- Email: [email protected]
Financial literacy of middle-aged and older Individuals: Comparison of Japan and the United States
Journal Pre-proofs Full length article Financial Literacy of Middle-Aged and Older Individuals: Comparison of Japan and the United States Satoshi Shimizutani, Hiroyuki Yamada PII: DOI: Reference:
S2212-828X(19)30101-X https://doi.org/10.1016/j.jeoa.2019.100214 JEOA 100214
To appear in:
The Journal of the Economics of Ageing
Received Date: Revised Date: Accepted Date:
12 December 2018 17 September 2019 19 September 2019
Please cite this article as: S. Shimizutani, H. Yamada, Financial Literacy of Middle-Aged and Older Individuals: Comparison of Japan and the United States, The Journal of the Economics of Ageing (2019), doi: https://doi.org/ 10.1016/j.jeoa.2019.100214
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2019 Published by Elsevier B.V.
Financial Literacy of Middle-Aged and Older Individuals: Comparison of Japan and the United States*
by
Satoshi Shimizutani and Hiroyuki Yamada** September 2019 Abstract Financial literacy holds growing interest for managing assets/savings during the longer retirement period currently experienced in rapidly aging countries. We examine and compare levels and determinants of financial literacy as well as its association with asset allocation among middle-aged and older individuals of Japan and the United States. We present some interesting findings. First, financial literacy is generally influenced by educational attainment, cognitive skills, coursework in economics or finance, and income level. Second, financial literacy is associated with household asset allocation; individuals with higher literacy also have investment in stocks or securities. These patterns are commonly observed both in Japan and the United States. Keywords: financial literacy, Japan, U.S., JSTAR, HRS, household asset allocation. JEL Classification Codes: D14, G11, J26.
*
This work was presented at the Gateway to Global Aging Data Research Conference, funded by the National Institute on Aging (2R01 AG030153). We thank our discussant, Silvia Barcellos, conference participants, one anonymous referee, and the editors for their constructive comments. The views expressed in this paper are completely personal and unrelated to those of any organizations with which we are affiliated or those who commented on this study. **
Satoshi Shimizutani (corresponding author); Visiting Research Fellow, Nakasone Yasuhiro Peace Institute (NPI); Postal Address: Toranomon 30 Mori Building 6F, 3-2-2 Toranomon, Minato-ku, Tokyo 105-0001 Japan; E-mail: [email protected]. Hiroyuki Yamada; Professor, Faculty of Economics, Keio University; Postal address: 2-15-45 Mita, Minato-ku, Tokyo 108-8345 Japan, Tel (Direct): +81-(0)3-5427-1271 E-mail: [email protected]. 1
1. Introduction Financial literacy holds growing interest for managing assets/savings during the longer retirement period currently experienced in rapidly aging countries. Indeed, the case is pronounced in and relevant to Japan, where population aging is taking place at the most rapid rate in the world in an environment that combines a persistent long-term low level fertility rate and continuing rise in life expectancy.1 These factors have resulted in the older generation occupying a higher proportion of the total population at 27.3% while putting pressure on the sustainability of the pay-as-you-go public pension program (Shimizutani, et al. 2016). Under these circumstances, there are two forces that act to lower the adequacy of retirement savings and thus obstruct the management of economic resources in later life for middle-aged and older individuals. One is a reduction in the amount of pension benefits to receive and the other is a longer time horizon that has to be compensated. A higher level of individual financial literacy may serve as a device to ensure economic well-being or even enhance the standard of living at older ages.2 Research on financial literacy has developed significantly since the 2000s (Lusardi and Mitchell, 2014) for the U.S. and other Western countries, but is still relatively scarce in Japan. Currently a widely used, standard measure of financial literacy is the number of correct answers to a 3-question quiz inquiring about compound interest, inflation, and risk/diversification. These questions were first asked 1
According to National Institute of Population and Social Security Research, the latest TFR (Total Fertility Rate) of Japan is 1.44 in 2016. The figure declined to below 1.3 in 2003-2005 and slightly recovered afterwards. The post-war development of TFR is described in detail by Shimizutani (2015). The latest life expectancy at birth in Japan is 80.98 year for males and 87.14 year for females in 2016, both of which represent the second longest life expectancy in the world. 2 Individuals with higher levels of financial literacy tend to possess greater retirement wealth, other things equal (Lusardi and Mitchell, 2011a, 2014; Clark et al. 2015). The reasons why more financially literate individuals are wealthier include (1) they plan and save more; (2) they may earn higher returns on their investments (Lusardi et al., 2017); (3) they may select better diversified investments. While the study of retirement wealth is of great importance, the subject is beyond the scope of this paper which focuses on a comparison of Japan and U.S. financial literacy. 2
in 2004 in the Health and Retirement Study. Using these and other metrics of financial literacy, Lusardi and Mitchell (2014) conclude that in the U.S. the level of financial literacy is low. Moreover, the low level of financial literacy is found to be prevalent in other countries including Japan (Lusardi and Mitchell, 2011b). When looking at the data by subgroup, the level of financial literacy is hump-shaped along with age and is higher for males, the educated, employed individuals, or those who are higher earners. Furthermore, a large volume of empirical research documented that the standard measure of financial literacy is indeed associated with a variety of financial decisions or economic behaviors such as stock market participation (van Rooij et.al, 2011) and retirement saving (Lusardi and Mitchell (2007)), and financial literacy is confirmed to play a non-negligible role in those decisions/behaviors if instrument variable (IV) estimation is employed to address endogeneity. This study examines factors influencing financial literacy and the association of financial literacy with household financial decisions among middle-aged and older individuals aged 52-79. Using harmonized micro-level data, we perform parallel analyses in Japan and the U.S., highlighting differences and similarities between the two countries as far as empirical findings are concerned. Research on financial literacy in Japan has been relatively scarce, let alone studies of individuals approaching or entering retirement age. Because of the demographic and economic changes mentioned above, understanding heterogeneity in financial literacy and decision-making within this segment of the population is especially important and a major contribution of this paper. Among the few previous studies, Clark et al. (2013) revealed that a higher level of financial literacy is found in males, residents in urban areas, the higher educated, or 3
high income earners in Japan.3 The study utilized a micro-level dataset on individuals from the 2010 National Survey on Work and Family, focusing on individuals in the sample who were aged 40 to 59 and employed at the time of the survey. Furthermore, they found that financial literacy is partially linked with the demand for human capital investment, measured by the need for additional skills to be competitive at the current job.4 As regards financial literacy for younger individuals, Cameron et al. (2013) showed that financial literacy is higher among Japanese high school students than in the U.S. and New Zealand. In a more recent work, Yoshino et al. (2017) used a dataset collected by the Central Council for Financial Service Information, which is not based on a standard two stage random sampling, to examine factors influencing financial literacy and the relationship between financial literacy and household asset allocation. While the data used by Yoshino et al. (2017) contains rich variables related with financial literacy, they are not internationally comparable and thus are not suitable for an international comparison study. This study performs two sets of analysis: the first is on the factors that influence financial literacy and the second is the association between financial literacy and household asset allocation. Our contribution to the literature features two new enhancements. First, we perform a Japan-U.S. comparison using internationally standardized datasets. While we share the spirit of the analytical approach with Yoshino et al. (2017), we use world-standard longitudinal datasets on the middle-aged and older individuals, namely Japanese Study on Aging and Retirement (JSTAR) for Japan, and Health and Retirement Study (HRS) for the U.S. These surveys are “sisters” in international harmonization efforts for studies on population aging, and most of the variables collected, including the three metrics of financial literacy, are standardized 3 4
Sekita (2011) performed a similar analysis and obtained similar results. Clark et al. (2013) did not explore an association between financial literacy and asset holdings. 4
across countries. Thus, these datasets are most suitable for a Japan-U.S. comparison in financial literacy, as described in the next section. Second, this paper advances the research in Japan. We focus on middle-aged and older individuals aged 52 to 79 who are sensitive to retirement preparation or who are indeed retired. Moreover, we include some new variables as covariates which were not included in the previous literature in Japan, specifically, measures of cognitive ability and experience of taking an economics/accounting coursework.5 We believe that these additional variables will help to uncover a more accurate assessment of the impact of socio-economic status on financial literacy or household asset allocation.6 This paper is organized as follows. Section 2 describes the dataset used in the empirical analysis and presents descriptive statistics of financial literacy. Section 3 performs regression analysis to associate the level of financial literacy with a variety of factors including household demographics, socio-economic status, and measures of cognitive ability, and so on. Further, Section 4 investigates the relationship between financial literacy and household asset allocation. Section 5 concludes.
2. Data description This study uses microdata from two sets of comparable datasets, JSTAR and
5
Those variables are not available in the dataset used by Yoshino et al. (2017). Other candidates for new variables include those related to behavioral economics (i.e., risk attitude and time discounting), those that capture current mental state (symptoms of depression), and expectations for the future (survival probability to measure the life horizon). While HRS contains variables regarding symptoms of depression and expectations for the future (survival probability to measure the life horizon), the styles used in HRS and JSTAR vary. For symptoms of depression, HRS uses an 8-item version of the Center for Epidemiologic Studies Depression Scale (CESD) while JSTAR uses the 20-item version of that the CESD. For survival probability, the target age of survival differs depending on the respondent’s age in HRS while it is the same for all respondents in JSTAR. Therefore, we did not use those variables in the comparison. We confirmed that most of those variables indeed significantly influence financial literacy and hold an expected sign in the regression. The estimation results are reported in detail in Shimizutani and Yamada (2018) and some that are related to factors influencing financial literacy are in Appendix Table 1. The measurement of risk attitude and time discounting is also explained in the appendix of Shimizutani and Yamada (2018). 5 6
HRS. JSTAR is a world-standard longitudinal dataset on middle-aged and older individuals and holds the position as the Japanese counterpart of HRS in the U.S., English Longitudinal Study of Ageing (ELSA) in the U.K., the Survey of Health, Ageing and Retirement in Europe (SHARE) in continental Europe. Similar to those surveys, JSTAR’s questionnaire covers a wide variety of variables related to health, employment, economic status, and other life features. The unit of the sample of JSTAR at baseline is individuals aged 50 to 75 who were randomly chosen from household registration within each municipality. If a respondent is married, selected questions are also asked of the spouse. JSTAR began collection of data in 2007 in five municipalities in the first wave, followed by four subsequent waves at two-year intervals.7 The second JSTAR wave of data, collected in 2009, contains the standard measure of financial literacy. In addition, the wave includes a related question on whether the respondent has ever studied accounting or economics. In the second wave, the total sample size is 2,852 persons aged 52 to 79. HRS is the pioneer and the eldest sister in the field of international data collection and harmonization efforts on middle-aged and older individuals.8 The HRS survey was implemented in 1992 and tracks the same people every two years. HRS covers a wide range of information that is also covered by JSTAR. Further, the timing of the surveys coincides. In 2010, HRS includes the standard measure of financial literacy. Note that the special module is asked of only a subset of the sample (one-tenth that are randomly sampled). The sample size is 1,473 persons aged 52 to 79 in the same age range as those in JSTAR. 7
The total number of municipalities from the third wave is 10 and the total number of individuals in the sample is about 8,000 at baseline. The five municipalities in the first wave are Takikawa city, Sendai city, Adachi ward (metropolitan Tokyo), Kanazawa city, and Shirakawa town. 8 The detailed description of the dataset is available at https://hrs.isr.umich.edu/about?_ga=2.94165525.941904344.1535601989-1380807771.153560198 9 6
Both JSTAR and HRS pose the financial literacy question to an individual in a household (i.e., either husband or wife if a household is a “couple”) but whether or not the surveyed individual is the most financially knowledgeable in the household can be known only in HRS. This information is important because the most financially knowledgeable individual in the household responded to the financial asset questions in HRS. Thus, we use HRS data from all individuals when examining the association between the level of financial literacy and a variety of covariates (henceforth, “U.S. full sample”) but use the data from only the most financially knowledgeable in the household when examining the association between household asset allocation and the covariates. The sample size of respondents to this question is reduced to 1,011 persons (henceforth, “investment behavior sample”). Another difference between JSTAR and HRS lies in the questions regarding asset holding. In JSTAR, if a respondent has a spouse (or a partner), the question regarding holding of each type of asset is asked separately of the respondent and the spouse. On the other hand, HRS asks questions only of “household” level asset holdings. In other words, we cannot identify whether an asset is held by the respondent or his/her spouse (or both).9 For the sake of a Japan-U.S. comparison, we construct variables of household level asset holdings from JSTAR. We interpret a household as holding a positive value of assets in the case that at least either the respondent or the partner holds the corresponding asset. Table 1 presents simple descriptive statistics of each measure of financial literacy. Note that the HRS sample is limited to people aged between 52 and 79 to correspond to the age of JSTAR respondents. At a glance, the level of financial literacy is much lower in Japan than that in the U.S. For example, Panel A shows that 9
For the U.S., we use RAND HRS Detailed Imputation File 2014 (V2), released in February 2018, for the information on each type of asset holding. 7
the proportion of correct responses to “compound interest” is 38.6% in Japan, which is substantially lower than that in the U.S. (71.8%). This is also the case for “inflation” in Panel B; the proportion of correct response is 38.7%, much lower than 82.4% in the U.S.; for “stock risk” in Panel C, the proportion of correct response is 30.9%, much lower than 66.4% in the U.S. The bottom of Table 1 reports some summary statistics across questions. Only 13.9% give correct answers to all three questions in Japan while 44.6% do so in the U.S. In contrast, about 42% give no correct answers in Japan while this proportion is only 4.1% in the U.S. About 70% choose “Don’t Know” or “Not Sure” to at least one out of the three questions in Japan while this proportion is much lower in the U.S. at 18.8%. Moreover, about 30% choose “Don’t Know” or “Not Sure” to all three questions, while the proportion of such individuals in the U.S. is negligible. It might be tempting to conclude that the level of financial literacy is lower in Japan, but we should be careful how to interpret these figures since the proportion of “Don’t Know” or “Not Sure” is much higher (42.6%, 47.9% and 53.9%, respectively, for each of the three questions) in Japan.10 These high proportions of “Don’t Know” or “Not Sure” are not necessarily specific to this dataset (i.e. JSTAR). The proportion of “Don’t Know” or “Not Sure” for each of the three questions, respectively, is 12.5%, 28.6%, and 56.1% in Sekita (2011), 14.3%, 27.2%, and 53.8% in Clark et al. (2013), and 24.0%, 34.1%, and 49.0% in the Central Council for Financial Service Information (2016).11 Note that the samples of Sekita (2011), Clark et al. (2013), and
10
Indeed, the proportion of “incorrect” answers is larger in the U.S. than in Japan for all three questions. 11 Note that the question wording and choice set of answers are not exactly the same across these datasets. For example, the question words of the “compounding interest” and “stock risk” questions are the same in the Central Council for Financial Service Information (2016) and JSTAR but the choice set of answers is different between them. Neither question words nor choice set of answers of the “inflation” question are the same between the Central Council for Financial Service Information (2016) and JSTAR. 8
the Central Council for Financial Service Information (2016) are not restricted to middle-aged and older individuals. Thus, the proportions of “Don’t Know” or “Not Sure” among the Japanese samples are in general much higher than that of the HRS sample. This might indicate that Japanese are very cautious and answer only when confident in their response (Sekita 2011). Further, we compare the basic characteristics of those who chose (a) “Don’t Know” or “Not Sure,” (b) the incorrect answer(s), and (c) the correct answer for each of the three questions in JSTAR. The sample of (a) are older, higher number of females, less likely to work and be currently married, less educated (in general education and economics/accounting courses), and have lower cognitive skills than the samples of (b) and (c).1213 Table 2 shows summary statistics for the variables used in the analysis below. We focus on the comparison between the Japan study and the U.S. full sample below.14 First, the number of correct answers is different between Japan (1.08) and the U.S. (2.20). The average age is 65.6 years old in Japan and 64.0 in the U.S. In Japan, more than one half of the respondents were working while the figure was under 40% in the U.S. More than 80% were currently married in Japan but the proportion was lower in the U.S. Educational attainment is higher in the U.S. than in Japan. In Japan, the proportion of senior high school graduates or under is three quarters and the proportion of university graduates or higher is slightly larger than 10%. In the U.S., the proportion of senior high school graduates or under is slightly less than one half 12
The detailed descriptive statistics for this comparison are available upon request from the authors. 13 One could argue, for instance, that in countries with a history of high inflation there might be more knowledge of its consequences. However, this concern could apply only to the second question on inflation and it could be a minor issue in our context because neither Japan nor the U.S. has experienced high inflation rates for many years. We acknowledge that the populations are different due to many varying characteristics between the two countries but confined our analysis to the variables which were commonly used in the previous literature. 14 In comparing the number of correct answers between “U.S. full sample” and “U.S. investment behavior sample.” that of the former is very slightly higher than that of latter although the latter is the sample of financially responsible individuals. 9
and the proportion of university graduates or higher is close to one quarter. Looking at the cognitive measure, the proportion of correct answers to serial sevens is 45.3% and the average number of 10 word immediate recall is 4.2 words in Japan. The proportion of correct responses to serial sevens is slightly lower in the U.S. while the number of word recall is higher. The proportion of those who have ever studied accounting or economics is slightly more than 10% in Japan while the figure in the U.S. is three-fold (34.2%). Turning to household income, the dominant bracket of household income is 20,000-50,000 dollars in both countries (31.6% in Japan and 31.9% in the U.S.). Lastly, household asset holding reveals that the proportion of risky financial instruments is low in both countries; bond holders make up about 20% and stockholders occupy less than 20% in Japan. The proportion of bond holders is much smaller in the U.S. but that of stockholders is larger, exceeding 20%. The lower share of bond holders may be due to the different definition or perception on the part of respondents. Indeed, the proportion of CD/government saving bonds/T-bills, which are not included in bonds, is 16.0% in the U.S. The share of either bond or stockholders is 26.9% in Japan and 23.1% in the U.S. On the flip side, about 90% have saving accounts in Japan, suggesting that Japanese middle-aged and older individuals are generally conservative in asset investment. The proportion of bank account holders is lower at 77.1% in the U.S. but it is noted that more than one-third of households have IRA accounts. In Japan, IRA-type accounts were scarcely available at the time of the second wave survey in 2009. Therefore, no corresponding question regarding this type of asset was included in JSTAR in 2009.
3. Factors that influence financial literacy In this section, we examine factors influencing financial literacy by regression 10
analysis. The basic specification for the analysis is described as follows.
(1) 𝐹𝐿𝑖 = 𝛽𝑋𝑖 + 𝜀𝑖
where 𝐹𝐿𝑖 measures the level of financial literacy referring to the number of correct answers to the set of three financial literacy questions (OLS estimation). We also run the regressions using a logit model by replacing the dependent variable to a binary variable to take 1 if all questions were answered correctly and 0 otherwise. This measure may be less noisy since there is some randomness as to whether people get one or two questions correct. The vector of explanatory variables, Xi, includes age, sex, marital status, educational attainment, cognitive skills (serial sevens and immediate word recall test score), an indicator for whether the individual has taken courses in economics or accounting, household income, and region fixed-effects. Columns (1) and (3) in Table 3 report the estimated coefficients using OLS for Japan and the U.S. when taking the number of correct answers as dependent variables, respectively and Columns (2) and (4) show marginal effects using a logit model for each country when taking a binary variable as the dependent variable. First, looking at individual characteristics, the coefficients and marginal effects on age are negative and significant in Japan in Columns (1) and (2) but they are not significant in the U.S. in Columns (3) and (4). The coefficients and marginal effects on sex are positive and significant in both countries, showing that the level of financial literacy is higher for males. These results on sex are common and consistent with findings in previous studies. Our results show that being male increases the number of correct answers by 0.24 in both countries while it improves the likelihood of answering all questions correctly by 13 percentage points in the U.S. but only 5 percentage points in Japan. 11
The coefficients and marginal effects on current work status and marital status are not significant. The coefficients and marginal effects are significant and larger for higher educational attainment, which again conforms to the result in previous studies. The coefficients on each school attainment level are positive and significant (the default case is primary/junior high school graduates) for both countries, but it is noted that the gradient is steeper in the U.S. Columns (1) and (3) show that the number of correct answers increases by 0.4 in Japan and by 0.5 in the U.S. in the case of completion of graduate school. Columns (2) and (4) show a larger gap between Japan and the U.S.; the likelihood of answering all questions correctly increases by 28 percentage points in the U.S. and only 10 percentage points in Japan if graduate school has been completed. The coefficients and marginal effects are positive and significant for serial sevens and immediate word recall again for both countries. This is natural since respondents with higher cognitive skills are more likely to retain a higher level of financial literacy. While the coefficients in Columns (1) and (3) are roughly comparable, the likelihood to answer all questions correctly is larger in the U.S. in the case of serial sevens; specifically, it is 7 percentage points in the U.S. and 2 percentage points in Japan. This is also the case for taking economics/accounting coursework and the coefficients and marginal effects are positive and generally significant. While the likelihood of answering all questions correctly is similar in both countries, the coefficient in Column (1) is larger than in Column (3), showing that the experience of having studied accounting or economics increases the number of correct answers by 0.23 in Japan and by only 0.12 in the U.S. Turning to household income, the coefficients and marginal effects are larger for a higher level of income (the default case is 0 to 20,000 dollars), showing that higher earners are more likely to enjoy a higher level of financial literacy. The gradient is steeper in Japan when 12
considering the number of correct answers which increase by 0.34 if household income is larger than 100,000 dollars and by only 0.22 in the U.S. In contrast, the gradient is steeper in the U.S. when considering the likelihood of answering all questions correctly which increases by 13 percentage points in the U.S. if household income is higher than 100,000 dollars and by only 8 percentage points in Japan. In sum, we found a higher level of financial literacy in individuals who are male, more educated, with higher cognitive skills, or a higher annual income. These patterns are confirmed in both countries.15
4.
Financial literacy and household asset allocation Next, we aim to understand the level of financial literacy that associates with
household asset holding. By “household” assets, we mean bonds, stocks, bank accounts (deposits) in the name of either the respondent or his/her spouse (or partner) both in Japan and the U.S. The basic specification for the analysis is described as follows.
(2) 𝐴𝑗𝑖 = α𝐹𝐿𝑖 + 𝛽𝑋𝑖 + 𝜀𝑖
where 𝐴𝑗𝑖 is an indicator taking value 1 if household i holds asset type j in its portfolio and 0 otherwise. We run separate regressions for each asset type j. The main
15
We present the estimation results by adding some new covariates in Appendix 1 which are available only in Japan. First, the coefficient on time discounting is positive and significant for the 1-6% range, showing that slightly patient individuals are more likely to have a higher level of financial literacy. Second, the coefficients on an incidence of depression symptoms by the CESD measure are negative and significant, showing that respondents with depression symptoms are likely to have lower financial literacy. Third, the coefficients on the subjective probability to live to age 85 are now positive (but marginally insignificant), showing that a subjective assessment of a longer life time horizon is linked to a higher level of financial literacy. We obtain a similar pattern on the coefficients if we use individual income instead of household income in Japan (Shimizutani and Yamada 2018). 13
independent variable 𝐹𝐿𝑖 is the level of financial literacy of the surveyed individual in the household in JSTAR and of the most financially knowledgeable in the household in HRS measured by the number of correct answers to the three questions, which connect financial literacy to household asset holding. The other covariates (a vector of Xi) refer to individual characteristics which are exactly the same in specification (1): age, sex, work status, marital status, educational attainment, cognitive skills (serial sevens and immediate word recall), taking coursework in economics or accounting, household income, and dummy variables to indicate each region. We present the marginal effects using logit estimation. In what follows, we will focus on the marginal effects of the main variable, the level of financial literacy. Table 4 reports the results for bond holding in Japan (Column (1)) and the U.S. (Column (2)). The marginal effects on financial literacy are positive and significant, showing that financial literacy is indeed associated with bond holding. Concretely, one additional correct answer to the financial literacy questions increases the likelihood of bond holding by about 5 percentage points in Japan and 2 percentage points in the U.S. In addition, we observe that older, non-working individuals with higher educational attainment or higher household income level are associated with bond holding in both Japan and the U.S. Table 5 shows the marginal effects for stock holding in Japan (Column (1)) and the U.S. (Column (2)). The marginal effects on financial literacy are positive and significant, again showing that financial literacy is associated with stock holding. Concretely, one additional correct answer to the financial literacy questions increases the likelihood of bond holding by about 3 percentage points in Japan and 5 percentage points in the U.S. Interestingly, the size of the marginal effect on financial literacy is smaller in stock holding than bond 14
holding in Japan but the pattern is reverse in the U.S. The marginal effects on other covariates are generally the same as in the bond-holding.16 Table 6 reports the marginal effects for bank account holders in Japan (Column (1)) and the U.S. (Column (2)).Contrary to the results on bonds and stocks, the marginal effects contrast between Japan and the U.S.; those on financial literacy are not significant in Japan but are positive and significant in the U.S. probably because bank account holders are very prevalent in Japan. The marginal effect on age is not significant in Japan but is positive and significant in the U.S. due to the same reason. In contrast, the marginal effects on being married are positive and significant in Japan but they are not significant in the U.S. While the marginal effects on each educational attainment are positive and significant in both countries, the gradient along education is steeper in the U.S. than in Japan. This is also the case for income. Most of the marginal effects on income quartiles are positive and significant and the gradient along income is steeper in the U.S. than in Japan. In sum, financial literacy is associated with household asset allocation; households with a higher literacy also have more investment in bonds or stocks, apart from their savings.17 These patterns are commonly observed in Japan and the U.S.18,19
5. 16
Conclusions
Table 2 shows the proportion of bond holding is very different between Japan and the U.S. As discussed, the difference may come from a different definition or perception of respondents. Thus, we perform the same analyses when the dependent variable takes 1 if a respondent is a bond and/or stockholder and 0 otherwise. Appendix 2 shows that the marginal effects on financial literacy are positive and significant and those on other covariates are generally the same. 17 The obtained association between financial literacy and stock market participation is qualitatively consistent with the past literature (van Rooij et al. 2011, Yoong 2011, Almenberg and Dreber 2015, and Arrondel et al 2012). 18 We also conduct regression analysis using the principal component of the answers of the three questions as a variable of financial literacy and the results are qualitatively the same. The results are available upon request from the authors. 19 We conducted the same analyses using the “U.S. full sample” and the patterns of coefficients of the level of financial literacy are qualitatively very similar to when using “investment behavior sample.” The results are available upon request from the authors. 15
We examine and compare the levels and determinants of financial literacy as well as its association with asset allocation among middle-aged and older individuals in Japan and the U.S. In both countries, we found a higher level of financial literacy in individuals who are male, more educated, have higher cognitive skills, or a higher level of annual income, results which conform to findings by previous studies. Moreover, we found that the level of financial literacy is significantly associated with household asset holding and the impact is larger for bond and stock holdings than for savings deposits. Financial literacy holds growing interest for managing assets/savings during the longer retirement period currently experienced in rapidly aging countries, a societal feature particularly relevant to Japan. It is interesting to find the same pattern in the factors influencing financial literacy and the association between financial literacy and asset holding in Japan and the U.S. since it is widely perceived that Japan lags in financial education and the U.S. is more advanced. Further research should develop in two directions. One direction is to investigate a causal relationship between financial literacy and asset holding. Based on the examination, the other direction of further study should put the cases of Japan and the U.S. in an international perspective to extract policy implications. While improvement of financial literacy in middle-aged and older individuals seems to be a common policy agenda, consensus on concrete policy intervention is lacking. One reason is that there are too many factors influencing financial literacy and the effectiveness of intervention, which depends on policy instrument, is ambiguous. An international research project using comparable datasets would contribute toward revealing policy features that may more effectively increase financial literacy in different country-specific contexts.
16
References Almenberg, Johan, and Anna Dreber. (2015) “Gender, stock market participation and financial literacy.” Economics Letters, vol.137, pp140-142. Arrondel, Luc, Majdi Debbich, and Frédérique Savignac. (2012) “Stockholding and financial literacy in the French population.” International Journal of Social Sciences and Humanity Studies, vol.4, no.2, pp 285-294. Cameron, Michael, Richard Calderwood, Ashleigh Cox, Steven Lim and Michio Yamaoka (2013). “Personal Financial Literacy among High School Students in New Zealand, Japan and the USA,” Citizenship, Social and Economics Education, vol.12, no.3, pp.200-215. Central Council for Financial Service Information (2016). “Financial Literacy Survey: 2016 Results.” Available at https://www.shiruporuto.jp/e/survey/kinyulite/pdf/16kinyulite.pdf (accessed on December 6, 2018). Clark, Robert, Annamaria Lusardi, and Olivia S. Mitchell (2017). “Financial knowledge and 401 (k) investment performance: a case study.” Journal of Pension Economics & Finance, vol.16, no.3, pp.324-347. Clark, Robert, Rikiya Matsukura and Naohiro Ogawa (2013). “Low Fertility, Human Capital, and Economic Growth: The Importance of Financial Education and Job Retraining,” Demographic Research, vol.29, Article 32, pp.865-884. Lusardi, Annamaria, Pierre-Carl Michaud, and Olivia S. Mitchell (2017). “Optimal financial knowledge and wealth inequality.” Journal of Political Economy, vol.125, no.2, pp.431-477. Lusardi, Annamaria, and Olivia S. Mitchell (2007). “Baby Boomer Retirement Security: The Roles of Planning, Financial Literacy, and Housing Wealth.” Journal of Monetary Economics, vol.54, no.1, pp.205–224. Lusardi, Annamaria and Olivia Mitchell (2011a). “Financial literacy and planning: implications for retirement wellbeing.” In O. S. Mitchell and A. Lusardi (eds), Financial Literacy: Implications for Retirement Security and the Financial Marketplace. Oxford: Oxford University Press, pp.17–39. Lusardi, Annamaria and Olivia Mitchell (2011b). “Financial Literacy around the World: An Overview,” Journal of Pension Economics and Finance, vol.10, no.4, pp.497-508. Lusardi, Annamaria and Olivia Mitchell (2014). “The Economic Importance of Financial Literacy: Theory and Evidence,” Journal of Economic Literature, vol.52, no.1, pp.5-44. Sekita, Shizuka (2011). “Financial Literacy and Retirement Planning in Japan,” 17
Journal of Pension Economics and Finance, vol.10, no.4, pp.637-656. Shimizutani, Satoshi (2015). “Population Aging in Postwar Japan: Processes and Prospects,” Asia-Pacific Review, vol.22, no.2, pp. 53-76. Shimizutani, Satoshi, Takashi Oshio and Mayu Fujii (2016). “Option Value of Work, Health Status, and Retirement Decisions in Japan: Evidence from the Japanese Study on Aging and Retirement (JSTAR),” in David Wise eds. Social Security Programs and Retirement around the World: Disability Insurance Programs and Retirement, The University of Chicago Press, pp.497-535. Shimizutani, Satoshi and Hiroyuki Yamada (2018). “Financial Literacy of Middle and Older Generations in Japan,” Keio-IES Discussion Paper Series DP2018-010. van Rooij, Maarten, Annamaria Lusardi, and Rob Alessie. (2011) “Financial literacy and stock market participation.” Journal of Financial Economics, vol.101, no.2, pp.449-472. Yoong, Joanne. (2011) “Financial Illiteracy and Stock Market Participation: Evidence from the RAND American Life Panel.” In Financial Literacy: Implications for Retirement Security and the Financial Marketplace, edited by Olivia S. Mitchell and Annamaria Lusardi, 76–97. Oxford and New York: Oxford University Press. Yoshino, Naoyuki, Peter J. Morgan, and Long Q. Trinh (2017). “Financial Literacy in Japan: Determinants and Impacts” ADBI Working Paper Series No.796.
18
Table 1 Financial literacy in Japan with a comparison with the U.S. A. Compound interest "Suppose you had $100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow? More than $102, exactly $102, or less than $102?" JSTAR HRS (52
13.88 22.41 42.04 69.25 29.42
19
HRS (52
Table 2 Descriptive statistics Variables # of correct answers
Japan (1) N Obs. Average S.D. Min Max 2,852 1.081 1.093 0 3
US: full sample (2) N Obs. Average S.D. Min 1,436 2.205 0.853 0
age (years) sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise) work: missing dummy current_married (=1 if yes, 0 otherwise) current_married: missing dummy
2,852 2,852 2,852 2,852 2,852 2,852
65.646 0.512 0.526 0.002 0.808 0.002
6.974 0.500 0.499 0.046 0.394 0.046
52 0 0 0 0 0
79 1 1 1 1 1
1,473 1,473 1,473 1,473 -
63.999 0.451 0.392 0.684 -
8.156 0.498 0.488 0.465 -
primary/junior high schoool senior_high junior_college university graduate Education variable: missing dummy
2,852 2,852 2,852 2,852 2,852 2,852
0.353 0.402 0.122 0.107 0.007 0.004
0.478 0.490 0.328 0.310 0.086 0.065
0 0 0 0 0 0
1 1 1 1 1 1
1,473 1,473 1,473 1,473 1,473 1,473
0.189 0.305 0.255 0.136 0.109 0.005
0.392 0.461 0.436 0.343 0.312 0.074
0 0 0 0 0 0
Serial 7 (=1 if correct, 0 otherwise) Immediate word recall ever studied accounting/economics? (=1 if yes, 0 otherwise) ever studied accounting/economics?: missing dummy
2,852 2,852 2,852 2,852
0.453 4.211 0.111 0.102
0.498 2.486 0.314 0.303
0 0 0 0
1 9 1 1
1,473 1,473 1,473 1,473
0.430 5.479 0.341 0.002
0.495 1.668 0.474 0.045
Income (household): $0-$20,000 Income (household): $20,000-$50,000 Income (household): $50,000-$100,000 Income (household): more than $100,000 Income (household): missing dummy
2,852 2,852 2,852 2,852 2,852
0.269 0.316 0.099 0.020 0.296
0.443 0.465 0.299 0.140 0.457
0 0 0 0 0
1 1 1 1 1
1,473 1,473 1,473 1,473 -
0.240 0.319 0.260 0.181 -
0.427 0.466 0.439 0.385 -
Having bonds (household level) Having stocks (household level) Having bond or/and stocks (household level) Having bank accounts (household level) Having CD/government saving bond/T-bill (household level) Having IRA accounts (household level)
2,430 2,412 2,397 2,412
0.202 0.157 0.269 0.918
0.402 0.364 0.443 0.275
0 0 0 0
1 1 1 1
1,473 1,473 1,473 1,473 1,473 1,473
0.038 0.219 0.232 0.771 0.160 0.382
0.191 0.413 0.422 0.420 0.366 0.486
Max 3
52 0 0
1,011 64.260 1,011 0.473 1,011 0.389 1,011 0.540 -
8.246 0.500 0.488 0.499 -
1 1 1 1 1 1
1,011 1,011 1,011 1,011 1,011 1,011
0.197 0.292 0.253 0.146 0.107 0.005
0.398 0.455 0.435 0.354 0.309 0.070
0 0 0 0 0 0
1 1 1 1 1 1
0 0 0 0
1 10 1 1
1,011 1,011 1,011 1,011
0.425 5.474 0.345 0.003
0.495 1.677 0.476 0.054
0 0 0 0
1 10 1 1
0 0 0 0
1 1 1 1
1,011 1,011 1,011 1,011 -
0.305 0.315 0.233 0.147 -
0.460 0.465 0.423 0.355 -
0 0 0 0
1 1 1 1
1,011 1,011 1,011 1,011 1,011 1,011
0.037 0.181 0.193 0.750 0.140 0.337
0.188 0.385 0.395 0.433 0.348 0.473
-
79 1 1
US: investment behavior sample (3) N Obs. Average S.D. Min Max 1,011 2.201 0.857 0 3
0
-
1 -
-
0 0 0 0 0 0
1 1 1 1 1 1
(Note) Authors' calculations. Incomes of Japanese households are converted to US$ using purchasing power parities in 2009 published by OECD. Source: https://data.oecd.org/conversion/purchasing-power-parities-ppp.htm
20
52 0 0 -
79 1 1 -
0 -
1 -
-
0 0 0 0 0 0
1 1 1 1 1 1
Table 3 Factors influencing financial literacy Japan
US (2) Logit
(1) OLS
-0.004*** -0.015** [0.004] [0.001] sex (=1 if male, 0 otherwise) 0.242*** 0.052*** [0.009] [0.042] 0.007 -0.016 work (=1 if working, 0 otherwise) [0.007] [0.049] -0.172 work: missing dummy [0.176] current_married -0.056 -0.005 [0.035] [0.012] current_married: missing dummy -0.481 [0.389] Education category (default: primary/junior high schoool) 0.036 0.156** senior_high [0.022] [0.047] junior_college 0.200* 0.055* [0.033] [0.086] 0.102*** 0.396** university [0.119] [0.031] 0.100*** 0.419*** graduate [0.079] [0.033] -0.010 Education variable: missing dummy -0.091 [0.062] [0.123] Serial 7 (=1 if correct, 0 otherwise) 0.146*** 0.017*** [0.004] [0.026] 0.016*** 0.092*** Immediate word recall [0.005] [0.013] ever studied accounting/economics? 0.227* 0.056*** [0.012] [0.109] (=1 if yes, 0 otherwise) 0.034** ever studied accounting/economics? 0.047 missing dummy [0.036] [0.017] Household income category (default: $0-$20,000) $20,000-$50,000 0.243** 0.058*** [0.054] [0.004] $50,000-$100,000 0.271** 0.058*** [0.088] [0.011] more than $100,000 0.343* 0.077*** [0.154] [0.026] income: missing dummy -0.119 -0.016 [0.066] [0.026] Constant 1.488*** [0.288] age
(3) OLS
(4) Logit
-0.004 [0.002] 0.241*** [0.040] 0.032 [0.044]
-0.001 [0.001] 0.129*** [0.015] 0.029 [0.022]
0.035 [0.044]
0.006 [0.030]
0.189*** [0.035] 0.278*** [0.068] 0.476*** [0.056] 0.507*** [0.048] 0.168 [0.443] 0.164*** [0.040] 0.056*** [0.011] 0.124** [0.041] 0.161 [0.320]
0.118*** [0.024] 0.153*** [0.034] 0.275*** [0.028] 0.278*** [0.045] 0.087 [0.237] 0.072*** [0.020] 0.032*** [0.005] 0.067*** [0.023] -0.059 [0.157]
0.093** [0.039] 0.167** [0.053] 0.223** [0.079]
0.026 [0.032] 0.078** [0.039] 0.126** [0.053]
1.473*** [0.179]
Observations 2,852 2,840 1,473 1,473 R-squared 0.201 0.191 Region dummies Yes Yes Yes Yes The dependent variable in (1) and (3) is the number of correct answers. The dependent variable in (2) and (4) takes 1 if a respondent got all correct answers and 0 otherwise. Marginal effects are reported for logit models ((2) and (4)). Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
21
Table 4 Financial literacy and investment behaviors: Bond
Number of correct answers age sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise)
Japan
US
(1) Logit
(2) Logit
0.049*** [0.008] 0.003*** [0.001] -0.047*** [0.012] -0.060*** [0.018]
0.021*** [0.004] 0.002*** [0.001] 0.002 [0.009] -0.016 [0.010]
-0.004 [0.019]
0.010 [0.014]
work: missing dummy current_married current_married: missing dummy Education category (default: primary/junior high schoool) senior_high 0.096*** [0.031] junior_college 0.110*** [0.034] university 0.083*** [0.015] graduate 0.213*** [0.067] Education variable: missing dummy 0.185** [0.086] Serial 7 (=1 if correct, 0 otherwise) 0.007 [0.013] Immediate word recall -0.005* [0.003] ever studied accounting/economics? 0.086*** (=1 if yes, 0 otherwise) [0.006] ever studied accounting/economics? -0.021 missing dummy [0.037] Household income category (default: $0-$20,000) $20,000-$50,000 0.044* [0.024] $50,000-$100,000 0.120*** [0.032] more than $100,000 0.174*** [0.018] income: missing dummy -0.010 [0.011]
-0.012 [0.030] 0.009 [0.028] 0.028 [0.020] 0.047** [0.019]
0.000 [0.005] 0.004 [0.003] -0.012 [0.016]
0.048 [0.040] 0.059 [0.039] 0.092** [0.042]
Observations 2,420 1,001 Region dummies Yes Yes The dependent variable takes 1 for a bond holder and 0 otherwise. Marginal effects are reported. Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
22
Table 5 Financial literacy and investment behaviors: Stock
Number of correct answers age sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise)
Japan
US
(1) Logit
(2) Logit
0.030*** [0.002] 0.002** [0.001] -0.014 [0.014] -0.039*** [0.015]
0.051*** [0.012] 0.005*** [0.002] 0.004 [0.029] -0.038** [0.017]
0.026 [0.028]
-0.002 [0.016]
work: missing dummy current_married current_married: missing dummy Education category (default: primary/junior high schoool) senior_high 0.110*** [0.017] junior_college 0.146*** [0.028] university 0.174*** [0.017] graduate 0.219*** [0.043] Education variable: missing dummy 0.166*** [0.060] Serial 7 (=1 if correct, 0 otherwise) -0.012 [0.011] Immediate word recall -0.002 [0.002] ever studied accounting/economics? 0.033*** (=1 if yes, 0 otherwise) [0.011] ever studied accounting/economics? -0.093 missing dummy [0.057] Household income category (default: $0-$20,000) $20,000-$50,000 0.044*** [0.013] $50,000-$100,000 0.095*** [0.023] more than $100,000 0.167*** [0.049] income: missing dummy 0.008 [0.021]
0.079* [0.047] 0.115** [0.054] 0.174*** [0.050] 0.240*** [0.039]
0.014 [0.024] 0.004 [0.007] -0.011 [0.031]
0.126*** [0.039] 0.131*** [0.047] 0.230*** [0.052]
Observations 2,403 1,003 Region dummies Yes Yes The dependent variable takes 1 for a stock holder and 0 otherwise. Marginal effects are reported. Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
23
Table 6 Financial literacy and investment behaviors: Bank accounts
Number of correct answers age sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise)
Japan
US
(1) Logit
(2) Logit
0.006 [0.007] 0.001 [0.001] -0.005 [0.011] -0.002 [0.013]
0.026** [0.010] 0.006*** [0.001] -0.027* [0.016] -0.031 [0.027]
0.046*** [0.014]
-0.004 [0.021]
work: missing dummy current_married current_married: missing dummy Education category (default: primary/junior high schoool) senior_high 0.051*** [0.015] junior_college 0.053** [0.027] university 0.036** [0.015] graduate Education variable: missing dummy Serial 7 (=1 if correct, 0 otherwise)
0.025*** [0.008] Immediate word recall 0.005* [0.003] ever studied accounting/economics? 0.029* (=1 if yes, 0 otherwise) [0.015] ever studied accounting/economics? 0.026 missing dummy [0.042] Household income category (default: $0-$20,000) $20,000-$50,000 0.044*** [0.013] $50,000-$100,000 0.133*** [0.039] more than $100,000 income: missing dummy
0.168*** [0.014] 0.208*** [0.020] 0.213*** [0.028] 0.308*** [0.078] 0.070 [0.306] 0.038** [0.016] 0.016* [0.010] -0.042* [0.024] 0.043 [0.249] 0.146*** [0.041] 0.208*** [0.047] 0.246*** [0.034]
0.030* [0.016]
Observations 2,321 980 Region dummies Yes Yes The dependent variable takes 1 for a bank account holder and 0 otherwise. Marginal effects are reported. Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
24
Appendix 1 Factors influencing financial literacy in Japan (with a full set of the covariates) (1) OLS age
-0.018** [0.005] sex (=1 if male, 0 otherwise) 0.244*** [0.043] work (=1 if working, 0 otherwise) -0.034 [0.045] work: missing dummy -0.122 [0.196] current_married -0.059 [0.034] current_married: missing dummy -0.539 [0.373] Education category (default: primary/junior high schoool) senior_high 0.156** [0.051] 0.198* junior_college [0.074] university 0.393** [0.114] graduate 0.423*** [0.091] -0.104 Education variable: missing dummy [0.133] 0.105** Serial 7 (=1 if correct, 0 otherwise) [0.025] Immediate word recall 0.068** [0.019] ever studied accounting/economics? 0.232* (=1 if yes, 0 otherwise) [0.107] ever studied accounting/economics? 0.099 [0.052] missing dummy Household income category (default: $0-$20,000) $20,000-$50,000 0.226** [0.051] $50,000-$100,000 0.252** [0.089] more than $100,000 0.318* [0.137] income: missing dummy -0.116 [0.074] Risk preference (default: no uncertainty) risk: 10-30% -0.112 [0.056] risk: 40-60% -0.052 [0.033] risk: 70-90% 0.076 [0.149] risk preference: missing dummy -0.164*** [0.032] Time discounting preference (default: 0-0.5%) time: 1-6% 0.155** [0.043] time: 10-40% 0.089 [0.056] time discounting preference: -0.011 missing dummy [0.066] CESD (=1 if score>=20, 0 otherwise) -0.092 [0.045] CESD: missing dummy -0.067 [0.045] subjective probability living up to 85 0.002 [0.001] Constant 1.741*** [0.310]
(2) Logit -0.004*** [0.001] 0.050*** [0.009] 0.003 [0.006]
-0.005 [0.011]
0.034 [0.025] 0.052 [0.033] 0.100*** [0.031] 0.097*** [0.036] -0.019 [0.072] 0.011** [0.005] 0.015*** [0.005] 0.056*** [0.013] 0.044* [0.023] 0.053*** [0.006] 0.052*** [0.013] 0.072*** [0.024] -0.016 [0.030] -0.022 [0.023] -0.025 [0.015] 0.009 [0.040] -0.035** [0.015] 0.056*** [0.015] 0.032* [0.017] 0.047*** [0.011] -0.033** [0.015] -0.022 [0.021] 0.000 [0.000]
Observations 2,852 2,840 R-squared 0.214 Region dummies Yes Yes The dependent variable in (1) is the number of correct answers. The dependent variable in (2) takes 1 if a respondent got all correct answers and 0 otherwise. Marginal effects are reported for (2). Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
25
Appendix Table 2 Financial literacy and investment behaviors: Bond or/and Stock
Number of correct answers age sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise)
Japan
US
(1) Logit
(2) Logit
0.055*** [0.007] 0.004*** [0.001] -0.035** [0.014] -0.052*** [0.020]
0.055*** [0.009] 0.006*** [0.002] -0.009 [0.029] -0.042*** [0.015]
-0.000 [0.016]
0.009 [0.013]
work: missing dummy current_married current_married: missing dummy Education category (default: primary/junior high schoool) senior_high 0.130*** [0.028] junior_college 0.173*** [0.042] university 0.168*** [0.025] graduate 0.236*** [0.065] Education variable: missing dummy 0.174* [0.100] Serial 7 (=1 if correct, 0 otherwise) 0.003 [0.009] Immediate word recall -0.004* [0.002] ever studied accounting/economics? 0.085*** (=1 if yes, 0 otherwise) [0.016] ever studied accounting/economics? -0.043 missing dummy [0.056] Household income category (default: $0-$20,000) $20,000-$50,000 0.058** [0.024] $50,000-$100,000 0.159*** [0.027] more than $100,000 0.233*** [0.073] income: missing dummy -0.007 [0.021]
0.074 [0.057] 0.109* [0.060] 0.172*** [0.052] 0.251*** [0.041]
0.013 [0.025] 0.005 [0.007] -0.015 [0.022]
0.141*** [0.035] 0.147*** [0.045] 0.251*** [0.047]
Observations 2,388 1,003 Region dummies Yes Yes The dependent variable takes 1 for a bond and/or stock holder and 0 otherwise. Marginal effects are reported for logit models. Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
26
S2212-828X(19)30101-X https://doi.org/10.1016/j.jeoa.2019.100214 JEOA 100214
To appear in:
The Journal of the Economics of Ageing
Received Date: Revised Date: Accepted Date:
12 December 2018 17 September 2019 19 September 2019
Please cite this article as: S. Shimizutani, H. Yamada, Financial Literacy of Middle-Aged and Older Individuals: Comparison of Japan and the United States, The Journal of the Economics of Ageing (2019), doi: https://doi.org/ 10.1016/j.jeoa.2019.100214
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2019 Published by Elsevier B.V.
Financial Literacy of Middle-Aged and Older Individuals: Comparison of Japan and the United States*
by
Satoshi Shimizutani and Hiroyuki Yamada** September 2019 Abstract Financial literacy holds growing interest for managing assets/savings during the longer retirement period currently experienced in rapidly aging countries. We examine and compare levels and determinants of financial literacy as well as its association with asset allocation among middle-aged and older individuals of Japan and the United States. We present some interesting findings. First, financial literacy is generally influenced by educational attainment, cognitive skills, coursework in economics or finance, and income level. Second, financial literacy is associated with household asset allocation; individuals with higher literacy also have investment in stocks or securities. These patterns are commonly observed both in Japan and the United States. Keywords: financial literacy, Japan, U.S., JSTAR, HRS, household asset allocation. JEL Classification Codes: D14, G11, J26.
*
This work was presented at the Gateway to Global Aging Data Research Conference, funded by the National Institute on Aging (2R01 AG030153). We thank our discussant, Silvia Barcellos, conference participants, one anonymous referee, and the editors for their constructive comments. The views expressed in this paper are completely personal and unrelated to those of any organizations with which we are affiliated or those who commented on this study. **
Satoshi Shimizutani (corresponding author); Visiting Research Fellow, Nakasone Yasuhiro Peace Institute (NPI); Postal Address: Toranomon 30 Mori Building 6F, 3-2-2 Toranomon, Minato-ku, Tokyo 105-0001 Japan; E-mail: [email protected]. Hiroyuki Yamada; Professor, Faculty of Economics, Keio University; Postal address: 2-15-45 Mita, Minato-ku, Tokyo 108-8345 Japan, Tel (Direct): +81-(0)3-5427-1271 E-mail: [email protected]. 1
1. Introduction Financial literacy holds growing interest for managing assets/savings during the longer retirement period currently experienced in rapidly aging countries. Indeed, the case is pronounced in and relevant to Japan, where population aging is taking place at the most rapid rate in the world in an environment that combines a persistent long-term low level fertility rate and continuing rise in life expectancy.1 These factors have resulted in the older generation occupying a higher proportion of the total population at 27.3% while putting pressure on the sustainability of the pay-as-you-go public pension program (Shimizutani, et al. 2016). Under these circumstances, there are two forces that act to lower the adequacy of retirement savings and thus obstruct the management of economic resources in later life for middle-aged and older individuals. One is a reduction in the amount of pension benefits to receive and the other is a longer time horizon that has to be compensated. A higher level of individual financial literacy may serve as a device to ensure economic well-being or even enhance the standard of living at older ages.2 Research on financial literacy has developed significantly since the 2000s (Lusardi and Mitchell, 2014) for the U.S. and other Western countries, but is still relatively scarce in Japan. Currently a widely used, standard measure of financial literacy is the number of correct answers to a 3-question quiz inquiring about compound interest, inflation, and risk/diversification. These questions were first asked 1
According to National Institute of Population and Social Security Research, the latest TFR (Total Fertility Rate) of Japan is 1.44 in 2016. The figure declined to below 1.3 in 2003-2005 and slightly recovered afterwards. The post-war development of TFR is described in detail by Shimizutani (2015). The latest life expectancy at birth in Japan is 80.98 year for males and 87.14 year for females in 2016, both of which represent the second longest life expectancy in the world. 2 Individuals with higher levels of financial literacy tend to possess greater retirement wealth, other things equal (Lusardi and Mitchell, 2011a, 2014; Clark et al. 2015). The reasons why more financially literate individuals are wealthier include (1) they plan and save more; (2) they may earn higher returns on their investments (Lusardi et al., 2017); (3) they may select better diversified investments. While the study of retirement wealth is of great importance, the subject is beyond the scope of this paper which focuses on a comparison of Japan and U.S. financial literacy. 2
in 2004 in the Health and Retirement Study. Using these and other metrics of financial literacy, Lusardi and Mitchell (2014) conclude that in the U.S. the level of financial literacy is low. Moreover, the low level of financial literacy is found to be prevalent in other countries including Japan (Lusardi and Mitchell, 2011b). When looking at the data by subgroup, the level of financial literacy is hump-shaped along with age and is higher for males, the educated, employed individuals, or those who are higher earners. Furthermore, a large volume of empirical research documented that the standard measure of financial literacy is indeed associated with a variety of financial decisions or economic behaviors such as stock market participation (van Rooij et.al, 2011) and retirement saving (Lusardi and Mitchell (2007)), and financial literacy is confirmed to play a non-negligible role in those decisions/behaviors if instrument variable (IV) estimation is employed to address endogeneity. This study examines factors influencing financial literacy and the association of financial literacy with household financial decisions among middle-aged and older individuals aged 52-79. Using harmonized micro-level data, we perform parallel analyses in Japan and the U.S., highlighting differences and similarities between the two countries as far as empirical findings are concerned. Research on financial literacy in Japan has been relatively scarce, let alone studies of individuals approaching or entering retirement age. Because of the demographic and economic changes mentioned above, understanding heterogeneity in financial literacy and decision-making within this segment of the population is especially important and a major contribution of this paper. Among the few previous studies, Clark et al. (2013) revealed that a higher level of financial literacy is found in males, residents in urban areas, the higher educated, or 3
high income earners in Japan.3 The study utilized a micro-level dataset on individuals from the 2010 National Survey on Work and Family, focusing on individuals in the sample who were aged 40 to 59 and employed at the time of the survey. Furthermore, they found that financial literacy is partially linked with the demand for human capital investment, measured by the need for additional skills to be competitive at the current job.4 As regards financial literacy for younger individuals, Cameron et al. (2013) showed that financial literacy is higher among Japanese high school students than in the U.S. and New Zealand. In a more recent work, Yoshino et al. (2017) used a dataset collected by the Central Council for Financial Service Information, which is not based on a standard two stage random sampling, to examine factors influencing financial literacy and the relationship between financial literacy and household asset allocation. While the data used by Yoshino et al. (2017) contains rich variables related with financial literacy, they are not internationally comparable and thus are not suitable for an international comparison study. This study performs two sets of analysis: the first is on the factors that influence financial literacy and the second is the association between financial literacy and household asset allocation. Our contribution to the literature features two new enhancements. First, we perform a Japan-U.S. comparison using internationally standardized datasets. While we share the spirit of the analytical approach with Yoshino et al. (2017), we use world-standard longitudinal datasets on the middle-aged and older individuals, namely Japanese Study on Aging and Retirement (JSTAR) for Japan, and Health and Retirement Study (HRS) for the U.S. These surveys are “sisters” in international harmonization efforts for studies on population aging, and most of the variables collected, including the three metrics of financial literacy, are standardized 3 4
Sekita (2011) performed a similar analysis and obtained similar results. Clark et al. (2013) did not explore an association between financial literacy and asset holdings. 4
across countries. Thus, these datasets are most suitable for a Japan-U.S. comparison in financial literacy, as described in the next section. Second, this paper advances the research in Japan. We focus on middle-aged and older individuals aged 52 to 79 who are sensitive to retirement preparation or who are indeed retired. Moreover, we include some new variables as covariates which were not included in the previous literature in Japan, specifically, measures of cognitive ability and experience of taking an economics/accounting coursework.5 We believe that these additional variables will help to uncover a more accurate assessment of the impact of socio-economic status on financial literacy or household asset allocation.6 This paper is organized as follows. Section 2 describes the dataset used in the empirical analysis and presents descriptive statistics of financial literacy. Section 3 performs regression analysis to associate the level of financial literacy with a variety of factors including household demographics, socio-economic status, and measures of cognitive ability, and so on. Further, Section 4 investigates the relationship between financial literacy and household asset allocation. Section 5 concludes.
2. Data description This study uses microdata from two sets of comparable datasets, JSTAR and
5
Those variables are not available in the dataset used by Yoshino et al. (2017). Other candidates for new variables include those related to behavioral economics (i.e., risk attitude and time discounting), those that capture current mental state (symptoms of depression), and expectations for the future (survival probability to measure the life horizon). While HRS contains variables regarding symptoms of depression and expectations for the future (survival probability to measure the life horizon), the styles used in HRS and JSTAR vary. For symptoms of depression, HRS uses an 8-item version of the Center for Epidemiologic Studies Depression Scale (CESD) while JSTAR uses the 20-item version of that the CESD. For survival probability, the target age of survival differs depending on the respondent’s age in HRS while it is the same for all respondents in JSTAR. Therefore, we did not use those variables in the comparison. We confirmed that most of those variables indeed significantly influence financial literacy and hold an expected sign in the regression. The estimation results are reported in detail in Shimizutani and Yamada (2018) and some that are related to factors influencing financial literacy are in Appendix Table 1. The measurement of risk attitude and time discounting is also explained in the appendix of Shimizutani and Yamada (2018). 5 6
HRS. JSTAR is a world-standard longitudinal dataset on middle-aged and older individuals and holds the position as the Japanese counterpart of HRS in the U.S., English Longitudinal Study of Ageing (ELSA) in the U.K., the Survey of Health, Ageing and Retirement in Europe (SHARE) in continental Europe. Similar to those surveys, JSTAR’s questionnaire covers a wide variety of variables related to health, employment, economic status, and other life features. The unit of the sample of JSTAR at baseline is individuals aged 50 to 75 who were randomly chosen from household registration within each municipality. If a respondent is married, selected questions are also asked of the spouse. JSTAR began collection of data in 2007 in five municipalities in the first wave, followed by four subsequent waves at two-year intervals.7 The second JSTAR wave of data, collected in 2009, contains the standard measure of financial literacy. In addition, the wave includes a related question on whether the respondent has ever studied accounting or economics. In the second wave, the total sample size is 2,852 persons aged 52 to 79. HRS is the pioneer and the eldest sister in the field of international data collection and harmonization efforts on middle-aged and older individuals.8 The HRS survey was implemented in 1992 and tracks the same people every two years. HRS covers a wide range of information that is also covered by JSTAR. Further, the timing of the surveys coincides. In 2010, HRS includes the standard measure of financial literacy. Note that the special module is asked of only a subset of the sample (one-tenth that are randomly sampled). The sample size is 1,473 persons aged 52 to 79 in the same age range as those in JSTAR. 7
The total number of municipalities from the third wave is 10 and the total number of individuals in the sample is about 8,000 at baseline. The five municipalities in the first wave are Takikawa city, Sendai city, Adachi ward (metropolitan Tokyo), Kanazawa city, and Shirakawa town. 8 The detailed description of the dataset is available at https://hrs.isr.umich.edu/about?_ga=2.94165525.941904344.1535601989-1380807771.153560198 9 6
Both JSTAR and HRS pose the financial literacy question to an individual in a household (i.e., either husband or wife if a household is a “couple”) but whether or not the surveyed individual is the most financially knowledgeable in the household can be known only in HRS. This information is important because the most financially knowledgeable individual in the household responded to the financial asset questions in HRS. Thus, we use HRS data from all individuals when examining the association between the level of financial literacy and a variety of covariates (henceforth, “U.S. full sample”) but use the data from only the most financially knowledgeable in the household when examining the association between household asset allocation and the covariates. The sample size of respondents to this question is reduced to 1,011 persons (henceforth, “investment behavior sample”). Another difference between JSTAR and HRS lies in the questions regarding asset holding. In JSTAR, if a respondent has a spouse (or a partner), the question regarding holding of each type of asset is asked separately of the respondent and the spouse. On the other hand, HRS asks questions only of “household” level asset holdings. In other words, we cannot identify whether an asset is held by the respondent or his/her spouse (or both).9 For the sake of a Japan-U.S. comparison, we construct variables of household level asset holdings from JSTAR. We interpret a household as holding a positive value of assets in the case that at least either the respondent or the partner holds the corresponding asset. Table 1 presents simple descriptive statistics of each measure of financial literacy. Note that the HRS sample is limited to people aged between 52 and 79 to correspond to the age of JSTAR respondents. At a glance, the level of financial literacy is much lower in Japan than that in the U.S. For example, Panel A shows that 9
For the U.S., we use RAND HRS Detailed Imputation File 2014 (V2), released in February 2018, for the information on each type of asset holding. 7
the proportion of correct responses to “compound interest” is 38.6% in Japan, which is substantially lower than that in the U.S. (71.8%). This is also the case for “inflation” in Panel B; the proportion of correct response is 38.7%, much lower than 82.4% in the U.S.; for “stock risk” in Panel C, the proportion of correct response is 30.9%, much lower than 66.4% in the U.S. The bottom of Table 1 reports some summary statistics across questions. Only 13.9% give correct answers to all three questions in Japan while 44.6% do so in the U.S. In contrast, about 42% give no correct answers in Japan while this proportion is only 4.1% in the U.S. About 70% choose “Don’t Know” or “Not Sure” to at least one out of the three questions in Japan while this proportion is much lower in the U.S. at 18.8%. Moreover, about 30% choose “Don’t Know” or “Not Sure” to all three questions, while the proportion of such individuals in the U.S. is negligible. It might be tempting to conclude that the level of financial literacy is lower in Japan, but we should be careful how to interpret these figures since the proportion of “Don’t Know” or “Not Sure” is much higher (42.6%, 47.9% and 53.9%, respectively, for each of the three questions) in Japan.10 These high proportions of “Don’t Know” or “Not Sure” are not necessarily specific to this dataset (i.e. JSTAR). The proportion of “Don’t Know” or “Not Sure” for each of the three questions, respectively, is 12.5%, 28.6%, and 56.1% in Sekita (2011), 14.3%, 27.2%, and 53.8% in Clark et al. (2013), and 24.0%, 34.1%, and 49.0% in the Central Council for Financial Service Information (2016).11 Note that the samples of Sekita (2011), Clark et al. (2013), and
10
Indeed, the proportion of “incorrect” answers is larger in the U.S. than in Japan for all three questions. 11 Note that the question wording and choice set of answers are not exactly the same across these datasets. For example, the question words of the “compounding interest” and “stock risk” questions are the same in the Central Council for Financial Service Information (2016) and JSTAR but the choice set of answers is different between them. Neither question words nor choice set of answers of the “inflation” question are the same between the Central Council for Financial Service Information (2016) and JSTAR. 8
the Central Council for Financial Service Information (2016) are not restricted to middle-aged and older individuals. Thus, the proportions of “Don’t Know” or “Not Sure” among the Japanese samples are in general much higher than that of the HRS sample. This might indicate that Japanese are very cautious and answer only when confident in their response (Sekita 2011). Further, we compare the basic characteristics of those who chose (a) “Don’t Know” or “Not Sure,” (b) the incorrect answer(s), and (c) the correct answer for each of the three questions in JSTAR. The sample of (a) are older, higher number of females, less likely to work and be currently married, less educated (in general education and economics/accounting courses), and have lower cognitive skills than the samples of (b) and (c).1213 Table 2 shows summary statistics for the variables used in the analysis below. We focus on the comparison between the Japan study and the U.S. full sample below.14 First, the number of correct answers is different between Japan (1.08) and the U.S. (2.20). The average age is 65.6 years old in Japan and 64.0 in the U.S. In Japan, more than one half of the respondents were working while the figure was under 40% in the U.S. More than 80% were currently married in Japan but the proportion was lower in the U.S. Educational attainment is higher in the U.S. than in Japan. In Japan, the proportion of senior high school graduates or under is three quarters and the proportion of university graduates or higher is slightly larger than 10%. In the U.S., the proportion of senior high school graduates or under is slightly less than one half 12
The detailed descriptive statistics for this comparison are available upon request from the authors. 13 One could argue, for instance, that in countries with a history of high inflation there might be more knowledge of its consequences. However, this concern could apply only to the second question on inflation and it could be a minor issue in our context because neither Japan nor the U.S. has experienced high inflation rates for many years. We acknowledge that the populations are different due to many varying characteristics between the two countries but confined our analysis to the variables which were commonly used in the previous literature. 14 In comparing the number of correct answers between “U.S. full sample” and “U.S. investment behavior sample.” that of the former is very slightly higher than that of latter although the latter is the sample of financially responsible individuals. 9
and the proportion of university graduates or higher is close to one quarter. Looking at the cognitive measure, the proportion of correct answers to serial sevens is 45.3% and the average number of 10 word immediate recall is 4.2 words in Japan. The proportion of correct responses to serial sevens is slightly lower in the U.S. while the number of word recall is higher. The proportion of those who have ever studied accounting or economics is slightly more than 10% in Japan while the figure in the U.S. is three-fold (34.2%). Turning to household income, the dominant bracket of household income is 20,000-50,000 dollars in both countries (31.6% in Japan and 31.9% in the U.S.). Lastly, household asset holding reveals that the proportion of risky financial instruments is low in both countries; bond holders make up about 20% and stockholders occupy less than 20% in Japan. The proportion of bond holders is much smaller in the U.S. but that of stockholders is larger, exceeding 20%. The lower share of bond holders may be due to the different definition or perception on the part of respondents. Indeed, the proportion of CD/government saving bonds/T-bills, which are not included in bonds, is 16.0% in the U.S. The share of either bond or stockholders is 26.9% in Japan and 23.1% in the U.S. On the flip side, about 90% have saving accounts in Japan, suggesting that Japanese middle-aged and older individuals are generally conservative in asset investment. The proportion of bank account holders is lower at 77.1% in the U.S. but it is noted that more than one-third of households have IRA accounts. In Japan, IRA-type accounts were scarcely available at the time of the second wave survey in 2009. Therefore, no corresponding question regarding this type of asset was included in JSTAR in 2009.
3. Factors that influence financial literacy In this section, we examine factors influencing financial literacy by regression 10
analysis. The basic specification for the analysis is described as follows.
(1) 𝐹𝐿𝑖 = 𝛽𝑋𝑖 + 𝜀𝑖
where 𝐹𝐿𝑖 measures the level of financial literacy referring to the number of correct answers to the set of three financial literacy questions (OLS estimation). We also run the regressions using a logit model by replacing the dependent variable to a binary variable to take 1 if all questions were answered correctly and 0 otherwise. This measure may be less noisy since there is some randomness as to whether people get one or two questions correct. The vector of explanatory variables, Xi, includes age, sex, marital status, educational attainment, cognitive skills (serial sevens and immediate word recall test score), an indicator for whether the individual has taken courses in economics or accounting, household income, and region fixed-effects. Columns (1) and (3) in Table 3 report the estimated coefficients using OLS for Japan and the U.S. when taking the number of correct answers as dependent variables, respectively and Columns (2) and (4) show marginal effects using a logit model for each country when taking a binary variable as the dependent variable. First, looking at individual characteristics, the coefficients and marginal effects on age are negative and significant in Japan in Columns (1) and (2) but they are not significant in the U.S. in Columns (3) and (4). The coefficients and marginal effects on sex are positive and significant in both countries, showing that the level of financial literacy is higher for males. These results on sex are common and consistent with findings in previous studies. Our results show that being male increases the number of correct answers by 0.24 in both countries while it improves the likelihood of answering all questions correctly by 13 percentage points in the U.S. but only 5 percentage points in Japan. 11
The coefficients and marginal effects on current work status and marital status are not significant. The coefficients and marginal effects are significant and larger for higher educational attainment, which again conforms to the result in previous studies. The coefficients on each school attainment level are positive and significant (the default case is primary/junior high school graduates) for both countries, but it is noted that the gradient is steeper in the U.S. Columns (1) and (3) show that the number of correct answers increases by 0.4 in Japan and by 0.5 in the U.S. in the case of completion of graduate school. Columns (2) and (4) show a larger gap between Japan and the U.S.; the likelihood of answering all questions correctly increases by 28 percentage points in the U.S. and only 10 percentage points in Japan if graduate school has been completed. The coefficients and marginal effects are positive and significant for serial sevens and immediate word recall again for both countries. This is natural since respondents with higher cognitive skills are more likely to retain a higher level of financial literacy. While the coefficients in Columns (1) and (3) are roughly comparable, the likelihood to answer all questions correctly is larger in the U.S. in the case of serial sevens; specifically, it is 7 percentage points in the U.S. and 2 percentage points in Japan. This is also the case for taking economics/accounting coursework and the coefficients and marginal effects are positive and generally significant. While the likelihood of answering all questions correctly is similar in both countries, the coefficient in Column (1) is larger than in Column (3), showing that the experience of having studied accounting or economics increases the number of correct answers by 0.23 in Japan and by only 0.12 in the U.S. Turning to household income, the coefficients and marginal effects are larger for a higher level of income (the default case is 0 to 20,000 dollars), showing that higher earners are more likely to enjoy a higher level of financial literacy. The gradient is steeper in Japan when 12
considering the number of correct answers which increase by 0.34 if household income is larger than 100,000 dollars and by only 0.22 in the U.S. In contrast, the gradient is steeper in the U.S. when considering the likelihood of answering all questions correctly which increases by 13 percentage points in the U.S. if household income is higher than 100,000 dollars and by only 8 percentage points in Japan. In sum, we found a higher level of financial literacy in individuals who are male, more educated, with higher cognitive skills, or a higher annual income. These patterns are confirmed in both countries.15
4.
Financial literacy and household asset allocation Next, we aim to understand the level of financial literacy that associates with
household asset holding. By “household” assets, we mean bonds, stocks, bank accounts (deposits) in the name of either the respondent or his/her spouse (or partner) both in Japan and the U.S. The basic specification for the analysis is described as follows.
(2) 𝐴𝑗𝑖 = α𝐹𝐿𝑖 + 𝛽𝑋𝑖 + 𝜀𝑖
where 𝐴𝑗𝑖 is an indicator taking value 1 if household i holds asset type j in its portfolio and 0 otherwise. We run separate regressions for each asset type j. The main
15
We present the estimation results by adding some new covariates in Appendix 1 which are available only in Japan. First, the coefficient on time discounting is positive and significant for the 1-6% range, showing that slightly patient individuals are more likely to have a higher level of financial literacy. Second, the coefficients on an incidence of depression symptoms by the CESD measure are negative and significant, showing that respondents with depression symptoms are likely to have lower financial literacy. Third, the coefficients on the subjective probability to live to age 85 are now positive (but marginally insignificant), showing that a subjective assessment of a longer life time horizon is linked to a higher level of financial literacy. We obtain a similar pattern on the coefficients if we use individual income instead of household income in Japan (Shimizutani and Yamada 2018). 13
independent variable 𝐹𝐿𝑖 is the level of financial literacy of the surveyed individual in the household in JSTAR and of the most financially knowledgeable in the household in HRS measured by the number of correct answers to the three questions, which connect financial literacy to household asset holding. The other covariates (a vector of Xi) refer to individual characteristics which are exactly the same in specification (1): age, sex, work status, marital status, educational attainment, cognitive skills (serial sevens and immediate word recall), taking coursework in economics or accounting, household income, and dummy variables to indicate each region. We present the marginal effects using logit estimation. In what follows, we will focus on the marginal effects of the main variable, the level of financial literacy. Table 4 reports the results for bond holding in Japan (Column (1)) and the U.S. (Column (2)). The marginal effects on financial literacy are positive and significant, showing that financial literacy is indeed associated with bond holding. Concretely, one additional correct answer to the financial literacy questions increases the likelihood of bond holding by about 5 percentage points in Japan and 2 percentage points in the U.S. In addition, we observe that older, non-working individuals with higher educational attainment or higher household income level are associated with bond holding in both Japan and the U.S. Table 5 shows the marginal effects for stock holding in Japan (Column (1)) and the U.S. (Column (2)). The marginal effects on financial literacy are positive and significant, again showing that financial literacy is associated with stock holding. Concretely, one additional correct answer to the financial literacy questions increases the likelihood of bond holding by about 3 percentage points in Japan and 5 percentage points in the U.S. Interestingly, the size of the marginal effect on financial literacy is smaller in stock holding than bond 14
holding in Japan but the pattern is reverse in the U.S. The marginal effects on other covariates are generally the same as in the bond-holding.16 Table 6 reports the marginal effects for bank account holders in Japan (Column (1)) and the U.S. (Column (2)).Contrary to the results on bonds and stocks, the marginal effects contrast between Japan and the U.S.; those on financial literacy are not significant in Japan but are positive and significant in the U.S. probably because bank account holders are very prevalent in Japan. The marginal effect on age is not significant in Japan but is positive and significant in the U.S. due to the same reason. In contrast, the marginal effects on being married are positive and significant in Japan but they are not significant in the U.S. While the marginal effects on each educational attainment are positive and significant in both countries, the gradient along education is steeper in the U.S. than in Japan. This is also the case for income. Most of the marginal effects on income quartiles are positive and significant and the gradient along income is steeper in the U.S. than in Japan. In sum, financial literacy is associated with household asset allocation; households with a higher literacy also have more investment in bonds or stocks, apart from their savings.17 These patterns are commonly observed in Japan and the U.S.18,19
5. 16
Conclusions
Table 2 shows the proportion of bond holding is very different between Japan and the U.S. As discussed, the difference may come from a different definition or perception of respondents. Thus, we perform the same analyses when the dependent variable takes 1 if a respondent is a bond and/or stockholder and 0 otherwise. Appendix 2 shows that the marginal effects on financial literacy are positive and significant and those on other covariates are generally the same. 17 The obtained association between financial literacy and stock market participation is qualitatively consistent with the past literature (van Rooij et al. 2011, Yoong 2011, Almenberg and Dreber 2015, and Arrondel et al 2012). 18 We also conduct regression analysis using the principal component of the answers of the three questions as a variable of financial literacy and the results are qualitatively the same. The results are available upon request from the authors. 19 We conducted the same analyses using the “U.S. full sample” and the patterns of coefficients of the level of financial literacy are qualitatively very similar to when using “investment behavior sample.” The results are available upon request from the authors. 15
We examine and compare the levels and determinants of financial literacy as well as its association with asset allocation among middle-aged and older individuals in Japan and the U.S. In both countries, we found a higher level of financial literacy in individuals who are male, more educated, have higher cognitive skills, or a higher level of annual income, results which conform to findings by previous studies. Moreover, we found that the level of financial literacy is significantly associated with household asset holding and the impact is larger for bond and stock holdings than for savings deposits. Financial literacy holds growing interest for managing assets/savings during the longer retirement period currently experienced in rapidly aging countries, a societal feature particularly relevant to Japan. It is interesting to find the same pattern in the factors influencing financial literacy and the association between financial literacy and asset holding in Japan and the U.S. since it is widely perceived that Japan lags in financial education and the U.S. is more advanced. Further research should develop in two directions. One direction is to investigate a causal relationship between financial literacy and asset holding. Based on the examination, the other direction of further study should put the cases of Japan and the U.S. in an international perspective to extract policy implications. While improvement of financial literacy in middle-aged and older individuals seems to be a common policy agenda, consensus on concrete policy intervention is lacking. One reason is that there are too many factors influencing financial literacy and the effectiveness of intervention, which depends on policy instrument, is ambiguous. An international research project using comparable datasets would contribute toward revealing policy features that may more effectively increase financial literacy in different country-specific contexts.
16
References Almenberg, Johan, and Anna Dreber. (2015) “Gender, stock market participation and financial literacy.” Economics Letters, vol.137, pp140-142. Arrondel, Luc, Majdi Debbich, and Frédérique Savignac. (2012) “Stockholding and financial literacy in the French population.” International Journal of Social Sciences and Humanity Studies, vol.4, no.2, pp 285-294. Cameron, Michael, Richard Calderwood, Ashleigh Cox, Steven Lim and Michio Yamaoka (2013). “Personal Financial Literacy among High School Students in New Zealand, Japan and the USA,” Citizenship, Social and Economics Education, vol.12, no.3, pp.200-215. Central Council for Financial Service Information (2016). “Financial Literacy Survey: 2016 Results.” Available at https://www.shiruporuto.jp/e/survey/kinyulite/pdf/16kinyulite.pdf (accessed on December 6, 2018). Clark, Robert, Annamaria Lusardi, and Olivia S. Mitchell (2017). “Financial knowledge and 401 (k) investment performance: a case study.” Journal of Pension Economics & Finance, vol.16, no.3, pp.324-347. Clark, Robert, Rikiya Matsukura and Naohiro Ogawa (2013). “Low Fertility, Human Capital, and Economic Growth: The Importance of Financial Education and Job Retraining,” Demographic Research, vol.29, Article 32, pp.865-884. Lusardi, Annamaria, Pierre-Carl Michaud, and Olivia S. Mitchell (2017). “Optimal financial knowledge and wealth inequality.” Journal of Political Economy, vol.125, no.2, pp.431-477. Lusardi, Annamaria, and Olivia S. Mitchell (2007). “Baby Boomer Retirement Security: The Roles of Planning, Financial Literacy, and Housing Wealth.” Journal of Monetary Economics, vol.54, no.1, pp.205–224. Lusardi, Annamaria and Olivia Mitchell (2011a). “Financial literacy and planning: implications for retirement wellbeing.” In O. S. Mitchell and A. Lusardi (eds), Financial Literacy: Implications for Retirement Security and the Financial Marketplace. Oxford: Oxford University Press, pp.17–39. Lusardi, Annamaria and Olivia Mitchell (2011b). “Financial Literacy around the World: An Overview,” Journal of Pension Economics and Finance, vol.10, no.4, pp.497-508. Lusardi, Annamaria and Olivia Mitchell (2014). “The Economic Importance of Financial Literacy: Theory and Evidence,” Journal of Economic Literature, vol.52, no.1, pp.5-44. Sekita, Shizuka (2011). “Financial Literacy and Retirement Planning in Japan,” 17
Journal of Pension Economics and Finance, vol.10, no.4, pp.637-656. Shimizutani, Satoshi (2015). “Population Aging in Postwar Japan: Processes and Prospects,” Asia-Pacific Review, vol.22, no.2, pp. 53-76. Shimizutani, Satoshi, Takashi Oshio and Mayu Fujii (2016). “Option Value of Work, Health Status, and Retirement Decisions in Japan: Evidence from the Japanese Study on Aging and Retirement (JSTAR),” in David Wise eds. Social Security Programs and Retirement around the World: Disability Insurance Programs and Retirement, The University of Chicago Press, pp.497-535. Shimizutani, Satoshi and Hiroyuki Yamada (2018). “Financial Literacy of Middle and Older Generations in Japan,” Keio-IES Discussion Paper Series DP2018-010. van Rooij, Maarten, Annamaria Lusardi, and Rob Alessie. (2011) “Financial literacy and stock market participation.” Journal of Financial Economics, vol.101, no.2, pp.449-472. Yoong, Joanne. (2011) “Financial Illiteracy and Stock Market Participation: Evidence from the RAND American Life Panel.” In Financial Literacy: Implications for Retirement Security and the Financial Marketplace, edited by Olivia S. Mitchell and Annamaria Lusardi, 76–97. Oxford and New York: Oxford University Press. Yoshino, Naoyuki, Peter J. Morgan, and Long Q. Trinh (2017). “Financial Literacy in Japan: Determinants and Impacts” ADBI Working Paper Series No.796.
18
Table 1 Financial literacy in Japan with a comparison with the U.S. A. Compound interest "Suppose you had $100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow? More than $102, exactly $102, or less than $102?" JSTAR HRS (52
13.88 22.41 42.04 69.25 29.42
19
HRS (52
Table 2 Descriptive statistics Variables # of correct answers
Japan (1) N Obs. Average S.D. Min Max 2,852 1.081 1.093 0 3
US: full sample (2) N Obs. Average S.D. Min 1,436 2.205 0.853 0
age (years) sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise) work: missing dummy current_married (=1 if yes, 0 otherwise) current_married: missing dummy
2,852 2,852 2,852 2,852 2,852 2,852
65.646 0.512 0.526 0.002 0.808 0.002
6.974 0.500 0.499 0.046 0.394 0.046
52 0 0 0 0 0
79 1 1 1 1 1
1,473 1,473 1,473 1,473 -
63.999 0.451 0.392 0.684 -
8.156 0.498 0.488 0.465 -
primary/junior high schoool senior_high junior_college university graduate Education variable: missing dummy
2,852 2,852 2,852 2,852 2,852 2,852
0.353 0.402 0.122 0.107 0.007 0.004
0.478 0.490 0.328 0.310 0.086 0.065
0 0 0 0 0 0
1 1 1 1 1 1
1,473 1,473 1,473 1,473 1,473 1,473
0.189 0.305 0.255 0.136 0.109 0.005
0.392 0.461 0.436 0.343 0.312 0.074
0 0 0 0 0 0
Serial 7 (=1 if correct, 0 otherwise) Immediate word recall ever studied accounting/economics? (=1 if yes, 0 otherwise) ever studied accounting/economics?: missing dummy
2,852 2,852 2,852 2,852
0.453 4.211 0.111 0.102
0.498 2.486 0.314 0.303
0 0 0 0
1 9 1 1
1,473 1,473 1,473 1,473
0.430 5.479 0.341 0.002
0.495 1.668 0.474 0.045
Income (household): $0-$20,000 Income (household): $20,000-$50,000 Income (household): $50,000-$100,000 Income (household): more than $100,000 Income (household): missing dummy
2,852 2,852 2,852 2,852 2,852
0.269 0.316 0.099 0.020 0.296
0.443 0.465 0.299 0.140 0.457
0 0 0 0 0
1 1 1 1 1
1,473 1,473 1,473 1,473 -
0.240 0.319 0.260 0.181 -
0.427 0.466 0.439 0.385 -
Having bonds (household level) Having stocks (household level) Having bond or/and stocks (household level) Having bank accounts (household level) Having CD/government saving bond/T-bill (household level) Having IRA accounts (household level)
2,430 2,412 2,397 2,412
0.202 0.157 0.269 0.918
0.402 0.364 0.443 0.275
0 0 0 0
1 1 1 1
1,473 1,473 1,473 1,473 1,473 1,473
0.038 0.219 0.232 0.771 0.160 0.382
0.191 0.413 0.422 0.420 0.366 0.486
Max 3
52 0 0
1,011 64.260 1,011 0.473 1,011 0.389 1,011 0.540 -
8.246 0.500 0.488 0.499 -
1 1 1 1 1 1
1,011 1,011 1,011 1,011 1,011 1,011
0.197 0.292 0.253 0.146 0.107 0.005
0.398 0.455 0.435 0.354 0.309 0.070
0 0 0 0 0 0
1 1 1 1 1 1
0 0 0 0
1 10 1 1
1,011 1,011 1,011 1,011
0.425 5.474 0.345 0.003
0.495 1.677 0.476 0.054
0 0 0 0
1 10 1 1
0 0 0 0
1 1 1 1
1,011 1,011 1,011 1,011 -
0.305 0.315 0.233 0.147 -
0.460 0.465 0.423 0.355 -
0 0 0 0
1 1 1 1
1,011 1,011 1,011 1,011 1,011 1,011
0.037 0.181 0.193 0.750 0.140 0.337
0.188 0.385 0.395 0.433 0.348 0.473
-
79 1 1
US: investment behavior sample (3) N Obs. Average S.D. Min Max 1,011 2.201 0.857 0 3
0
-
1 -
-
0 0 0 0 0 0
1 1 1 1 1 1
(Note) Authors' calculations. Incomes of Japanese households are converted to US$ using purchasing power parities in 2009 published by OECD. Source: https://data.oecd.org/conversion/purchasing-power-parities-ppp.htm
20
52 0 0 -
79 1 1 -
0 -
1 -
-
0 0 0 0 0 0
1 1 1 1 1 1
Table 3 Factors influencing financial literacy Japan
US (2) Logit
(1) OLS
-0.004*** -0.015** [0.004] [0.001] sex (=1 if male, 0 otherwise) 0.242*** 0.052*** [0.009] [0.042] 0.007 -0.016 work (=1 if working, 0 otherwise) [0.007] [0.049] -0.172 work: missing dummy [0.176] current_married -0.056 -0.005 [0.035] [0.012] current_married: missing dummy -0.481 [0.389] Education category (default: primary/junior high schoool) 0.036 0.156** senior_high [0.022] [0.047] junior_college 0.200* 0.055* [0.033] [0.086] 0.102*** 0.396** university [0.119] [0.031] 0.100*** 0.419*** graduate [0.079] [0.033] -0.010 Education variable: missing dummy -0.091 [0.062] [0.123] Serial 7 (=1 if correct, 0 otherwise) 0.146*** 0.017*** [0.004] [0.026] 0.016*** 0.092*** Immediate word recall [0.005] [0.013] ever studied accounting/economics? 0.227* 0.056*** [0.012] [0.109] (=1 if yes, 0 otherwise) 0.034** ever studied accounting/economics? 0.047 missing dummy [0.036] [0.017] Household income category (default: $0-$20,000) $20,000-$50,000 0.243** 0.058*** [0.054] [0.004] $50,000-$100,000 0.271** 0.058*** [0.088] [0.011] more than $100,000 0.343* 0.077*** [0.154] [0.026] income: missing dummy -0.119 -0.016 [0.066] [0.026] Constant 1.488*** [0.288] age
(3) OLS
(4) Logit
-0.004 [0.002] 0.241*** [0.040] 0.032 [0.044]
-0.001 [0.001] 0.129*** [0.015] 0.029 [0.022]
0.035 [0.044]
0.006 [0.030]
0.189*** [0.035] 0.278*** [0.068] 0.476*** [0.056] 0.507*** [0.048] 0.168 [0.443] 0.164*** [0.040] 0.056*** [0.011] 0.124** [0.041] 0.161 [0.320]
0.118*** [0.024] 0.153*** [0.034] 0.275*** [0.028] 0.278*** [0.045] 0.087 [0.237] 0.072*** [0.020] 0.032*** [0.005] 0.067*** [0.023] -0.059 [0.157]
0.093** [0.039] 0.167** [0.053] 0.223** [0.079]
0.026 [0.032] 0.078** [0.039] 0.126** [0.053]
1.473*** [0.179]
Observations 2,852 2,840 1,473 1,473 R-squared 0.201 0.191 Region dummies Yes Yes Yes Yes The dependent variable in (1) and (3) is the number of correct answers. The dependent variable in (2) and (4) takes 1 if a respondent got all correct answers and 0 otherwise. Marginal effects are reported for logit models ((2) and (4)). Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
21
Table 4 Financial literacy and investment behaviors: Bond
Number of correct answers age sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise)
Japan
US
(1) Logit
(2) Logit
0.049*** [0.008] 0.003*** [0.001] -0.047*** [0.012] -0.060*** [0.018]
0.021*** [0.004] 0.002*** [0.001] 0.002 [0.009] -0.016 [0.010]
-0.004 [0.019]
0.010 [0.014]
work: missing dummy current_married current_married: missing dummy Education category (default: primary/junior high schoool) senior_high 0.096*** [0.031] junior_college 0.110*** [0.034] university 0.083*** [0.015] graduate 0.213*** [0.067] Education variable: missing dummy 0.185** [0.086] Serial 7 (=1 if correct, 0 otherwise) 0.007 [0.013] Immediate word recall -0.005* [0.003] ever studied accounting/economics? 0.086*** (=1 if yes, 0 otherwise) [0.006] ever studied accounting/economics? -0.021 missing dummy [0.037] Household income category (default: $0-$20,000) $20,000-$50,000 0.044* [0.024] $50,000-$100,000 0.120*** [0.032] more than $100,000 0.174*** [0.018] income: missing dummy -0.010 [0.011]
-0.012 [0.030] 0.009 [0.028] 0.028 [0.020] 0.047** [0.019]
0.000 [0.005] 0.004 [0.003] -0.012 [0.016]
0.048 [0.040] 0.059 [0.039] 0.092** [0.042]
Observations 2,420 1,001 Region dummies Yes Yes The dependent variable takes 1 for a bond holder and 0 otherwise. Marginal effects are reported. Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
22
Table 5 Financial literacy and investment behaviors: Stock
Number of correct answers age sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise)
Japan
US
(1) Logit
(2) Logit
0.030*** [0.002] 0.002** [0.001] -0.014 [0.014] -0.039*** [0.015]
0.051*** [0.012] 0.005*** [0.002] 0.004 [0.029] -0.038** [0.017]
0.026 [0.028]
-0.002 [0.016]
work: missing dummy current_married current_married: missing dummy Education category (default: primary/junior high schoool) senior_high 0.110*** [0.017] junior_college 0.146*** [0.028] university 0.174*** [0.017] graduate 0.219*** [0.043] Education variable: missing dummy 0.166*** [0.060] Serial 7 (=1 if correct, 0 otherwise) -0.012 [0.011] Immediate word recall -0.002 [0.002] ever studied accounting/economics? 0.033*** (=1 if yes, 0 otherwise) [0.011] ever studied accounting/economics? -0.093 missing dummy [0.057] Household income category (default: $0-$20,000) $20,000-$50,000 0.044*** [0.013] $50,000-$100,000 0.095*** [0.023] more than $100,000 0.167*** [0.049] income: missing dummy 0.008 [0.021]
0.079* [0.047] 0.115** [0.054] 0.174*** [0.050] 0.240*** [0.039]
0.014 [0.024] 0.004 [0.007] -0.011 [0.031]
0.126*** [0.039] 0.131*** [0.047] 0.230*** [0.052]
Observations 2,403 1,003 Region dummies Yes Yes The dependent variable takes 1 for a stock holder and 0 otherwise. Marginal effects are reported. Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
23
Table 6 Financial literacy and investment behaviors: Bank accounts
Number of correct answers age sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise)
Japan
US
(1) Logit
(2) Logit
0.006 [0.007] 0.001 [0.001] -0.005 [0.011] -0.002 [0.013]
0.026** [0.010] 0.006*** [0.001] -0.027* [0.016] -0.031 [0.027]
0.046*** [0.014]
-0.004 [0.021]
work: missing dummy current_married current_married: missing dummy Education category (default: primary/junior high schoool) senior_high 0.051*** [0.015] junior_college 0.053** [0.027] university 0.036** [0.015] graduate Education variable: missing dummy Serial 7 (=1 if correct, 0 otherwise)
0.025*** [0.008] Immediate word recall 0.005* [0.003] ever studied accounting/economics? 0.029* (=1 if yes, 0 otherwise) [0.015] ever studied accounting/economics? 0.026 missing dummy [0.042] Household income category (default: $0-$20,000) $20,000-$50,000 0.044*** [0.013] $50,000-$100,000 0.133*** [0.039] more than $100,000 income: missing dummy
0.168*** [0.014] 0.208*** [0.020] 0.213*** [0.028] 0.308*** [0.078] 0.070 [0.306] 0.038** [0.016] 0.016* [0.010] -0.042* [0.024] 0.043 [0.249] 0.146*** [0.041] 0.208*** [0.047] 0.246*** [0.034]
0.030* [0.016]
Observations 2,321 980 Region dummies Yes Yes The dependent variable takes 1 for a bank account holder and 0 otherwise. Marginal effects are reported. Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
24
Appendix 1 Factors influencing financial literacy in Japan (with a full set of the covariates) (1) OLS age
-0.018** [0.005] sex (=1 if male, 0 otherwise) 0.244*** [0.043] work (=1 if working, 0 otherwise) -0.034 [0.045] work: missing dummy -0.122 [0.196] current_married -0.059 [0.034] current_married: missing dummy -0.539 [0.373] Education category (default: primary/junior high schoool) senior_high 0.156** [0.051] 0.198* junior_college [0.074] university 0.393** [0.114] graduate 0.423*** [0.091] -0.104 Education variable: missing dummy [0.133] 0.105** Serial 7 (=1 if correct, 0 otherwise) [0.025] Immediate word recall 0.068** [0.019] ever studied accounting/economics? 0.232* (=1 if yes, 0 otherwise) [0.107] ever studied accounting/economics? 0.099 [0.052] missing dummy Household income category (default: $0-$20,000) $20,000-$50,000 0.226** [0.051] $50,000-$100,000 0.252** [0.089] more than $100,000 0.318* [0.137] income: missing dummy -0.116 [0.074] Risk preference (default: no uncertainty) risk: 10-30% -0.112 [0.056] risk: 40-60% -0.052 [0.033] risk: 70-90% 0.076 [0.149] risk preference: missing dummy -0.164*** [0.032] Time discounting preference (default: 0-0.5%) time: 1-6% 0.155** [0.043] time: 10-40% 0.089 [0.056] time discounting preference: -0.011 missing dummy [0.066] CESD (=1 if score>=20, 0 otherwise) -0.092 [0.045] CESD: missing dummy -0.067 [0.045] subjective probability living up to 85 0.002 [0.001] Constant 1.741*** [0.310]
(2) Logit -0.004*** [0.001] 0.050*** [0.009] 0.003 [0.006]
-0.005 [0.011]
0.034 [0.025] 0.052 [0.033] 0.100*** [0.031] 0.097*** [0.036] -0.019 [0.072] 0.011** [0.005] 0.015*** [0.005] 0.056*** [0.013] 0.044* [0.023] 0.053*** [0.006] 0.052*** [0.013] 0.072*** [0.024] -0.016 [0.030] -0.022 [0.023] -0.025 [0.015] 0.009 [0.040] -0.035** [0.015] 0.056*** [0.015] 0.032* [0.017] 0.047*** [0.011] -0.033** [0.015] -0.022 [0.021] 0.000 [0.000]
Observations 2,852 2,840 R-squared 0.214 Region dummies Yes Yes The dependent variable in (1) is the number of correct answers. The dependent variable in (2) takes 1 if a respondent got all correct answers and 0 otherwise. Marginal effects are reported for (2). Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
25
Appendix Table 2 Financial literacy and investment behaviors: Bond or/and Stock
Number of correct answers age sex (=1 if male, 0 otherwise) work (=1 if working, 0 otherwise)
Japan
US
(1) Logit
(2) Logit
0.055*** [0.007] 0.004*** [0.001] -0.035** [0.014] -0.052*** [0.020]
0.055*** [0.009] 0.006*** [0.002] -0.009 [0.029] -0.042*** [0.015]
-0.000 [0.016]
0.009 [0.013]
work: missing dummy current_married current_married: missing dummy Education category (default: primary/junior high schoool) senior_high 0.130*** [0.028] junior_college 0.173*** [0.042] university 0.168*** [0.025] graduate 0.236*** [0.065] Education variable: missing dummy 0.174* [0.100] Serial 7 (=1 if correct, 0 otherwise) 0.003 [0.009] Immediate word recall -0.004* [0.002] ever studied accounting/economics? 0.085*** (=1 if yes, 0 otherwise) [0.016] ever studied accounting/economics? -0.043 missing dummy [0.056] Household income category (default: $0-$20,000) $20,000-$50,000 0.058** [0.024] $50,000-$100,000 0.159*** [0.027] more than $100,000 0.233*** [0.073] income: missing dummy -0.007 [0.021]
0.074 [0.057] 0.109* [0.060] 0.172*** [0.052] 0.251*** [0.041]
0.013 [0.025] 0.005 [0.007] -0.015 [0.022]
0.141*** [0.035] 0.147*** [0.045] 0.251*** [0.047]
Observations 2,388 1,003 Region dummies Yes Yes The dependent variable takes 1 for a bond and/or stock holder and 0 otherwise. Marginal effects are reported for logit models. Standard errors are clustered at regional level. *** p<0.01, ** p<0.05, * p<0.1
26