male respondents in their 40s and 50s. People

in these categories are more likely to be the breadwinners in Japanese households. This

reduces sample size to only 3603 respondents. However, the results are basically similar

to other regressions.

In general, the results indicate that higher financial literacy is displayed by male,

senior, highly educated, and wealthier respondents. The results are primarily consistent

with previous studies, with the exception of age, which previous studies report as having

an adverse relationship with financial literacy, while this current study finds a positive

relationship for Japanese people.

The effects of financial literacy on awareness of financial needs

In this subsection, probit regressions are employed to test the first hypothesis by

estimating the effect of financial literacy on the dependent variables, which indicate if the

respondent is aware of his or her living expenses, in the present or for the future.

Exogenous control variables include the respondent’s socioeconomic characteristics. I

report the results with and without using the instrumental variable estimating method. As

described in the preceding section, the instrument is an indicator variable for those who

answered in the survey that their “parents or guardians teach them how to manage

finances”. Table 6 reports the results for both probit and IV probit regressions with the

maximum likelihood estimates based on robust standard errors.

Table 6

Columns 1 and 2 of Table 6 show the results for the dependent variable of being
19
aware of one-month expenses. The estimates in both columns show that people with

higher financial literacy are more aware of their short-term financial needs. The IV probit

estimates highlight a downward bias in the probit estimates. Also reported in the bottom

of the table is the estimate of a financial education instrument variable from the first-stage

regression, which indicates a positive and statistically significant impact on financial

literacy. The robust F-statistic is larger than the critical value of 16.38 for an actual Wald

test size of 10% suggested by Stock and Yogo (2005), rejecting the null hypothesis of

weak instruments. Conditional on the validity of the instruments, the second-stage

estimate implies that a one standard-deviation increase in the financial literacy index

increases the probability of being aware of short-term financial needs by 5.5%.

Columns 3 and 4 of Table 6 present the results for the dependent variable of

being aware of post-retirement living expenses. The results are similar to those in the first

two columns in that the IV probit estimate for financial literacy remains positive,

statistically significant, and larger than the probit estimate. The IV probit estimate for

financial literacy implies that a one standard-deviation increase in the financial literacy

index increases the probability of awareness of post-retirement financial needs by 7.1%.

As for other control variables, women seem more aware of financial needs than

men. The effects of other control variables are inconclusive since the coefficients are

somewhat different between probit and IV probit regressions.

The effects of financial literacy on financial prudence

In this subsection, probit regressions are employed to test the second

hypothesis—whether financially literate people behave more wisely in their financial

behaviors, i.e., search for information and comparing alternatives when making financial

investments. Two dependent variables are examined—the dummy variable for comparing
20
financial investment products and for comparing loans. Table 7 reports the results for both

probit and IV probit regressions, with the maximum likelihood estimates based on robust

standard errors.

Table 7

Both probit and IV probit regressions indicate that people with higher financial

literacy are more likely to compare financial products (Column 1 and 2) or loans (Column

3 and 4). In both models, the IV probit estimate for financial literacy is larger than the

probit estimate. The first-stage regression shows that the IV is relevant in both IV

regressions (Column 2 and 4). Although in Column 2, the robust F (=9.14) fails the Stock

and Yogo test for a Wald test size of 10%, the null hypothesis of weak instrument can still

be rejected if we tolerate a test size of 15% at most, for which the critical value is 8.96.

The marginal effects of financial literacy suggest that a one standard deviation increase

in financial literacy index increases the probability of comparing financial products

(loans) by 12.7% (7.8%).

On control variables, women also behave with more rationality than males, and,

surprisingly, older people are less likely to compare investment and insurance alternatives.

One possible explanation is the decline in cognitive ability as people get older. The effects

of other control variables are inconclusive since the coefficients are somewhat different

between probit and IV probit regressions.

The effects of financial literacy on financial biases

This subsection investigates whether financial literacy can alleviate finance-

related biases in relation to Hypothesis 4. Probit and IV probit methods are performed

when loss aversion bias is used as the dependent variable, while the ordered probit and

IV-ordered probit methods are performed for present-time bias (which is based on a 5-
21
point Likert scale). Table 8 reports the maximum likelihood estimates based on robust

standard errors.

Table 8

The results regarding loss aversion are inconclusive. In column 1, the probit

regression shows that financial literacy has a negative and significant association with

loss aversion bias, while in column 2, the IV probit regression shows that the negative

coefficient is not statistically significant at the 10% level. Still, in both models, the

negative coefficients of financial literacy are consistent with the hypothesis.

The results in column 3 and 4 suggest a statistically significant mitigating effect

of financial literacy on present-time bias. The IV-ordered probit estimate of financial

literacy is stronger than the ordered probit estimate bias. 2 The first-stage regression

shows that the IV is relevant and that the null hypothesis of weak instrument is rejected.

Conditional on the validity of the instruments, a one standard deviation increase in the

financial literacy index reduces the probability of belonging to the highest present-time

bias category by 7.5%.

Regarding control variables, the estimates show that women are more loss averse

than men but are less biased toward present consumption, while older people are more

loss averse than younger ones and are also more biased toward present consumption.

Occupation appears to be associated with biases, but the coefficients are difficult to

interpret. Finally, individuals with a higher household income and education present

lower behavioral biases.

2IV order probit regression is estimated by performing the mixed-process models suggested by
Roodman (2011).

22
 The effects of financial literacy on preparation for post-retirement life

This subsection tests the effects of financial literacy in relation to Hypothesis 3.

Reported here are results for using “securing funds for retirement” as the dependent

variable, while the results for using “having a plan” remains qualitatively the same. The

regressions include an additional explanatory variable regarding whether one has a

mortgage, as the ability to save for post-retirement life may be hindered by the obligation

of mortgage payment. Table 9 reports the probit and IV probit estimation results, with

maximum likelihood estimates based on robust standard errors. In the first two columns,

the estimates show that people with higher financial literacy are more likely to secure

retirement funds. The IV probit estimate for financial literacy is larger than the probit

estimate. The first-stage regression shows that the IV is relevant and that the null

hypothesis of weak instrument is rejected. The marginal effect of a one standard deviation

increase in financial literacy has a higher probability of securing funds for retirement by

1.8%. Additionally, both columns suggest that people with higher income are more likely

to secure funds for retirement, while people with mortgages are less prepared. For the

educational attainment variable, the coefficients change signs in different regressions and

are difficult to interpret.

Table 9

Furthermore, to test if financial literacy affects preparation for retirement

through mitigating behavioral biases, the same regressions are performed again on a

smaller set of sample with low-bias and high-bias respondents, separately. Since Table 8

indicates that financial literacy reduces present-time bias, I divide respondents into high-

bias group (those with above-median bias scores) and low-bias group (below-median). If

financial literacy affects preparation for retirement through mitigating bias, then the effect
23
of financial literacy on preparation for retirement should be more pronounced for the

high-bias group. Columns 3 and 4 of Table 9 report the IV probit regression results for

low-bias and high-bias respondents, respectively, to accounting for potential endogeneity

of financial literacy. In both columns, financial literacy has a significantly positive effect

on preparation for retirement, while it has a greater magnitude for high-bias individuals

(=1.112) than for low-bias individuals (=0.935). The estimates imply that a one standard

deviation increase in financial literacy raises the probability of securing funds for

retirement by 2.70% for high-bias individuals and by 1.14% for low-bias individuals.

Furthermore, structural equation modeling (SEM) can be employed to test the

magnitude and significance of the direct and indirect effect of financial literacy. SEM has

the advantage of using full information even in the absence of missing values, as well as

of being able to treat mediation models with multiple mediation factors (which applies to

this current study). SEM is performed to estimate the following equations, where M1 and

M2 denote present-time bias and loss aversion bias, respectively, FL financial literacy,

SAVE the dummy for those who having a plan (or secured funds) for retirement, and

controls are a set of covariates including gender, age, having a loan, educational

attainment, occupation, annual income, and areas. Bootstrapped standard errors are

obtained by performing 200 replications.

𝑀1 = 𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡 + 𝑎1 𝐹𝐿 + 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝑒𝑟𝑟𝑜𝑟 (1)

𝑀2 = 𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡 + 𝑎2 𝐹𝐿 + 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝑒𝑟𝑟𝑜𝑟 (2)

𝑆𝐴𝑉𝐸 = 𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡 + 𝑐𝐹𝐿 + 𝑏1 𝑀1 + 𝑏2 𝑀2 + 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝑒𝑟𝑟𝑜𝑟 (3)

In Table 10, the first column reports the results by using “having a retirement”

for the dependent variable SAVE. The table only presents results for the variables of

24
interest, due to limitations of space. Financial literacy is associated with lower present-

time bias and loss aversion and at a significant level, with coefficients of −0.181 and

−0.111, respectively. At the same time, present-time bias and loss aversion bias have a

significantly negative effect on having a plan, with coefficients of −0.018 and −0.081,

respectively. The indirect effect of financial literacy accounts for 16.48% of the total

effect, while the direct effect is 83.52%. The second column reports the results by using

“securing funds” for the dependent variable SAVE. The results remain similar; however,

the magnitudes of the coefficients all become lower than in the first column, suggesting

the more challenging task of securing funds for retirement than having a plan. The indirect

effect of financial literacy through mitigating behavioral biases now accounts for a larger

proportion (75.77%) of the total effect. This suggests that restraining behavioral biases

may play a more important role in order to implement the retirement and achieve the goal.

Finally, it should be noted that the present-time bias is measured using a Likert scale from

1 to 5, which do not have a cardinal meaning.

Table 10

Additional robust tests

As those working for companies or government agencies in Japan are statutorily

subject to defined benefit pension plans provided by their employers, they can expect to

receive pension payments after retirement. Therefore, they may plan to save for their post-

retirement life less proactively, regardless of their financial literacy level. The first

additional test repeats analyses similar to those in Table 9 by further investigating a subset

of samples by excluding company employees and civil servants. The results are

qualitatively similar. SEM analysis is also performed for this sample subset, with the

results similar to those reported in Table 10.

25
 The second set of additional tests employ the propensity score matching method

to compare respondents with high financial literacy and those with low financial literacy.

In the first test, respondents are divided into a high (low) financial literacy group if their

literacy index is larger (smaller) than the median. Then, using a propensity score matching

method (with replacement), each individual in the high literacy group is matched with a

nearest-neighbor peer from the low literacy group, based on a set of covariates, such as

gender, age group, education attainments, occupation, income level, and residing areas.

Multiple matches are allowed when they have the same propensity score and are tied with

the nearest-neighbor. Comparison is then made between the high and low literacy groups

regarding their financial behaviors. In Table 11, the first column reports the average

treatment effect on the treated (ATET). The standard error is computed by taking into

account the fact that the propensity score is estimated, relying on the work of Abadie and

Imbens (2016).

Table 11

The difference in financial literacy is statistically significant. The high financial

literacy group is more likely to compare financial products or loans than the low literacy

group, at significant levels. The former group is also less subject to behavioral biases than

the low literacy group at significant levels. Finally, the high financial literacy group is

more likely to be prepared for retirement.

In column (2), a high literacy group only contains those individuals in the top

25% of financial literacy. Each individual in the high literacy group is then matched with

a nearest-neighbor peer from the low literacy group (below the median) in the same

manner as previously described. The ATET results for this pair comparison provide the

same conclusion as those reported in the first column; however, the differences (i.e.,

26
ATETs) are becoming larger than those in the first column. Since the high financial

literacy group in column 2 contains only the top 25% individuals with high financial

literacy, the more pronounced ATETs further confirm the effects of financial literacy.

Conclusion

While the existing literature has extensively reported positive effects of financial

literacy on substantive household financial matters such as saving for retirement, the

mechanisms through which financial literacy plays a role are yet fully explored. This

study fills the gap by investigating the effects of financial literacy on the three phases of

decision-making processes: information perception, information search and evaluation,

and decision making and implementation. The primary financial setting is

saving/investing for retirement, for which the majority of people have failed to prepare

sufficiently, as reported in previous studies conducted in different countries. Analyzing a

large sample of Japanese adults, the empirical results indicate that, even after accounting

for various control variables, financial literacy has significantly positive effects on one’s

awareness of necessary post-retirement living expenditures, the ability to compare

alternative products when making significant financial decisions, displaying fewer

behavioral biases relating to financial decisions, and successfully preparing for retirement.

The results also indicate that financial literacy can not only directly contribute to saving

for retirement, but also indirectly by reducing behavioral biases. The results are consistent

with the findings of previous studies that people with higher financial literacy are better

off in their financial outcomes, such as accumulating greater net wealth (Behrman et al.,

2012; Gustman et al., 2012; Jappelli and Padula, 2013; Lusardi and Mitchell, 2008; van

Rooij et al., 2012).

27
 The results of this study may have implications for the importance of financial

education. Even though there is little consensus as to the efficacy of financial education

(Hastings et al., 2013), the results of this study suggest that it can be effective in terms of

mitigating behavioral biases commonly observed in financial decisions, or guiding one to

make an informed decision via more information acquisition and appropriate evaluation.

Whether financial education programs designed to that effect can change one’s saving

behaviors awaits further research in the future.

28
Appendix: Survey questions used to construct an individual’s financial literacy.
1. Financial Literacy on Inflation
Q20. Imagine that the interest rate on your savings account was 1% per year and inflation was
2% per year. After 1 year, how much would you be able to buy with the money in this account?
Choose only one answer. (1) More than today. (2) Exactly the same. (3) Less than today. (4)
Don't know.
Q21_1. Please indicate whether you think the following statements are true or false. “High
inflation means that the cost of living is increasing rapidly.” Choose one answer for each item.
(1) True. (2) False. (3) Don’t Know.

2. Financial Literacy on Insurance

Q25. Which of the following statements on the basic function of insurance is appropriate? Choose
only one answer. (1) Insurance is effective when a risk occurs with high frequency, causing a
large loss. (2) Insurance is effective when a risk occurs with low frequency, causing a large
loss. (3) Insurance is effective when a risk occurs with high frequency, causing a small loss.
(4) Insurance is effective when a risk occurs with low frequency, causing a small loss. (5) Don't
know.
Q26. When a 50-year-old man reviews his life insurance policy (whole life insurance) after his
children have become financially independent, which of the following statements is
appropriate? Suppose that other circumstances have not changed. Choose only one answer. (1)
He should consider increasing the death benefit. (2) He should consider decreasing the death
benefit. (3) There is no need to review the policy in particular. (4) Don't know.
Q28. Which of the following statements on insurance is inappropriate? Choose only one answer.
(1) You need to pay national pension contributions if you are aged 20 or over, even if you are
a student. (2) The damage caused by an automobile accident will be fully covered by the
automobile liability insurance. (3) You should review the necessity of life insurance and the
amount of coverage of insurance according to changes in circumstances of family members
and yourself. (4) Health insurance may not cover pre-existing medical conditions that you had
before purchasing the insurance policy. (5) Don't know.

3. Financial Literacy on Interest Rate

Q12. Taro and Hanako are the same age. At age 25 Hanako began saving 100,000 yen per year
and continued to save the same amount annually thereafter. Meanwhile, Taro did not save
money at age 25, but began saving 200,000 yen per year at age 50. When they are aged 75,
which of them will have more money saved? Choose only one answer. (1) They would each
have the same amount because they put away exactly the same amount. (2) Taro, because he
saved more each year. (3) Hanako, because she has put away more money. (4) Hanako, because
her money has grown for a longer time at compound interest. (5) Don't know.
Q18. Suppose you put 1 million yen into a savings account with a guaranteed interest rate of 2%
per year. If no further deposits or withdrawals are made, how much would be in the account
after 1 year, once the interest payment is made? Disregard tax deductions. Answer with a whole
number.
Q19. Then, how much would be in the account after 5 years? Disregard tax deductions. Choose
only one answer. (1) More than 1.1 million yen. (2) Exactly 1.1 million yen. (3) Less than 1.1
million yen. (4) Impossible to tell from the information given. (5) Don't know.
Q21_2. Please indicate whether you think the following statements are true or false. “When
compared, a 15-year mortgage typically requires higher monthly payments than a 30-year loan,
but the total interest paid over the life of the loan will be less.” Choose one answer for each
item. (1) True. (2) False. (3) Don’t Know.
Q22. If interest rates rise, what will typically happen to bond prices? Choose only one answer. (1)
They will rise. (2) They will fall. (3) They will stay the same. (4) There is no relationship

29
 between bond prices and the interest rate. (5) Don't know.
Q23. Which of the following is appropriate as an action to take when investing (making deposits,
etc.) or borrowing funds at a time of interest rate rise? Choose only one answer. (1) Investing
and borrowing at fixed interest rates. (2) Investing at a fixed interest rate and borrowing at a
floating interest rate. (3) Investing at a floating interest rate and borrowing at a fixed interest
rate. (4) Investing and borrowing at floating interest rates. (5) Don't know.
Q30. Which of the following statements on mortgages is appropriate? Choose only one answer.
(1) It is far less costly to continue living in a rented house for your whole life than buying a
house with a loan. (2) Mortgages can be repaid by either the equal payment method or the
equal principal payment method, but the total repayment is the same for both methods. (3)
Mortgages are offered with either a floating interest rate or a fixed interest rate, and those with
a fixed interest rate are always more advantageous than those with a floating interest rate. (4)
In order to decrease the total mortgage repayment, it is effective to prepare as much down
payment as possible and make advanced repayments to the extent possible. (5) Don't know.
Q31. Suppose you owe 100,000 yen on a loan and the interest rate you are charged is 20% per
year compounded annually. If you didn't pay anything off, at this interest rate, how many years
would it take for the amount you owe to double? Choose only one answer. (1) Less than 2
years. (2) At least 2 years but less than 5 years. (3) At least 5 years but less than 10 years. (4)
At least 10 years. (5) Don't know.

4. Financial Literacy on Pension

Q27_1. About public pensions you are qualified to receive, are you aware of the type of public
pension that covers you? Choose (1) Yes. (2) No.
Q27_2. About public pensions you are qualified to receive, are you aware of the category of
insured person you fall into? Choose (1) Yes. (2) No.
Q27_3. About public pensions you are qualified to receive, are you aware of the required number
of years of paying contributions in order to qualify for pension benefits.? Choose (1) Yes. (2)
No.
Q27_4. About public pensions you are qualified to receive, are you aware of the amounts of
pension you are qualified to receive.? Choose (1) Yes. (2) No.
Q27_5. About public pensions you are qualified to receive, are you aware of the age at which you
will start receiving the pension? Choose (1) Yes. (2) No.

5. Financial Literacy on Risk

Q14. Which of the following is inappropriate as an action to take when concluding a contract?
Choose only one answer. (1) Reconsidering whether the contract is truly necessary. (2)
Checking whether cancellation of the contract is possible and whether a penalty is charged for
doing so. (3) Concluding a contract based on a detailed explanation from the service provider,
and carefully reading the contract document later. (4) Seeking advice from a third party as
needed when concluding a contract. (5) Don't know.
Q15. Which of the following is inappropriate as a behavior to avoid being involved in financial
trouble? Choose only one answer. (1) Avoiding disclosing your personal information as much
as possible. (2) Making an effort to acquire financial and economic knowledge. (3) Trusting
and leaving the entire matter to the service provider when it is difficult to make a decision. (4)
Checking the user reviews of the product you are planning to purchase. (5) Don't know.
Q16. Which of the following is inappropriate as an action related to Internet transactions? Choose
only one answer. (1) I updated the security software to the latest version. (2) I received an e-
mail, but I did not open it since it was sent from an unknown address. (3) I made a bank transfer
by using a computer at an Internet café. (4) I checked many times to make sure that the
information I entered had no errors. (5) Don't know.
Q21_3. Please indicate whether you think the following statements are true or false. “An

30
 investment with a high return is likely to be high risk”. Choose one answer for each item. (1)
True. (2) False. (3) Don’t Know.
Q21_4. Please indicate whether you think the following statements are true or false. “Buying a
single company's stock usually provides a safer return than a stock mutual fund.” Choose one
answer for each item. (1) True. (2) False. (3) Don’t Know.
Q33. Which of the following statements on the types of deposits protected up to 10 million yen
under Japan's deposit insurance system is appropriate? Choose only one answer. (1) Only
ordinary deposits are protected. (2) Ordinary deposits and time deposits are protected. (3) All
types of deposits including ordinary deposits, time deposits, and foreign currency deposits are
protected. (4) No deposit is protected due to the principle of self-responsibility. (5) Don't know.
Q36. Which of the following is inappropriate as behavior or attitude when determining whether
to purchase an unfamiliar financial product? Choose only one answer. (1) Collecting
information to make sure that the product is not frequently causing trouble and no warning has
been issued by a public institution. (2) Collecting information from the Internet, books, and
several sellers and comparing the product with other products. (3) Consulting with an
institution, agency, etc., that provides information from a neutral standpoint and receiving
advice. (4) Purchasing the product if the seller tells you that you can expect a high return. (5)
Don't know.
Q37. Which of the following is appropriate as an action to take when considering purchase of a
financial product with a complicated structure? Choose only one answer. (1) Purchasing the
product if it is selling well, even if you do not understand its structure clearly. (2) Purchasing
the product if you can trust the financial institution providing the product, even if you do not
understand its structure clearly. (3) Purchasing the product if you can expect a high return,
even if you do not understand its structure clearly. (4) Purchasing the product if you understand
its structure and find no problem. (5) Don't know.
Q38. Which of the following is inappropriate as a consultant office or a system to be used when
trouble occurs in relation to a contract for a financial product? Choose only one answer. (1)
Consumer center. (2) Financial alternative dispute resolution (ADR) system. (3) Rating
company. (4) Attorney at law.

6. Financial Literacy on Spending

Q4. Which of the following statements on household behavior is inappropriate? Choose only one
answer. (1) Managing income and expenditure by keeping a household account book or the
like. (2) Deciding on expenditure after considering whether it is truly necessary and whether
there is enough income. (3) Saving some money out of income by transferring a fixed amount
of income into a savings account or the like. (4) Frequently using installment payment plans
of credit cards in order to defer payment. (5) Don't know.
Q5. Which of the following statements on family budget management and credit cards is
inappropriate? Choose only one answer. (1) Using credit cards in a well-planned manner
according to income. (2) Any unsettled credit card payment is practically a debt. (3) A credit
card fee (interest) is charged for revolving payments but not for installment payments. (4)
Failure to pay the credit card charge may cause credit card transactions to be declined. (5)
Don't know.
Q13. What are the so-called three major expenses in life? Choose only one answer. (1) Living
expenses for your lifetime, children's educational expenses, and your medical expenses. (2)
Children's educational expenses, costs of buying a house, and living expenses for your
retirement. (3) Costs of buying a house, your medical expenses, and costs of nursing care for
your parents. (4) Don't know.

31
References

Abadie, A., Imbens, G. 2016. Matching on the estimated propensity score. Econometrica.
84(2), 781–807.
Agarwal, S., Driscoll, J., Gabaix, X., Laibson, D. 2009. The age of reason: financial
decisions over the lifecycle and implications for regulation. Brookings Papers on
Economic Activity. 2, 51–117.
Agnew, J. R., 2011. Pension participant behavior. In H. K. Baker and J. R. Nofsinger (eds)
Behavioral Finance. John Wiley & Sons, Inc.
Agnew, J. R., Szykman, L.R., 2005. Asset allocation and information overload: The
influence of information display, asset choice, and investor experience. Journal of
Behavioral Finance. 6(2), 57–70.
Behrman, J., Mitchell, O., Soo, C.K., Bravo, D., 2012. How financial literacy affects
household wealth accumulation. American Economic Review. 102, 300–304.
Calvet, L., Campbell, J., Sodini, P., 2009. Measuring the financial sophistication of
households. American Economic Review. 99, 393–398.
Choi, J.J., Laibson, D., Brigitte M., 2010. Why does the law of one price fail? An
experiment on index mutual funds. Review of Financial Studies. 23(4), 1405–1432.
Choi, J.J., Laibson, D., Metrick, A., 2002. How does the internet affect trading? Evidence
from investor behavior in 401(k) plans. Journal of Financial Economics. 64(3), 397–421.
Clark, R., d’Ambrosio, M., 2008. Adjusting retirement goals and savings behavior: The
role of financial education. In: A. Lusardi (ed) Overcoming the Saving Slump: How to
Increase the Effectiveness of Financial Education and Savings Programs. University of
Chicago Press.
Daxhammer, R. J., Facsar, M., 2012. Behavioral finance. Verhaltenswissenschaftliche
Finanzmarktforschung im Lichte begrenzt rationaler Marktteilnehmer. Constance: UVK.
Dhar, R., Zhu, N., 2006. Up close and personal: investor sophistication and the disposition
effect. Management Science. 52(5), 726–740.
Disney, R., Gathergood, J., 2013. Financial literacy and consumer credit portfolios.
Journal of Banking and Finance. 37, 2246–2254.
Duflo, E., Saez, E., 2002. Participation and investment decisions in a retirement plan: The
influence of colleagues’ choices. Journal of Public Economics. 85(1), 121–48.

32
Gathergood, J., Weber, J., 2017. Financial literacy, present bias and alternative mortgage
products. Journal of Banking and Finance. 78, 58–83.
Gaudecker, Hans-Martin, 2015. How does household portfolio diversification vary with
financial literacy and financial advice? Journal of Finance. 70 (2), 489-507.
Gerardi, K., Goette, L., Meier, S., 2010. Financial literacy and subprime mortgage
delinquency: evidence from a survey matched to administrative data. Federal Reserve
Bank of Atlanta Working paper.
Gustman, A.L., Steinmeier, T.L., Tabatabai, N., 2012. Financial knowledge and financial
literacy at the household level. American Economic Review: Papers & Proceedings.
102(3), 309–313.
Hastings, J.S., Madrian, B.C., Skimmyhorn, W.L., 2013. Financial literacy, financial
education and economic outcomes. Annual Review of Economics. 5, 347–373.
Jappelli, T., 2010. Economic literacy: An international comparison. Economic Journal.
120(548), 429–451.
Jappelli, T., Padula, M., 2013. Investing in financial literacy and saving decisions. Journal
of Banking and Finance. 37, 2779–2792.
Kahneman, D., Tversky, A., 1979. Prospect theory: An analysis of decision under risk.
Econometrica. 47(2), 263–91.
Klapper, L., Lusardi, A., Panos, G.A., 2013. Financial literacy and its consequences:
evidence from Russia during the financial crisis. Journal of Banking and Finance. 37,
3904–3923.
Laibson, D., 1997. Golden eggs and hyperbolic discounting. Quarterly Journal of
Economics. 112 (2), 443–478.
Loerwald, D., Stemmann, A., 2016. Behavioral finance and financial literacy:
Educational implications of biases in financial decision making. In: Aprea C. et al. (eds)
International Handbook of Financial Literacy. Springer, Singapore
Lusardi, A., Mitchell, O.S., 2007. Financial literacy and retirement preparedness:
evidence and implications for financial education. Business Economics. 42 (1), 35–44.
Lusardi, A., Mitchell, O.S., 2008. Planning and financial literacy: how do women fare?
The American Economic Review. 98(2), Papers and Proceedings of the One Hundred
Twentieth Annual Meeting of the American Economic Association. 413–417.
Lusardi, A., Tufano, P., 2009. Debt literacy, financial experience and over-indebtedness.

33
NBER Working Paper.
MacFarland, D.M., Marconi, C.D., Utkus, S.P., 2004. Money attitudes’ and retirement
plan design: One size does not fit all,” In: Mitchell O.S. and Utkus S.P. (eds) Pension
Design and Structure: New Lessons from Behavioral Finance. Oxford University Press,
New York.
Meier, S., Sprenger, C., 2010. Present-biased preferences and credit card borrowing.
American Economic Journal. 2 (1), 193–210.
Moreno-Herrero, D., Salas-Velasco, M., Sanchez-Campillo, J., 2018. Factors that
influence the level of financial literacy among young people: The role of parental
engagement and students' experiences with money matters. Children and Youth Services
Review. 95, 334–351.
Moore, D., 2003. Survey of financial literacy in Washington State: knowledge, behavior,
attitudes, and experiences. Washington State University Social and Economic Sciences
Research Center Technical Report. 03–39.
Roodman, D., 2011. Fitting fully observed recursive mixed-process models with cmp.
The Stata Journal 11(2), 159-206.
Samuelson, W., Zeckhauser, R., 1988. Status quo bias in decision making. Journal of Risk
and Uncertainty. 1(1), 7–59.
Stock, J.H., Yogo, M., 2005. Testing for weak instruments in linear IV regression. In Stock
J.H. and Andrews D.W.K. (eds) Identification and Inference for Econometric Models:
Essays in Honor of Thomas J. Rothenberg, Cambridge University Press.
Tang, N., Peter, P.C., 2015. Financial knowledge acquisition among the young: the role
of financial education, financial experience, and parents' financial experience. Financial
Services Review, 24, 119–137.
Thaler, R., Shefrin, H.M., 1981. An economic theory of self-control. Journal of Political
Economy. 89(2), 392–406.
The Economist, “Financial education. Poor young things” July 12th, 2007.
Von Gaudecker, H., 2015. How does household portfolio diversification vary with
financial literacy and financial advice? Journal of Finance. 70(2), 489–507.
van Rooij, M., Lusardi, A., Alessie, R.J.M., 2011. Financial literacy and stock market
participation. Journal of Financial Economics. 101 (2), 449–472.
van Rooij, M., Lusardi, A., Alessie, R.J.M., 2012. Financial literacy, retirement planning

34
and household wealth. Economic Journal. 122, 449–478.

35
Table: 1: The socioeconomic characteristics and financial behaviors of the respondents to the financial
literacy survey conducted by the Bank of Japan in 2015. Respondents are non-student adults over 20 years
of age. The column titled “item” refers to the corresponding question number in the original survey. Mean
and median are reported for the items, most of which pertain to binary information with the exception of
Q43 (age) and Q1-10 (present-time bias).
Item Description No. Mean Median
Q42 Female respondents 23714 0.51 1
Q43 Age 23714 50.2 50
Q44 Occupation employed by a company 23714 0.34 0
Q44 Occupation civil servant 23714 0.04 0
Q44 Occupation self-employed 23714 0.07 0
Q44 Occupation part-timers 23714 0.14 0
Q44 Occupation house-work 23714 0.22 0
Q44 Occupation unemployed 23714 0.16 0
Q46 Education junior high school 23714 0.03 0
Q46 Education senior high school 23714 0.32 0
Q46 Education vocational school 23714 0.11 0
Q46 Education 2-year college 23714 0.12 0
Q46 Education 4-year college 23714 0.38 0
Q46 Education graduate school 23714 0.04 0
Q50 Household annual income zero 19267 0.03 0
Q50 Household annual income <2.5 million yen 19267 0.17 0
Q50 Household annual income >2.5 and <5 million 19267 0.37 0
Q50 Household annual income >5 and <7.5 million 19267 0.22 0
Q50 Household annual income >7.5 and <10 million 19267 0.13 0
Q50 Household annual income >10 and <15 million 19267 0.06 0
Q50 Household annual income >15 million 19267 0.02 0
Q51 Household financial asset zero 15824 0.18 0
Q51 Household financial asset <2.5 million yen 15824 0.22 0
Q51 Household financial asset >2.5 and <5 million 15824 0.16 0
Q51 Household financial asset >5 and <7.5 million 15824 0.08 0
Q51 Household financial asset >7.5 and <10 million 15824 0.07 0
Q51 Household financial asset >10 and <20 million 15824 0.11 0
Q51 Household financial asset >20 million 15824 0.18 0
Q3-2 Respondents aware of one-month expenses 13820 0.73 1
Q8-1 Respondents aware of necessary living expenses for retirement 13820 0.50 1
Q9-1 Respondents planning for post-retirement 13820 0.36 0
Q10-1 Respondents securing post-retirement living expenses 13820 0.27 0
Q29 Respondents comparing alternatives taking out loans 5150 0.54 1
Q32 Respondents comparing alternatives buying financial products 6099 0.63 1
Q1-10 Present-time bias (1 for the lowest and 5 highest) 23714 3.21 3
Q6 Respondents indicating loss aversion bias 23714 0.79 1
Table 2: Distributions of the respondents’ financial literacy for all non-student respondents over 20 years of
age. The survey questions, presented in the Appendix at the end of this paper, are divided into six categories:
“inflation,” “interest,” “pension,” “insurance,” “spending,” and “risk.” The rightmost column reports the
percentage of respondents correctly answering the questions. The item column refers to the corresponding
question number in the original survey.
Item No. %
% respondents correctly answering questions about inflation
Q20 Question on inflation and purchasing power 23714 0.57
Q21-1 Question on inflation 23714 0.62
% respondents correctly answering questions about insurance
Q25 Question on insurance mechanism 23714 0.47
Q26 Question on adjustment of insurance in response to family structure change 23714 0.52
Q28 Question on insurance 23714 0.61
% respondents correctly answering questions about interest rate
Q12 Question on compound interest 23714 0.54
Q18 Question on deposit interest 23714 0.66
Q19 Question on compound interest 23714 0.43
Q21-2 Question on mortgage interest 23714 0.70
Q22 Question on bond price and interest rate 23714 0.24
Q23 Question on changes in interest rate 23714 0.45
Q31 Question on compound interest 23714 0.41
Q30 Question on mortgage 23714 0.52
% respondents correctly answering questions about spending
Q4 Question on income/spending management 23714 0.55
Q5 Question on use of credit card 23714 0.47
Q13 Question on the three primary categories of living expenses 23714 0.49
% respondents correctly answering questions about risk
Q14 Question on entering contracts 23714 0.67
Q15 Question on keeping away from financial troubles 23714 0.73
Q16 Question on troubles involving transactions on internet 23714 0.81
Q21-3 Question on risk/return relationship 23714 0.76
Q21-4 Question on diversification of asset allocation 23714 0.47
Q33 Question on deposit insurance 23714 0.44
Q36 Question on avoidance of financial troubles 23714 0.60
Q38 Question on hot lines for people with financial troubles 23714 0.74
Q37 Questions on buying complicated financial products 23714 0.64
% respondents correctly answering questions about pension
Q27-1 Question on one's public pension type 23714 0.66
Q27-2 Question on one's public pension status 23714 0.44
Q27-3 Question on the eligibility for pension payment 23714 0.47
Q27-4 Question on one's pension payment amount 23714 0.38
Q27-5 Questions on one's pension payment age 23714 0.47
Table 3: Financial literacy indices and pairwise correlations. Panel A reports the performance measures for
a respondent’s financial literacy. Financial literacy for each category is measured by the standardized number
of questions correctly answered by the respondent in that given category. The financial literacy composite
index is obtained by performing a factor analysis on the six standardized measures using the iterative
principal factor method. Panel B reports the pairwise correlation coefficients for these financial literacy
measures, with * indicating significance at 1%.
Panel A No. Mean Std. Dev. Min Max
Financial literacy on pension 23714 0.052 0.994 −1.221 1.413
Financial literacy on inflation 23714 0.024 0.996 −1.376 0.989
Financial literacy on insurance 23714 0.031 0.996 −1.381 1.251
Financial literacy on interest rate 23714 0.027 0.995 −1.560 1.644
Financial literacy on spending 23714 0.015 0.999 −1.394 1.400
Financial literacy on risk 23714 0.025 0.990 −2.114 1.180
Financial literacy composite index 23714 0.024 0.824 −1.952 1.876
Panel B
Pairwise correlations pension inflation insurance interest spending risk Composite
index
Literacy on pension 1
inflation 0.443* 1
insurance 0.471* 0.498* 1
interest rate 0.510* 0.641* 0.677* 1
spending 0.315* 0.302* 0.472* 0.468* 1
risk 0.502* 0.541* 0.681* 0.718* 0.580* 1
composite index 0.533* 0.850* 0.599* 0.910* 0.249* 0.578* 1
Table 4: The descriptive statistics of financial literacy for all non-student respondents over 20 years of age. The average financial literacy composite index is reported and
stratified by the respondent’s gender, age, occupation, educational attainment, household financial wealth, and household income. The lower part reports the results for the
analysis of variance testing and whether there are no differences among the stratified groups.
By gender By age By occupation By education By financial wealth By household income
Mean No. Mean No. Mean No. Mean No. Mean No. Mean No.
Male 0.209 11,622 20s −0.430 2,759 1 0.029 8,044 1 −0.506 673 1 −0.364 2,844 1 −0.353 506
Female −0.153 12,092 30s −0.230 4,557 2 0.289 874 2 −0.085 7,748 2 −0.031 3,416 2 −0.098 3,321
40s −0.086 4,245 3 0.141 1,746 3 −0.227 2,545 3 0.129 2,458 3 0.068 7,193
50s 0.178 4,162 4 −0.243 3,473 4 −0.088 2,778 4 0.187 1,316 4 0.135 4,149
60s 0.309 4,848 5 −0.061 5,219 5 0.226 8,956 5 0.301 1,178 5 0.267 2,414
70s 0.299 3,143 6 0.238 3,898 6 0.390 978 6 0.394 1,749 6 0.327 1,278
7 0.160 460 7 −0.353 36 7 0.582 2,863 7 0.411 406
Total 0.024 23,714 Total 0.024 23,714 Total 0.024 23,714 Total 0.024 23,714 Total 0.135 15,824 Total 0.092 19,267
Analysis of variance testing the differences among groups
F P-value F P-value F P-value F P-value F P-value F P-value
1199.3 0.000 538.5 0.000 141.7 0.000 254.5 0.000 464.4 0.000 109.6 0.000

20s in the 20s 1 employed 1 < senior high school 1 (million yen) 1 zero (million yen)
30s in the 30s 2 civil servant 2 senior high school 2 <2.5 2 <2.5
40s in the 40s 3 self-employed 3 vocational 3 >2.5 and <5 3 >2.5 and <5
50s in the 50s 4 part-timers 4 2-year college 4 >5 and <7.5 4 >5 and <7.5
60s in the 60s 5 house-work 5 4-year college 5 >7.5 and <10 5 >7.5 and <10
70s in the 70s 6 unemployed 6 graduate 6 >10 and <20 6 >10 and <15
7 others 7 others 7 >20 7 >15
Table 5: Ordinary least squares regressions of a respondent’s composite financial literacy index on the
socioeconomic characteristics, including age, gender, household annual income, occupation, educational
attainment, and areas of residence. Column (1) reports results for all non-student respondents aged over 20
years; column (2) reports results for a smaller subset for those without experience in investing in financial
assets such as stocks, funds, or foreign exchanges; column (3) for those without loans in their household;
and column (4) for male respondents in their 40s and 50s. The coefficients are estimated based on robust
standard errors.
(1) (2) (3) (4)
Dependent variable Financial literacy Financial literacy Financial literacy Financial literacy
Coef. p-value Coef. p-value Coef. p-value Coef. p-value
Female −0.266 0.000 −0.214 0.000 −0.267 0.000
Age 0.801 0.000 0.663 0.000 0.831 0.000 0.757 0.000
Occupation
company employee
civil servant 0.083 0.003 0.133 0.001 0.127 0.001 0.043 0.289
self-employed −0.015 0.504 0.040 0.189 −0.038 0.179 0.022 0.576
part-timers −0.051 0.004 −0.026 0.219 −0.032 0.139 −0.033 0.645
house-work −0.003 0.887 0.001 0.972 0.023 0.288 0.317 0.000
unemployed 0.049 0.006 0.047 0.065 0.050 0.016 0.213 0.001

Household income
<2.5
>2.5 and <5 0.139 0.000 0.124 0.000 0.143 0.000 0.114 0.018
>5 and <7.5 0.221 0.000 0.206 0.000 0.235 0.000 0.200 0.000
>7.5 and <10 0.276 0.000 0.243 0.000 0.285 0.000 0.287 0.000
>10 and <15 0.305 0.000 0.170 0.000 0.272 0.000 0.381 0.000
>15 0.337 0.000 0.209 0.003 0.344 0.000 0.434 0.000
Education
< college
2-year college 0.086 0.000 0.047 0.042 0.098 0.000 0.089 0.182
4-year college 0.284 0.000 0.248 0.000 0.278 0.000 0.304 0.000
graduate 0.451 0.000 0.414 0.000 0.481 0.000 0.358 0.000

Residence dummies Yes Yes Yes Yes Yes Yes Yes Yes
Constant −3.246 0.000 −2.856 0.000 −3.394 0.000 −3.151 0.000

No. of observations 19,267 10,804 13,105 3,603

F statistic 251.75 0.000 88.11 0.000 184.74 0.000 19.55 0.000
R-squared 0.215 0.153 0.225 0.107
Table 6: The effects of financial literacy on awareness of living expenses for non-student respondents aged
between 20 and 60 years. Probit and IV probit regressions are utilized to estimate the effect of financial
literacy on the dependent variables, which take the value of 1 if the respondent is aware of his or her one-
month living expenses (columns 1, 2), or post-retirement living expenses (columns 3, 4). The instrumental
variable for financial literacy is a dummy indicating receiving financial education at home. Exogenous
control variables include the respondent’s demographic and socioeconomic characteristics. The coefficients
are maximum likelihood estimates with robust standard errors. In the bottom of table, the estimate of the
instrument from the first stage regression is reported, as well as the test of weak instruments, where the
critical value is the 2SLS estimator for the Wald test size of 10%, based on Stock and Yogo (2005).
(1) Probit (2) IV probit (3) Probit (4) IV probit
Dependent variable: Aware of one-month living expenses Aware of post-retirement living
expenses
Coef. p-value Coef. p-value Coef. p-value Coef. p-value
Financial literacy 0.226 0.000 1.068 0.000 0.241 0.000 0.916 0.000
Female 0.237 0.000 0.422 0.000 0.039 0.284 0.266 0.000
Ln(Age) 0.074 0.129 −0.592 0.000 0.723 0.000 0.130 0.424
Occupation
company employee
civil servant −0.156 0.003 −0.199 0.000 0.088 0.138 0.032 0.583
self-employed 0.195 0.000 0.145 0.002 −0.048 0.446 −0.046 0.444
part-timers −0.007 0.852 0.069 0.048 −0.079 0.086 −0.016 0.733
house-work −0.016 0.694 0.032 0.375 0.001 0.988 0.030 0.493
unemployed −0.375 0.000 −0.317 0.000 −0.143 0.109 −0.154 0.078

Household annual
income (million yen)
<2.5
>2.5 and <5 0.043 0.238 −0.056 0.099 −0.075 0.118 −0.105 0.022
>5 and <7.5 0.018 0.652 −0.155 0.000 0.022 0.651 −0.085 0.110
>7.5 and <10 0.025 0.583 −0.226 0.000 0.169 0.002 −0.003 0.963
>10 and <15 0.015 0.778 −0.250 0.000 0.265 0.000 0.039 0.644
>15 −0.015 0.867 −0.319 0.000 0.400 0.000 0.168 0.142
Education
<college
2-year college 0.013 0.753 −0.072 0.055 −0.029 0.537 −0.094 0.040
4-year college 0.031 0.271 −0.256 0.000 0.032 0.338 −0.200 0.001
graduate 0.078 0.178 −0.381 0.000 0.126 0.054 −0.261 0.014

Residence dummies Yes Yes Yes Yes

Constant 0.319 0.087 2.877 0.000 −2.983 0.000 −0.654 0.310

No. of observations 12,903 12,903 8,620 8,620

Wald chi-squared 378.09 0.000 867.03 0.000 620.26 0.000 926.63 0.000
R-squared 0.026 0.057
First-stage regression of financial literacy
Instrument Coef. p-value Coef. p-value
Financial education at
0.224 0.000 0.153 0.000
home
Robust F 215.5 0.000 75.069 0.000
Stock & Yogo test’s
16.38 16.38
critical value
Table 7: The effects of financial literacy on the comparing behaviors of non-student respondents aged
between 20 and 60 years. Probit and IV probit regressions are employed to test whether financial literacy
leads to comparing alternatives when purchasing financial products (columns 1, 2) or when taking out a
mortgage (columns 3, 4). The instrumental variable for financial literacy is a dummy indicating receiving
financial education at home. Exogenous control variables include the respondent’s geographical
characteristics. The coefficients are maximum likelihood estimates with robust standard errors. In the bottom
of table, the estimate of the instrument from the first stage regression is reported, as well as the test of weak
instruments, where the critical value is the 2SLS estimator for the Wald test size of 10%, based on Stock and
Yogo (2005).
(1) Probit (2) IV probit (3) Probit (4) IV probit
Dependent variable: Comparing financial products Comparing loans
Coef. p-value Coef. p-value Coef. p-value Coef. p-value
Financial literacy 0.458 0.000 1.586 0.000 0.269 0.000 1.337 0.000
Female 0.148 0.031 0.376 0.000 0.091 0.153 0.363 0.000
Ln(Age) −0.334 0.001 −0.973 0.000 −0.091 0.374 −0.878 0.000
Occupation
company employee
civil servant −0.064 0.513 −0.218 0.001 0.018 0.850 −0.041 0.620
self-employed 0.112 0.216 0.014 0.856 −0.056 0.516 −0.059 0.440
part-timers 0.038 0.690 0.028 0.725 −0.077 0.350 0.108 0.148
house-work 0.064 0.512 −0.012 0.883 0.137 0.164 0.146 0.093
unemployed 0.162 0.226 −0.161 0.151 −0.487 0.000 −0.360 0.004

Household annual
income (million yen)
<2.5
>2.5 and <5 0.079 0.376 −0.093 0.219 0.082 0.295 0.076 0.280
>5 and <7.5 0.177 0.055 −0.123 0.137 0.223 0.007 0.026 0.745
>7.5 and <10 0.301 0.002 −0.189 0.044 0.345 0.000 0.000 1.000
>10 and <15 0.372 0.001 −0.244 0.020 0.407 0.000 −0.030 0.795
>15 0.269 0.071 −0.307 0.010 0.456 0.005 −0.195 0.212
Education
< college
2-year college 0.067 0.477 −0.057 0.484 −0.011 0.902 −0.071 0.364
4-year college 0.069 0.242 −0.311 0.000 0.039 0.469 −0.297 0.000
graduate 0.233 0.022 −0.435 0.000 0.111 0.349 −0.419 0.000

Residence dummies Yes Yes Yes Yes

Constant 1.205 0.003 3.469 0.000 0.203 0.604 3.264 0.000

No. of observations 2,922 2,922 2,963 2,963

Wald chi-squared 220.29 0.000 5080.8 0.000 170.41 0.000 911.52 0.000
R-squared 0.060 0.043
First-stage regression of financial literacy
Instrument Coef. p-value Coef. p-value
Financial education at
0.083 0.002 0.167 0.000
home
Robust F 9.139 0.003 34.122 0.000
Stock & Yogo test’s
16.38 16.38
critical value
Table 8: The effects of financial literacy on behavioral biases for non-student respondents aged between 20
and 60 years. (Ordered) probit and IV (ordered) probit regressions are employed to test whether financial
literacy reduces behavioral biases. In columns (1) and (2), the dependent variable is the dummy for loss
aversion, while in columns (3) and (4), the dependent variable is the degree of present-time bias (from 1 to
5). The instrumental variable for financial literacy is a dummy indicating receiving financial education at
home. Exogenous control variables include the respondent’s geographical characteristics. The coefficients
are maximum likelihood estimates with robust standard errors. In the bottom of table, the estimate of the
instrument from the first stage regression is reported, as well as the test of weak instruments, where the
critical value is the 2SLS estimator for the Wald test size of 10%, based on Stock and Yogo (2005).
(1) Probit (2) IV probit (3) Ordered (4) IV ordered
probit probit
Dependent variable: Loss aversion Present-time bias
Coef. p-value Coef. p-value Coef. p-value Coef. p-value
Financial literacy −0.440 0.000 −0.199 0.174 −0.115 0.000 −0.269 0.010
Female 0.510 0.000 0.565 0.000 −0.272 0.000 −0.312 0.000
Ln(Age) 0.648 0.000 0.470 0.000 0.287 0.000 0.397 0.000
Occupation
company employee
civil servant 0.166 0.002 0.142 0.012 −0.043 0.293 −0.029 0.512
self-employed −0.026 0.612 −0.026 0.610 0.158 0.000 0.157 0.000
part-timers 0.163 0.000 0.180 0.000 0.131 0.000 0.118 0.000
house-work 0.231 0.000 0.240 0.000 −0.055 0.092 −0.062 0.055
unemployed 0.213 0.001 0.201 0.002 0.054 0.262 0.060 0.207

Household annual
income (million yen)
<2.5
>2.5 and <5 −0.010 0.817 −0.033 0.455 −0.067 0.024 −0.051 0.105
>5 and <7.5 −0.069 0.132 −0.112 0.029 −0.117 0.000 −0.087 0.021
>7.5 and <10 −0.130 0.011 −0.192 0.002 −0.103 0.004 −0.060 0.198
>10 and <15 −0.244 0.000 −0.309 0.000 −0.118 0.007 −0.072 0.179
>15 −0.244 0.005 −0.321 0.001 −0.132 0.062 −0.078 0.314
Education
<college
2-year college −0.099 0.042 −0.119 0.016 −0.115 0.001 −0.100 0.004
4-year college −0.048 0.110 −0.121 0.019 −0.154 0.000 −0.105 0.010
graduate −0.116 0.042 −0.229 0.008 −0.287 0.000 −0.209 0.003

Residence dummies Yes Yes Yes Yes

Constant −1.726 0.000 −1.015 0.033

No. of observations 12,903 . 12,903 12,903 12,903

Wald chi-squared 1440.5 0.000 986.06 0.000 485.70 0.000 3856.2 0.000
R-squared 0.125 0.012
First-stage regression of financial literacy
Instrument Coef. p-value Coef. p-value
Financial education at
0.224 0.000 0.224 0.000
home
Robust F 215.5 0.000 215.5 0.000
Stock & Yogo test’s
16.38 16.38
critical value
Table 9: The effect of financial literacy on preparation for retirement for non-student respondents aged
between 20 and 60 years. Probit (IV probit) regressions are employed to estimate the effect of financial
literacy on the dependent variable of having secured funds for retirement. The instrumental variable for
financial literacy is a dummy indicating receiving financial education at home. Exogenous control variables
include the respondent’s geographical characteristics. The coefficients are maximum likelihood estimates
with robust standard errors. In the bottom of table, the estimate of the instrument from the first stage
regression is reported, as well as the test of weak instruments, where the critical value is the 2SLS estimator
for the Wald test size of 10%, based on Stock and Yogo (2005).
(1) Probit (2) IV probit (3) IV probit (4) IV probit
Low present time High present time
All respondents All respondents
bias respondents bias respondents
Dependent variable: Securing funds Securing funds Securing funds Securing funds
Coef. p-value Coef. p-value Coef. p-value Coef. p-value
Financial literacy 0.111 0.000 1.011 0.000 0.935 0.000 1.112 0.000
Female −0.018 0.699 0.292 0.000 0.202 0.011 0.421 0.000
Ln(Age) 1.019 0.000 0.120 0.505 0.352 0.180 −0.150 0.550
Having loans −0.261 0.000 −0.204 0.000 −0.182 0.000 −0.229 0.000
Occupation
company employee
civil servant 0.237 0.000 0.121 0.062 −0.005 0.954 0.239 0.042
self-employed 0.055 0.475 0.037 0.599 0.135 0.197 −0.055 0.571
part-timers −0.001 0.982 0.068 0.179 −0.008 0.909 0.150 0.044
house-work 0.277 0.000 0.245 0.000 0.225 0.000 0.267 0.001
unemployed 0.113 0.307 0.044 0.633 −0.024 0.844 0.126 0.376
Household annual
income (million yen)
<2.5
>2.5 and <5 0.082 0.208 0.011 0.842 −0.096 0.202 0.140 0.121
>5 and <7.5 0.266 0.000 0.065 0.347 0.007 0.939 0.139 0.212
>7.5 and <10 0.530 0.000 0.205 0.024 0.096 0.404 0.353 0.017
>10 and <15 0.778 0.000 0.337 0.004 0.272 0.055 0.416 0.040
>15 1.147 0.000 0.633 0.000 0.557 0.005 0.730 0.006
Education
<college
2-year college 0.059 0.295 −0.048 0.346 0.002 0.977 −0.115 0.121
4-year college 0.088 0.032 −0.233 0.000 −0.245 0.003 −0.202 0.016
graduate 0.219 0.005 −0.322 0.002 −0.322 0.025 −0.303 0.053
Residence dummies Yes Yes Yes Yes
Constant −5.269 0.000 −1.445 0.069 −2.241 0.047 −0.523 0.651

No. of observations 8,620 8,620 4,850 3,768

Wald chi-squared 720.80 0.000 1791.3 0.000 921.40 0.000 1052.5 0.000
R-squared 0.112
First-stage regression of financial literacy
Instrument Coef. p-value Coef. p-value Coef. p-value
Financial education at
0.153 0.000 0.158 0.000 0.144 0.000
home
Robust F 75.044 0.000 45.191 0.000 28.579 0.000
Stock & Yogo test’s
16.38 16.38 16.38
critical value
Table 10: Mediation Analysis of the direct and indirect effect of financial literacy using structural equation
modeling (SEM). Sample contains non-student respondents aged between 20 and 60. SEM is performed to
estimate three equations, where the dependent variables are present-time bias, loss aversion bias, and
preparation for retirement, respectively. Unreported here due to space limitation are a set of control variables
including gender, age, having a loan, educational attainment, occupation, annual income, and areas of
residence. The dependent variable for preparation for retirement in column (1) is having a retirement plan,
while in column (2) having secured the funds. Tests of significance are based on bootstrapped standard errors.
(1) (2)
Equations Coef. p-value Coef. p-value
Dependent variable: Present-time bias (M1)
Financial literacy (FL) −0.181 0.000 -0.181 0.000
Dependent variable: Loss aversion bias (M2)
Financial literacy (FL) −0.111 0.000 −0.111 0.000
Dependent variable (SAVE)
for (1): Having a saving plan for retirement
for (2): Securing funds for retirement
Present-time bias (M1) −0.018 0.000 −0.007 0.008
Loss aversion bias (M2) −0.081 0.000 −0.044 0.000
Financial literacy (FL) 0.062 0.000 0.019 0.000
No. of observations 8,620 8,620
Direct vs. indirect effect on saving for retirement Coef. p-value Coef. p-value
Direct effect of financial literacy 0.062 0.000 0.019 0.000
Indirect effect of financial literacy 0.012 0.000 0.006 0.000
Total effect of financial literacy 0.074 0.025 0.000
Proportion of financial literacy’s direct effect 83.5% 75.8%
Proportion of financial literacy’s indirect effect 16.5% 24.2%
 Table 11: Comparison of high versus low financial literacy respondents. Using a propensity score matching
method, each individual in the high literacy group is matched with a nearest-neighbor peer from the low
literacy group, based on a set of covariates, gender, age group, educational attainment, occupation, income
level, and area of residence. In Column (1), the high literacy group contains those with higher than median
financial literacy index, with the matching peers selected from the below-median group. In Column (2), the
high literacy respondents are in the top 25th of the distribution of financial literacy, with matching peers
selected from the below-median group. Multiple matches are allowed when they have the same propensity
score and tied for nearest-neighbor. Comparison is then made between the high literacy and the matching by
computing the average treatment effect on the treated (ATET).
(1) Difference between (2) Difference between
The top 50% vs. the The top 25% vs. the matching
matching group group
ATET No. Difference p-value No. Difference p-value
Financial literacy 12,903 1.375 0.000 9,531 1.735 0.000
Comparing investment products 2,922 0.250 0.000 2,150 0.308 0.000
Comparing loans 2,963 0.143 0.000 2,108 0.211 0.000
Present-time bias 12,903 −0.184 0.000 9,531 −0.257 0.000
Loss aversion bias 12,903 −0.156 0.000 9,531 −0.218 0.000
Having a plan for retirement 8,620 0.078 0.000 6,213 0.126 0.000
Securing funds for retirement 8,620 0.024 0.066 6,213 0.031 0.078

View publication stats