The effect of Financial Literacy Overconfidence
on Stock Market Participation
Kaylie Maathuis1 Thesis MSc Finance
Supervisor: prof. dr. B.W. (Robert) Lensink June, 2017
ABSTRACT
This study examines the effect of financial literacy overconfidence on stock market participation. Financial literacy overconfidence is measured using combined measures of the actual and perceived financial literacy of households. The data used in this study is from the 2005 DNB Household Survey. The results show that actual and perceived financial literacy both have a positive relation on stock market participation. Furthermore, the results imply a negative relationship between financial literacy overconfidence and stock market participation, however, these results should be interpreted with caution due to possible endogeneity and multicollinearity problems. Additionally, this study finds a positive effect of underconfidence on stock market participation.
Key words: Household Finance, Overconfidence, Financial Literacy, Stock Market Participation
JEL Classification numbers: D14, D83, G11
1 University of Groningen
Faculty of Economics and Business Student number: s2233037
2 Table of contents
1. Introduction ... 3
2. Literature review ... 6
2.1. Financial literacy and stock market participation ... 6
2.2. Financial literacy overconfidence and stock market participation ... 7
3. Data and methodology ... 9
3.1. Data ... 9
3.2. Key variables ... 11
3.2.1. Stock market participation ... 11
3.2.2. Financial literacy ... 11
3.2.3. Perceived financial literacy ... 14
3.2.4. Financial literacy overconfidence ... 15
3.2.5. Control variables ... 16
3.3. Methodology ... 17
4. Results ... 19
4.1. Financial literacy ... 19
4.2 Financial literacy overconfidence ... 22
4.3. Interaction effects ... 25
4.4. Discussion ... 27
5. Conclusion and limitations ... 29
5.1. Conclusion ... 29
5.2. Limitations and future research ... 30
Appendices ... 31
Appendix A Financial literacy questions designed by van Rooij et al. (2011) ... 31
Appendix B Factor Analysis ... 34
Appendix C Factor analysis risk tolerance ... 36
Appendix D The logit model ... 38
3 1. Introduction
In the Netherlands, only about one fifth of the households participated in the stock market in 2015.2 This seems like a small amount but the stock market participation rate has increased significantly over time. Prior literature states that the rate of stock market participation can have a direct effect on the equity premium as introduced by Mehra and Prescott (1985). This stems from the little correlation of the aggregate consumption growth with stock returns which cannot justify the large risk premium observed on stocks. For this reason, the determinants of stock market participation can help to improve the understanding of the equity-premium puzzle (Mankiw and Zeldes, 1991; Heaton and Lucas, 2000; and Vissing-Jorgensen, 2000; and Hong et al., 2004).
Additionally, investigating the determinants for stock market participation is even more important since the lack of stock market participation can have economic implications as well. Cocco et al. (2005) find that non-participation can cause considerable welfare loss. They state that not participating in the stock market can result in a welfare loss of more than 2% of the annual consumption. Moreover, Hong et al. (2004) argue that the rate of stock market participation can also have policy implications considering proposals whereby the government invests a portion of the social security tax proceeds in the stock market. However, in the perspective of a frictionless model with full information and optimizing households, Hong et al. (2004) argue that there is no gain in having the government invest in the stock market on behalf of the households. There will only be a gain if households do not participate in the stock market due to lack of information or frictional costs.
According to Guiso et al. (2003), households have more instruments to smooth their consumption and manage their risk by holding stocks. It could reduce the inequality due to higher returns, but also has more risk. When an individual is trading excessively or is ill-advised, it can considerably reduce its returns.
Some determinants of stock market participation are already well known and studied. Prior studies find that stock market participation increases with the level of wealth and income (Bertaut, 1998; Guiso et al., 2003; and Vissing-Jorgensen, 2000) since wealthier individuals and individuals with higher income are better able to pay for the participation costs and have more money to invest in stocks. Another important driver of stock market participation is the level of education. Haliassos and Bertaut (1995) and Campbell (2006) find that stock market participation increases with the level of education. Other drivers of stock market participation
4 are trust (Guiso, Sapienza and Zingales, 2008), awareness (Guiso and Japelli, 2005), social interaction (Hong et al., 2004), and IQ (Grinblatt et al., 2011).
Guiso et al. (2003) point out that differences in stockholding exist due to the lower participation costs, which causes new entrants to participate in the stock market. These new entrants can have a poor judgement or a limited ability to withstand participation, which creates a higher volatility in the stock market. An increasing part of the population is getting more actively involved in the financial markets, while the financial markets are becoming increasingly complex and difficult to understand due to several new financial products and services. These new products and services can be difficult to understand, especially for individuals who are less financially literate. Van Rooij et al (2011, pp 450) state that: “market liberalization and structural reforms to social security and pensions have caused an ongoing shift in decision-making responsibility away from the government and employers and toward private individuals. Thus individuals have to assume more responsibility for their own financial well-being”. This leaves the question if people are financially literate and capable enough to have this responsibility, or do they only think they are?
Several studies indicate that financial literacy is widespread among the general population (van Rooij et al., 2011; Lusardi and Mitchell, 2007, 2008), even so in well-developed financial markets (Lusardi and Mitchel, 2011). Especially women, people with low education, African Americans and Hispanics show low levels of financial literacy (Lusardi, 2008). Among the young, financial literacy shows a strong relation to sociodemographic characteristics and family financial literacy (Lusardi et al., 2010).
Financial literacy is an important factor in financial decision-making since failure to plan for retirement, lack of stock market participation, poor borrowing behaviour and problems with debt can all be linked to financial illiteracy (Lusardi 2008, Lusardi et al. 2010). Lusardi (2008) reveals that most individuals lack knowledge of basic financial concepts and cannot perform simple economic calculations.
5 Van Rooij et al (2011) examined the effect of financial literacy on stock market participation. This study extends their research with perceived financial literacy and financial literacy overconfidence. To my knowledge, this is the first study that examines the effect of perceived financial literacy and financial literacy overconfidence on stock market participation in the Netherlands. There is one other study that also examined the effect of financial literacy overconfidence on stock market participation (Xia et al.; 2014), however this study used the 2012 Chinese Survey of Consumer Finance and hence includes only the Chinese population. Whereas Xia et al. (2014) only use six financial literacy questions and do not distinguish between basic and advanced financial literacy, this paper uses the financial literacy module designed by van Rooij et al. (2011) of the 2005 Nederlandsche Bank’s Household Survey (DHS) which consist of five basic financial literacy questions and 11 advanced financial literacy questions.
Furthermore, there is extensive literature on the concept of overconfidence, however, this is often not defined as overconfidence in financial literacy. There is still little research on financial literacy overconfidence and most of the existing studies examine the effect of financial literacy overconfidence on advice seeking (Kramer, 2016; Porto and Xiao, 2016). Since Xia et al. (2014) find that overconfident individuals are more likely to participate in the stock market and several other studies (Odean, 1998; Odean, 1999; Statman et al., 2006; Glaser and Weber, 2007) find that overconfident investors trade more, it is expected that financial literacy overconfidence has a positive impact on stock market participation. Yet, the results show the opposite effect and indicate a negative correlation between overconfidence and stock market participation. However, the results should be interpreted with caution due to endogeneity and multicollinearity problems. The contradicting findings can also be partly attributable to the way financial literacy is measured since most studies measure financial literacy using questionnaires, however, these questionnaires vary among the different studies (van Rooij et al., 2011; Xia et al., 2014; Gentile et al., 2016; Porto and Xiao, 2016). This makes it difficult to compare.
6 2. Literature review
2.1. Financial literacy and stock market participation
There is still little research about the effect of financial literacy on stock market participation. However, Yoong (2010) examines the effect of financial illiteracy on stock market participation using multiple waves of the American Life Panel (ALP). The ALP is an ongoing internet panel with approximately 2,500 respondents which is representative of the general U.S. population. She focuses on adults over the age of 50 and uses a novel instrumental variables strategy to examine the relationship between financial illiteracy and stock market participation. Yoong (2010) finds a negative relationship of financial illiteracy on stock market participation whereby an increase of one standard deviation above the mean level of financial illiteracy decreases stock market participation with 10%.
Another study that examines the effect of financial literacy on stock market participation is the study of van Rooij et al (2011). The authors use the Nederlandsche Bank’s Household Survey (DHS) of CentERdata, which collects information of financial behaviour from over 2,000 households. Van Rooij et al. (2011) designed a financial literacy module with five basic financial literacy questions and 11 advanced financial literacy questions and is completed by 1,508 households. The authors created a financial literacy index using the advanced financial literacy questions to examine the possible relationship with stock market participation. Their results show a significant positive effect of the advanced financial literacy index and stock market participation indicating that a person who is more financially literate has a higher propensity to hold stocks.
The first hypothesis is mainly based on the studies of Yoong (2010) and van Rooij et al. (2011) since there is limited prior literature on the effect of financial literacy on stock market participation. However, Yoong (2010) and van Rooij et al. (2011) have similar results using the actual financial literacy of respondents. These studies both find that individuals with higher actual financial literacy are more likely to hold stocks. Based on these results, the following hypothesis of the effect of financial literacy on stock market participation is formed:
H1: Actual financial literacy has a positive effect on stock market participation
7 Allgood and Walstad (2012) a person makes decisions relying on what they think they know and their actual objective knowledge. Since van Rooij et al. (2011) find a strong positive correlation between the actual and perceived financial literacy, it can be expected that perceived financial literacy also has a positive effect on stock market participation. Furthermore, Xia et al. (2014) examine the effect of financial literacy overconfidence on stock market participation whereby they take both financial literacy measures into account. They find that financial literacy overconfidence has a positive significant effect on stock market participation, which could imply that perceived financial literacy also has a positive effect on stock market participation. Therefore, another hypothesis is formed:
H2: Perceived financial literacy has a positive effect on stock market participation
2.2. Financial literacy overconfidence and stock market participation
Overconfidence is an important psychological concept that has been studied extensively. DeBondt and Thaler (1995, pp 6) even state that “perhaps the most robust finding in the psychology of judgement is that people are overconfident”. Several studies indicate that men are more overconfident than women (Barber and Odean, 2001; Pirinisky, 2013; van Rooij et al., 2011; Almenberg and Dreber, 2015), overconfidence increases with income and education (Pirinsky, 2013) and overconfidence increases with age (Pirinsky, 2013; Pak and Chatterjee, 2016).
According to Moore and Healy (2008), there are three definitions of overconfidence. The first one is when individuals overestimate their actual ability, performance, level of control, or chance of success, which is referred to as overestimation. The second definition is when people believe themselves to be better than others. This definition is called overplacement. The last definition of Moore and Healy (2008) is the excessive certainty regarding the accuracy of one’s beliefs and is referred to as overprecision. This paper uses the definition of overestimation whereby individuals overestimate their actual financial literacy by having a higher perceived financial literacy.
Overconfidence is linked to individuals in various contexts and is also considered as an important concept in financial decision-making (Daniel et al. 1998, Daniel and Hirshleifer, 2015).3 Individuals who are overconfident have a tendency to make suboptimal financial decisions. Malmendier and Tate (2005) find that overconfident CEOs can lead to corporate
8 investment distortions and argue in a later paper (Malmendier and Tate, 2008) that CEOs who are overconfident tend to overestimate their ability to generate returns and overpay for target companies and hence undertake value-destroying mergers. Furthermore, Chuang and Lee (2006) state that overconfident investors trade more in riskier securities.
Additionally, overconfident investors trade more than rational investors (Odean, 1998; Statman et al., 2006; Glaser and Weber, 2007). Odean (1998) argues that overconfidence leads to increases in expected trading volume, increases market depth, and decreases in the expected utility of those traders that are overconfident. In a later paper, Odean (1999) also finds that overconfident investors even trade when their expected gains do not offset their trading costs and thus lower their returns.
Although there is extensive literature that studied overconfidence, there is limited research on financial literacy overconfidence. Porto and Xiao (2016), Kramer (2016) and Gentile et al. (2016) have investigated the impact of financial literacy overconfidence on advice seeking. Gentile et al. (2016) and Porto and Xiao (2016) find that overconfidence is negatively related to the demand for financial advice. Additionally, Porto and Xiao (2016) find that overconfident individuals are generally less likely to seek advice in the areas of saving/investment and mortgages, which could actually help them to grow, however, they are more likely to have received debt advice in the last five years in the period of the survey and to exhibit demand for tax planning. Individuals who are overconfident about their financial literacy just pick and choose when to use financial advice.
Kramer (2016) on the other hand finds no significant result of overconfidence on advice seeking, although he does find that confident investors (investors with high perceived financial literacy) are less likely to seek advice. Kramer (2016) also argues that the negative relation between confidence and advice seeking is even more pronounced among wealthy households. Asaad (2015) also measures financial literacy overconfidence combining actual and perceived financial literacy using the national survey data from the United States. She finds that overconfident individuals are more likely to engage in risky financial behaviours. A key contribution of her study is a better understanding of the influence of confidence on financial behaviours. She argues that confidence is good, however, overconfidence is not and, more importantly, can actually be self-injurious.
9 risks associated with those investments. This, in turn, can then lead to suboptimal or even worse outcomes.
Apparently, there is only one study that examines the effect of financial literacy overconfidence on stock market participation. Xia et al. (2014) define financial literacy overconfidence as the difference between the subjective and objective financial literacy scores and investigated the relationship between their definition of financial literacy overconfidence and stock market participation. They used data from the 2012 Chinese Survey of Consumer Finance and find that financial literacy overconfidence is positively correlated with stock market participation. An overconfident individual is according to Xia et al. (2014) around 20% more likely to participate in the stock market.
Based on the fact that overconfident investors trade more than rational investors (Odean, 1998; Statman et al., 2006; Glaser and Weber, 2007) and the finding of Xia et al. (2014) that financial literacy overconfidence is positively correlated with the stock market, the following hypothesis is formed:
H3: Financial literacy overconfidence has a positive effect on stock market participation.
3. Data and methodology
Although the effect of financial literacy on stock market participation is already studied by van Rooij et al. (2011), the effects of actual financial literacy and perceived financial literacy are first re-examined in this paper to see whether the results of the studies are equal. This is done in line with van Rooij et al. (2011), however, instead of OLS regressions, logistic regressions are used. Thereafter, the paper is extended with the effect of financial literacy overconfidence on stock market participation.
3.1. Data
The same dataset as the study of van Rooij et al. (2011) is used, which is from the 2010 Nederlandsche Bank’s Household Survey (DHS). The DHS is part of CentERdata, a research institute specialized in internet surveys at the Tilburg University.4 The DHS is a unique set of data of psychological and economic aspects of financial behaviour and is collected from over 2,000 households participating in the CentERpanel, which was launched in 1993. CentERdata
10 provided the data of the financial literacy module designed by van Rooij, Lusardi and Alessie (2011) as well.5
Table 1 provides an overview of the distribution of the sample and their stockholdings within different demographic factors. In this sample, 25.8% of the respondents holds stocks and/or mutual funds. Although the respondents are relatively evenly distributed among the education categories, the degree of stock market participation increases with the level of education. However, not even half of the people with the highest level of education participate in the stock market (36.94%). When looking at the occupation of the respondents, the largest group of respondents that participate in the stock market are the individuals who are retired (34.00%). Furthermore, stock market participation is more than twice as much among men compared to women, there is only a small difference in stock market participation between respondents who are married or not and whether they have children or not. Additionally, stock market participation gradually increases with age and increases strongly with net monthly household income and wealth.6
Table 1
Distribution of the sample.
DHS Households (%) Stock market participation (%)
Education
Primary school 4.45 18.37
Pre-vocational education 22.89 16.36
Pre-university education 13.74 24.68
Senior vocational training 19.51 19.27
Vocational colleges 26.34 33.33 University education 13.07 36.94 Occupation Employee 52.72 27.17 Self-employed 3.85 25.00 Retired 20.03 34.00 Other 23.41 14.94 Gender Male 55.31 32.98 Female 44.69 16.26 Marital status Married 68.44 27.16
5 I would like to thank Maarten van Rooij, Senior Economist at De Nederlandsche Bank for providing the correct answers of the financial literacy module.
11 Not married 31.56 22.97 Children Children (1 or more) 36.07 26.91 No children 63.93 28.10 Age ≤ 30 11.87 13.64 31 – 40 20.29 19.62 41 – 50 22.08 26.53 51 – 60 20.62 29.48 61 – 70 14.39 29.78 > 70 10.74 34.09
Net monthly household income
Income ≤ €1,150 15.91 7.83 €1,150< Income ≤ 1,800 27.95 17.23 1,800< Income ≥2,600 29.0 24.62 Income ≥2,600 34.2 39.68 Wealth (N = 2,084) Wealth <= €2,300 15.91 3.83 €2,300 < Wealth <= €45,500 27.95 12.48 €45,501 < Wealth <= €197,300 7.80 29.94 Wealth > €197,300 48.33 30.22
This table represents the distribution of the sample and the degree of stock market participation across demographics. Stock market participation is defined in line with van Rooij et al (2011) as owning stocks and/or mutual funds. The data are from the 2005 DNB Household Survey.
3.2. Key variables
3.2.1. Stock market participation
The dependent variable in this paper is stock market participation. A dummy variable is created for stock market participation based on whether a respondent is holding stocks or not, where a [1] is assigned if a respondent is participating in the stock market and a [0] otherwise. Stock market participation is defined in line with van Rooij et al. (2011) as owning stocks or mutual funds.
3.2.2. Financial literacy
12 questions are validated by van Rooij et al. (2011). For the exact wording of the financial literacy questions, please refer to appendix A.
Table 2 provides an overview of the distribution of the financial literacy scores and the relationship with stock market participation. Panel A shows the distribution of the basic financial literacy questions. On average, respondents answered four out of the five basic financial literacy questions correctly. Almost half of the respondents (43.57%) have answered all basic financial literacy questions correctly and around one third (32.63%) have answered only one out of the five questions wrong. More than 75% of the respondents answered all questions correct or only one question wrong, consequently, the majority of the respondents have good basic financial knowledge and less than 25% have a basic financial literacy below average. For this reason, the basic financial literacy questions are not representative for the actual financial literacy of the respondents. Nevertheless, the table shows that, as the measured basic financial literacy increases, there is also an increase in stock market participation implying a positive relation between basic financial literacy and stock market participation.
13
Table 2
Distribution financial literacy
DHS Households (%) Stock market participation (%)
Panel A: Measured basic financial literacy
None correct 2.39 16% 1 2.19 7% 2 5.11 12% 3 14.12 14% 4 32.63% 23% 5 43.57% 35% Mean score (#) 4.03 N 1,508 1,147
Panel B: Measured advanced financial literacy
None correct 6.70% 6.85% 1 3.85% 4.44% 2 5.24% 1.89% 3 5.11% 8.62% 4 7.10% 5.33% 5 8.02% 13.04% 6 11.54% 24.06% 7 15.25% 25.42% 8 12.47% 31.08% 9 13.86% 44.52% 10 8.02% 53.92% 11 2.85% 55.56% Mean score (#) 6.1 N 1,508 1,147
Panel C: Perceived financial literacy
Not knowledgeable 9.19% 9.78%
More or less knowledgable 62.10% 24.27%
Knowledgeable 25.48% 34.98%
Very knowledgeable 3.23% 51.42%
Mean 2.23
N 1,393 1,061
Panel D: Combined measured and perceived financial literacy
High perceived - High actual 21.62% 43.72%
Low perceived - Low actual 35.01% 12.11%
High perceived - Low actual 12.53% 9.93%
Low perceived - High actual 30.84% 33.80%
N 1,508 851
14 In addition to this calculation of actual financial literacy, a factor analysis similar to van Rooij et al. (2011) and Kramer (2016) is performed as a robustness check, granting that the results of the regression are correct using the financial literacy score instead of the factor loadings used in van Rooij et al. (2011) and Kramer (2016). With the factor analysis, an index of the advanced financial literacy is created. In line with van Rooij et al. (2011), the difference between the “incorrect” answers and the answers in which respondents indicate “I don’t know” are explicitly taken into account. Consequently, two dummy variables are created for each of the 11 advanced financial literacy questions: one dummy variable for the correct answers which is assigned a [1] if the respondent answered the question correct and a [0] otherwise and another dummy variable for the “I don’t know” answers which is assigned [1] if a respondent answered “I don’t know” and [0] otherwise. Therefore the factor analysis is performed on 22 variables. From this factor analysis one factor is retained. For the details of the factor analysis, see appendix B. The Kaiser-Meyer-Olkin (KMO) test of sampling adequacy shows a value of 0.9219 which indicates that the performed factor analysis is appropriate. The index obtained from the factor analysis and the score of advanced financial literacy are highly correlated (0.9102) as can be seen in table 3 of appendix B. Because of this high correlation, the factor loadings as well as the financial literacy score can be used as a variable for the actual financial literacy in the regressions. In this paper the financial literacy score is used as the variable for actual financial literacy.
3.2.3. Perceived financial literacy
After creating the variable for actual financial literacy, the variable for perceived financial literacy is created. The perceived financial literacy of the respondents is measured based on a question about people’s understanding of financial matters. The question is as follows: “How would you rate your understanding of financial matters (based on a scale of 1
to 7, whereby 1 is equal to ‘very bad’ and 7 is equal to ‘very good’)? According to van Rooij
15 Panel C of table 2 presents the distribution of the perceived financial literacy of the respondents. Only 9.19% of the respondents think of themselves as “not knowledgeable”, the majority of the respondents (62.1%) placed themselves in the second group of “more or less knowledgeable”, about a quarter of the respondents (25.48%) placed themselves in the group “knowledgeable” and only a small fraction of the respondents (3.23%) indicate themselves to be “very knowledgeable”. Notably, table 2 shows an increase in stock market participation when the perceived financial literacy increases suggesting that perceived financial literacy has a positive effect on stock market participation as well.
3.2.4. Financial literacy overconfidence
For the measure of financial literacy overconfidence the method of Allgood and Walstad (2012) is used, which is in line with Xia et al (2014), Kramer (2016) and Porto and Xiao (2016). To be able to construct the variable for financial literacy overconfidence the actual and perceived financial literacy measures are used. For these measures the mean is calculated which is 6.1 for the actual financial literacy and 2.23 for the perceived financial literacy. With the average of both financial measures, the respondents are divided over four groups. For each of the four groups, a dummy variable is created which is assigned a [1] a respondent belongs to that group and a [0] otherwise.
Panel D of table 2 presents the distribution of the combined financial literacy measures. Individuals with a lower than average perceived financial literacy and a lower than average actual financial literacy are placed in the “Low perceived – Low actual” group which consists of 35.01% of the respondents.7 Individuals with higher than average perceived financial literacy and higher than average actual financial literacy are placed in the group “High perceived – High actual” which consists of 21.62% of the respondents. Consequently, more than half of the respondents estimated their financial literacy correctly. Individuals that have a higher than average perceived financial literacy but a lower than average actual financial literacy think that they are more financially literate than they actually are. Therefore, these individuals are defined as overconfident and placed in the group “Overconfidence”. Only 12.53% of the respondents are defined as overconfident. Individuals with lower than average perceived financial literacy
16 but higher than average actual financial literacy are the opposite of overconfident individuals. They are defined as underconfident since they think that they have less financial knowledge than they actually have. Therefore they are placed in the group “Underconfidence”. Almost a third of the respondents (30.84%) are underconfident. For all four groups, dummy variables are created where a [1] is assigned to the dummy variable when a person belongs to that group and a [0] otherwise.
3.2.5. Control variables
As previously discussed, stock market participation can have several determinants. It is therefore necessary to control for possible endogeneity problems and to obtain a more reliable estimate of the main independent variable, in this case financial literacy overconfidence. To control for this bias, several control variables are included in the model. Previous studies find that stock market participation gradually increases with household income as well as with wealth due to the participation costs of stock market participation. (Bertaut, 1998; Guiso et al., 2003; Vissing-Jorgensen, 2000; van Rooij et al., 2011). Wealthier households and households with higher income are better able to pay the participation costs, which lowers the barrier to participate in the stock market. For this reason, dummies for wealth and household categories are created and added as control variables to the regressions. Haliassos and Bertaut (1995) and Campbell (2006) indicate that higher education also has a positive effect on stock market participation. Haliassos and Bertraut (1995, pp 1122) report that “education and the free acquisition of information are important in overcoming the barrier to stockholding erected by ignorance and misperceptions”. Campbell (2006) hereby adds that higher education can possibly reduce the objective costs of stock market participation. He states that educated households are better at diversifying their portfolios and thereby increase their returns. As in the previous studies, this paper also controls for demographic differences including age, gender, marital status, number of children and occupation (Haliassos and Bertaut, 1995; Campbell, 2006; van Rooij et al., 2011; Xia et al., 2014).
To reduce the possibility of endogeneity problems even further, some additional control variables are defined and added to the regressions. Christiansen et al. (2008) indicate that it is more important to control for the education of economics than for educational levels in general. Therefore dummy variables are created for the attention spent on economics during an individual’s education. The respondents could point out the attentions spent on economics by a four-point scale based on the following question: “How much attention is spent on economics
17 all four possible answers, a dummy is created which is assigned a [1] to the answer corresponding to the dummy and a [0] otherwise. These dummies are included as additional controls whereby the dummy for ‘a lot’ is used as the base group.
Moreover, van Rooij et al. (2011) state that individuals who need a good understanding of economics during their daily activities are more likely to participate in the stock market. For this reason, the understanding of economics during the daily activities of the respondents is added as an extra control variable as well. The respondents could also point out the understanding of financial matters needed on a four-point scale which is based on the following question: “How much understanding of financial matters do you need during your daily
occupation (work, hobbies, etc.)?” whereby a 1 is equal to ‘a lot’ and a 4 is equal to ‘almost
none’. Again, for all four possible answers, dummy variables are created which are assigned a [1] to the answer corresponding to the dummy and a [0] otherwise. These dummies are also included as additional controls whereby the dummy for ‘a lot’ is used as the base group.
Finally, Xia et al. (2014) state that risk aversion has a negative impact on stock market participation. Although Haliassos and Bertaut (1995) state that risk aversion alone cannot provide an explanation for the low rate of stock market participation, it does have some influence on stock market participation. Therefore, a control variable for risk tolerance is also included as additional control. For the variable of risk tolerance, a factor analysis is performed on six statements about the risk tolerance of the respondents based on agreement, whereby a 1 corresponds to ‘totally disagree’ and a 7 to ‘totally agree’. For the exact wording of the statements and factor analysis, see appendix C. From the factor analysis, one factor is retained and used to predict the variable of risk tolerance which is then included as additional control variable. The KMO test of sampling adequacy shows a value of 0.6620, which is not very high, but still indicates that the performed factor analysis is appropriate.
3.3. Methodology
Measuring stock market participation gives two possible outcomes and is therefore a binary dependent variable. It equals one when a person is holding stocks and zero otherwise. To test the effect of financial literacy overconfidence on stock market participation, the following model is used:
18 where Pi is the probability that the dependent variable stock market participation = 1 (yi = 1),
βx are the regression coefficients, FLOV is the variable of financial literacy overconfidence, CV
corresponds to a vector of control variables and 𝜀i is the error term. A vector of control variable
is included to limit endogeneity problems and get more reliable estimates of the main independent variables.
Since stock market participation is a binary dependent variable, a probit or logit model can be used to overcome the limitation of the linear probability model whereby it can produce estimated probabilities that are negative or greater than 1. In this paper, the logit model is used so that the fitted values are within the (0,1) interval. For the exact wording of the logit model, see appendix D.
The error term of the logistic regressions may suffer from heteroscedasticity. For this reason, the heteroskedastic-robust standard errors are used in this paper. Furthermore, the direct interpretation of the coefficients of the logit model is rather difficult. Since the model is not linear, it is incorrect to say that, for example, a 1-unit increase in FLOV causes a β2% increase in the probability that yi = 1. To interpret the parameter estimates more appropriately, the marginal effects are calculated. The marginal effects show how much a change in the explanatory variables changes the probability that yi = 1 and can be interpreted as follows: As, for example, the marginal effect of FLOV is 0.3, it can be said that a 1-unit increase in FLOV causes a 30% increase in the probability yi = 1.
In prior literature it is stated that men are more overconfident than women (Barber and Odean, 2001; Pirinisky, 2013; van Rooij et al., 2011; Almenberg and Dreber, 2015) and that overconfidence increases with age, wealth and income (Pirinsky, 2013; Pak and Chatterjee, 2016). Accordingly, it is examined if there are any interaction effects of overconfidence with gender, age, wealth, and income. An interaction effect is a joint effect of two independent variables. It could be that the effect of one independent variable is dependent on another independent variable. A combination of these independent variables can then create a special effect that could bias the results. To accurately understand the results of this study, it is examined if there are any interaction effects. To capture such an interaction effect, an extra term is added to the logistic regression, namely, financial literacy overconfidence multiplied by the variable of which we want to know whether there is an interaction term. For example overconfidence multiplied by gender. The model is therefore as follows:
19 where Male is the variable for gender which is assigned a [1] if a respondent is male and a [0] if a respondent is female, and 𝜀i is the error term.
4. Results
4.1. Financial literacy
As previously discussed, the effects of actual financial literacy and perceived financial literacy on stock market participation is first re-examined in this paper to check whether the results are similar to van Rooij et al. (2011).
Multiple logistic regressions are performed with the financial literacy measures. The first model that is estimated is a model without any control variables, hence only has actual financial literacy as explanatory variable. The marginal effects are presented in table 3. The results imply that actual financial literacy has a significant positive effect on stock market participation and indicates that an individual with higher actual financial literacy has a higher propensity to participate in the stock market.
In the second logistic regression, only the perceived financial literacy variable is included and again no control variables. The results indicate that perceived financial literacy has a significant positive effect on stock market participation as well. A 1-unit increase in perceived financial literacy increases the probability of stock market participation with 12.2%. The third model in table 3 includes both financial literacy measures and again the results indicate a significant positive effect for actual financial literacy as well as for perceived financial literacy. However, the likelihood of participating in the stock market of individuals with high perceived financial literacy has decreased to less than half of the effect when only perceived financial literacy is included in the regression. This indicates that adding both financial literacy measures to the regression provides more reliable estimates.
20 significant result, indicating that individuals who are married are less likely to participate in the stock market.
In the fifth model, even more control variables are added to the logistic regression. When also controlling for these additional controls, the actual financial literacy and perceived financial literacy measures still give a significant positive effect on stock market participation. According to the results, a 1-unit increase in actual financial literacy increases the probability of stock market participation with 4.65%. Furthermore, a 1-unit increase in perceived financial literacy causes an increase in the probability of stock market participation of 5.03%. However, this model also estimates a sudden negative significant result of senior vocational training, being a male respondent and whenever a respondent is self-employed on stock market participation. Furthermore, the model estimates a significant positive result for risk tolerance on stock market participation, indicating that a 1-unit increase in risk tolerance causes a 16% increase in the probability of stock market participation.
The results from the regressions presented in table 3 confirm the findings of van Rooij et al. (2011) and Xia et al. (2014) who find that actual perceived financial literacy has a positive effect on stock market participation. The first two hypotheses of this paper are accepted on at least the 5% and 10% confident levels as perceived financial literacy also shows a significant positive effect on stock market participation.
Table 3
Multivariate analysis of financial literacy on stock market participation
I II III IV V
Advanced financial literacy 0.0601*** 0.0609*** 0.0454*** 0.0465*** (0.00471) (0.00531) (0.00565) (0.00670) Perceived financial literacy 0.122*** 0.0521*** 0.0615*** 0.0503** (0.0205) (0.0200) (0.0187) (0.0251) Age dummies (Base group: age ≤
30) >30 age ≤ 40 0.0492 0.00671 (0.0606) (0.0700) >40 age ≤ 50 0.0981 0.0666 (0.0632) (0.0767) >50 age ≤ 60 0.133** 0.0745 (0.0662) (0.0754) >60 age ≤ 70 0.206** 0.185* (0.0926) (0.109) Age > 70 0.269** 0.231* (0.108) (0.130) Education dummies (Base group:
Primary school)
21
(0.0602) (0.0615)
Pre-university education -0.0440 -0.0903
(0.0627) (0.0589)
Senior vocational training -0.0440 -0.0978*
(0.0625) (0.0578) Vocational colleges -0.00491 -0.0558 (0.0674) (0.0679) University education 0.0311 -0.0119 (0.0776) (0.0769) Male 0.00813 -0.0665* (0.0302) (0.0395) Married -0.0807** -0.0857* (0.0388) (0.0453) Children 0.0148 0.0147 (0.0152) (0.0178) Occupation dummies (Base group:
Other) Employee 0.0293 0.000215 (0.0430) (0.0463) Self-employed -0.0390 -0.126*** (0.0625) (0.0470) Retired -0.0107 0.0299 (0.0533) (0.0682) Household net monthly income
dummies (Base group: ≤ €1150)
€1151 - €1800 0.102 0.0520 (0.0798) (0.0961) €1801 - €2600 0.169* 0.0995 (0.0863) (0.104) >€2600 0.239*** 0.170 (0.0873) (0.108) Wealth dummies (Base group:
Wealth ≤ €2300) €2300 < Wealth ≤ €45500 0.178 0.214 (0.123) (0.172) €45501 < Wealth ≤ €197300 0.376** 0.518*** (0.161) (0.191) Wealth > €197300 0.215** 0.266** (0.105) (0.131) Risk Tolerance 0.160*** (0.0198) Attention dummies (Base group:
22 4.2 Financial literacy overconfidence
The correlation of financial literacy overconfidence and stock market participation is presented in table 4. This correlation is negative and hence suggests the opposite effect of the hypothesis of this paper that financial literacy overconfidence has a positive effect on stock market participation.
Table 4
Correlation matrix overconfidence and stock market participation
Stock market participation Overconfidence Stock market participation 1.0000
Overconfidence -0.1358 1.0000
This table presents the correlation between overconfidence and stock market participation. The data are from the 2005 DNB Household Survey.
Multiple logistic regressions of financial literacy overconfidence on stock market participation are performed to examine if this negative effect of financial literacy overconfidence is significant. In the first model, no control variables are added to the logistic regression and therefore financial literacy overconfidence is the only explanatory variable in this regression. The marginal effects are presented in table 5. The results show that financial literacy overconfidence has a significant negative effect on stock market participation suggesting that a 1-unit increase in financial literacy overconfidence causes the likelihood of stock market participation to decrease with 18.1%. This finding rejects the hypothesis of this
Dummies understanding economics daily occupation (Base group: A lot) Some -0.0158 (0.0449) Little 0.0248 (0.0530) None 0.0297 (0.0809) Observations 1,147 1,061 1,061 1,060 875 Pseudo R2 0.1264 0.0287 0.1378 0.1980 0.2836
23 paper and contradicts the findings of Xia et al. (2014), who find a significant positive effect of financial literacy overconfidence on stock market participation.
To decrease the possibility of endogeneity problems, the control variables are included in the second model of the logistic regression. This model shows that there is still a significant negative effect of financial literacy overconfidence on stock market participation. To test the robustness of the estimates even more, the additional control variables are also added to the regression. The results still indicate a significant negative effect of financial literacy overconfidence on stock market participation.
Finally, the other combined measures of financial literacy are added to the logistic regression in the fourth model using the low perceived – low actual group as base group. Suddenly, the negative effect of financial literacy overconfidence on stock market participation becomes insignificant. The results are therefore not robust. However, the results should be interpreted with caution due to the fact that they can possibly be biased by multicollinearity since the combined financial literacy measures are all negatively correlated with each other (see table 6). Multicollinearity occurs when two explanatory variables are very highly correlated with each other. However, when the relationship involves more than two variables that are collinear, multicollinearity is difficult to detect.
Table 5
Combined financial literacy measures
I II III VI
Overconfidence -0.181*** -0.121*** -0.102** 0.00921
(0.0289) (0.0393) (0.0405) (0.0656)
Underconfidence 0.174***
(0.0460)
High perceived – High actual 0.240***
(0.0589)
Control variables (see table 5) No Yes Yes Yes
Additonal control variables (see table 5) No No Yes Yes
Observations 1,147 1,147 946 946
Pseudo R2 0.0191 0.1246 0.2144 0.2365
24
Table 6
Correlation matrix combined financial literacy measures High perceived – high actual Low perceived – low actual Overconfidence Underconfidence High perceived – high actual 1.0000
Low perceived – low actual
-0.3855 1.0000
Overconfidence -0.1988 -0.2779 1.0000
Underconfidence -0.3507 -0.4901 -0.2528 1.0000
This table presents the correlation between the combined financial literacy measures. The data are from the 2005 DNB Household Survey.
Furthermore, the fourth model shows a significant positive effect of the high perceived – high actual group on stock market participation. An individual in this group is 24% more likely to participate in the stock market compared to the low perceived – low actual group. This finding is similar with the findings of van Rooij et al. (2011) and Xia et al. (2014), who find that individuals with high actual financial literacy are more likely to participate in the stock market and a positive correlation between the perceived financial literacy and actual financial literacy.
Moreover, table 5 shows a positive result for individuals that are underconfident, however, Xia et al. (2014) find that underconfidence has a negative impact on stock market participation. Underconfidence decreases stock market participation with 10.09% in their study while in this study an individual that is underconfident has a 17.4% higher likelihood of participating in the stock market compared to the base group of individuals with low perceived and low actual financial literacy. This result could suggest that actual financial literacy has a stronger effect on stock market participation than perceived financial literacy and hence overconfident individuals participate less in the stock market due to their lower actual financial literacy and underconfident individuals more due to their higher actual financial literacy.
25 4.3. Interaction effects
In prior literature it is stated that men are more overconfident than women and that overconfidence increases with age, wealth and income. Therefore, it is examined if there are any interaction effects of overconfidence with gender, age, wealth, and income. The marginal effects are presented in table 7. In the first logistic regression, the interaction term for gender is included. As can be seen in table, the estimated interaction term between gender and financial literacy overconfidence has a significant negative effect. This indicates that there is indeed an interaction effect between financial literacy overconfidence and gender, implying that the marginal effect of an increase in financial literacy overconfidence on stock market participation is dependent on gender. The negative sign of the interaction effect suggests that the marginal effect of financial literacy overconfidence on stock market participation is smaller for men compared to women. This could be an explanation for the fact that financial literacy overconfidence is negatively correlated with stock market participation since an increase in financial literacy overconfidence has a larger marginal effect for women compared to men, however, women are less likely to participate in the stock market (Haliassos and Bertaut, 1995; van Rooij et al., 2011).
Table 7 Interaction effects I II III IV overconfidence -0.0715 -0.130** -0.162 -0.0600 (0.0549) (0.0584) (0.102) (0.101) Male 0.171*** (0.0259) Overconfidence * male -0.188*** (0.0437) Age 0.0366*** (0.00845) Overconfidence * age -0.0297 (0.0298) Household income 0.109*** (0.0139)
Overconfidence * household income -0.00195
26
This table reports the marginal effects of the interaction terms of gender, age, household income and wealth with overconfidence on stock market participation using a logit model. The data are from the 2005 DNB Household Survey. The robust standard errors are reported in parentheses; *** p<0.01, ** p<0.05, * p<0.1
As can be seen from table 7 the variable male shows a significant positive effect, meaning that men are 17.1% more likely to participate in the stock market than women. Table 8 shows the distribution of men and women in this study who are overconfident and the degree of stock market participation. Even though almost twice as much women are overconfident compared to men, men participate in the stock market almost three times as much as women. The negative correlation of financial literacy overconfidence on stock market participation in this paper could therefore be explained by this interaction effect since women are more overconfident than men but participate less in the stock market.
In the other models of table 7, the interaction term between age, income and wealth and financial literacy overconfidence is examined. The results, however, show no interaction effects between these variables and financial literacy overconfidence. Hence, there is only one interaction effect, which is between gender and financial literacy overconfidence.
To test for the robustness of this interaction effect, the control variables are added to the logistic regressions. The results are presented in table 9. By including the control variables in the first model, the interaction term of gender and financial literacy overconfidence remains significant. In the second model, the additional control variables are also added to the regression to get even more reliable estimates. Again, the interaction term between gender and financial literacy overconfidence remains significant. At last, the other combined measures of financial literacy are included in the regression with the low perceived – low actual group as base group. Even after controlling for these measures, the interaction term between gender and financial literacy overconfidence remains significant. Hence, it can be concluded that the marginal effect of an increase in financial literacy overconfidence on stock market participation is indeed dependent on the variable gender and that this effect is smaller for men than for women.
Table 8
Distribution of men and women that are overconfident and participate in the stock market
Male (%) Female (%)
Overconfidence 4.84 7.69
Participation 14.32 5.3
27
Table 9
Interaction effect gender
I II III Overconfidence -0.0389 0.0252 0.144 (0.0546) (0.0766) (0.0964) Male 0.0981*** 0.0174 -0.0103 (0.0286) (0.0355) (0.0366) Overconfidence * male -0.172*** -0.183*** -0.165*** (0.0377) (0.0351) (0.0396) Underconfidence 0.167*** (0.0458)
High perceived – high actual 0.229***
(0.0587)
Control variables Yes Yes Yes
Additional control variables No Yes Yes
Observations 1,146 946 946
Pseudo R2 0.1346 0.2217 0.2420
This table reports the marginal effects of the interaction terms of gender on overconfidence, controlling for the control variables in the first model and for the additional control variables in the second and third model. In the third model, the combined measures of financial literacy are also included. The data are from the 2005 DNB Household Survey. The robust standard errors are reported in parentheses; *** p<0.01, ** p<0.05, * p<0.1
4.4. Discussion
Van Rooij et al. (2011) studied the effect of financial literacy on stock market participation. They find that actual financial literacy, measured with the advanced financial literacy questions, has a positive relationship with stock market participation. The findings of this paper confirm these findings, accepting the first two hypotheses and extents the paper with the effect of perceived financial literacy and financial literacy overconfidence on stock market participation.
28 disappears when the other combined financial literacy measures are added to the logistic regression and no unambiguous conclusion can be drawn from this result.
Another contradicting finding in this study is the fact that individuals who are underconfident have a positive significant effect on stock market participation since Xia et al. (2014) find that an individual who is underconfident is less likely to participate in the stock market.
Furthermore, this paper finds that individuals with high perceived – high actual financial literacy have a significant positive effect on stock market participation which is in line with van Rooij et al. (2011) and Xia et al. (2014). Individuals with higher actual financial literacy as well as persons with higher perceived financial literacy are more likely to participate in the stock market. Combining these two measures gives the result that it increases the likelihood of participating in the stock market which makes intuitively sense.
There are several possible explanations for these contradicting findings. One possible explanation could be attributable to the way in which the financial literacy is measured. Most studies use questionnaires to measure the financial literacy, however, the questions often vary among different studies (van Rooij et al., 2011; Xia et al., 2014; Gentile et al., 2016; Porto and Xiao, 2016). This makes it hard to compare studies on the effects of financial literacy.
Another possible explanation could be a endogeneity problem. Endogeneity exists when an explanatory variable is correlated with the error term. It is possible that the measurements of financial literacy and hence financial literacy overconfidence are endogenous. For example, the sample in this study could be biased by selectivity which means that the sample is not properly randomized. Nonetheless, the DHS household survey is run by CentERdata which specializes in internet surveys and contains over 2,000 households which can be regarded as representative for the Dutch population. However, by merging the data for stock market participation and financial literacy the sample is reduced considerably, increasing the possibility of selection bias since the exclusion of data could influence the statistical significance of the logistic regression.
Endogeneity may also be caused by omitted variables. This is the case when an omitted variable affects the dependent as well as the independent variables separately. Problems of endogeneity are limited in this study as much as possible by controlling for a large set of demographic factors and some other extra control variables. However, looking at all the possible variables, a bias caused by omitted variables cannot be ruled out. Another possible cause for endogeneity could be a measurement error.
29 experiences of siblings and parents as instrumental variables since respondents are not in control over these factors. They find that using instrumental variables enhances the positive impact of financial literacy on stock market participation. However, the instrumental approach is beyond the scope of this study.
Additionally, the results could be biased by multicollinearity. With multicollinearity, two or more independent variables correlate with each other. Since financial literacy (and hence also financial literacy overconfidence) can be influenced by many factors, for example gender, education, daily occupation, family etc. there is a possibility of multicollinearity. Especially the four combined measures of financial literacy correlate with each other, however, since this includes more than two independent variables, multicollinearity is difficult to detect.
Since prior literature states that men are more overconfident than women, and that overconfidence increases with age, income and wealth, this study also examines the interaction effects between these variables and financial literacy overconfidence. However, this study finds only one significant interaction term, which is the effect of gender on overconfidence. The marginal effect of an increase in financial literacy overconfidence on stock market participation is dependent on gender. This effect is smaller for men compared to woman, even though women participate less in the stock market. This interaction effect can therefore have implications for the outcome of this study.
5. Conclusion and limitations 5.1. Conclusion
Investigating the drivers for stock market participation is of great value since not participating in the stock market can have important economic implications. Cocco et al. (2005) find that considerable welfare loss can arise from non-participation. According to them not participating in the stock market can result in a welfare loss of more than 2% of the annual consumption. Moreover, prior literature shows that overconfident individuals are likely to make suboptimal decisions which questions if overconfident individuals are able to invest efficiently.
30 financial literacy (Allgood and Walstad, 2012). Especially when these two measures are both high, an individual is more likely to participate in the stock market. However, for financial literacy overconfidence, this paper has contradicting findings with prior literature. A negative effect of financial literacy overconfidence on stock market participation is found, while the study of Xia et al. (2014) find a positive relationship. Furthermore, this paper finds that individuals who are underconfident are more likely to participate in the stock market even though Xia et al. (2014) find that underconfidence has a negative impact on stock market participation. These contradicting findings could suggest a stronger effect of actual financial literacy on stock market participation compared to perceived financial literacy in this study, however, there is no literature available that can support this theory. The differences between this study and the study of Xia et al. (2014) can also be caused by endogeneity problems or multicollinearity and could be partially attributed to the way financial literacy is measured.
5.2. Limitations and future research
This study encounters several limitations that could be taken into account for future research. First, the sample consist of respondents who are willing to give their information about their stock holdings. It could be that individuals who are willing to give this information are more likely to participate in the stock market. Furthermore, the financial literacy score may be measured with substantial error. As van Rooij et al. (2011) also point out, respondents may guess several questions, whereby some questions are regarded as correct although this resulted from guessing. This could especially be the case for the advanced financial literacy questions which are used as the financial literacy measure. Van Rooij et al. (2011) also note that respondents can learn from participating in the stock market, get familiar with the questions asked in the module, and improve their knowledge.
31 Furthermore, the measure for perceived financial literacy in this study is based on only one question in the 2005 DNB Household Survey. Future research could expand this measure by creating more questions concerning perceived financial literacy making the measure of perceived financial literacy and consequently the financial literacy overconfidence measure more reliable.
Another limitation is the problem of endogeneity. This study has limited this by controlling for a large set of demographic factors and some additional control variables. However, endogeneity cannot be ruled out. Future research could further reduce the possibility of endogeneity by using an instrumental variable approach. Several studies on financial literacy have used instrumental variables estimation to control for endogeneity and estimate the effect of financial literacy on financial behaviour (Lusardi and Mitchell, 2014). This approach allows for a consistent estimation when the independent variables are correlated with the error term. A valid instrument has to be created that can be added to the regression which induces changes in the independent variable and has no independent effect on the dependent variable.
Finally, this study is limited to the Dutch population. Future research could expand the effects of financial literacy overconfidence on stock market participation in other countries or for an area of countries, for example, Europe and investigate the differences between countries.
Appendices
Appendix A Financial literacy questions designed by van Rooij et al. (2011)
Basic financial literacy questions designed by van Rooij et al (2011)
1. Numeracy 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? i. More than €102
ii. Exactly €102 iii. Less than €102 iv. Do not know v. Refusal
2. Interest compounding Suppose you had €100 in a savings account and the interest rate is 20% per year and you never withdraw money or interest payments. After 5 years, how much would you have on this account in total?
32
3. Inflation 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? i. More than today
ii. Exactly the same iii. Less than today iv. Do not know v. Refusal
4. Time value of money Assume a friend inherits €10,000 today and his sibling inherits €10,000 3 years from now. Who is richer because of the inheritance?
i. My friend ii. His sibling
iii. They are equally rich iv. Do not know
v. Refusal
5. Money illusion Suppose that in the year 2010, your income has doubled and prices of all goods have doubled too. In 2010, how much will you be able to buy with your income?
i. More than today ii. Exactly the same iii. Less than today iv. Do not know v. Refusal
Advanced financial literacy questions designed by van Rooij et al (2011)
1. Which of the following statements describes the main function of the stock market? i. The stock market helps to predict stock earnings
ii. The stock market results in an increase in the price of stocks
iii. The stock market brings people who want to buy stocks together with those who want to sell stocks iv. None of the above
v. Do not know vi. Refusal
2. Which of the following statements is correct? If somebody buys the stock of firm B in the stock market: i. He owns a part of firm B
ii. He has lent money to firm B iii. He is liable for firm B’s debts iv. None of the above
33
3. Which of the following statements is correct?
i. Once one invests in a mutual fund, one cannot withdraw the money in the first year ii. Mutual funds can invest in several assets, for example invest in both stocks and bonds iii. Mutual funds pay a guaranteed rate of return which depends on their past performance iv. None of the above
v. Do not know vi. Refusal
4. Which of the following statements is correct? If somebody buys a bond of firm B in the stock market: i. He owns a part of firm B
ii. He has lent money to firm B iii. He is liable for firm B’s debts iv. None of the above
v. Do not know vi. Refusal
5. Considering a long time period (for example 10 or 20 years), which asset normally gives the highest return? i. Savings accounts
ii. Bonds iii. Stocks iv Do not know v. Refusal
6. Normally, which asset displays the highest fluctuations over time? i. Savings accounts
ii. Bonds iii. Stocks iv Do not know v. Refusal
7. When an investor spreads his money among different assets, does the risk of losing money: i. Increase
ii. Decrease iii. Stay the same iv Do not know v. Refusal
8. If you buy a 10-year bond, it means you cannot sell it after 5 years without incurring a major penalty. True or false?
i. True ii. False
iii. Do not know iv. Refusal
9. Stocks are normally riskier than bonds. True or false? i. True
ii. False
iii. Do not know iv. Refusal
10. Buying a company stock usually provides a safer return than a stock mutual fund. True or false? i. True
ii. False
iii. Do not know iv. Refusal
11. If the interest rate falls, what should happen to bond prices? i. Rise
34
iii. Stay the same iv. None of the above v. Do not know vi. Refusal
Appendix B Factor Analysis
Table 1
Factor analysis financial literacy
N = 1508
Retained factors = 1 Number of params = 22
Factor Eigenvalue Difference Proportion Cumulative
Factor 1 8.55596 7.83275 1.0000 1.0000 Factor 2 0.72321 0.12543 0.0845 1.0845 Factor 3 0.59778 0.23718 0.0699 1.1544 Factor 4 0.36060 0.04192 0.0421 1.1965 Factor 5 0.31868 0.09068 0.0372 1.2338 Factor 6 0.22800 0.07606 0.0266 1.2604 Factor 7 0.15194 0.01246 0.0178 1.2782 Factor 8 0.13948 0.03376 0.0163 1.2945 Factor 9 0.10572 0.05868 0.0124 1.3068 Factor 10 0.04704 0.04481 0.0055 1.3123 Factor 11 0.0023 0.00999 0.0003 1.3126 Factor 12 -0.00776 0.18492 -0.0009 1.3117 Factor 13 -0.19267 0.02035 -0.0225 1.2892 Factor 14 -0.21302 0.02740 -0.0249 1.2643 Factor 15 -0.24042 0.00894 -0.0289 1.2362 Factor 16 -0.24936 0.00858 -0.0291 1.2070 Factor 17 -0.25794 0.00737 -0.0301 1.1769 Factor 18 -0.26532 0.02164 -0.0310 1.1459 Factor 19 -0.28695 0.02247 -0.0335 1.1123 Factor 20 -0.30942 0.01057 -0.0362 1.0762 Factor 21 -0.31999 0.01179 -0.0374 1.0388 Factor 22 -0.33179 -0.0388 1.0000
LR test: independent vs. saturated: chi2 (231) = 1.7e+04 Prob>chi2 = 0.0000
This table reports the eigenvalues, differences, proportions and cumulative proportions of the factor analysis used for the advanced financial literacy questions. The iterated principal component method is used to obtain the factor loadings. The factor analysis is performed over 22 dummy variables. 11 dummy variables are created for the correct answers of the advanced financial literacy questions, assigning a [1] when the answer is correct, and a [0] otherwise. The 11 other dummies are created for the “I don’t know” answers which are assigned a [1] if the respondent answered “I don’t know” and a [0] otherwise. The data are from the 2005 DNB Household Survey. Table 2
Factor loadings and unique variances
Variable Factor1 Uniqueness
Adv1 0.5976 0.6429
DK -0.7144 0.4896
Adv2 0.4241 0.8201
DK -0.6065 0.6322
35 DK -0.7613 0.4205 Adv4 0.6050 0.6340 DK -0.7504 0.4369 Adv5 0.5057 0.7443 DK -0.7448 0.4452 Adv6 0.6400 0.5904 DK -0.7673 0.4112 Adv7 0.5543 0.6928 DK -0.7502 0.4373 Adv8 0.4704 0.7787 DK -0.5190 0.7306 Adv9 0.2978 0.9113 DK -0.7494 0.4384 Adv10 0.4646 0.7842 DK -0.7600 0.4224 Adv11 0.3671 0.8652 DK -0.6924 0.5206
This table reports the factor loadings and the uniqueness of the factor analysis used for the advanced financial literacy questions. The iterated principal component method is used to obtain the factor loadings. The factor analysis is performed over 22 dummy variables. 11 dummy variables are created for the correct answers of the advanced financial literacy questions, assigning a [1] when the answer is correct, and a [0] otherwise. The 11 other dummies are created for the “I don’t know” answers which are assigned a [1] if the respondent answered “I don’t know” and a [0] otherwise. Adv corresponds to a correct answer of the advanced financial literacy question with the number of the question followed. DK stands for “Don’t know” and corresponds to the advanced financial literacy question when it is answered with “I don’t know”. The data are from the 2005 DNB Household Survey.
Table 3
Kaiser-Meyer-Olkin measure of sampling adequacy
