Consumer financial education and risky financial asset holding in China

Risky financial asset holding is arguably a desirable financial behavior that contributes to consumer financial wellbeing. However, studies about associations between consumer financial education and risky financial asset holding in China remain limited. To fill this gap, using data from the 2015 China Household Finance Survey, this study examined the association between financial education and risky financial asset holdings and explored its mediators. Results from Probit regressions showed that financial education was positively associated with the household risky financial asset holding. Further analyses based on the mediating model found that financial literacy, economic and financial information search, and risk tolerance were mediating factors in the association between financial education and risky financial asset holding. The results have policy implications for improving consumer financial education and financial market participation.

and trust (Georgarakos and Pasini, 2011). To our knowledge, no previous research has looked at the relationship between financial education and household risky financial asset holding, and explored the mediating roles of financial literacy, economic and financial information search, and risk tolerance. To fill this gap, we examined the association between financial education and risky financial asset holding using data from the 2015 China Household Finance Survey (CHFS). We also examined the mediating roles of consumer financial literacy, economic and financial information search, and risk tolerance in the association between financial education and risky financial asset holding. In addition, we explored heterogeneous associations in terms of different household characteristics. Finally, we used the 2013 CHFS, different definitions of risky financial asset holding and different types of risky financial assets to conduct robust tests.
Financial market participation refers to investor's general participation by holding and investing in specific financial products in a financial market (Chen et al., 2020). Participation in financial markets can be divided in two ways according to different risk level of assets: holding risk-free assets and risky assets. In this paper, we focus on financial market participation by holding risky financial assets.
It is important to explore the relationship between financial education and risky financial asset holding, which would explore whether financial education promotes participation in financial market.
On the one hand, risky financial assets have higher return than other risk-free assets. Consumers who have received financial education could invest risky financial assets and allocate assets properly to gain high returns. On the other hand, high returns always accompany high risks in financial markets.
Consumers who have received financial education should understand the risk and be more careful in holding risky financial assets. Furthermore, it is also important to examine specific channels leading to financial risky asset holding to better understand this consumer behavior. This paper makes the following contributions: First, it enriches the literature of financial education by examining the relationship between financial education and risky financial asset holding. Previous research did not focus on the relationship between financial education and risky financial asset holding directly or indirectly. Many studies examined the relationship between financial literacy and financial market participation (Arrondel, Debbich & Savignac, 2012;Christelis, Jappelli & Padula, 2010;Li, Li & Wei, 2020). However, we estimate the relationship between financial education, a factor never used in previous research, and risky financial asset holding, and find the positive association between them.
Second, we explore possible mediating variables between financial education and risky financial asset holding, such as financial literacy, economic and financial information search, and risk tolerance. We estimate both direct and indirect associations between financial education and risky financial asset holding.
The remainder of this paper is arranged as follows. Section 2 reviews relevant existing studies and proposes hypotheses. Section 3 describes the dataset, variables, and analytic models. Results and discussion about benchmark regression, intermediate mechanisms and robustness checks are presented in Section 4. Section 5 includes limitations and future directions. Finally, Section 6 presents the conclusion and policy implications. sophisticated financial products to raise higher returns. It can be socially optimal to raise financial knowledge for everyone early in life, such as by mandating financial education in high school, gaining knowledge from parents and so on. Delavande, Rohwedder and Willis (2008) presented a simple twoperiod model of consumers portfolio allocation across safe bond and risky stock with the financial knowledge as human capital. They found that it is optimal for individuals to invest in financial knowledge.
This could help them identify better-performing assets and save on the cost of hiring financial advisers.
Further, it could increase the expected return from their portfolio of risky assets without incurring additional risk. As a formal financial socialization process, financial education plays a prominent role in obtaining financial knowledge (Fan & Chatterjee, 2019). Therefore, from a theoretical point of view, these predictions suggest that investment in financial knowledge, in other words receiving financial education, plays a very important role in increasing consumers ability to manage their money and financially perform better. Previous studies also show financial education contributes to financial literacy, behavior, capability, and wellbeing (Kaiser & Menkhoff, 2017, 2019Lusardi & Mitchell, 2014;Xiao & O'Neill, 2016;Xiao & Porto, 2017). Based on the literature discussed above, we propose the conceptual framework shown in Figure 1 that is similar to Xiao and Porto (2017)'s conceptual model of "educationcapability-wellbeing." In this conceptual framework, financial education is directly associated with risky financial asset holding, which is regarded as a desirable financial behavior leading to financial wellbeing (Chen et al., 2020). What's more, as consumer financial ability indicators, financial literacy, information search, and risk tolerance are mediators between financial education and risky financial asset holding.
This conceptual framework is used to guide the development of hypotheses for this study.

Financial Education and Risky Financial Asset Holding
Benefits of financial education for consumers are documented in previous research (Asarta, Hill & Meszaros 2014;Brown et al., 2016;Clark, Lusardi & Mitchell, 2017;Kaiser & Menkhoff, 2019;Kim & Xiao, 2020;Lusardi, 2019;Miller et al., 2015;Urban et al. 2020). Recent meta-analysis verified the positive effects of financial education on financial literacy and behaviors (Kaiser & Menkhoff, 2017, 2019Miller et al., 2015). Kaiser and Menkhoff (2019) examined the literature on school financial education programs for children and youth which contain 37 total independent (quasi-) experimental studies and found sizeable impacts of financial education treatment on financial knowledge and behaviors. Empirical studies also show evidence on positive effects of financial education on financial literacy, financial behaviors, financial capability, and financial wellbeing. Urban et al. (2020) found that rigorous financial education programs, coupled with teacher training and high school financial education requirements, are correlated with fewer defaults and higher credit scores among young adults in the US.
However, the effects of financial education are controversial. Some empirical or review research suggest that financial education has limited effects on financial outcomes (Collins & O'Rourke, 2010;Fernandes, Lynch & Netemeyer, 2014;Mandell & Klein, 2009;Willis, 2011). However, these studies did not deny the importance of financial education, and believed that financial education can be improved to become more effective.
Previous research mainly focuses on one type of financial education, such as high school (Bernheim, Garrett & Maki, 2001;Brown et al., 2014;Brown et al., 2016;McCormick, 2009;Walstad, Rebeck & MacDonald, 2010), college (Lyons, 2004;Lyons, 2008;Xiao, Serido & Shim, 2012;Xiao et al., 2014) or among adults (Bayer, Bernheim & Scholz, 2009;Bernheim & Garrett, 2003;Clark, Lusardi & Mitchell, 2014;Joo & Grable, 2005;Kim, Garman & Quach, 2005), but few studies have looked at all three (Wagner, 2019). Xiao and O'Neill (2016) estimated the effects of high school, college, and adult financial education simultaneously on financial capability. They found that receiving financial education from high school, college, and the workplace are positively associated with financial capability. In this paper, we also use financial education which refers to any financial education received from high school, college, the workplace, or any other sources, and link it with an important economic outcome: risky financial asset holding.
According to the Merton model (Merton, 1969), all investors, independent of their wealth and of their preferences towards risk, should participate in all risky assets markets and should invest in the market portfolio to take advantage of the equity premium (Guiso & Sodini, 2013). Theoretical and empirical research on household financial asset holding has been the focus of research in the past half century (Campbell, 2006). Many previous research studies focus on household participation in the financial market, especially in the stock market (Poterba & Samwick, 2003;Vissing-Jorgensen, 2002).
Previous studies also have addressed the role of financial market participation by holding risky financial assets in consumers subjective wellbeing (Frijters et al., 2015;Murgea & Reisz, 2013). Chen et al. (2020) investigated the relationships between holding risky financial assets and subject wellbeing by using the China Household Financial Survey (CHFS), and the results indicate that holding risky assets contribute negatively to subjective wellbeing. However, Murgea and Reisz (2013) found positive relationships between subjective wellbeing and stock market participation in the USA. In this paper, we focus on household participation in financial markets by observing whether households hold risky financial assets or not, other than the share of risky financial assets held.
Based on the above discussion, financial education would increase financial knowledge. More knowledgeable consumers are more likely to perform desirable financial behaviors. Holding risky financial assets is considered such a behavior that helps increase wealth. Then, we propose the following hypothesis: H1. Financial education is positively associated with risky financial asset holding.

Financial Literacy as the Mediator
Besides the direct associations between financial education and risky financial asset holding, we also propose three hypotheses about mediating factors in the association between financial education and risk financial asset holding.
First, financial literacy has been found to affect both saving and investment behavior and debt management and borrowing practices (Hamid & Loke, 2020;Hastings & Mitchell, 2020;Lusardi & Tufano, 2009, 2015Lusardi & Mitchell, 2014). Financial literacy mainly focuses on the understanding of economic and financial concepts and knowledge about financial instruments (Xiao, 2015). Consumers with higher financial literacy have better performance in financial behavior, such as lower probability of bank loan overdue (Fedorova, Nekhaenko & Dovzhenko, 2015), lower credit card and mortgage costs (Huston 2012), better financial planning (Arrondel, Debbich & Savignac, 2013), and a better chance of receiving a positive investment return (Chu et al., 2017;Jiang et al., 2020). As for risky financial asset holding, Liao et al. (2017) found that consumers with higher financial literacy are more likely to hold risky financial assets than those with lower financial literacy in China. By studying the portfolio allocation decisions of Australian households, Cardak and Wilkins (2009) found that financial awareness and knowledge play important roles in determining risky asset holdings. Several studies also found that those who are more financially literate are more likely to participate in financial markets, invest in stocks Based on those previous studies, we infer that there are positive associations between both financial education and financial literacy, and between financial literacy and risky financial asset holding in Chinese households. That is, financial literacy is the mediator between financial education and risky financial asset holding. Therefore, we propose the following hypothesis: H2. Financial education is positively associated with financial literacy, and financial literacy in turn is positively associated with risky financial asset holding.

Information Search as the Mediator
Second, we also assume that with financial education, consumers pay more attention to the economic or financial information in their daily life. Consumers usually make financial decisions based on limited knowledge and incomplete information. Improving the access to information would facilitate more effective investment decisions (Li, 2014). Previous studies have emphasized the importance of acquiring information on financial market participation from social networks (Chiteji & Stafford, 1999;Guiso, Sapienza & Zingales, 2004;Hong, Kubik & Stein, 2004;Li, 2014) and the Internet (Bogan, 2008;Liang & Guo, 2015;Markus & Alexander, 2013). Therefore, it is important for consumers to capture related information actively. After knowing about the economic situation and the financial markets, consumers are more likely to invest in risky financial assets. Hence, we propose the following hypothesis: H3. Financial education positively contributes to economic and financial information search, and economic and financial information search in turn is positively associated with the risky financial asset holding.

Risk Tolerance as the Mediator
Finally, financial education has been found to increase consumers financial risk tolerance (Ryack, 2011), and risk attitude which affects the choice of financial assets. High risk aversion is associated with a lesser likelihood of making investments in the stock market (Dimmock & Kouwenberg, 2010;Lim, Soutar & Lee, 2013). Barasinska, Schfer and Stephan (2012) examined the effect of personal risk attitude on financial portfolios among German households, and found that there is a significant effect of risk attitude on holding risky assets. The higher level of risk aversion, the higher proportion of risk-free assets households prefer to hold. In addition, the more educated investors are, the more they value diversification and hold risky assets (Mitchell & Moore, 1998). Several studies also found there are positive effects of financial literacy on the level of risk tolerance (Bajo, Barbi & Sandri, 2015;Mishra, 2018). Based on this correlation, financial education may increase financial confidence which may lead to higher levels of risk tolerance. Thus, we propose the following hypothesis: H4. Financial education is positively associated with risk tolerance, and risk tolerance in turn is positively associated with risky financial asset holding.

Variables
CHFS chooses the family member who has the best knowledge of the family's financial situation as the respondent in each household. Therefore, this paper used the respondent as the representative of each household.
Financial education. This is a dummy variable in CHFS2015. Based on the survey question "Have you ever taken an economic or financial class?" if the respondent has received economic or financial education, then =1, otherwise, =0.
Risky financial asset holding. In CFHS, the risky financial assets include stocks, funds, wealth management products (a popular investment product issued through banks), derivatives, bonds, non-RMB assets and precious metal. The CFHS has detailed information about whether the household holds each of these assets or not. In this paper, the variable regarding holding risky financial assets is generated whether households have any types of risky financial assets or not. The dummy variable _ℎ=1 if the household holds risky financial assets, otherwise, _ℎ=0. Risk tolerance. The risk attitude was measured using a 5-point scale in CHFS. The respondents were asked "Which of the choices below do you want to invest in most if you have adequate money?" The answers were "Unwilling to carry any risk", "Project with slight risk and return", "Project with average risk and return", "Project with slightly high-risk and slightly high-return" and "Project with highrisk and high-return", with corresponding scales ranging from 1 to 5. The higher the scale, the higher level of risk tolerance.
Control variables. Control variables at individual, household, and regional level were used following previous research. Several demographic variables were included such as age, education years, political status, marriage, gender, the registered residence type and if working in a financial sector at an individual level (respondent's situation); and income and household size at household level. At the regional level, the following control variable were used: whether the household is in a rural region or not and whether it is in the eastern or western region of China (middle region is the reference category). The specific definitions of variables are shown in Table 1.

Data Analysis
The basic model is to estimate the relationship between financial education and risky financial asset holding. Because the main dependent variable, risky financial asset holding ( _ℎ), is a binary variable, the Probit model was used to estimate the relationship between financial education and risky financial asset holding. Suppose that for household , the decision _ℎ to hold risky financial assets can only assume two values, 0 or 1. Its value is determined by the latent variable _ℎ * .
When _ℎ * > 0 , _ℎ = 1 ; when _ℎ * = 0 , _ℎ = 0 . Therefore, the following Probit regression was used with assuming obeys standard normal distribution: where represent the control variables including individual level, household level and region level. In control variables, income is transformed to logarithms. And represents province fixed effect and is the error term. Equation one is used to test H1.
After estimating the relationships between financial education and risky financial asset holding, this study explored the intermediate mechanisms.
The following three mediators were considered in this study: financial literacy, economic and financial information search, and risk tolerance. To test H2, H3, and H4, the mediating model was used as follows (Baron & Kenny, 1986), where the first formula is the same as formula (1): where the include three variables, , _ and _ .
According to previous hypotheses, the coefficients of 1 in equation (2) and 2 in equation (3) where n represent the number of observations in our sample, and ̂1 is the estimated value of β 1 .
Finally, heterogeneity analysis was estimated in terms of several household characteristics and robustness tests for the basic results and the intermediate mechanisms were conducted by using data from the 2013 China Household Finance Survey, using different definition of risky financial asset holding and different types of risky financial assets. STATA 15 was used for data analyses.

Descriptive Statistics of the Sample
Descriptive statistics of the sample are shown in Table 2. During the data cleansing process, following Li, Wu and Xiao (2020), we eliminated observations with missing values of used variables, and only kept households in which the respondent's age is 18 or older. 32,554 observations remained in the data analyses which accounts for about 87.3% of the initial whole sample. As seen in  and only 17.7% of households hold risky financial assets which means financial market participation by holding risky assets is low. In households that hold financial assets, the most popular financial asset is demand deposit which has the lowest risk. Moreover, the most popular risky financial asset is stock, followed by wealth management product.
In the 2015 CHFS, the respondents were asked "Have you ever taken an economic or financial class?" We did not know from which channel the respondent received financial education (in school, the workplace, community, or internet). Therefore, we did a simple cross tabulation between the education years and financial education (see its row percentage in Table 4). This table shows that financial education and education years are correlated with one other. Higher education levels have higher probabilities of receiving financial education. However, we cannot conclude that the financial education is from school.
In China, financial education is not compulsory at each education level. Some people in economics, business, or related majors may receive financial education at colleges or universities. Others who are interested in economics and finance may receive related knowledge from various sources such as financial education programs offered by financial institutions, government agencies, or nonprofit organizations.
We also did the Pearson Chi-square test between financial education and risky financial asset holding. The results are significant which means financial education and risky financial asset holding are not mutually independent. This suggests consumers receiving financial education are more likely to hold risky financial assets than those not receiving financial education (52% vs. 15%, χ 2 =2.1e+03, p<0.001). Table A1 shows the correlation matrix among all variables we used. The correlation coefficients between any two independent variables are small which means a weak correlation and will not lead to multicollinearity problems in regression.

Benchmark Regression Results
The Probit regression was used to estimate the relationship between financial education and risky financial asset holding and the results are shown in Table 5. Column one only has one independent variable: financial education. Column two adds control variables and column three adds both control variables and province fixed effects. Column four is the average marginal effects of each independent variables on risky financial asset holding. Because the coefficients in the Probit models do not show marginal effects, we presented marginal effects in column four calculated by equation four. The coefficients of financial education in all four columns are positive and significant. Consumers receiving financial education experience an 8.5% increase in the probability of holding risky financial assets compared to those without financial education. According to the results, financial education is positively associated with the probability of risky financial asset holding, supporting H1.
As for the control variables, most of the correlations are with expected signs.

Mechanisms Results
This section presents results of possible mediating factors between financial education and risky financial asset holding. We considered three mediators: financial literacy, economic and financial information search, and risk tolerance. Equations 1-3 were used to identify the mediating effects.
H2 states that financial education is positively associated with financial literacy, and financial literacy in turn is positively associated with risky financial asset holding. Table 6 presents the results.
Column one of Table 6 is the result of equation one which is the same with column three in Table 5.
Column two is its marginal effects results. Column three of Table 6 is the result of equation two and shows that financial education would increase consumers' financial literacy significantly. In addition, in column four and column five, the coefficients of financial literacy are significant and the signs of coefficients are the same with the coefficients of financial education. This means that financial literacy would increase the probability of holding risky financial assets, which is consistent with Liao et al. (2017).
After adding financial literacy as an independent variable in columns four and five, the coefficient of the association between financial education and risky financial asset holding probability is still significant and the marginal effect is lower than the coefficients in column two. This means that financial literacy has a certain mediating effect, supporting H2. Table 7 shows the results for which economic and financial information search is used as the mediating variable in the association between financial education and risky financial asset holding.
Columns one and two of Table 7 is the result of equations one and four. This is the same with columns three and four in Table 5. In column three, the coefficient of financial education is positive and significant, suggesting that financial education would increase the consumers' attention to economic and financial information. The results in columns four and five show that the coefficient and marginal effect of economic and financial information search are both positive and significant. The marginal effect of financial education is significant and its value is smaller than that in column two. Thus, economic and financial information search is a mediating variable in the association between financial education and risky financial asset holding, supporting H3. It is very important for investors to know information about the economic situation and the financial market situation before they make investment choices (Li, 2014).
The knowledge about the economy and market will help consumers make the right choices and allocate their assets into promising products. Nowadays, consumers have multiple ways to obtain the information about the economy and financial markets, such as traditional media (TV, newspaper), apps related to finance and economics, and websites through the Internet and mobile phone. Financial education will help consumers to understand professional vocabularies and theories, which could also encourage consumers to economic and financial information search. More economic and financial information search will let them know more about the markets and invest the risky financial assets in proper time and ways.
Risk tolerance is an important factor that affects risky financial asset holding (Barasinska, Schfer & Stephan, 2012). Table 8 presents the results for which risk tolerance is the mediator. Columns one and two are the results of equation one which are the same with columns three and four in Table 5. Firstly, in column three, results suggest that financial education may increase consumers' risk tolerance. Then, in columns four and five, the coefficient and marginal effect of risk tolerance and financial education are both positive and significant. The value of marginal effect of financial education is smaller than that in baseline results, which means that risk tolerance is a mediating variable in the association between financial education and risky financial asset holding, supporting H4. After receiving financial education, consumers will know more about the economy and financial markets, and understand the risks of various types of financial assets. Then, an individual's risk tolerance in financial markets will change (Ryack, 2011) and lead to diversified risky financial asset holding choices.
In addition, we conducted Sobel (1986) test for our mediating analyses. The results are shown in Table 9. The coefficients of Sobel in Table 9 are significant which means all three variables are mediators between financial education and probability of risky financial asset holding. The results also show that the mediating effects of financial literacy, economic and financial information search, and risk tolerance account for 7.2%, 26.6% and 15.0% of total effects of financial education on the probability of risky financial asset holding, respectively.

Heterogeneity Results
This section reports the heterogeneity of the relationships between financial education and risky financial asset holding in terms of several household characteristics: household respondents' education, political status, and age. For households with different characteristics, financial education may play different roles in risky financial asset holding. The regression results are shown in Table 10. However, for consumers with lower education levels, they receive financial education more likely for investment purposes and the marginal benefits of financial education for them are larger than those with higher education levels.
We also explore the heterogeneous associations on different political statuses in columns three and four. The coefficient and marginal effect of the interaction terms are significant and negative. This means that for members of the Communist Party, financial education has a smaller association with the probability to hold risky financial assets than their counterparts. Party membership is a promising indicator of associational social capital in China (Knight & Yueh, 2008). Therefore, the findings could be explained in that most party members are highly educated and have abundant social network resources and have basic financial knowledge and market information without receiving financial education. The marginal benefits of financial education for non-party members are larger than their counterparts.
Additionally, we explored the heterogeneous associations on age in columns five and six. The coefficient and marginal effect of the interaction term are significant and positive. This means that for the older household respondents, financial education has higher associations with the probability to hold risky financial assets than their counterparts. For young people, they are curious about something new and can accept and understand them quickly, and can actively participate in the financial market. Our cross tabulation between age and risky financial asset holding in Table A2 show that there is higher probability of holding risky financial assets in younger consumers. Although the Chinese financial market developed with rapid space, the financial market participation of Chinese consumers by holding risky assets is still low, with only about 16.34% households in 2017 2 . Financial markets and related financial knowledge are new things for older people which they know little about and don't invest without financial education. In addition, according to the life cycle hypothesis, older consumers accumulated their income for retirement and they have greater wealth than younger consumers. This means the marginal benefits of financial education for them is larger, which leads to a higher association between financial education and risky financial asset holding than younger consumers. the mediating variable regression (equations one to three). The results are shown in the Appendix (Table   A3). Columns 1 and 2 are the baseline results. The coefficient and marginal effect of financial education are significant and the signs are the same as those in Table 5. In addition, we tested the robustness of mediating regression in Section 4.3. Columns three to five are the regressions with financial literacy as the mediator. The results show that financial literacy has a certain mediating effect through which financial education is associated with the probability of risky financial asset holding. Columns six to eight and columns nine to eleven test the mediating effects of economic and financial information search, and risk tolerance, respectively. The results show that economic and financial information search and risk tolerance have certain mediating effects between financial education and the probability of risky financial asset holding. The above results are consistent with our previous findings which means that our previous results are robust.  (Table A4). We also show a strong and significant associations between financial education and risky financial asset holding.

Different Types of Risky Financial Assets
Considering that different types of risky financial assets have different risk, we estimate the heterogeneous responses of different type of risky financial assets. Column three to ten of Table A4 consider four main types: stock, funds, wealth management products and bonds which exclude government bonds. Financial education has positive associations with all these four types of risky financial assets, especially with higher associations with stock market participation.

Limitations and Directions for Future Research
Several limitations exist in this study. First, we only used a simple conceptual framework but didn't propose a formal theoretical model that describes the association between financial education and risky financial asset holding and its mechanisms. Second, only survey data was used. Third, this paper only considered whether or not consumers hold risky financial assets. Fourth, we only considered three mediators between financial education and risky financial asset holding. Fifth, financial education was measured by only a single item.
Following the approach used by Paul and Mas (2020), we propose three directions for future research in terms of theories, constructs, and methods. First, a more formal theory on the association between financial education and risky financial asset holding and its mechanisms could be developed.
Relevant theories in economics, finance, psychology, and education could be considered for theory building. Existing theories such as the life cycle consumption theory (Modigliani & Brumberg, 1954), theory of planned behavior (Ajzen,1991), the financial literacy model proposed by Lusardi, Michaud and Mitchell (2017), portfolio theory (Markowitz, 1952)

Conclusion
This study has used the 2015 China Household Financial Survey to examine the association between financial education and risky financial asset holding. The results suggest that financial education may increase the probability to invest in risky financial assets, especially for consumers being at lower education levels, non-party members, and older.

Income
The total amount of household income in last year, including the wage income with after tax wage, after tax bonus, after-tax subsidies or subsidy in-kind and money obtained from the second job, agricultural income, business income and rental income.