Confidence, financial literacy and investment in risky assets: Evidence from the Survey of Consumer Finances

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Confidence, financial literacy and investment in risky assets: Evidence from the Survey of Consumer Finances Andrej Cup´ak, Pirmin Fessler, Joanne W. Hsu, Piotr R. Paradowski 2020-004 Please cite this paper as: Cup´ak, Andrej, Pirmin Fessler, Joanne W. Hsu, Piotr R. Paradowski (2020). “Confidence, financial literacy and investment in risky assets: Evidence from the Survey of Consumer Finances,” Finance and Economics Discussion Series 2020-004. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2020.004. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Confidence, financial literacy and investment in risky assets: ∗ Evidence from the Survey of Consumer Finances Andrej Cupák a, Pirmin Fessler b, Joanne W. Hsu c, and Piotr R. Paradowski d January 2020 Abstract We employ recent Survey of Consumer Finances (SCF) microdata from the US to analyze the impacts of confidence in one’s own financial knowledge, confidence in the economy, and objective financial literacy on investment in risky financial assets (equity and bonds) on both the extensive and intensive margins. Controlling for a rich set of covariates including risk aversion, we find that objective financial literacy is positively related to investment in risky assets as well as debt securities. Moreover, confidence in own financial skills additionally increases the probability of holding risky assets and bonds. While these relationships are rather robust for the extensive margin, they break down with regard to the conditional share of financial wealth in risky assets of those who actually hold them. The relevance of financial literacy as well as confidence varies considerably with the distribution of wealth as well as across several socio-economic dimensions such as age, education and race. JEL Classifications: D12, D14, D31, D91, I20, G11. Key Words: financial literacy, confidence, risky assets, household finance, survey data, US. ∗ The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or members of the Board of Governors of Federal Reserve System, the Oesterreichische Nationalbank, the National Bank of Slovakia, the Eurosystem, or the LIS Cross-National Data Center in Luxembourg. Any omissions and remaining errors in the text are solely ours. a LIS Cross-National Data Center in Luxembourg; National Bank of Slovakia, e-mail: andrej.cupak@gmail.com. b Foreign Research Division, Oesterreichische Nationalbank, e-mail: pirmin.fessler@oenb.at. c Corresponding author. Research and Statistics, Board of Governors of the Federal Reserve System, e-mail: joanne.w.hsu@frb.gov. d LIS Cross-National Data Center in Luxembourg; Gdańsk University of Technology, e-mail: piotr.paradowski@pg.edu.pl. 1

1 Introduction The composition of household portfolios differs substantially across households and countries (e.g. Christelis et al., 2013; Badarinza et al., 2016). While economic theory predicts that household portfolios should generally be diversified and include more sophisticated financial instruments like equities, empirical evidence shows that few households actually hold such assets (Haliassos and Bertaut, 1995). Low engagement of households in equity markets could yield significant welfare losses — nonparticipation can generate welfare losses of up to 2% of annual household consumption (Cocco et al., 2005). Similarly, “financial ignorance” can also be costly. For example, one might under-diversify one’s financial assets, generating unnecessary portfolio risk, or choose unfavorable mortgages (Campbell, 2006; Calvet et al., 2007). With emerging high-quality micro-data on household balance sheets, researchers have related stock market participation in particular and investment in financial assets in general to household characteristics including gender, education, income, or risk aversion (e.g. Campbell, 2006). More recently, the role of financial literacy has been highlighted as a key input for sound financial behavior (see Lusardi and Mitchell, 2014 for a comprehensive overview). Few empirical studies have shown that confidence in one’s own financial knowledge matters for investment decisions (e.g. Xia et al., 2014; Allgood and Walstad, 2016; Bannier and Schwarz, 2018). While the empirical literature on the importance of financial literacy is already well established, evidence on the role of confidence remains rather limited. In this paper, we investigate the influence of financial literacy, confidence in one’s own financial knowledge, and confidence in the future development of the economy on household portfolio allocation using recent data from the Survey of Consumer Finances (SCF). We particularly focus on investment in equity and investment in bonds. First, we analyze the relationship between financial literacy together with the two confidence measures and the probability of holding risky assets (i.e. direct and indirect investment in stocks) and the probability of holding bonds. Second, we analyze the determinants of the intensive margin of investing in equities and bonds. To do so, we employ standard regression techniques as well as unconditional quantile regressions based on the recentered influence function (Firpo et al., 2009) to analyze the effects at different points of the distribution. To the best of our knowledge, our study is the first to utilize the SCF to analyze the role of financial literacy and confidence as determinants of household investment behavior. The SCF is a comprehensive household wealth survey, containing 2

detailed information on risky assets and bonds, together with a rich set of covariates including financial literacy, confidence, and risk aversion providing a rare opportunity. Our analysis makes an important contribution to the empirical literature on understanding the “equity premium puzzle” from the household perspective. Furthermore, an analysis of US households provides an interesting addition to the previous empirical literature on the role of financial literacy and confidence in household investment behavior and also helps to understand generalizability of the previous research results: for instance, in China (Xia et al., 2014; Chu et al., 2017), Germany (Bannier and Schwarz, 2018), and the Netherlands, (Kramer, 2016). Furthermore, individual financial decision-making might be even more important in the US context, where the government Social Security plan is intended to support retirees alongside employer-linked pension plans and personal savings. Finally, in our empirical analysis we are able to control for confidence in the development of the economy. To the best of our knowledge, none of the previous empirical studies have controlled for confidence in the economy when analyzing confidence in one’s own financial literacy with respect to household investment behavior. Our descriptive results show that overall around 52% of households in the US participate in the market for risky assets and around 10% of households hold bonds and other secured financial assets. The conditional share of total financial assets held in equities is above 40%, with nearly 9% in bonds. At the same time, only about 50% of the US households can correctly answer all three (relatively basic) financial literacy questions on the survey. Around 78% of the respondents can correctly answer the questions related to interest rates and inflation. The concept of riskiness of financial instruments is less-understood, with around 63% of respondents correctly answering this question. In addition, US households seem to be fairly confident in their financial knowledge, with an average self-rating of 7.35 on a scale to 0 - 10. We find a weak correlation between confidence and objective financial literacy (0.173), consistent with previous empirical studies. On top of it, nearly 40% of households are confident that the US economy will be doing much better in 5 years as compared to the year of the survey, 2016. Results from regression analyses indicate that, controlling for a rich set of family characteristics, both financial literacy and confidence in own financial knowledge positively affect the extensive-margin decision of participating in equity markets. Participation in the market for secured financial assets such as bonds is mostly driven by the level of financial literacy. While the confidence in own financial knowledge exert less influence over the intensive-margin decision for 3

holding risky assets and bonds, the impact of confidence in the economy turns out to be positive and statistically significant, especially in the case of bonds. Despite the fact that households are not professional forecasters, trust in the economy appears to help shape their investment decisions. Furthermore, recentered influence function regressions show considerable heterogeneity in the estimated impacts of the covariates across the distribution of both equity and bonds. Our results support recent work arguing that structural changes in financial attitudes (including literacy and sophistication) might be behind the growing stock-market participation in the US (Bilias et al., 2017). Yet, the overall engagement of household in markets for risky assets (especially direct participation) is still far from what economic theory predicts. Our findings could be seen as a potential policy tool to further foster stock-market participation and investment behavior in the US. The paper proceeds as follows. The previous literature is reviewed and testable hypotheses are developed in section 2. The data and variables are described in section 3 and section 4 presents the implemented empirical framework and strategies. Results of the univariate and multivariate analysis, along with several robustness checks, are presented and discussed in section 5. Finally, section 6 concludes and offers policy implications. 2 Related literature and conceptual framework 2.1 Financial literacy Financial literacy has been highlighted as a key input for sound financial decision-making for both participation in financial assets as well as liabilities. The positive impact of financial literacy and cognitive abilities on the participation in risky asset markets in particular and financial wealth accumulation in general has been documented in several empirical studies. For example (see Lusardi and Mitchell, 2014, for a comprehensive overview), Christelis et al. (2010) studied the relationship between cognitive abilities and stockholding using survey microdata from 11 European countries for a population aged 50+, and found that cognitive abilities positively relate to direct and indirect stockholding. In another influential study, Van Rooij et al. (2011) found that in a sample of Dutch households, those with lower levels of financial literacy are less likely to invest in stocks, and vice versa. Furthermore, Von Gaudecker (2015) found that households in The Netherlands that score high on financial literacy or rely on professional financial 4

advice achieve better investment outcomes and more diversified portfolios.1 Research has also documented the positive impact of financial literacy on pension savings and retirement planning (Van Rooij et al., 2012; Cupák et al., 2019). Based on the robust empirical evidence of positive impacts of financial literacy on financial wealth accumulation and participation in financial instruments, we hypothesize that the ability to understand financial concepts of interest rates, inflation, and risk diversification (captured by the financial literacy index) is associated with a higher propensity to hold sophisticated financial instruments such as equity and bonds as well as larger conditional values invested in these financial instruments. 2.2 Confidence In addition to the importance of financial literacy for household and personnel financial outcomes, other research has focused on more psychological aspects of household investment behavior. Barber and Odean (2000, 2001), in their seminal works, stressed the importance of confidence of retail investors for high levels of trading. The authors conclude that overconfidence can explain a large part of the high trading levels, but at the same time can result in poor performance of individual investors. The authors also highlight the gender difference in confidence with respect to investing. Recent empirical work includes studies using Chinese survey data demonstrating that (over)-confidence (proxied by a difference between subjective and objective financial literacy) is positively related to stock market participation (Xia et al., 2014; Chu et al., 2017). More confident households have been shown to have higher levels of financial wealth in a sample of German households (Bannier and Schwarz, 2018). Kramer (2016) in addition found that, in the sample of Dutch households, people confident in their own financial literacy are less likely to seek financial advice, which in turn has implications for their wealth accumulation. Finally, Allgood and Walstad (2016) examined web-based survey data of US households and found that confidence in one’s own financial knowledge matters on top of measured financial literacy, especially for investment decisions. While the impact of financial literacy and cognitive abilities on household financial outcomes is well documented in the empirical literature, there is significantly fewer studies on the importance of confidence in financial assets holding and investment decisions, perhaps due to lack 1 Similarly, Shin, et al. (2020) find that, controlling for financial literacy and confidence, using financial planners is associated in more diversified portfolios. 5

of the available data. Given the available empirical evidence from multiple countries we hypothesize that confidence in one’s own financial knowledge is associated with a higher propensity to hold risky financial assets and – conditional on holding them – to a larger share of these assets in overall financial wealth. Finally, the SCF contains information on the expectations of households about the development of the US economy in 5 years, which we use as proxy for the confidence in the macroeconomic environment. Although households are generally neither professional traders nor forecasters, their beliefs about the future economic development and business cycles likely shape their financial decisions. Indeed, the available literature shows that household expectations play an important role in determining economic behaviors such as life-cycle consumption (e.g. Jappelli and Pistaferri, 2000) or choosing an optimal level of debt (e.g. Brown et al., 2005). Regarding investment in financial assets, Guiso et al. (2008) found that in Italy and The Netherlands, individuals who trust less the stock market hold less stocks, and they also invest less into such assets conditional on participation. Our variable captures the confidence in the overall performance of the economy and not stock markets specifically, so the confidence we measure may be secondary to the decision to participate in financial markets. That said, the empirical evidence suggests that the interdependence between stock markets and economic growth is positive (e.g. Levine and Zervos, 1988). We therefore hypothesize that households that believe that the economy in 5 years will perform better than today are more likely to invest in risky assets, and also to invest more into these assets conditional on participation. 3 Data and variables We use microdata from the 2016 wave of the Survey of Consumer Finances (SCF), covering 6248 US households. The SCF is a US nationally representative survey of households conducted triennially by the Board of Governors of the Federal Reserve System to gather comprehensive data on household wealth, and describes both the ownership and magnitude of particular assets and debt. The survey also includes information on families’ pensions, income, and demographic 6

characteristics.2 Our unit of analysis is the household, which is represented by the original respondent to the SCF survey questions.3 The main advantage of our data compared to other wealth micro data is that the SCF includes comprehensive measures of household balance sheets, as well as rich set of covariates. Our data provide the rare opportunity to analyze standard measures for financial literacy, confidence, and risk aversion alongside a full wealth survey. Please refer to the appendix for a detailed description of all variables, including the text of the survey questions and how our measures were constructed. 3.1 Dependent variables The primary outcome of interest is investment in risky financial assets, measured at the household level. While the level of risk differs substantially across different types of financial instruments, we group them into two broader categories, which we refer to as equities and bonds. These categories includes investments in financial instruments, which are directly or indirectly linked to stocks or investment in bonds and other debt securities.4 Our first set of outcome variables are a dummy variable taking a value of 1 if household holds equities and 0 otherwise, and a dummy variable taking a value of 1 a household holds bonds and 0 otherwise. Our second set of outcome variables are then the share of total financial assets (including retirement accounts) held in equities out of the as well as the share of total financial assets held in bonds, conditional on holding those financial assets. 3.2 The main explanatory variables We focus on levels of financial literacy and confidence, which were added to the SCF in the most recent wave in 2016. The financial literacy score is created by counting the number of questions 2 For more information about the survey, visit: https://www.federalreserve.gov/econres/scfindex.htm. For a detailed overview of the main results from the 2016 SCF data, see Bricker et al. (2017). 3 In the original survey household unit is composed of all individuals who are financially interdependent and referred to as the "primary economic unit" (PEU). For all individual-level characteristics, we use the respondent, who is generally the most financially knowledgeable person in the PEU. We do not use in our analysis the notion of head as it is described in the SCF codebook, namely "the head is taken to be the single core individual in a PEU without a core couple; in a PEU with a central couple, the head is taken to be either the male in a mixed-sex couple or the older individual in the case of a same-sex couple". 4 A full overview of all the considered assets is listed in Appendix A. 7

answered correctly. The financial literacy questions are the standard three questions proposed by Lusardi and Mitchell (2014) covering inflation, interest rates, and riskiness. The confidence measure is created in two ways. In general, the most direct way to measure subjective confidence is a survey, but it is not trivial to do so.5 Here we rely on the approach of Bannier and Schwarz (2018) and proxy for the level of confidence in own skills by using the self-assessed level of financial knowledge. In the SCF, the household respondents were asked the following question: “On a scale from zero to ten, where zero is not at all knowledgeable about personal finance and ten is very knowledgeable about personal finance, what number would you (and your husband/wife/partner) be on the scale?” The SCF also includes another variable on expectations about the future macroeconomic development of the US economy in 5 years’ horizon, which we apply as a measure of a different kind of confidence. Respondents were asked the following: “I'd like to start this interview by asking you about your expectations for the future. Over the next five years, do you expect the U.S. economy as a whole to perform better, worse, or about the same as it has over the past five years?” Dominitz and Manski (2004) describe this kind of variable as confidence in the economy. See, again, Appendix A for the creation of financial literacy and confidence scores. Figure 1 shows the response behavior with regard to these three variables. Figure 2 shows both measured financial literacy as well as self-assessed financial literacy jointly. The size of each bubble is proportional to the number of respondents with such a combination of measured and selfassessed financial literacy. We see a positive, albeit imprecisely measured correlation between measured and self-assessed financial literacy. A considerable proportion of observations (𝜌𝜌 = 0.173) are also observed in the lower right area, namely individuals with rather high self-assessed financial literacy but rather low measured financial literacy. These individuals could be described as “overconfident” with regard to their financial literacy. We will further investigate the impact of (over)confidence and financial literacy on equity and bond holdings in our multivariate analysis. [Figure 1 about here] 5 See Dominitz and Manski (2004) for a detailed discussion on measuring consumers’ confidence with micro and macro data. 8

[Figure 2 about here] 3.3 Control variables We also control for a rich set of covariates that could also drive the investment choices of households. Following prior literature, we include variables capturing economic resources of the household, both net real estate wealth and income.6 We also consider whether or not a household owns a closely-held business or received an inheritance. We also control for an extensive set of demographic characteristics (age, gender, education, marital status, ethnicity, and employment status of the respondent) and family composition, defined as the household size, and presence of children under the age of 18. Furthermore, we include a subjective measure of risk aversion, which has been identified as an important factor determining household risky behavior. Detailed explanation of all control variables construction is shown in Appendix B, while the descriptive statistics are presented in Table 1. [Table 1 about here] 4 Empirical strategy Our empirical strategy is motivated by standard theories of household portfolios (see Guiso et al., 2002). A household faces a two-step decision: the binary participation decision (probability of holding an asset), and a continuous allocation decision (share of invested amounts in the particular asset relative to total financial wealth). We model these two decisions, with a particular focus on financial literacy and confidence, by several econometric techniques. We employ standard regression models as well as unconditional quantile regressions based on recentered influence function to assess the relative importance of financial literacy and confidence over the distribution of the equity shares. 6 As our outcome variable covers financial wealth, our control variable uses non-financial wealth to avoid endogeneity due to simultaneity. Furthermore, we subtract liabilities to generate a measure of net real estate wealth. Since our income and wealth variables have highly skewed distributions, and a significant share of households possesses negative wealth, we employ an Inverse-hyperbolic-sine (IHS) transformation to these two variables. Note that the interpretation of the coefficients of the IHS transformed variables is very similar to that of a natural log. 9

4.1 Participation decision In our baseline model, we estimate the relationship between participation in equity / bonds and financial literacy, and confidence measures (controlling for a large set of covariates) by a Linear Probability Model7, which takes the following form: , (1) ′ wh𝑃𝑃e𝑃𝑃re( 𝐷𝐷_𝑒𝑒𝑒𝑒 𝑒𝑒is𝑒𝑒 𝑒𝑒𝑒𝑒th;e𝐷𝐷 _le𝑏𝑏v𝑏𝑏e𝑏𝑏l 𝑏𝑏o𝑏𝑏f= m1e|a𝑋𝑋s)ur=ed𝛽𝛽 0 fi+na𝛽𝛽n 1 ci 𝐹𝐹al𝐹𝐹 l+ite𝛽𝛽r 2 a c𝐶𝐶y𝑏𝑏 𝑏𝑏a𝐶𝐶n 1 d +𝛽𝛽3𝐶𝐶𝑏𝑏 𝑏𝑏a𝐶𝐶n 2 d+ 𝛽𝛽4 𝑋𝑋 + re𝑒𝑒present the individu 𝐹𝐹 a 𝐹𝐹 l’s confidence in own financial knowledge and c 𝐶𝐶 on 𝑏𝑏 f 𝑏𝑏 id 𝐶𝐶1ence in 𝐶𝐶 𝑏𝑏 t 𝑏𝑏 h 𝐶𝐶 e2 US economy, respectively, as described in Section 3. Vector includes a large set of control variables usually used to predict financial literacy and generally financial behavior of households such as education, 𝑋𝑋 gender, age, race, employment status, income, wealth, closely-held business ownership, inheritance, risk tolerance, household size and number of children, etc., and is the error term. 4.2 Allocation decision 𝑒𝑒 For those households holding equities and bonds ( ), we model the relationship between the share of financial wealth invested in the asset type and our explanatory (𝐷𝐷_𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 = 1) 𝐷𝐷_𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 = 1 variables of interest by the following linear regression model: , (2) ′ 𝑊𝑊_𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒;𝑊𝑊_𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 = 𝜋𝜋0 +𝜋𝜋1 𝐹𝐹𝐹𝐹+𝜋𝜋2 𝐶𝐶𝑏𝑏𝑏𝑏𝐶𝐶1 +𝜋𝜋3𝐶𝐶𝑏𝑏𝑏𝑏𝐶𝐶2 +𝜋𝜋4 𝑋𝑋 +𝑣𝑣 Where the notation of right-hand-side variables remains the same as in the participation equation (1), except the new error term . Next, we estimate the influence of our regressors at different parts of the distribution of 𝑣𝑣 our outcome variable , the share of equity and bonds held in financial assets for those household 𝑋𝑋 who hold equity and bonds. We employ unconditional quantile regressions (UQR) developed by 𝑌𝑌 Firpo et al. (2009), which are based on recentered influence functions (RIF). This framework has several advantages over the standard conditional quantile regressions and is an attractive tool for researchers studying distributional impacts. The UQR estimator can be written as follows: 7 As a robustness check, we re-estimated the baseline participation-stage equations (Tables 2a and 2b) by means of Probit models. The estimated marginal effects at the mean of explanatory variables did not differ significantly from the Linear Probability Model estimates. Since we consider interaction terms in some of our estimates (Tables E.2a and E.2b), and because of its straightforward interpretation, we prefer the Linear Probability Model as the baseline estimates. Results of the Probit estimates are available upon request. 10

(3) 𝑈𝑈𝑈𝑈𝑈𝑈 W 𝑅𝑅 e 𝑅𝑅 e 𝐹𝐹 st ( i 𝑌𝑌 m , a 𝜏𝜏 t ) e = UQ 𝑋𝑋 R 𝛽𝛽 s for + t 𝜀𝜀 a . king values of 0.10, 0.25, 0.5, 0.75, and 0.9.8 Finally, we would like to point out that missing values in some of the SCF variables (mostly 𝜏𝜏 related to assets, debts and incomes) were imputed and replaced 5 times.9 Multiple-imputed data allow us to consider imputation uncertainty related to item non-response while obtaining statistical inference. We follow the standard procedure suggested by Rubin (1987) to obtain correct point estimates and variance estimation of the statistics of interest.10 We describe the applied multipleimputation technique in Appendix D. 5 Results 5.1 Descriptive analysis Figures 3 and 4 show the relationship between confidence (self-assessed financial literacy) and the probability of holding the two types of assets. The figures summarize our main findings: confidence is positively related to the probabilities of holding equities and bonds (left panels of Figures 3 and 4). The result remains when controlling for measured financial literacy and confidence in the economy (right panel of Figures 3 and 4). Therefore, confidence in one’s own financial knowledge matters for holding equities as well as bonds beyond the effect of objective financial knowledge and the confidence in the country’s economy. As we will see in section 5.2, this relationship holds even if we control for a larger set of socio-economic variables relevant for household financial behavior. [Figure 3 about here] [Figure 4 about here] Analogously, Figures 5 and 6 show the relationship between confidence in own financial knowledge and the share of financial assets held in equity and in bonds for those who actually hold those assets. This positive relationship breaks down when we control for measured financial literacy and confidence in the economy. While confidence in own skills seems to be rather relevant 8 We briefly describe the UQR framework in Appendix C. 9 The imputation procedure in the SCF data is described in detail by Kennickell (1998). 10 We use the package to do so. STATAmi 11

for the participation decision in the market of risky assets and bonds, it seems to matter much less with regard to the allocation decision for those who actually participate. [Figure 5 about here] [Figure 6 about here] 5.2 Regression analysis The baseline results of our multivariate analyses are presented through tables 2 to 4. Table 2a and Table 2b show the results of the Linear Probability Model estimates for participation in equity and bonds as outlined in section 4 (equation 1). We find that financial literacy, confidence in the economy and confidence in own skills are all positively related to the probability of holding equity (Table 2a). However, only the effects of financial literacy and self-confidence are precisely measured. The effect of financial literacy remains stable over the various specifications, all of which control for a large set of covariates relevant for household financial behavior. Most importantly, in specifications (4) and (7), we see that the effect of self-confidence remains positive and significant even when controlling for objective financial literacy. That means, that the finding that confidence in skills matters above the actual skills (graphically demonstrated in Figure 3) holds even if we control for a large set of covariates. Results from the linear probability model estimates for participation in bonds and other debt securities (Table 2b) reveal that only the impact of objective financial literacy is positive and precisely measured. The estimated effects of other covariates are in line with the previous literature. The probability of holding equities and bonds significantly rises with economic resources: wealth, income (only for equities), employment, as well as inheritance receipt. It also rises generally with age, decreasing somewhat at oldest ages, and rises with education. Owners of privately-held businesses are less likely to participate in equity markets, which is unsurprising given that equities and privately-held businesses can both be considered risky assets. Non-Hispanic white respondents are more likely to own stocks and bonds than non-white groups. This result is in line with Thomspon and Suarez (2015), who document substantial financial wealth gaps by race in the US. Our results also show that self-rated risk tolerance is positively associated with holding equities, but has no effect with respect to holding bonds, and hence suggests that households are quite consistent in assessing their risk aversion and actual risky behavior. 12

[Table 2a about here] [Table 2b about here] Tables 3a and 3b display estimates of regressions of the determinants of the share of total financial assets held in equities and bonds, respectively, conditional on participating in those markets. The results confirm the descriptive analysis from Figures 5 and Figure 6. Estimated results reveal little to no relationship between the portfolio allocation measures and financial literacy or confidence. One interesting exception is that confidence in the economy has positive and precisely measured (p<0.05) relationship with the share of financial wealth invested in bonds. Owners of privately-held businesses invest a greater share of their financial wealth in bonds but less in stocks, which improves the diversification of their overall portfolios. The overall precision of other coefficient estimates decreases relative to the participation regressions. [Table 3a about here] [Table 3b about here] However, further investigation using UQRs reveals that the lack of a measurable relationship for the conditional mean masks substantial heterogeneity across the distribution of the share of equity and bonds in total financial assets. For example, in the case of the share of financial wealth held in equity (Table 4a): while the 25th percentile of equity holdings exhibits a positive relationship with measured financial literacy and no relationship with either confidence measures, we find a significant and negative impact of financial literacy on the 90th percentile of equity holdings. That means, that for households generally holding riskier assets – given their set of covariates – those with higher literacy and higher confidence actually hold a little less of their financial wealth in stocks. That is, for those with comparably risky portfolios, higher financial literacy and confidence is associated with mean-reverting allocation behavior. Note that this differential behavior at top of the distribution holds in spite of the fact that we control for business ownership, so it is not specifically driven by business owners with substantial other risky assets. On the other hand, measured financial literacy and self-confidence have imprecisely estimated effects across the distribution of the share of financial wealth held in bonds (Table 4b). The impact of confidence in the performance of the US economy is positive and precisely measured between the 50th and 75th percentile of the distribution. Our results show that confidence 13

in the economy—including the performance of the financial markets—shapes household investment decisions, especially in the case of bonds. [Table 4a about here] [Table 4b about here] 5.3 Robustness Checks and Extensions We estimate several additional models serving as robustness checks and extensions. First, we estimate our model using a more explicit measure of overconfidence. Following Xia et al. (2014) and Allgood and Walstad (2016), we combined the objective (measured) financial literacy score with the subjective 0-10 self-confidence score to create a four-category measure of overconfidence. We create dummy variables for different degrees of confidence with respect to the measured financial literacy: above-average self-rated literacy and below-average measured literacy (over-confident), above-average self-rated literacy and above-average measured literacy (soundly-confident), below-average self-rated literacy and above-average measured literacy (under-confident), and below-average self-rated literacy and below-average measured-literacy, visualized in the following diagram: 14

As shown in Table E.1a and Table E.1b, the results from this robustness check suggest that discretized measures of confidence matter on top of the actual measured literacy, especially for participation in risky assets. [Table E.1a about here] [Table E.1b about here] Next, we analyze the role of gender in financial literacy and confidence for ownership of equities and bonds. This extension is motivated by a body of literature finding gender differences in financial literacy and attitudes towards finance, including differences in confidence and risk aversion (see Cupák et al., 2018; Schwarz and Bannier, 2018). Here, we estimate several models interacting financial literacy and gender, as well as confidence measures and gender, in addition to the standard set of covariates considered in baseline models. The interaction terms allow for gender-specific impacts of financial literacy or confidence on financial behavior. Results from this exercise are presented in Table E.2a and Table E.2b. Similarly to Schwarz and Bannier (2018), we find that financial literacy has a similar effect on risky asset behavior for both men and women. We find that the relationship between self-confidence and participation in the stock market is weaker for women, though this is precisely measured only at the 0.1 level (Table E.2a columns 3 and 5). The influence of self-confidence on the intensive margin does not appear to be different for men and women (Table E.2a columns 8 and 10). We find no differential effect by gender of financial literacy or either confidence measure on the extensive margin for bonds (Table E.2b columns 1-5). We also find that the effect of being confident into the economy on the share of financial wealth in bonds is weaker for women (columns 9 and 10 in Table E.2b), although not precisely measured. Note that Schwarz and Bannier (2018) find much stronger gender-related effects of confidence on financial wealth accumulation in the sample of German households. [Table E.2a about here] [Table E.2b about here] 5.4 Limitations Given that we were unable to address the possible endogeneity of financial literacy and confidence, we caution against a strict causal interpretation of our results. Financial literacy and confidence have been identified in some previous empirical studies as endogenous determinants with respect 15

to investment in financial wealth. The SCF data contain some variables, which could be potentially employed as relevant instruments for financial literacy, yet they would not fulfill the exclusion restriction condition of valid instruments.11 Confidence is also possibly an endogenous variable. We are not aware of any relevant instruments in the SCF data for the possibly endogenous confidence variable. Given the previous research (see, e.g. Lusardi and Mitchell, 2014), it is very likely that the effect of measured financial literacy would strengthen after addressing its endogeneity. We are not so confident in saying what would happen to the effect of confidence, as the previous empirical literature addressing endogenous confidence is very limited. Even in the presence of valid instruments, models with multiple endogenous variables at the same time (measured literacy and confidence in our case) are extremely hard to identify and to interpret (Angrist and Pischke, 2008). The measured effects hence should be interpreted as conditional correlations or predictive effects rather than causal effects. Despite of this limitation, we believe that our rich empirical results offer interesting insights into the financial behavior of the US households. 6 Conclusion In this paper we analyzed unique data on financial literacy, confidence in one owns financial knowledge as well as confidence in the economy and investment into risky assets (equity and bonds) using the recent 2016 SCF wealth microdata. The 2016 survey included, for the first time, the three standard financial literacy questions (Lusardi and Mitchell, 2014). Therefore, it allows to jointly analyze financial literacy of the respondent as well as the according household balance sheets the respondent lives in. The 2016 survey also asked respondents to provide a self-assessment with regard to their own financial knowledge, and a question asking households about their future economic expectations which we employ as a proxies for confidence. We contribute to the literature on portfolio allocation by linking a novel set of behavioral variables on financial literacy and confidence with detailed, disaggregated information on financial assets held in stocks or bonds. We show that measures of objective financial literacy, confidence in one’s own financial knowledge, and confidence in economic conditions are all positively related 11 For example, mother’s and father’s education variables are indeed positively correlated with financial literacy; however, it is unlikely they would influence the outcome only via endogenous financial literacy and not by any other channel. 16

to the probability of investing in equities, whether analyzed separately or included together. Confidence in one’s own financial skills is relevant, even holding those skills constant. However, while this relationship holds even when controlling for a large set of socio-economic characteristics highly predictive for financial literacy and participation in markets for risky assets, it breaks down in the analysis of the intensive margins. Conditional on actually holding risky assets, the share of total financial assets held in risky assets still rises with financial literacy but no longer with confidence. Using unconditional quantile regressions, we find evidence that financial literacy has a positive and significant effect on the 25th percentile of equity shares, but a negative effect on the 75th. Among households typically holding higher shares of risky assets the financially literate hold relatively lower shares. Financial literacy therefore dampens risk appetite within the comparably risk-tolerant households. The impact of confidence in the economy mostly matters for the allocation decision of financial wealth into bonds, especially at the median and the 75th percentile of the bonds’ distribution. Our results are consistent with past findings on the effect of race and other socioeconomic characteristics on the holdings of financial assets. Most interestingly we reproduce the finding that white families tend to have larger shares of risky assets. We also confirm that lower risk aversion is associated with a higher probability of stock market participation, but no measurable effect for participation in bond markets. Results from robustness checks and extensions suggest that the effects are likely decreasing somewhat over the distribution of net wealth. In sum, measured financial literacy together with confidence in own skills as well as confidence in the economy seem to be yet another powerful determinants explaining participation and allocation decisions of households in markets for risky assets and bonds. This result confirms the recent findings of Bilias et al. (2017) who argue that structural changes in financial attitudes (including literacy and sophistication) of the US households might be behind the growing stockmarket participation. Yet, the overall engagement of household in market for risky assets is still far from what the economic theory predicts. Our results could be seen to inform potential policy tools, perhaps through initiatives to increase financial knowledge, to foster stock-market participation in the US. 17

References Allgood, S., and Walstad, W. B. (2016). The effects of perceived and actual financial literacy on financial behaviors. Economic Inquiry, 54(1), 675–697. Angrist, J. D., and Pischke, J. S. (2008). Mostly harmless econometrics: An empiricist's companion. New Jersey: Princeton University Press. Badarinza, C., Campbell, J. Y. and Ramadorai, T. (2016). International comparative household finance. Annual Review of Economics, 8, 111–144. Bannier, C. E., and Milena Schwarz, M. (2018). Gender-and education-related effects of financial literacy and confidence on financial wealth. Journal of Economic Psychology, 67, 66–86. Barber, B. M., and Odean, T. (2001). Boys will be boys: Gender, overconfidence, and common stock investment. The Quarterly Journal of Economics, 116(1), 261–292. Barber, B. M., and Odean. T. (2000). Trading is hazardous to your wealth: The common stock investment performance of individual investors. The Journal of Finance, 55(2), 773–806. Bilias, Y., Georgarakos, D., and Haliassos, M. (2017). Has Greater Stock Market Participation Increased Wealth Inequality in the US?. Review of Income and Wealth, 63(1), 169–188. Bricker, J., Dettling, L. J., Henriques, A., Hsu, J. W., Jacobs, L., Moore, K. B., Pack, S., Sabelhaus, J., Thompson, J., and Windle, R. A. (2017). Changes in US family finances from 2013 to 2016: Evidence from the Survey of Consumer Finances. Federal Reserve Bulletin, 103(3), 1–42. Brown, S., Garino, G., Taylor, K., and Price, S. W. (2005). Debt and Financial Expectations: An Individual‐ and Household‐Level Analysis. Economic Inquiry, 43(1), 100–120. Calvet, L. E., Campbell, J. Y., and Sodini, P. (2007). Down or out: Assessing the welfare costs of household investment mistakes. Journal of Political Economy, 115(5), 707–747. Campbell, J. Y. (2006). Household finance. The Journal of Finance, 61(4), 1553–1604. Christelis, D., Georgarakos, D., and Haliassos, M. (2013). Differences in portfolios across countries: Economic environment versus household characteristics. Review of Economics and Statistics, 95(1), 220–236. Christelis, D., Jappelli, T., and Padula, M. (2010). Cognitive abilities and portfolio choice. European Economic Review, 54(1), 18–38. Chu, Z., Wang, Z., Xiao, J. J., and Zhang, W. (2017). Financial literacy, portfolio choice and financial well-being. Social Indicators Research, 132(2), 799–820. 18

Cocco, J. F., Gomes, F. J., and Maenhout, P. J. (2005). Consumption and portfolio choice over the life cycle. The Review of Financial Studies, 18(2), 491–533. Cupák, A., Fessler, P., Schneebaum, A., and Silgoner, M. (2018). Decomposing gender gaps in financial literacy: New international evidence. Economics Letters, 168, 102–106. Cupák, A., Kolev, G. I., and Brokešová, Z. (2019). Financial literacy and voluntary savings for retirement: novel causal evidence. The European Journal of Finance, 25(16), 1606–1625. Dominitz, J., and Manski, C. F. (2004). How should we measure consumer confidence?. Journal of Economic Perspectives, 18(2), 51–66. Firpo, S., Fortin, N. M., and Lemieux, T. (2009). Unconditional quantile regressions. Econometrica, 77(3), 953–973. Guiso, L., Haliassos, M., and Jappelli, T. (Eds.). (2002). Household portfolios. Cambridge: MIT press. Guiso, L., Sapienza, P., and Zingales, L. (2008). Trusting the stock market. The Journal of Finance, 63(6), 2557–2600. Haliassos, M., and Bertaut, C. C. (1995). Why do so few hold stocks?. The Economic Journal, 105(432), 1110–1129. Jappelli, T. and Pistaferri, L. (2000). Using Subjective Income Expectations to Test for Excess Sensitivity of Consumption to Predicted Income Growth. European Economic Review, 44(2), 337–358. Kennickell, A. B. (1998). Multiple imputation in the Survey of Consumer Finances. In Proceedings of the Section on Survey Research Methods. Kramer, M. M. (2016). Financial literacy, confidence and financial advice seeking. Journal of Economic Behavior & Organization, 131, 198–217. Levine, R., and Zervos, S. (1998). Stock markets, banks, and economic growth. American Economic Review, 88(3), 537–558. Little, R., and Rubin, D. B. (2002). Statistical Analysis with Missing Data, 2nd Edition. New York: John Wiley. Lusardi, A., and Mitchell, O. S. (2014). The economic importance of financial literacy: Theory and evidence. Journal of Economic Literature, 52(1), 5–44. Rubin, D. B. (1987). Multiple Imputation for Nonresponse in Surveys, New York: Wiley. 19

Shin, S. H., Kim, K. T., and Seay, M. (2020). Sources of information and portfolio allocation. Journal of Economic Psychology, 76, 102212. Thompson, J.P., and Suarez, G. A. (2015). Exploring the Racial Wealth Gap Using the Survey of Consumer Finances. Finance and Economics Discussion Series 2015-076. Washington: Board of Governors of the Federal Reserve System. Van Rooij, M., Lusardi, A., and Alessie, R. (2011). Financial literacy and stock market participation. Journal of Financial Economics, 2011, 101(2), 449–472. Van Rooij, M. C., Lusardi, A., and Alessie, R. J. (2012). Financial literacy, retirement planning and household wealth. The Economic Journal, 122(560), 449–478. Von Gaudecker, H. M. (2015). How does household portfolio diversification vary with financial literacy and financial advice?. The Journal of Finance, 70(2), 489–507. Xia, T., Wang, Z., and Li, K. (2014). Financial Literacy Overconfidence and Stock Market Participation. Social Indicators Research, 119, 1233–1245. 20

Figures Figure 1: Distributions of financial literacy score and confidence measures Data source: Survey of Consumer Finances 2016 21

Figure 2: Correlation between measured financial literacy score and confidence in one’s own financial knowledge Notes: This graph shows a scatter plot of measured and self-rated financial knowledge, where dots are weighted by frequency of observation; as well as a linear fit based on a standard ordinary least squares regression. Data source: Survey of Consumer Finances 2016 22

Figure 3: Correlation of confidence in own knowledge and participation in equity markets Notes: Figure (a) shows the share of households holding equity across confidence in own financial knowledge. Figure (b) shows the residualized share of households holding stocks and other equity across confidence in own financial knowledge, where measured financial literacy and confidence in the economy are controlled for. Data source: Survey of Consumer Finances 2016 23

Figure 4: Correlation of confidence in own knowledge and ownership of bonds Notes: Figure (a) shows the share of households holding bonds and other secured debt assets across confidence in own financial knowledge. Figure (b) shows the residualized share of households holding bonds and other secured debt assets across confidence in own financial knowledge, where measured financial literacy and confidence in the economy are controlled for. Data source: Survey of Consumer Finances 2016 24

Figure 5: Correlation of confidence in own knowledge and share of total financial assets held in equities Notes: Figure (a) shows the share of households’ financial wealth held in equities across confidence in own financial knowledge. Figure (b) shows the residualized share of households’ financial wealth held in equities across confidence in own financial knowledge, where measured financial literacy and confidence in the economy are controlled for. Data source: Survey of Consumer Finances 2016 25

Figure 6: Correlation of confidence in own knowledge and share of total financial assets held in bonds Notes: Figure (a) shows the share of households’ financial wealth held in bonds and other secured debt assets across confidence in own financial knowledge. Figure (b) shows the residualized share of households’ financial wealth held in bonds and other secured financial assets across confidence in own financial knowledge, where measured financial literacy and confidence in the economy are controlled for. Data source: Survey of Consumer Finances 2016 26

Tables Table 1: Summary statistics N Mean SD Min Max Holds equities* 6248 0.52 0.50 0 1 Share of total fin. assets in equities 6248 0.23 0.30 0 1 Share of total fin. assets in equities (conditional on participation) 3770 0.43 0.29 0 1 Holds bonds* 6248 0.10 0.30 0 1 Share of total fin. assets in bonds 6248 0.01 0.06 0 1 Share of total fin. assets in bonds (conditional on participation) 769 0.09 0.18 0 1 Financial literacy score 6248 2.18 0.86 0 3 Fin. literacy (risk) correct* 6248 0.63 0.48 0 1 Fin. literacy (interest) correct* 6248 0.77 0.42 0 1 Fin. literacy (inflation) correct* 6248 0.78 0.42 0 1 Confidence in own financial knowledge 6248 7.27 2.19 0 10 Confidence in the economy in 5 years 6248 2.16 0.75 1 3 Above-average risk tolerance* 6248 0.21 0.40 0 1 Net real estate wealth (in 1000 US $) 6248 208.10 1240.86 -1440 580000 Business ownership (privately held)* 6248 0.13 0.34 0 1 Disposable household income (in 1000 US $) 6248 66.31 219.40 -1284 176000 Inheritance/gift received* 6248 0.20 0.40 0 1 Number of household members 6248 2.35 1.36 1 10 Presence of children under 18* 6248 0.32 0.47 0 1 Age 6248 51.05 17.30 18 95 Female* 6248 0.53 0.50 0 1 Employed for wage* 6248 0.64 0.48 0 1 Primary or no education* 6248 0.11 0.31 0 1 High school graduate* 6248 0.40 0.49 0 1 College graduate* 6248 0.49 0.50 0 1 White* 6248 0.68 0.47 0 1 Black* 6248 0.16 0.37 0 1 Hispanic* 6248 0.11 0.32 0 1 Other race/ethnicity* 6248 0.05 0.21 0 1 Married* 6248 0.57 0.50 0 1 Single* 6248 0.17 0.38 0 1 Divorced/separated* 6248 0.18 0.38 0 1 Widowed* 6248 0.09 0.28 0 1 Notes: Descriptive statistics estimated using survey weights and multiple-imputation techniques. * denotes dummy variables. Data Source: Survey of Consumer Finances 2016 27

Table 2a: OLS estimates of determinants of participation in equity markets (1) (2) (3) (4) (5) (6) (7) Financial literacy score 0.060*** 0.059*** 0.060*** 0.059*** (0.006) (0.006) (0.006) (0.006) Confidence in own financial knowledge 0.006*** 0.005** 0.006*** 0.005** (0.002) (0.002) (0.002) (0.002) Confidence in the economy in 5 years 0.009* 0.009* 0.009* 0.009* (0.005) (0.005) (0.005) (0.005) Above-average risk tolerance 0.099*** 0.106*** 0.106*** 0.098*** 0.105*** 0.098*** 0.097*** (0.012) (0.011) (0.012) (0.012) (0.011) (0.012) (0.012) Net real estate wealth (IHS) 0.015*** 0.016*** 0.016*** 0.015*** 0.016*** 0.015*** 0.015*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Business ownership (privately held) -0.031** -0.030** -0.028* -0.032** -0.030** -0.031** -0.032** (0.014) (0.014) (0.014) (0.014) (0.014) (0.014) (0.014) Disposable household income (IHS) 0.024*** 0.025*** 0.025*** 0.024*** 0.025*** 0.024*** 0.024*** (0.006) (0.007) (0.007) (0.006) (0.007) (0.006) (0.006) Inheritance/gift received 0.042*** 0.045*** 0.046*** 0.041*** 0.045*** 0.042*** 0.041*** (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) Number of household members -0.020*** -0.020*** -0.020*** -0.020*** -0.020*** -0.019*** -0.020*** (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) Presence of children under 18 0.048*** 0.045*** 0.044*** 0.048*** 0.044*** 0.047*** 0.047*** (0.015) (0.015) (0.015) (0.015) (0.015) (0.016) (0.015) Age 0.010*** 0.011*** 0.011*** 0.010*** 0.011*** 0.010*** 0.010*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Age squared -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Female 0.005 -0.008 -0.007 0.005 -0.008 0.005 0.005 (0.009) (0.010) (0.010) (0.009) (0.010) (0.009) (0.009) Employed for wage 0.189*** 0.192*** 0.191*** 0.190*** 0.191*** 0.189*** 0.189*** (0.011) (0.011) (0.011) (0.011) (0.011) (0.011) (0.011) High school graduate 0.130*** 0.136*** 0.140*** 0.127*** 0.136*** 0.130*** 0.126*** (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) College graduate 0.290*** 0.314*** 0.320*** 0.286*** 0.314*** 0.290*** 0.286*** (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) Black -0.128*** -0.140*** -0.140*** -0.129*** -0.141*** -0.129*** -0.129*** (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) Hispanic -0.132*** -0.142*** -0.144*** -0.131*** -0.143*** -0.133*** -0.132*** (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) Other race/ethnicity -0.058*** -0.063*** -0.066*** -0.056*** -0.064*** -0.058*** -0.057*** (0.020) (0.021) (0.021) (0.020) (0.021) (0.020) (0.020) Married 0.082*** 0.082*** 0.084*** 0.080*** 0.082*** 0.082*** 0.080*** (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) Divorced/separated -0.063*** -0.071*** -0.069*** -0.064*** -0.070*** -0.062*** -0.064*** (0.017) (0.016) (0.016) (0.017) (0.016) (0.017) (0.017) Widowed -0.013 -0.018 -0.018 -0.013 -0.018 -0.013 -0.013 (0.027) (0.027) (0.027) (0.027) (0.027) (0.027) (0.027) Constant -0.611*** -0.572*** -0.559*** -0.635*** -0.591*** -0.630*** -0.654*** (0.070) (0.073) (0.074) (0.069) (0.073) (0.071) (0.069) R squared 0.368 0.358 0.358 0.368 0.358 0.368 0.368 N 6248 6248 6248 6248 6248 6248 6248 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Dummy variables for less than high school education, non-Hispanic white ethnicity, and being single are reference categories of the respective dummy variables sets. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 28

Table 2b: OLS estimates of determinants of ownership of bonds (1) (2) (3) (4) (5) (6) (7) Financial literacy score 0.014*** 0.014*** 0.014*** 0.014*** (0.004) (0.004) (0.004) (0.004) Confidence in own financial knowledge -0.000 -0.001 -0.000 -0.000 (0.001) (0.001) (0.001) (0.001) Confidence in the economy in 5 years -0.008** -0.008** -0.008** -0.008** (0.004) (0.004) (0.004) (0.004) Above-average risk tolerance 0.006 0.008 0.008 0.006 0.008 0.006 0.006 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) Net real estate wealth (IHS) 0.003*** 0.003*** 0.003*** 0.003*** 0.003*** 0.003*** 0.003*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Business ownership (privately held) -0.007 -0.006 -0.006 -0.007 -0.006 -0.007 -0.006 (0.012) (0.012) (0.011) (0.012) (0.012) (0.012) (0.012) Disposable household income (IHS) -0.004 -0.003 -0.003 -0.004 -0.003 -0.004 -0.004 (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) Inheritance/gift received 0.027*** 0.027*** 0.027*** 0.027*** 0.027*** 0.027*** 0.027*** (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) Number of household members 0.006* 0.006* 0.006* 0.006* 0.006* 0.006* 0.006* (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) Presence of children under 18 0.032*** 0.031*** 0.032*** 0.032*** 0.032*** 0.032*** 0.032*** (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) Age 0.000 0.001 0.001 0.000 0.001 0.000 0.000 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Age squared 0.000 0.000 0.000 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Female 0.006 0.003 0.002 0.006 0.002 0.005 0.005 (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) Employed for wage 0.017** 0.017** 0.018** 0.017** 0.018** 0.017** 0.017** (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) High school graduate 0.026*** 0.029*** 0.029*** 0.026*** 0.029*** 0.026*** 0.027*** (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) College graduate 0.070*** 0.078*** 0.078*** 0.071*** 0.078*** 0.070*** 0.071*** (0.008) (0.009) (0.008) (0.009) (0.009) (0.008) (0.009) Black -0.050*** -0.053*** -0.052*** -0.050*** -0.052*** -0.050*** -0.050*** (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) Hispanic -0.075*** -0.077*** -0.076*** -0.075*** -0.076*** -0.074*** -0.074*** (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) Other race/ethnicity -0.063*** -0.065*** -0.064*** -0.063*** -0.064*** -0.062*** -0.063*** (0.013) (0.012) (0.012) (0.013) (0.012) (0.013) (0.013) Married 0.016 0.016 0.016 0.016 0.016* 0.016 0.016 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) Divorced/separated -0.023** -0.024** -0.025** -0.023** -0.025** -0.023** -0.023** (0.011) (0.011) (0.011) (0.011) (0.011) (0.011) (0.011) Widowed 0.020 0.018 0.018 0.019 0.018 0.019 0.019 (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) Constant -0.029 -0.011 0.006 -0.026 0.007 -0.011 -0.008 (0.051) (0.052) (0.051) (0.051) (0.051) (0.050) (0.051) R squared 0.065 0.064 0.064 0.065 0.064 0.065 0.065 N 6248 6248 6248 6248 6248 6248 6248 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Dummy variables for less than high school education, non-Hispanic white ethnicity, and being single are reference categories of the respective dummy variables sets. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 29

Table 3a: OLS estimates of determinants of share of total financial assets held in equities (conditional on participation) (1) (2) (3) (4) (5) (6) (7) Financial literacy score 0.007 0.007 0.007 0.007 (0.007) (0.008) (0.007) (0.008) Confidence in own financial knowledge -0.001 -0.001 -0.001 -0.001 (0.003) (0.003) (0.003) (0.003) Confidence in the economy in 5 years 0.007 0.007 0.007 0.007 (0.007) (0.007) (0.007) (0.007) Above-average risk tolerance 0.069*** 0.070*** 0.070*** 0.069*** 0.070*** 0.069*** 0.069*** (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) Net real estate wealth (IHS) 0.001 0.001 0.001 0.001 0.001 0.001 0.001 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Business ownership (privately held) -0.025* -0.024* -0.025* -0.024* -0.024* -0.025* -0.025* (0.014) (0.014) (0.014) (0.014) (0.014) (0.014) (0.014) Disposable household income (IHS) 0.008*** 0.008*** 0.008*** 0.008*** 0.008*** 0.008*** 0.008*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Inheritance/gift received 0.017* 0.017* 0.017* 0.017* 0.017* 0.017* 0.017* (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) Number of household members -0.005 -0.005 -0.005 -0.005 -0.005 -0.005 -0.005 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) Presence of children under 18 0.018 0.018 0.017 0.018 0.017 0.017 0.017 (0.020) (0.020) (0.021) (0.020) (0.021) (0.021) (0.021) Age 0.006** 0.006** 0.006** 0.006** 0.006** 0.006** 0.006** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Age squared -0.000** -0.000** -0.000** -0.000** -0.000** -0.000** -0.000** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Female -0.050*** -0.051*** -0.051*** -0.050*** -0.051*** -0.050*** -0.050*** (0.011) (0.010) (0.011) (0.011) (0.010) (0.011) (0.011) Employed for wage -0.010 -0.011 -0.011 -0.011 -0.011 -0.011 -0.011 (0.014) (0.014) (0.014) (0.014) (0.014) (0.014) (0.014) High school graduate -0.062* -0.060 -0.061 -0.062* -0.061 -0.063* -0.062* (0.037) (0.036) (0.037) (0.036) (0.036) (0.037) (0.036) College graduate -0.043 -0.038 -0.039 -0.042 -0.039 -0.043 -0.043 (0.034) (0.033) (0.033) (0.033) (0.033) (0.033) (0.033) Black -0.025 -0.027 -0.027 -0.025 -0.027 -0.025 -0.025 (0.018) (0.018) (0.018) (0.018) (0.018) (0.018) (0.018) Hispanic -0.034* -0.036* -0.036* -0.033* -0.036* -0.034* -0.034* (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) Other race/ethnicity 0.007 0.006 0.006 0.007 0.006 0.007 0.007 (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) Married -0.000 0.000 -0.000 0.000 0.000 -0.001 -0.000 (0.021) (0.020) (0.021) (0.020) (0.020) (0.021) (0.020) Divorced/separated -0.014 -0.014 -0.014 -0.014 -0.014 -0.014 -0.014 (0.025) (0.025) (0.025) (0.025) (0.025) (0.025) (0.025) Widowed -0.012 -0.012 -0.012 -0.011 -0.012 -0.011 -0.011 (0.028) (0.028) (0.028) (0.028) (0.028) (0.028) (0.028) Constant 0.219*** 0.235*** 0.214*** 0.224*** 0.218*** 0.203*** 0.208*** (0.070) (0.073) (0.072) (0.072) (0.074) (0.072) (0.074) R squared 0.050 0.050 0.050 0.051 0.050 0.051 0.051 N 3770 3770 3770 3770 3770 3770 3770 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Dummy variables for less than high school education, non-Hispanic white ethnicity, and being single are reference categories of the respective dummy variables sets. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 30

Table 3b: OLS estimates of determinants of share of total financial assets held in bonds (conditional on participation) (1) (2) (3) (4) (5) (6) (7) Financial literacy score 0.011 0.010 0.012 0.011 (0.010) (0.010) (0.010) (0.010) Confidence in own financial knowledge 0.003 0.002 0.003 0.002 (0.004) (0.004) (0.004) (0.004) Confidence in the economy in 5 years 0.022** 0.022** 0.023*** 0.022*** (0.009) (0.009) (0.009) (0.009) Above-average risk tolerance 0.003 0.003 -0.000 0.003 -0.001 -0.001 -0.002 (0.014) (0.014) (0.015) (0.014) (0.015) (0.015) (0.015) Net real estate wealth (IHS) -0.004* -0.004* -0.004* -0.004** -0.004* -0.004* -0.004* (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Business ownership (privately held) 0.036** 0.037** 0.036** 0.036** 0.035** 0.034** 0.034** (0.017) (0.016) (0.017) (0.016) (0.017) (0.017) (0.017) Disposable household income (IHS) 0.002** 0.002** 0.002** 0.002** 0.002* 0.002** 0.002** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Inheritance/gift received 0.004 0.003 0.002 0.003 0.001 0.002 0.002 (0.013) (0.013) (0.013) (0.014) (0.014) (0.013) (0.014) Number of household members -0.000 -0.002 0.000 -0.001 -0.000 0.001 0.001 (0.008) (0.008) (0.007) (0.008) (0.007) (0.008) (0.008) Presence of children under 18 -0.021 -0.020 -0.023 -0.021 -0.023 -0.024 -0.024 (0.022) (0.022) (0.021) (0.022) (0.021) (0.021) (0.021) Age -0.008*** -0.008*** -0.007*** -0.008*** -0.007*** -0.008*** -0.008*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Age squared 0.000** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Female -0.008 -0.010 -0.010 -0.008 -0.009 -0.007 -0.007 (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) Employed for wage -0.039* -0.038* -0.037* -0.038* -0.037* -0.037* -0.037* (0.020) (0.020) (0.020) (0.020) (0.020) (0.020) (0.020) High school graduate 0.046 0.046 0.046 0.046 0.047 0.047 0.048 (0.042) (0.041) (0.040) (0.041) (0.039) (0.040) (0.039) College graduate 0.005 0.009 0.008 0.006 0.009 0.006 0.007 (0.038) (0.037) (0.036) (0.036) (0.035) (0.036) (0.035) Black 0.025 0.018 0.024 0.024 0.023 0.031 0.030 (0.035) (0.035) (0.034) (0.036) (0.035) (0.035) (0.036) Hispanic -0.073*** -0.070*** -0.071*** -0.072*** -0.071*** -0.073*** -0.073*** (0.020) (0.019) (0.017) (0.020) (0.018) (0.019) (0.019) Other race/ethnicity -0.028 -0.030 -0.029 -0.028 -0.029 -0.026 -0.026 (0.025) (0.025) (0.025) (0.025) (0.025) (0.025) (0.025) Married -0.076** -0.077** -0.077** -0.078** -0.079** -0.079** -0.080** (0.031) (0.031) (0.031) (0.031) (0.031) (0.031) (0.031) Divorced/separated -0.052 -0.055* -0.054* -0.053 -0.055* -0.053 -0.054 (0.033) (0.032) (0.032) (0.033) (0.032) (0.033) (0.033) Widowed -0.031 -0.033 -0.036 -0.032 -0.036 -0.034 -0.035 (0.039) (0.040) (0.038) (0.040) (0.039) (0.038) (0.039) Constant 0.379*** 0.384*** 0.345*** 0.370*** 0.333*** 0.323*** 0.317*** (0.087) (0.087) (0.081) (0.088) (0.083) (0.083) (0.084) R squared 0.094 0.098 0.096 0.098 0.099 0.096 0.100 N 769 769 769 769 769 769 769 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Dummy variables for less than high school education, non-Hispanic white ethnicity, and being single are reference categories of the respective dummy variables sets. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 31

Table 4a: Unconditional quantile regression estimates of determinants of share of total financial assets held in equities (conditional on participation) (1) (2) (3) (4) (5) p10 p25 p50 p75 p90 Financial literacy score 0.004 0.023*** 0.037*** 0.013 -0.027** (0.005) (0.006) (0.009) (0.010) (0.011) Confidence in own financial knowledge -0.006*** -0.002 -0.003 -0.002 -0.000 (0.002) (0.003) (0.003) (0.004) (0.004) Confidence in the economy in 5 years 0.001 0.011 0.011 0.007 0.008 (0.005) (0.007) (0.010) (0.011) (0.009) Above-average risk tolerance 0.033*** 0.058*** 0.085*** 0.124*** 0.046*** (0.007) (0.010) (0.016) (0.016) (0.016) Net real estate wealth (IHS) 0.001 0.003*** 0.002 0.002 0.001 (0.001) (0.001) (0.002) (0.002) (0.001) Business ownership (privately held) -0.021** -0.021* -0.040** -0.021 -0.065*** (0.009) (0.013) (0.017) (0.020) (0.017) Disposable household income (IHS) 0.008** 0.011*** 0.013*** 0.006* 0.003 (0.003) (0.003) (0.004) (0.004) (0.003) Inheritance/gift received 0.019*** 0.021* 0.002 0.009 0.012 (0.007) (0.012) (0.017) (0.018) (0.015) Number of household members -0.002 0.002 -0.019* 0.007 -0.007 (0.007) (0.008) (0.011) (0.009) (0.008) Presence of children under 18 0.018 -0.007 0.042 0.004 0.005 (0.017) (0.020) (0.028) (0.023) (0.022) Age 0.007*** 0.006** 0.006* 0.001 0.006** (0.002) (0.003) (0.003) (0.003) (0.003) Age squared -0.000*** -0.000** -0.000* -0.000 -0.000* (0.000) (0.000) (0.000) (0.000) (0.000) Female -0.003 -0.022* -0.070*** -0.070*** -0.072*** (0.008) (0.012) (0.014) (0.015) (0.014) Employed for wage -0.020* -0.019 -0.013 -0.017 -0.046** (0.011) (0.016) (0.017) (0.021) (0.021) High school graduate -0.016 0.004 -0.016 -0.086** -0.163*** (0.022) (0.037) (0.036) (0.037) (0.049) College graduate 0.002 0.029 0.001 -0.083** -0.177*** (0.022) (0.036) (0.035) (0.037) (0.051) Black -0.028** -0.044** -0.062*** -0.001 0.001 (0.013) (0.022) (0.022) (0.026) (0.027) Hispanic -0.011 -0.029 -0.056** -0.064** -0.026 (0.015) (0.019) (0.028) (0.029) (0.025) Other race/ethnicity 0.009 -0.003 -0.037 -0.028 0.068 (0.014) (0.022) (0.037) (0.040) (0.046) Married 0.017 -0.003 0.033 -0.045 -0.006 (0.015) (0.021) (0.029) (0.033) (0.025) Divorced/separated -0.001 -0.034* 0.006 -0.012 0.024 (0.017) (0.021) (0.028) (0.034) (0.027) Widowed 0.024 -0.018 0.021 -0.039 -0.037 (0.025) (0.032) (0.045) (0.044) (0.033) Constant -0.186*** -0.190** 0.053 0.595*** 0.966*** (0.063) (0.091) (0.103) (0.108) (0.102) R squared 0.021 0.039 0.043 0.037 0.019 N 3770 3770 3770 3770 3770 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Dummy variables for less than high school education, non-Hispanic white ethnicity, and being single are reference categories of the respective dummy variables sets. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 32

Table 4b: Unconditional quantile regression estimates of determinants of share of total financial assets held in bonds (conditional on participation) (1) (2) (3) (4) (5) p10 p25 p50 p75 p90 Financial literacy score -0.001 0.002* 0.002 0.007 0.022 (0.001) (0.001) (0.002) (0.010) (0.042) Confidence in own financial knowledge -0.000 -0.000 0.000 0.003 -0.002 (0.000) (0.001) (0.001) (0.004) (0.024) Confidence in the economy in 5 years 0.003** 0.002 0.005*** 0.031*** 0.057 (0.001) (0.001) (0.002) (0.011) (0.044) Above-average risk tolerance -0.001 -0.004* -0.005* -0.023 0.085 (0.002) (0.002) (0.003) (0.016) (0.077) Net real estate wealth (IHS) -0.000 0.000 -0.001*** -0.003* -0.008 (0.000) (0.000) (0.000) (0.002) (0.007) Business ownership (privately held) 0.003 0.001 0.008** 0.033 0.099 (0.002) (0.003) (0.003) (0.020) (0.070) Disposable household income (IHS) 0.001* 0.001 0.000 0.001 0.010* (0.001) (0.001) (0.001) (0.002) (0.006) Inheritance/gift received -0.001 0.000 0.006** 0.017 0.005 (0.002) (0.002) (0.003) (0.016) (0.069) Number of household members -0.001 -0.001 -0.000 -0.006 0.000 (0.002) (0.001) (0.002) (0.010) (0.034) Presence of children under 18 0.000 -0.000 -0.004 -0.027 -0.121 (0.003) (0.003) (0.004) (0.027) (0.102) Age -0.001*** -0.002*** -0.002*** -0.009*** -0.035** (0.000) (0.000) (0.000) (0.003) (0.015) Age squared 0.000*** 0.000*** 0.000*** 0.000** 0.000** (0.000) (0.000) (0.000) (0.000) (0.000) Female 0.001 0.000 -0.003 -0.033* 0.055 (0.001) (0.002) (0.002) (0.017) (0.074) Employed for wage -0.003 -0.005** -0.010*** -0.021 -0.080 (0.002) (0.002) (0.003) (0.021) (0.099) High school graduate 0.005 0.001 0.003 0.103** -0.023 (0.008) (0.008) (0.011) (0.050) (0.263) College graduate 0.003 -0.007 -0.003 0.048 -0.160 (0.008) (0.008) (0.010) (0.048) (0.249) Black 0.003 0.012*** 0.005 0.008 0.086 (0.003) (0.003) (0.005) (0.035) (0.134) Hispanic 0.002 0.000 -0.005 -0.091*** -0.323*** (0.003) (0.006) (0.006) (0.034) (0.079) Other race/ethnicity 0.001 0.007 0.008* -0.023 -0.111 (0.005) (0.005) (0.005) (0.035) (0.135) Married 0.002 0.003 -0.000 -0.056* -0.214 (0.004) (0.004) (0.005) (0.033) (0.153) Divorced/separated 0.007** 0.007* 0.010** -0.023 -0.044 (0.003) (0.004) (0.005) (0.033) (0.181) Widowed 0.006 -0.001 0.007 0.036 -0.020 (0.004) (0.006) (0.006) (0.042) (0.202) Constant 0.007 0.057*** 0.075*** 0.263*** 1.360*** (0.011) (0.013) (0.016) (0.089) (0.471) R squared 0.049 0.077 0.105 0.101 0.062 N 769 769 769 769 769 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Dummy variables for less than high school education, non-Hispanic white ethnicity, and being single are reference categories of the respective dummy variables sets. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 33

Appendices Appendix A: Description of variables used in empirical analysis Variable Description Dependent variables Dummy variable taking a value of 1 if household holds any equity of the following type: publically traded stocks, stock mutual funds, combination funds, IRAs/Keoghs invested in stocks, other managed D_equity assets w/equity interest (e.g. annuities, trusts, MIAs), thrift-type retirement accounts invested in stocks, savings accounts classified as 529 (educational savings accounts) W_equity Share of the equity value in total financial assets (including pensions) Dummy variable taking a value of 1 if household holds any instruments of the following type: government saving bonds, corporate bonds, commercial paper, state or municipal non-saving bonds, D_bonds foreign bonds and other non-saving bonds, debentures, mortgage-backed securities, negotiable certificates of deposit, treasury bills (T-bills), treasury certificates (T-certificates), treasury bonds (Tbonds), zero-coupon bonds, and similar instruments normally traded in financial markets W_bonds Share of bonds and other debt securities in total financial assets (including pensions) Explanatory variables FL_score Number of correct answers to the three financial literacy questions; ranging from 0 to 3 FL1 Dummy variable taking a value of 1 if the knowledge question on “risk” is answered correctly FL2 Dummy variable taking a value of 1 if the knowledge question on “interest rates” is answered correctly FL3 Dummy variable taking a value of 1 if the knowledge question on “inflation” is answered correctly Confidence1 Self-assessed level of knowledge about financial matters; ranging from 0 “very low” to 10 “very high” Respondent’s confidence in overall performance of the US economy in 5 years’ horizon as compared to Confidence2 today; ranging from 1 “worse” to 3 “better” Risk Dummy variable taking a value of 1 if the respondent reports positive risk attitude Wealth Net real estate wealth defined as a total value of properties minus the corresponding liabilities Dummy variable taking a value of 1 if household owns or shares ownership in any privately-held Business ownership businesses Total monetary and non-monetary household current income net of income taxes and social security Income contributions Inheritance Dummy variable taking a value of 1 if household had received an inheritance HH_size Number of household members Num_children Number of children in household aged below 18 Age Age of the respondent in years Female Dummy variable taking a value of 1 if the respondent is female Working Dummy variable taking a value of 1 if the respondent is employed for wage or self-employed Edu_low Dummy variable taking a value of 1 if the respondent has less than high school education Edu_middle Dummy variable taking a value of 1 if the respondent has secondary education Dummy variable taking a value of 1 if the respondent has university education (bachelor, graduate, Edu_high postgraduate) Dummy variable taking a value of 1 if the respondent declares Caucasian ethnicity (including Middle- White eastern / Arab whites) Black Dummy variable taking a value of 1 if the respondent declares African-American ethnicity Hispanic Dummy variable taking a value of 1 if the respondent declares Hispanic / Latin-American ethnicity Other_race Dummy variable taking a value of 1 if the respondent declares other ethnicity (e.g. Chinese, Indian) Married Dummy variable taking a value 1 if the respondent is married or having partner Single Dummy variable taking a value 1 if the respondent is living in a single-member household Divorced Dummy variable taking a value 1 if the respondent is divorced or separated Widowed Dummy variable taking a value 1 if the respondent is widowed Source: own processing 34

Appendix B: Framing of the financial literacy questions 1) Please tell me whether this statement is true or false. “Buying a single company’s stock usually provides a safer return than a stock mutual fund.” True False (correct answer) Do not know Refuse to answer 2) 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 (correct answer) Exactly $102 Less than $102 Do not know Refuse to answer 3) Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After 1 year, how much would you be able to buy with the money in this account? More than today Exactly the same Less than today (correct answer) Do not know Refuse to answer 35

Appendix C: Unconditional quantile regression To analyze the relationship between covariates and the share of risky assets in total financial assets beyond the mean (baseline model outlined in equation 2), we employ unconditional quantile regression (UQR) method. To do so, we use a concept of the recentered influence function similar to the standard regression technique, besides that the dependent variable is replaced by the recentered influence function of the statistics of interest (Fortin et al., 2009). For a reader’s convenience we summarize the Fortin’s et. al (2009) UQR framework. Let be the observed outcome variable of interest in the presence of a set of available covariates . and have a joint distribution . The unconditional distribution of can 𝑌𝑌 be defined 𝑋𝑋 as 𝑌𝑌 follow 𝑋𝑋 s: 𝐹𝐹𝑌𝑌,𝑋𝑋(.;.) 𝑋𝑋 𝐹𝐹 𝐹𝐹𝑌𝑌(𝑒𝑒) = � (𝑒𝑒|𝑋𝑋 = 𝑥𝑥)×𝑏𝑏𝐹𝐹𝑋𝑋(𝑥𝑥). � � � The UQR framework is developed by def𝑌𝑌in𝑋𝑋ing a recentered influence function (RIF), which is an extension to the concept of influence function (IF). Influence functions are used in statistics to extract the effect/influence of adding or removing a particular observation on the value of statistics . The standard IF can be written as: 𝑣𝑣(𝐹𝐹𝑌𝑌) �𝑣𝑣�(1−𝜀𝜀)∙𝐹𝐹 +𝜀𝜀 ∙𝛿𝛿𝑦𝑦�−𝑣𝑣(𝐹𝐹)� 𝑅𝑅𝐹𝐹�𝑒𝑒;𝑣𝑣(𝐹𝐹)� = l𝜀𝜀i→m0 ,0 ≤ 𝜀𝜀 ≤ 1, where is the cumulative distribution function for 𝜀𝜀 , and represents the probability measure putting 𝐹𝐹 mass 1 at the value . We obtain the RIF by a 𝑌𝑌 dding t 𝛿𝛿 h𝑦𝑦e IF to the statistics of interest : 𝑒𝑒 . 𝑣𝑣(𝐹𝐹) After substituting in the partic𝑅𝑅ul𝑅𝑅a𝐹𝐹r (s𝑒𝑒ta;t𝑣𝑣is,t𝐹𝐹ic 𝑌𝑌 )of= in𝑣𝑣t(e𝐹𝐹re)st+, r𝑅𝑅e𝐹𝐹gr(e𝑒𝑒s;s𝑣𝑣io,n𝐹𝐹 𝑌𝑌 q)uantile , it yields: . 𝑒𝑒𝜏𝜏 In the case of quantiles, we𝑅𝑅 c𝑅𝑅𝐹𝐹an( 𝑒𝑒w;r𝑒𝑒it 𝜏𝜏 e) t=he𝑒𝑒 i 𝜏𝜏 nf+lu𝑅𝑅e𝐹𝐹nc(e𝑒𝑒 ;fu𝑒𝑒n 𝜏𝜏)ction as follows: , 𝜏𝜏−𝐼𝐼[𝑌𝑌≤𝑞𝑞𝜏𝜏] 𝑅𝑅𝐹𝐹(𝑒𝑒;𝑒𝑒𝜏𝜏) = 𝑓𝑓𝑌𝑌(𝑞𝑞𝜏𝜏) where represents the probability density function of evaluated at quantile . is an in 𝐶𝐶𝑌𝑌di ( c 𝑒𝑒 a𝜏𝜏t ) or variable taking a value of 1 if the outcome 𝑌𝑌 variable is smaller t 𝑒𝑒 h𝜏𝜏an 𝑅𝑅 [𝑌𝑌 a ≤ nd 𝑒𝑒 𝜏𝜏0 ] otherwise. 𝑌𝑌 𝑒𝑒𝜏𝜏 36

Finally, the UQR estimator is defined as coefficient vector obtained from the following linear regression of the RIF on a set of explanatory variables: . 𝑈𝑈𝑈𝑈𝑈𝑈 UQR produces coefficients corresp𝑅𝑅o𝑅𝑅n𝐹𝐹di(n𝑌𝑌g, 𝜏𝜏t)o =th𝑋𝑋e 𝛽𝛽impac+t 𝜀𝜀on the -th quantile of the outcome variable , irrespective of the included set of explanatory variables, which makes the method an 𝜏𝜏 attractive tool, especially in the policy context analysis. 𝑌𝑌 37

Appendix D: Analyzing multiple imputed data A specific feature of wealth surveys (including the SCF) is that missing values in some of the variables due to responses of “Don’t know” and “Refused” have been imputed and replaced times (five in the case of SCF). This feature should be considered in an empirical analysis when 𝑚𝑚 obtaining statistical inferences. Estimations ignoring item-nonresponse can lead to substantial biases as well as efficiency loss due to the large number of covariates included in regressions (see Christelis et al., 2010). Therefore, we follow the rules suggested by Rubin (1987) and Little and Rubin (2002), and we use multiple-imputation techniques to obtain correct estimated coefficients along with the variance. Let be our point estimate of an interest – that is mean and regression parameter. For each of the imputed datasets , we obtain an empirical estimate of , denoted by . The final average 𝛽𝛽 point estimate of , give 𝑚𝑚 n by , is computed as follows: 𝛽𝛽 𝛽𝛽̂𝑚𝑚 𝛽𝛽 𝛽𝛽̅ . 1 5 𝛽𝛽̅ = 5∑𝑚𝑚=1𝛽𝛽̂𝑚𝑚 The final estimated is associated with a variance having two components: , 1 5 which is a within imp𝛽𝛽u̅ tation sampling variance, and the between imputations va𝑊𝑊ria=nc5e∑ g𝑚𝑚iv=e1n𝑉𝑉 �b𝑚𝑚y . The total variance-covariance matrix, , associated with is given by: 1 5 2 𝐵𝐵 = 4∑𝑚𝑚=1�𝛽𝛽̂𝑚𝑚 −𝛽𝛽̅� 𝑉𝑉� 𝛽𝛽̅ . Finally, to estimate accounting for th𝑉𝑉� e =co𝑊𝑊mp+lex(6it⁄y 5o)f𝐵𝐵 the SCF sampling design, we implement a bootstrap procedure taking into account 999 replicate weights provided in the SCF. 𝑊𝑊 38

Appendix E: Additional tables Table E.1a: OLS estimates of determinants of participation in equity markets and share of total financial assets held in equities (alternative confidence measures) Participation Share (1) (2) (3) (4) Financial literacy score 0.031*** -0.006 (0.009) (0.015) Dummy: over-confident 0.008 0.007 -0.007 -0.006 (0.012) (0.012) (0.015) (0.015) Dummy: sound-confident 0.114*** 0.071*** 0.018 0.026 (0.013) (0.020) (0.014) (0.025) Dummy: under-confident 0.106*** 0.062*** 0.008 0.017 (0.012) (0.017) (0.017) (0.022) Confidence in the economy in 5 years 0.009* 0.007 (0.005) (0.007) Above-average risk tolerance 0.097*** 0.096*** 0.068*** 0.067*** (0.012) (0.012) (0.010) (0.010) Net real estate wealth (IHS) 0.015*** 0.015*** 0.001 0.001 (0.001) (0.001) (0.001) (0.001) Business ownership (privately held) -0.033** -0.033** -0.025* -0.026* (0.014) (0.014) (0.014) (0.014) Disposable household income (IHS) 0.024*** 0.024*** 0.008*** 0.008*** (0.006) (0.006) (0.003) (0.003) Inheritance/gift received 0.041*** 0.041*** 0.017 0.016 (0.010) (0.010) (0.010) (0.010) Number of household members -0.019*** -0.019*** -0.004 -0.004 (0.005) (0.005) (0.008) (0.008) Presence of children under 18 0.043*** 0.045*** 0.018 0.016 (0.016) (0.016) (0.020) (0.021) Age 0.010*** 0.010*** 0.006** 0.006** (0.002) (0.002) (0.003) (0.003) Age squared -0.000*** -0.000*** -0.000** -0.000** (0.000) (0.000) (0.000) (0.000) Female 0.005 0.007 -0.049*** -0.049*** (0.009) (0.009) (0.011) (0.011) Employed for wage 0.190*** 0.189*** -0.010 -0.010 (0.011) (0.011) (0.014) (0.014) High school graduate 0.134*** 0.131*** -0.062* -0.062* (0.015) (0.015) (0.036) (0.036) College graduate 0.292*** 0.288*** -0.044 -0.043 (0.016) (0.016) (0.033) (0.033) Black -0.129*** -0.128*** -0.025 -0.026 (0.015) (0.015) (0.018) (0.018) Hispanic -0.133*** -0.132*** -0.033* -0.034* (0.015) (0.015) (0.019) (0.019) Other race/ethnicity -0.058*** -0.058*** 0.007 0.007 (0.020) (0.020) (0.029) (0.029) Married 0.081*** 0.081*** -0.002 -0.002 (0.016) (0.016) (0.020) (0.020) Divorced/separated -0.062*** -0.062*** -0.014 -0.014 (0.017) (0.017) (0.025) (0.025) Widowed -0.010 -0.011 -0.011 -0.011 (0.027) (0.027) (0.028) (0.028) Constant -0.532*** -0.590*** 0.232*** 0.225*** (0.071) (0.071) (0.070) (0.076) R squared 0.368 0.369 0.051 0.052 N 6248 6248 3770 3770 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Dummy variables for not confident nor literate, less than high school education, non-Hispanic white ethnicity, and being single are reference categories of the respective dummy variables sets. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 39

Table E.1b: OLS estimates of determinants of ownership of bonds and share of total financial assets held in bonds (alternative confidence measures) Participation Share (1) (2) (3) (4) Financial literacy score 0.012** 0.036 (0.006) (0.029) Dummy: over-confident -0.003 -0.004 0.008 0.000 (0.007) (0.007) (0.029) (0.031) Dummy: sound-confident 0.019** 0.003 0.052 0.004 (0.009) (0.013) (0.031) (0.056) Dummy: under-confident 0.020** 0.004 0.087*** 0.038 (0.008) (0.011) (0.031) (0.053) Confidence in the economy in 5 years -0.008** 0.020* (0.004) (0.012) Above-average risk tolerance 0.006 0.006 0.087*** 0.083*** (0.008) (0.008) (0.024) (0.024) Net real estate wealth (IHS) 0.003*** 0.003*** 0.003 0.002 (0.001) (0.001) (0.002) (0.002) Business ownership (privately held) -0.007 -0.006 0.002 0.001 (0.012) (0.012) (0.033) (0.033) Disposable household income (IHS) -0.004 -0.004 0.006 0.006 (0.004) (0.004) (0.006) (0.006) Inheritance/gift received 0.027*** 0.026*** 0.027 0.029 (0.008) (0.008) (0.023) (0.023) Number of household members 0.007* 0.006* 0.003 0.004 (0.004) (0.004) (0.014) (0.015) Presence of children under 18 0.031*** 0.032*** 0.023 0.023 (0.010) (0.010) (0.038) (0.038) Age 0.000 0.000 0.007** 0.007** (0.001) (0.001) (0.003) (0.003) Age squared 0.000 0.000 -0.000 -0.000 (0.000) (0.000) (0.000) (0.000) Female 0.005 0.005 -0.027 -0.024 (0.007) (0.007) (0.021) (0.022) Employed for wage 0.017** 0.017** 0.039 0.043* (0.007) (0.007) (0.025) (0.025) High school graduate 0.028*** 0.027*** 0.053 0.052 (0.007) (0.007) (0.068) (0.070) College graduate 0.072*** 0.071*** 0.134** 0.130* (0.008) (0.009) (0.066) (0.069) Black -0.050*** -0.049*** -0.075* -0.061 (0.007) (0.007) (0.044) (0.045) Hispanic -0.075*** -0.074*** -0.064 -0.064 (0.006) (0.006) (0.101) (0.102) Other race/ethnicity -0.064*** -0.063*** -0.021 -0.018 (0.013) (0.013) (0.066) (0.065) Married 0.016 0.016 0.030 0.028 (0.010) (0.010) (0.043) (0.043) Divorced/separated -0.023** -0.023** -0.080* -0.078* (0.011) (0.011) (0.044) (0.044) Widowed 0.020 0.020 -0.064 -0.067 (0.016) (0.016) (0.055) (0.054) Constant -0.010 -0.007 -0.213 -0.316** (0.051) (0.050) (0.132) (0.145) R squared 0.065 0.065 0.098 0.102 N 6248 6248 769 769 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Dummy variables for not confident nor literate, less than high school education, non-Hispanic white ethnicity, and being single are reference categories of the respective dummy variables sets. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 40

Table E.2a: OLS estimates of determinants of participation in equity markets and share of total financial assets held in equities (gender-related effects of financial literacy and confidence) Participation Share (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Financial literacy score 0.059*** 0.065*** 0.059*** 0.059*** 0.063*** 0.007 0.009 0.007 0.008 0.009 (0.006) (0.009) (0.006) (0.006) (0.010) (0.008) (0.010) (0.008) (0.008) (0.010) Confidence in own financial knowledge 0.005** 0.005** 0.009*** 0.005** 0.009*** -0.001 -0.001 -0.000 -0.001 -0.000 (0.002) (0.002) (0.003) (0.002) (0.003) (0.003) (0.003) (0.004) (0.003) (0.004) Confidence in the economy in 5 years 0.009* 0.009* 0.009* 0.010 0.010 0.007 0.007 0.007 0.003 0.003 (0.005) (0.005) (0.005) (0.009) (0.009) (0.007) (0.007) (0.007) (0.010) (0.010) Financial literacy x female -0.010 -0.007 -0.002 -0.002 (0.011) (0.011) (0.012) (0.012) Confidence1 x female -0.008* -0.007* -0.001 -0.001 (0.004) (0.004) (0.007) (0.007) Confidence2 x female -0.002 -0.002 0.008 0.008 (0.013) (0.013) (0.014) (0.014) Constant -0.654*** -0.666*** -0.685*** -0.657*** -0.695*** 0.208*** 0.205*** 0.204*** 0.217*** 0.210*** (0.069) (0.070) (0.074) (0.074) (0.078) (0.074) (0.078) (0.079) (0.075) (0.082) Other socio-economic characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R squared 0.368 0.368 0.368 0.368 0.368 0.051 0.051 0.051 0.051 0.051 N 6248 6248 6248 6248 6248 3770 3770 3770 3770 3770 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Other socio-economic variables included as described in the text. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 41

Table E.2b: OLS estimates of determinants of ownership of bonds and share of total financial assets held in bonds (genderrelated effects of financial literacy and confidence) Participation Share (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Financial literacy score 0.014*** 0.014** 0.014*** 0.014*** 0.013** 0.011 -0.005 0.011 0.011 -0.003 (0.004) (0.005) (0.004) (0.004) (0.005) (0.010) (0.014) (0.010) (0.010) (0.014) Confidence in own financial knowledge -0.000 -0.000 0.001 -0.000 0.001 0.002 0.002 -0.004 0.002 -0.003 (0.001) (0.001) (0.002) (0.001) (0.002) (0.004) (0.004) (0.005) (0.004) (0.006) Confidence in the economy in 5 years -0.008** -0.008** -0.008** -0.011* -0.011* 0.022*** 0.022*** 0.022*** 0.035*** 0.036*** (0.004) (0.004) (0.004) (0.006) (0.006) (0.009) (0.009) (0.009) (0.011) (0.011) Financial literacy x female 0.001 0.002 0.025 0.020 (0.007) (0.007) (0.017) (0.017) Confidence1 x female -0.002 -0.002 0.011 0.010 (0.003) (0.003) (0.009) (0.009) Confidence2 x female 0.005 0.005 -0.024 -0.025 (0.009) (0.010) (0.016) (0.016) Constant -0.008 -0.007 -0.017 -0.002 -0.009 0.317*** 0.351*** 0.357*** 0.282*** 0.343*** (0.051) (0.053) (0.054) (0.052) (0.056) (0.084) (0.088) (0.088) (0.078) (0.084) Other socio-economic characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R squared 0.065 0.065 0.065 0.065 0.065 0.100 0.102 0.100 0.100 0.100 N 6248 6248 6248 6248 6248 769 769 769 769 769 Notes: Estimations carried out using multiple-imputation techniques. Bootstrapped standard errors presented in parentheses are based on 999 replicate weights. Other socio-economic variables included as described in the text. * p < 0.10, ** p < 0.05, *** p < 0.01. Data Source: Survey of Consumer Finances 2016 42