Parental influence, financial literacy and stock market participation for households

(1)

j.wilms@student.rug.nl Phone: 06-43680422

Parental influence, financial literacy and

stock market participation for households

Jarno Wilms

S2551470

MSc Finance

Supervisor: Robert Lensink

University of Groningen – Faculty of Economics and Business

Key words: Stock market participation, parental influence, financial literacy, household finance.

Abstract

This paper expands on existing literature by adding in a parental influence variable into a model that explains

stock market participation using advanced financial literacy. We find no proof that we need to account for

endogeneity in our model. Furthermore, we find no direct evidence for a direct effect of parental influence

(2)

2

1. Introduction

In the field of financial literacy there is a general consensus that consumers lack the financial knowledge needed to make the financial decisions that are in their own best interest (Perry, 2008).

For example, Van Rooij et al. (2011) found that financial illiterate households are less likely to participate in the stock market than their financially literate counterparts. This lack of stock market participation for households is otherwise known as the non-participation puzzle (Heaton & Lucas, 2000). Guiso et al. (2003) show that a mere 15-25% of the

households in the Netherlands, Germany and France are active in the stock market, and that a majority of European households holds their financial wealth in liquid, safe and low-return assets. As Cocco et al. (2005) show, the lack of entry in the stock market for households can lead to a sizeable welfare loss. The loss of consumption as a result of non-participation in the stock market can easily reach 1.5%-2% of consumption annually. Besides annual

consumption loss, not participating in the stock market also implies that a household has a reduced toolkit to smooth income and to manage financial risks (Guiso et al., 2003). A wide range of determinants have already been found for stock market participation (Agarwal et al., 2009; Hong et al., 2004; Hilgerth and Hogarth, 2002, among others). Although parental influence has been linked to financial literacy (Clarke et al., 2005), there has been no previous research where parental influence is used to help explain stock market participation.

The research field concerning financial literacy and its relation to stock market participation is fairly new: Only since the beginning of the century it is actively researched what are the driving determinants for stock market participation. The most influential research relating financial literacy to stock market participation was conducted by van Rooij et al. (2011). In their paper, van Rooij et al. (2011) set out to find a relationship between financial literacy and stock market participation, by means of a questions set they have added into the National Household Survey (NHS) conducted by De Nederlandsche Bank (DNB). Van Rooij et al. (2011) find that most respondents in their survey possess some basic financial literacy. However, only the knowledge of few goes beyond these basic financial literacy questions; many do not know differences between bonds and stocks, for example. Furthermore, only 23.8% of households in their data set are in active in the stock market, which is defined as either owning mutual funds or having direct stock holdings.

Van Rooij et al. (2011) then construct an advanced financial literacy index comprised of all advanced financial literacy questions based on factor analysis in order to obtain a

(4)

4 stock market participation over advanced financial literacy and a wide range of control variables, in order to account for the effect on stock market participation of these variables as well. In order to account for possible effects of endogeneity, they also run a Generalized Method of Moments (GMM) regression, using the financial situation of the oldest sibling as an instrumental variable.

Eventually, van Rooij et al. (2011) find that financial illiteracy is a significant deterrent for stock market participation, even after accounting for already known effects of other variables on stock market participation.

Parental influence (in the context of this research) is defined as the effect that parents can have on the financial literacy of their children. Children learn from their environment through socialization, which is defined by Ward (1974) as ‘’the process by which young people acquire skills, knowledge and attitudes relevant to their effective functioning as consumers in the marketplace’’. Clarke et al. (2005) found that strong parenting practices such as explicitly teaching financial concepts helps increase financial literacy for later years. Rettig & Mortenson (1986) found that children learn to financially manage themselves through observation, participation and intentional instructions by socialization agents such as their parents. Several other influential papers have explicitly researched the relation between parental influence and financial literacy; Jorgensen & Salva (2010) found no effect of parental influence on financial knowledge, although they did find support for both direct and indirect mechanisms where parental influence affected financial behavior. However, another paper of Jorgensen (2007) did find support for parental influence affecting children financial knowledge, financial attitudes, and financial behavior. Lusardi et al. (2010) found that a disproportionate amount of financially literate college student had parents who also had a college degree. Finally, Clarke et al. (2005) found that in their sample a majority of financial literacy came from influences within the family sphere, and very little came from outside sources.

(5)

5 participation of households. The practical application of this paper would be to gain more understanding in the behavior of households and what leads to their lack of participation in the stock market. As van Rooij et al. (2011) have shown, lower financial literacy leads to a lack of participation in the stock market. Obtain more clarity about the role of parental influence with regard to financial literacy and stock market participation can lead to better insights in explaining the non-participation puzzle. Furthermore, significant results of parental influence towards financial literacy and stock market participation might to policy changes to make parents more aware of the effects they have on their children. An example of a policy change based on this research could be to shape financial education programs towards parents, teaching them how to educate their children financially.

The remainder of the paper is structured as follows: In section II, I will discuss the existing literature on parental influence and the relation between financial literacy and stock market participation. In section III, I will discuss the data used in the research and how the variables in the research are constructed. In section IV, the methodology of the paper will be

(6)

6

2. Literature review

2.1 Financial literacy

Hastings et al. (2013) says the following about financial literacy: “ Financial literacy has taken on a variety of meanings; it has been used to refer to knowledge of financial products (e.g., what is a stock vs. a bond; the difference between a fixed vs. an adjustable rate mortgage), knowledge of financial concepts (inflation, compounding, diversification, credit scores), having the mathematical skills or numeracy necessary for effective financial decision making, and being engaged in certain activities such as financial planning.” Since financial literacy is such a broad concept, many different measures are used in different countries to measure financial literacy (Lusardi & Mitchell, 2011). Initially, the consensus for measuring financial literacy was asking the ‘Big Three’ questions as can be seen in the Appendix, which asked questions about interest rate compounding, inflation and risk diversification. This later expanded with two additional questions in the ‘Big Five’. However, there is a lack of

evidence whether this was the correct measure for financial literacy. Van Rooij et al. (2011) designed an expanded survey for the DHS Household Survey to measure financial literacy, which contained 5 questions for measuring basic financial literacy and 11 more questions for measuring advanced financial literacy.

There is a substantial impact of financial literacy on financial behavior. For example, higher levels of financial literacy lead to less financial difficulties in later years (Danes & Hira, 1987). Furthermore, high financial literacy is associated with lowered chance of bankrupcy and receiving government aid ((Bauer et al., 2000; Huston et al., 2003; Blalock et al, 2004). Choi et al. (2001) show that lack of financial literacy can lead to non-trivial welfare losses when consumers have to make financial decisions. In their paper discussed in the introduction, Van Rooij et al. (2011) have found support for the theory that financial literacy positively influences stock market participation likeliness.

(7)

7 2.2 Stock market participation

The question what the driving determinants are for stock market participation is important since several papers (e.g. Heaton and Lucas, 1999 and Moskowitz and Jorgensen, 2002) argue that it has a direct effect on the equity premium, and thus is important for solving the equity-premium puzzle, since there must be a distinction in the consumption levels of stockholders compared to non-stockholders, according to Mankiw and Zeldes (1991). There have been several papers that found a partial explanation for the non-participation puzzle: Vissing-Jorgensen (2004) points towards transaction costs as an explanation. Guiso and Jappelli (2005) found that many households had a lack of asset awareness, which also led to households being less likely to participate in the stock market. Grinnblatt and

Keloharju (2011) found a similar effect for IQ, as well as Christellis et al. (2010) did for limited numeracy. There has also been some research on social effects: Hong et al. (2004) have shown that a social interaction effect between households exists, meaning households that are more social are more likely to be stockholders. Furthermore, Guiso et al. (2008) show that a lack of trust in the stock market also contributes to the low participation of

households in the stock market.

Stock market participation is generally measured in the same way: It includes both holdings in single stocks as well as holdings in mutual funds. If a household possesses one of these two kinds of stock, they are participating in the stock market. No distinction is being made between possessing one kind or both kinds of stock. Further, differences in stock market participation between first-world countries are fairly small: In the sample used by Van Rooij et al. (2011), only 23.8% of Dutch households hold stocks. According to Campbell (2006) and Haliassos and Bertaut (1995), 60-70 percent of US households do not hold stocks. Also, Guiso et al. (2003) show that 15-25% of households in the Netherlands, Germany and France participate in the stock market.

2.3 Parental influence

Parental influence is widely considered to be an important predictor for the level of financial literacy that an individual or household has. For example, Shim et al. (2010) find that the effect of parents is more significant than that of having early work experience and high school financial education combined.

(8)

8 in the marketplace’’. Since this concept is very broad, the measurement of parental

influence differs between studies. Bucciol and Veronesi (2013) use the same survey as is used in this paper. They measure different aspects of parental influence by taking several youth questions asked in the survey (see Table 3). The factors of parental influence they measure are giving pocket money, controlling money usage and giving advice about savings and budgeting. Jorgenson (2007) measures parental influence by asking 6 questions about to which degree participants learned from their parents or peers about managing their money, among other factors. As of now, there are still many different approaches that can be taken for measuring parental influence.

Clarke et al. (2005) found that strong parenting practices such as explicitly teaching financial concepts helps increase financial literacy for later years. Rettig & Mortenson (1986) found that children learn to financially manage themselves through observation, participation and intentional instructions by socialization agents such as their parents. Several other influential papers have explicitly researched the relation between parental influence and financial literacy; Jorgensen & Salva (2010) found no effect of parental influence on financial

knowledge, although they did find support for both direct and indirect mechanisms where parental influence affected financial behavior. However, another paper of Jorgensen (2007) did find support for parental influence affecting children financial knowledge, financial attitudes, and financial behavior. Lusardi et al. (2010) found that a disproportionate amount of financially literate college student had parents who also had a college degree.

However, it should be noted that most of the research conducted on the effect of parental influence on financial literacy is measured on college-students or other young adolescents. Therefore, we do not know about the long-term effects of parental influence on financial literacy.

2.4 Hypotheses

(9)

9 participation.

Considering the literature discussed, we arrive at the following set of hypotheses:

H1: Parental influence will have a positive direct significant effect on stock market

participation.

H2: Parental influence will have a positive indirect effect on stock market participation,

through financial literacy.

3. Data and descriptives

3.1 Data

All data used in this paper is obtained from the De Nederlandsche Bank Household Survey (DHS). The DHS survey collects a wide range of information concerning the financial situation of households, as well as a range of psychological and economic concepts. Van Rooij et al. (2011) designed an additional module for this survey to measure financial literacy, which is what will be used for measuring financial literacy in this research as well. This data was collected in 2005, and the data from the DHS survey that we will use will therefore be from the 2005 data wave as well. The data for the DHS survey is collected by means of an internet survey, which decreased potential bias when compared to surveys conducted by phone (Chang and Krosnick, 2009). The internet survey is conducted by CentERdata, a research institute based in Tilburg which specializes in internet surveys.1_{The data set is}

representative of the Dutch population, and has about 2,000 respondents.

Since we use data from both the main survey of the DHS and also the financial literacy additional survey designed by Van Rooij et al. (2011), we have to combine the different datasets from the survey and use the overlapping sample of the group who answered questions about their youth and the financial literacy questions, who also participated in the main survey of the DHS. This leaves us with a sample set of n=383. In Table 1 below, we can see that there is some degree of sampling bias in our final sample. In the original sample, the stock market participation for the whole sample was 20%, whereas in our final sample the stock market participation of our sample has increased to 29.8%. It is important that this is kept in mind when considering the results of our regression.

(10)

10 Table 1: Stock market participation for the original sample and the final sample

Original sample, N=2084 Final sample, N=383

Percentage of households holding

stocks 10.8 16.2

Percentage of households invested

in mutual funds 14.3 19.8

Percentage of households active in

the stock market 20.0 29.8

In Table 2 is stock market participation shown for different subgroups of our sample. We can see that social status is positively correlated with stock market participation; if you are in a better social class you are also more likely to participate in the stock market. Furthermore, being employed (excluding self-employment) or retired, having completed a higher level of education, being married all seem to increase the likelihood of stock market participation. Also, having a higher net income and possession of larger amount of wealth seem to

contribute positively towards stock market participation, as well as having a higher advanced literacy score and a high self-score on financial literacy. These findings are in line with what was expected from the literature. However, age increase appears to have a positive

correlation with stock market participation, and not the inverted U-shape as Agarwal et al. (2009) found. This can be caused by wealthy people living longer than those who are poor, which could cause the high stock market participation for the upper age classes.

Table 2: Stock market participation for different sample subsets

This table shows stock market participation for different subgroups of social status, occupation, age, education level, gender, marital status, net household income, household wealth, advanced

literacy score and financial literacy self-score.

Sample subgroup % of sample % stock market participation

Social status (N=382)

Social class 1 (highest) 20.1 41.6

Social class 2 32.9 36.5

Social class 5 (lowest) 1.3 0.0

(11)

11

Age 50-60 29.8 24.6

Age 60-70 18.0 37.7

Age 70-80 5.2 45.0

Highest education completed (N=382)

Primary 3.1 25.0

Preparatory intermediate vocational 24.9 13.7

Intermediate vocational 13.1 26.0 Secondary pre-university 17.8 26.5 Higher vocational 27.7 40.6 University 13.4 47.1 Gender (N=383) Male 60.1 38.3 Female 40.0 17.0 Marital status (N=373) Not married 37.8 25.5 Married 62.2 31.9

Net household income, monthly (N=383)

EUR 1150 or less 9.9 10.5

EUR 1151 - EUR 1800 28.5 19.3

EUR 1801 - EUR 2600 28.5 32.1

EUR 2600 or more 33.2 42.5

Household wealth (N=359)

First quartile (poorest) 25.1 27.8

Second quartile 24.8 22.5

Third quartile 25.1 31.1

Fourth quartile (richest) 25.1 38.9

Advanced literacy score (N=383)

First quartile (lowest score) 25.1 5.2

Fourth quartile (highest score) 26.6 57.8

Self-score financial literacy (N=381)

1 (Lowest score) 0.0 0.0 2 2.4 11.1 3 7.1 7.4 4 25.9 19.2 5 32.6 35.2 6 26.1 37.0 7 (Highest score) 3.1 83.3 Don't know 2.4 11.1

1_{Quartile samples might have different size due to identical values.}

Note: Sample size may differ due to incomplete data. For wealth calculations, stock holdings are excluded, since this is already represented in the dependent variable.

(12)

12 3.2 Key variables

3.2.1 Stock market participation

Stock market participating is the dependent variable in our model. The stock market

participation of households is defined as a household either having stocks or investments in mutual funds (van Rooij et al., 2011). A dummy variable is created where a score of 1 is assigned to a household if they possess either stock or mutual funds, and a score of 0 when a household possesses neither of them. In our sample, stock market participation is 30%. 3.2.2 Financial literacy

(13)

13 the survey, it makes sense that the scores on advanced literacy of our sample is also higher, since we assume that higher financial literacy leads to stock market participation.

Table 3: Advanced financial literacy questions, showing proportions of households answering correct,

incorrect or don't know to each of the advanced literacy questions.

Advanced literacy question Correct (%)

Incorrect (%)

Don't know (%)

Which of the following statements describes the main

function of the stock market? 73.1 12.8 13.6

Which of the following statements is correct? If somebody

buys the stock of firm B in the stock market: 64.5 25.1 9.7

Which of the following statements is correct? 74.4 9.1 15.7

Which of the following statements is correct? If somebody

buys a bond of firm B in the stock market: 61.6 17.2 20.6

If the interest rate falls, what should happen to bond prices? 29.5 42.6 27.2

Buying a company stock usually provides a safer return than a

stock mutual fund. True or false? 58.2 20.4 20.9

Stocks are normally riskier than bonds. True or false? 66.6 13.1 19.8

Considering a long time period (for example 10 or 20 years),

which asset normally gives the highest return? 54.1 27.4 18.0

Normally, which asset displays the highest fluctuations over

time? 74.7 11.0 13.8

When an investor spreads his money among different assets, does the risk of losing money increase, decrease or stay the same?

71.0 13.8 14.9

If you buy a 10-year bond, it means you cannot sell it after 5

years without incurring a major penalty. True or false? 39.4 29.0 29.2

Note: Percentages may not add up to 100% due to 'refusal' answers. The full question set with all possible answers can be found in the Appendix

(14)

14 3.2.3 Parental influence

Parental influence was measured using six questions of the DHS survey which concerned parental influence in the youth of the respondents. Questions 1, 2 and 3 about youth are about the age of 8-12, and questions 4 and 6 are about age 12-16, whereas question 5 does not specify an age at all. The exact wording of these questions can be found in the Table 3. We inverted the scores of question 3, in order to create consistency among the questions. With the inversion for question 3, if the score on a question increases, there was less

parental help in that particular financial area. When we performed a factor analysis on these questions, we find that we receive a factor where the uniqueness of question 1 is 0, meaning that the underlying factor has exactly the same variance as question 1. Although

theoretically possible, it is very unlikely here that the parental influence variable that we are looking for here is so strongly similar to only question 1. When we perform the KMO-test, we find that the value is 0.55, so it is not appropriate to conduct factor analysis on the youth questions.Excluding Q5 and Q6 (with the lowest KMO-scores) does not significantly raise the KMO-score of our variables. Also excluding Q2 and Q4, as Bucciol and Veronesi (2013)

propose, does not increase our KMO-score significantly. Furthermore, the Cronbach alpha of our 6 questions about youth is 0.58, indicating that the internal consistency of this scale is not very high.

The results of our exploratory factor analysis show that the questions about the youth of the respondents do not measure 1 latent variable. However, it is possible that all questions measure different aspects of parental influence. This seems in line with how the questions are posed: If question 2 asks about age 8-12 and question 4 asks about age 12-16, it makes sense that they don’t measure the same latent variable. However, it is possible that they both fall under influence as a whole.

Table 4: Descriptive statistics for parental influence questions

Question Mean Scale Std. Dev.

Youth 1: When you were between 8 and 12 years of age,

did you receive an allowance from your parents then? 2.31 1-4 1.30

Youth 2: When you were between 8 and 12 years of age, did you do little household chores (like washing the car) for which you received some money from your parents?

3.54 1-5 1.38

Youth 3: When you were between 8 and 12 years of age,

could you spend your money as you pleased? 3.33 1-5 1.33

Youth 4: Did you have a job on the side (like a newspaper round, a job on Saturday etc.) when you were between 12 and 16 years of age?

(15)

15

Youth 5: Did your (grand)parents try to teach you how to

budget? 2.55 1-4 1.06

Youth 6: Did your (grand)parents stimulate you to save

money between the age of 12 and 16? 2.33 1-4 1.05

In order to compose 1 variable from the youth questions, we added all the scores of the questions and derive a parental influence variable by grouping the respondents in 4 PFA groups, with a quarter of the respondents in each group. We construct two different variables: One using all questions, and one which excludes question 2 and question 4,

following the argumentation used in Bucciol and Veronesi (2013). The lowest scoring quartile (quartile 1), has had the most parental influence and should be the most likely to participate in the stock market. When we consider the stock market participation for the different quartiles, we do find some evidence following our theory, most notably in quartile 4.

However, the other variables do not show much evidence for our theoretical construct. The two differently constructed variables only differ slightly. There is not much evidence in favor of one variable over the other, so we tried both variables in my regression.

Table 5: Stock market participation for different parental influence indices (quartiles)

% of sample % stock market participation

PFA Score, all questions

First quartile (had most advice from parents) 20.9 33.8

Fourth quartile (had least advice from parents) 27.7 23.6

PFA score, excluding Q2 and Q4

First quartile (had most advice from parents) 17.0 32.3

Fourth quartile (had least advice from parents) 25.9 22.2

3.2.4 Control variables

In the regression we also account for a large range of control variables, in order to increase to explanatory power of our regression and to find unbiased coefficients for our

(16)

16 the stock market then women. We account for this effect by creating a dummy variable for gender.

Social status is also used as a control variable, as Hong et al. (2004) show that if participation among peers is higher, stock market participation from this household will also be more likely. Where they use church attendance among other variables to determine social status, we will use social class. In the DHS survey, social class was self-reported between 1 (highest social class) and 5 (lowest social class). We account for this by creating a dummy variable for each of the 5 classes. There might be some bias involved with self-reporting social status, which we should keep in mind when interpreting our results. This theory is supported by our descriptives in Table 2, which shows that only 1.3% of our respondents place themselves in the lowest social class. We also account for wealth, as several papers (Campbell, 2006; Guiso et al., 2003), show a relation between wealth levels and stock market participation. Van Rooij et al. (2011) define wealth in their research as the following: ‘’Wealth is the sum of checking and savings accounts, employer-sponsored saving plans, cash value of life

insurance, home equity, additional real estate, and other financial assets minus total debt.’’ Stock and mutual funds wealth is not included, as stock market participation is already the dependent variable.

Furthermore, we also add a general education dummy. As Hilgerth and Hogarth (2002) have shown, lower lever of education is associated with a lower level of financial literacy. In the DHS survey, six answers are possible in regards to which education is followed: Primary education, Preparatory intermediate vocational, Intermediate vocational, Secondary pre-university, Higher vocational and University. In order to account for education level, we also create a dummy for each possible level of education in order to account for this effect in our regression. We also account for marital status and net household income by means of

dummies, which is in line with other research in the field of stock market participation (Hong et al., 2008; van Rooij et al., 2011; Kumar 2009). Also, an occupation dummy was also added, since self-employed persons are exposed to higher risk in the labor market and may

(17)

17

4. Methodology

The dependent variable, stock market participation, is a binary variable since it only has two possible answers, either yes or no. It can only take the value 0 or 1. To test our model whether the combination of parental influence and financial literacy has a predictive significance towards stock market participation, we use the following model.

𝑆𝑡𝑜𝑐𝑘 𝑚𝑎𝑟𝑘𝑒𝑡 𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑡𝑖𝑜𝑛 = 𝛼 + 𝛽1∗ 𝐹𝐿 + 𝛽2∗ 𝑃𝐼 + 𝛽3 ∗ 𝑃𝐹𝐴 ∗ 𝐹𝐿 + 𝛽4∗ 𝐶𝑉 + 𝜀𝑖

In the model, βx are the regression coefficients, FL is the independent variable financial

literacy, PI is the parental influence variable, CV stands for the various control variables, and 𝜀_𝑖 stands for the error term in the model.

Another regression than OLS which can account better for omitted variable bias instead of OLS-regressions should be used, a meta-analysis done by Fernandes et al. (2014) on 87 studies concerning financial literacy and financial education showed that the effects of those studies would be much less significant if it would account for omitted variable bias. We will try to account for this problem by adding a wide range of demographic control variables. However, we will also keep in mind that we won’t add too much control variables, which could lead to unnecessary variable bias.

An important problem with regressions concerning the effect of financial literacy on stock market participation is the endogeneity that might exist between the variables. This

endogeneity between financial literacy and stock market participation might lead to biased coefficients in our model. This bias could be in both directions: On one hand, unobservable characteristics not included in the model might lead to an underestimating of our coefficient of the relation between financial literacy and stock market participation. On the other hand, learning-by-doing on the stock market will lead to an overestimation of our relationship. Van Rooij (2011) tries to eliminate this effect by taking financial experiences of the oldest sibling of the respondent using a GMM approach. We take another approach: By using a two-stage least square (2SLS) regression we try to eliminate the existing endogeneity problem. In order to do so, we need a variable that interacts with both the independent variable (stock market participation) as well as the financial literacy. We try several variables for this; Yoong (2011) already showed using the financial experience of the sibling as an instrumental variable, as done in Van Rooij (2011), did not meet requirements for a proper instrumental variable. Besides the preferred instrumental variable in Yoong (2011), not knowing about bond pricing, we will test two other instrumental variables: The degree to which respondents think about their retirement, and the degree of budgeting of respondents.

(18)

18 We have our original OLS estimation:

𝑆𝑀𝑃 = 𝛼 + 𝛽₁∗ 𝐹𝐿 + 𝛽₂∗ 𝑃𝐼 + 𝛽₃∗ 𝑃𝐼 ∗ 𝐹𝐿 + 𝛽₄∗ 𝐶𝑉 + 𝜀_𝑖 Now, in order to combat our endogeneity problem, we first regress our instrumental variable, which is retirement planning, budgeting or not knowing about bond pricing depending on the model in combination with the availability of high school financial education, along with all the other variables on our advanced financial literacy variable: First stage regression (1): 𝐹𝐿̂ = 𝑦0+ 𝑦1∗ 𝐼𝑉 + 𝑦2∗ 𝑃𝐼 + 𝑦3∗ 𝑃𝐼 ∗ 𝐹𝐿 + 𝑦4∗ 𝐶𝑉 + 𝜀𝑦

However, since an interaction effect is also added in our model between parental influence and financial literacy, we have to account for potential endogeneity problems with this variables as well:

First stage regression (2): 𝐹𝐿 ∗ 𝑃𝐼̂ = 𝑦0 + 𝑦1∗ 𝐼𝑉 + 𝑦2∗ 𝐹𝐿 + 𝑦3∗ 𝑃𝐼 + 𝑦4∗ 𝐶𝑉 + 𝜀𝑦

In these models, IV is the instrumental variable set used, FL-hat is the substitute variable for financial literacy adjusted for endogeneity, and FL*PI-hat is the substitute variable of our interaction term adjusted for endogeneity. A more detailed explanation of how 2SLS works can be found in Appendix J.

Then, in the second stage of our regression we can plug in the fitted values for FL in our original estimation, in order to find the correct coefficients for all variables:

𝑆𝑀𝑃 = 𝛼 + 𝛽1∗ 𝐹𝐿̂ + 𝛽2∗ 𝑃𝐼 + 𝛽3 ∗ 𝑃𝐼 ∗ 𝐹𝐿̂ + 𝛽4∗ 𝐶𝑉 + 𝜀𝑦

Using this 2SLS method will help us produce more accurate results and show more realistic results than a normal OLS-estimation. Furthermore, to account for any heteroskedasticity issues in our regression, we use White standard errors in our regressions.

(19)

19

5. Results

5.1 Instrumental variable analysis

First we will test for our instrumental variables, in order to find our preferred instrumental variable for our 2SLS-regression. In all our tested models, we combine the proposed

instrumental variable with the self-reported degree of which high school financial education was available. Respondents were able to score availability of financial education in high-school on a scale of 1 (no attention given to financial education) to 4 (a lot of attention given to financial education). Our first proposed instrument is budgeting, which is self-reported by respondents how often they set budget goals for themselves, on a scale of 1 (always) to 4 (never). Our second proposed instrument is retirement consideration, which was also self-reported by respondents how much they plan ahead for their retirement on a scale from 1 (a lot of planning) to 4 (close to no planning). Our final proposed instrument is not knowing about bond pricing, which is a dummy variable constructed where respondents score a 1 if they answered ‘don’t know’ to the bond pricing question in our advanced financial literacy index (see Appendix A for wording), and a 0 otherwise.

Furthermore, since an interaction effect between parental influence and financial literacy is included in our model, we also have to account for the endogeneity that is present in this variable by constructing proper instruments for it. We do so by making adding instrumental variables that multiply parental influence with the already existing instrument set. In total we obtain a set of 4 instrumental variables for each testable model: High school financial education, high school financial education * parental influence, tested instrument and tested instrument * parental influence. We check the validity of our instrumental variable by checking for relevance and exogeneity. In order to check for exogeneity, we report the Sargan test for overidentifying restrictions. The null hypothesis of the test is that the additional instrument is uncorrelated with the error term, and that the excluded instrumental variable is excluded correctly. Relevance is accounted for by reporting the Craigg-Donald F-statistic for weak instruments. The F-statistics is checked against the critical values of this test provided by Stock & Yogo (2002), which varies with the number of

(20)

20 Table 6: Testing for appropriateness of instrumental variables

2SLS 2SLS 2SLS

IV(1) IV(2) IV(3)

Advanced financial literacy 0.283 0.502 0.242**

(0.262) (0.343) (0.117)

Advanced financial literacy * Parental influence -0.092 -0.288 -0.085

(0.140) (0.198) (0.068)

Craigg-Donald F-statistic (weak instruments) 5.79 1.98 57.13

Sargan J-statistic (overidentification) 4.96 2.62 0.36

p-value Sargan test 0.084 0.270 0.834

R-squared 0.214 0.139 0.217

N 334 334 334

Note: Standard errors of variables are in parentheses. All specifications include a constant and a wide range of control variables. Upper threshold for p-values: *** 0.01, **0.05, * 0.10. Instrumental variables used are a combination of high school financial availability and IV(1): Budgeting, IV(2) retirement consideration, IV(3) not knowing about bond pricing. Interaction terms of the

aforementioned IV's with parental influence are also included in the set of instrumental variables.

The results of Table 6 show that not knowing about bond pricing is our preferred

instrumental variable. In the first column, we can see that our budgeting IV does not pass the weak instrument test and we cannot reject the null hypothesis of the Sargan test, meaning that we cannot assume that our budgeting IV is exogenous. In the second column, we find that our retirement consideration IV does pass the Sargan test, but fails the weak instrument test. In the third and final column, we can observe that the bond pricing IV does pass both tests, and can therefore be used as instrumental variable in our 2SLS-regression. The bias of the improper instruments is reflected in the higher coefficients and higher standard errors in Column 1 and 2 compared to Column 3.

5.2 Results of regression models

In the table below we can find the results for our regressions. We end up with 334

(21)

21 heteroskedasticity in our model, we use White standard errors to correct for

heteroskedasticity.

Table 7: Regression results of the effect of parental influence and financial literacy on stock market

participation

Model I Model II Model III Model IV

Estimation method OLS OLS OLS 2SLS

Advanced financial literacy 0.153*** 0.244*** 0.190*** 0.242**

(0.016) (0.051) (0.054) (0.117)

Parental influence -0.035 -0.035 -0.048 -0.047

(0.045) (0.044) (0.049) (0.051)

Parental influence * Advanced financial

literacy -0.056* -0.052 -0.085

(0.032) (0.034) (0.068)

Control variables No No Yes Yes

Pseudo R-squared 0.115 0.117 0.155 0.153

p-value Durbin Hausman Wu test

(endogeneity) 0.783

N 334 334 334 334

Note: Control variables include controls for gender, age, income, wealth, education, daily use of economics, marital status, occupation, basic financial literacy score and social status. Upper threshold for p-values: *** 0.01, ** 0.05, *0.10. The instrumental variables used in model IV is a set

of instruments consisting of availability of high school financial education, not knowing about bond pricing, high school financial education * parental influence and not knowing about bond pricing *

parental influence.

In the first model we perform an OLS regression on our advanced financial literacy and parental influence variable. We find that advanced financial literacy has a positive significant effect on stock market participation, which is in line with our expectations. In this model, a 1 point increase in our advanced financial literacy index (which ranges from -2.8 to 0.8), leads to 15% more stock market participation. The coefficient for parental influence is negative and insignificant. A negative coefficient is expected because of the construction of our parental influence variable, since the value of the parental influence variable is lower when parental influence was higher. Therefore, more parental influence leads to higher stock market participation here. Since parental influence is either high or low, moving from the low to high group would lead to 3.5% more stock market participation.

(22)

22 high parental influence now leads to a 3.5%-5.6%*advanced financial literacy increase in stock market participation.

In the third model, we add all of our control variables in the model. Even though a large set of demographics and other controls is added, the coefficients of our variables of interest only change slightly. The effect of advanced financial literacy becomes slightly smaller and the effect of parental influence becomes slightly bigger.

In our fourth model, we perform our novel 2SLS-regression with our preferred instrument set. As a result of the 2SLS-regression, our standard errors increase for all variables. This is common for 2SLS-regressions where the sample is not large. However, the coefficients of both our interaction effect and of the advanced financial literacy variable increase in size as well. Since the coefficients of our suspected endogenous variables increase, this points to an underestimation of the coefficients in the OLS-estimation. In order to check whether there is actually endogeneity present in the advanced financial literacy variable and the interaction effect, we report a Durbin-Hausman-Wu test. The p-value of the test came back as 0.78, meaning that the null hypothesis of exogeneity cannot be rejected. The p-value is sufficiently high that we assume that the true coefficient difference between the OLS estimation and the 2SLS estimation is small enough that we can treat our suspected endogenous variables as exogenous. Therefore, we will use our OLS-estimation in Model III to interpret our results moving forward.

The full results of Model III, including all control variables coefficients and standard errors, are listed below in Table 8.

Table 8: Final OLS-regression (Model III) of describing the effect of advanced financial literacy and

parental influence on stock market participation.

Advanced financial literacy 0.192*** (0.055)

Parental influence -0.048 (0.049)

Parental influence* Advanced financial literacy index -0.052 (0.034)

Education (Base group: University)

Primary -0.023 (0.161)

Preparatory intermediate vocational -0.214* (0.118)

Intermediate vocational -0.121 (0.107)

Secondary pre-university -0.137 (0.101)

Higher vocational -0.070 (0.098)

Wealth (Base group: Lowest wealth quartile)

Second quartile -0.066 (0.064)

Third quartile 0.059 (0.068)

Fourth quartile (richest) 0.098 (0.066)

Marital status 0.057 (0.051)

Age (Base group: 20 - 30 years old)

(23)

23 Age 40-50 0.093 (0.102) Age 50-60 -0.033 (0.100) Age 60-70 0.131 (0.127) Age 70-80 0.166 (0.160) Gender 0.071 (0.056)

Occupation dummy (Base group: Employee)

Other -0.022 (0.070)

Retired -0.045 (0.107)

Self-employed -0.265* (0.137)

Log net household income 0.072* (0.038)

Social status -0.004 (0.031)

Daily use of economics 0.001 (0.002)

Basic financial literacy score 0.018 (0.029)

Observations 334

R-squared 0.155

Notes: Control variables include controls for gender, age, income, wealth, education, daily use of economics, marital status, occupation, basic financial literacy score and social status. Upper threshold for p-values: *** 0.01 , ** 0.05, *0.10

When we take a look at the signs of the coefficients of the control variables, we find that these are often in line with the theory discussed in Section 3. All education coefficients are negative, meaning that when the level of education is below university, likeliness of stock market participation also goes down. With our wealth coefficients in Model III, the second wealth quartile is negative but the third and fourth quartiles are positive, meaning that likeliness to participate in the stock market goes up if a household possesses more wealth. Our marital status dummy also has a positive coefficient, which makes sense since married people have more stability. Our age coefficients are all positive, which is partially in line with Agarwal et al. (2009). It was expected that the size of coefficients would decline for the highest age dummies, but this is not the case. Other research by Pirinsky (2013) does find similar results to our regression, that older people are more likely to participate in the stock market. A possible explanation is that wealthier people live longer than poorer people, which leads to a biased sample towards wealthy old people.

(24)

24 by a combination of wealth and education. When we consider the variance inflation factor (VIF) of social status, we find that this value indeed points to a high multicollinearity for social status (see Appendix G). Daily use of economics appears to be insignificant in our model, since the coefficient is only 0.001. Basic financial literacy score has a positive coefficient, which is in line with van Rooij et al. (2011). This indicates that as basic financial literacy goes up, stock market participation is more likely.

As can be seen in Table 6, we did not find a lot of significant variables in our model. However, for many of our dummies the Wald test shows that we cannot reject the hypothesis that all dummy coefficients of a control variable are zero (see Appendix H). Therefore, we keep the control variables in our model.

Furthermore, we find that multicollinearity might affect the significance of our variables of interest in this model. When we test for correlation between age and parental influence, we find that as age goes up, the degree of parental influence decreases. This could potentially lead to lower significance of our variables. We performed a robustness check considering interaction and age and parental influence, where we found that adding in interaction effects between these variables was not beneficial for our model, since it lowered

coefficients and reduced overall model strength. The full robustness check can be found in Appendix I.

It is also important to mention that, considering the large amount of dummy variables in our regression, our sample size is fairly small. Especially some dummies rely on a sample below 30 observations, which might lead to higher standard errors, thus causing lower levels of significance in our estimated models.

6. Conclusion

Van Rooij et al. (2011) found in their research that financial literacy had a significant effect on stock market participation. Similar to their findings, in our model we find that financial literacy has a significant effect on stock market participation, even when accounting for a large range of control variables. Yoong (2011) found that performing a 2SLS-regression to account for endogeneity within the financial literacy and stock market participation was not necessary, using a U.S. survey. In our regression, we found that their findings are also valid for the Netherlands. This indicates that we could treat both our financial literacy and our interaction variable as exogenous.

(25)

25 market participation. In our final model we found that although the coefficients of these factors did have the expected sign, but no significant effect for direct effects and only a significant effect for indirect effects when no control variables were added in our model. We can therefore not draw a conclusion whether the inclusion of parental influence as an

explanatory factor for stock market participation is beneficial or not.

A potential reason for not finding significant results could be the degree of multicollinearity in the model. For example, we find that parental influence and age are correlated (see Appendix), which can cause a decrease in significance of our variable.

Another thing that should be kept in mind when considering the results of our regression, is that the sample that we obtained has a significant bigger proportion participating in the stock market than the total sample of the DHS survey. Although this does not directly lead to biased results, it can’t be said with certainty that this wasn’t the case.

Moreover, it is possible that our results suffer from omitted variable bias. For example, financially illiterate persons might mitigate their illiteracy by using financial planners to help them with their financial matters. These planners might advice financially illiterate persons to participate in the stock market, which could lead to an underestimating of coefficients of financial literacy in our regression. We also were not able to include trust in stock markets, even though Guiso et al. (2008) have shown that this is significant towards stock market participation.

What we also have to consider is the sample size of our regression: Although a sample of 334 is generally sufficient, a 2SLS-regression needs a bigger sample when you have a large amount of dummy variables. This might have led to larger standard errors in both our OLS- and 2SLS-regression than they would be with a larger sample.

What also should be kept in mind when interpreting the results of our regression is that our dependent variable, stock market participation, is a binary variable. Therefore, a probit or logit model could lead to more accurate results, since using such a model is more

appropriate when dealing with a dichotomous dependent variable.

There are several matters in the research that invite further research: Several of the

(26)

26 different questions are asked for example. What also could be researched is the relation between other types of behavior and their relation to financial literacy, to expand upon the current literature. Although our findings did not give strong evidence of parental influence affecting financial literacy, other types of financial behavior might be.

References

Agarwal, S., Driscoll, J. C., Gabaix, X., & Laibson, D. (2009). The age of reason: Financial decisions over the life cycle and implications for regulation. Brookings Papers on Economic Activity, 2009(2), 51-117.

Almenberg, J., & Dreber, A. (2015). Gender, stock market participation and financial literacy. Economics Letters, 137, 140-142.

Bartlett, F. C. (1932). Remembering: An experimental and social study. Cambridge: Cambridge University.

Blalock, L. L., Tiller, V. R., & Monroe, P. A. (2004). “They get you out of courage:” Persistent deep poverty among former welfare-reliant women. Family Relations, 53, 127-137.

Bauer, J. W., Braun, B., & Olson, P. D. (2000). Welfare to well-being framework for research, education, and outreach. The Journal of Consumer Affairs, 34(1), 62-81.

Bernheim, B. D., & Garrett, D. M. (2003). The effects of financial education in the workplace: evidence from a survey of households. Journal of public Economics, 87(7-8), 1487-1519. Bucciol, A., & Veronesi, M. (2014). Teaching children to save: What is the best strategy for lifetime savings? Journal of Economic Psychology, 45, 1-17.

Campbell, J. Y. (2006). Household finance. The Journal of Finance, 61(4), 1553-1604. Chang, L., & Krosnick, J. A. (2009). National surveys via RDD telephone interviewing versus the Internet: Comparing sample representativeness and response quality. Public Opinion Quarterly, 73(4), 641-678.

Choi, J. J., Laibson, D., Madrian, B. C., & Metrick, A. (2002). Defined contribution pensions: Plan rules, participant choices, and the path of least resistance. Tax policy and the

economy, 16, 67-113.

(27)

27

Clarke, M. C., Heaton, M. B., Israelsen, C. L., & Eggett, D. L. (2005). The acquisition of family financial roles and responsibilities. Family and Consumer Sciences Research Journal, 33(4), 321-340.

Cocco, J. F., Gomes, F. J., & Maenhout, P. J. (2005). Consumption and portfolio choice over the life cycle. The Review of Financial Studies, 18(2), 491-533.

Danes, S. M., & Hira, T. K. (1987). Money management knowledge of college students. Journal of Student Financial Aid, 17(1), 1.

Duffee, G. R. (1996). Idiosyncratic variation of Treasury bill yields. The Journal of Finance, 51(2), 527-551.

Fernandes, D., Lynch Jr, J. G., & Netemeyer, R. G. (2014). Financial literacy, financial education, and downstream financial behaviors. Management Science, 60(8), 1861-1883. Grinblatt, M., Keloharju, M., & Linnainmaa, J. (2011). IQ and stock market participation. The Journal of Finance, 66(6), 2121-2164.

Guiso, L., & Jappelli, T. (2005). Awareness and stock market participation. Review of Finance, 9(4), 537-567.

Guiso, L., Sapienza, P., & Zingales, L. (2003). People's opium? Religion and economic attitudes. Journal of monetary economics, 50(1), 225-282.

Guiso, L., Sapienza, P., & Zingales, L. (2008). Trusting the stock market. The Journal of Finance, 63(6), 2557-2600.

Haliassos, M., & Bertaut, C. C. (1995). Why do so few hold stocks?. The Economic Journal, 1110-1129.

Hastings, J. S., Madrian, B. C., & Skimmyhorn, W. L. (2013). Financial literacy, financial education, and economic outcomes.

Heaton, J., & Lucas, D. (2000). Portfolio choice in the presence of background risk. The Economic Journal, 110(460), 1-26.

Hong, H., Kubik, J. D., & Stein, J. C. (2004). Social interaction and stock‐market participation. The journal of finance, 59(1), 137-163.

(28)

28 Huston, A. C., Miller, C., Richburg-Hayes, L., Duncan G. J., Eldred, C. A., Weisner, T. S., et al. (2003). New hope for families and children: Five-year results of a program to reduce poverty and reform welfare.

Jorgensen, B. L. (2007). Financial literacy of college students: Parental and peer influences (Doctoral dissertation, Virginia Tech).

Jorgensen, B. L., & Savla, J. (2010). Financial literacy of young adults: The importance of parental socialization. Family relations, 59(4), 465-478.

Kaiser, H. F. (1974). An index of factorial simplicity. Psychometrika, 39(1), 31-36.

Kramer, M. M. (2016). Financial literacy, confidence and financial advice seeking. Journal of Economic Behavior & Organization, 131, 198-217.

Kumar, A. (2009). Who gambles in the stock market?. The Journal of Finance, 64(4), 1889-1933.

Lusardi, A., Mitchell, O. S., & Curto, V. (2010). Financial literacy among the young. Journal of Consumer Affairs, 44(2), 358-380.

Mandell, L., & Klein, L. (2009). The impact of financial literacy education on subsequent financial behavior.

Mankiw, N. G., & Zeldes, S. P. (1991). The consumption of stockholders and nonstockholders. Journal of financial Economics, 29(1), 97-112.

Moskowitz, T. J., & Vissing-Jørgensen, A. (2002). The returns to entrepreneurial investment: A private equity premium puzzle?. American Economic Review, 92(4), 745-778.

Rettig, K. D., & Mortenson, M. (1986). Household production of financial management competence. Human resources research, 1887-1987.

Perry, V.G. (2008). Is ignorance bliss? Consumer accuracy in judgments about credit ratings. The Journal of Consumer Affairs, 42(2), 189-205.

Pirinsky, C. (2013). Confidence and economic attitudes. Journal of Economic Behavior & Organization, 91, 139-158.

Shim, S., Barber, B. L., Card, N. A., Xiao, J. J., & Serido, J. (2010). Financial socialization of first-year college students: The roles of parents, work, and education. Journal of Youth and Adolescence, 39(12), 1457-1470.

(29)

29

Van Rooij, M., Lusardi, A., & Alessie, R. (2011). Financial literacy and stock market participation. Journal of Financial Economics, 101(2), 449-472.

Vissing-Jorgensen, A. (2004). Perspectives on Behavioural Finance: Does Irrationality Disappear with Wealth.

Perry, V.G. (2008). Is ignorance bliss? Consumer accuracy in judgments about credit ratings. The Journal of Consumer Affairs, 42(2), 189-205

Walstad, W. B., Rebeck, K., & MacDonald, R. A. (2010). The effects of financial education on the financial knowledge of high school students. Journal of Consumer Affairs, 44(2), 336-357. Ward, S. (1974). Consumer socialization. Journal of consumer research, 1(2), 1-14.

Xia, T., Wang, Z., & Li, K. (2014). Financial literacy overconfidence and stock market participation. Social Indicators Research, 119(3), 1233-1245.

Xu, L., & Zia, B. (2012). Financial literacy around the world: an overview of the evidence with practical suggestions for the way forward.

(30)

30

Appendix

Appendix A

Financial literacy questions as posed in the financial literacy appendix for the DHS Survey, designed by van Rooij et al. (2011)

Basic financial literacy questions

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?

i. More than €200 ii. Exactly €200 iii. Less than €200 iv. Do not know v. Refusal

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?

(31)

31 Advanced financial literacy questions

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

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

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

5. If the interest rate falls, what should happen to bond prices? i. Rise

ii. Fall

iii. Stay the same iv. None of the above v. Do not know vi. Refusal

6. 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

7. Stocks are normally riskier than bonds. True or false? i. True

ii. False

8. Considering a long time period (for example 10 or 20 years), which asset normally gives the highest return? i. Savings accounts

(32)

32

iii. Stocks iv Do not know v. Refusal

9. Normally, which asset displays the highest fluctuations over time? i. Savings accounts

ii. Bonds iii. Stocks iv Do not know v. Refusal

10. 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

11. 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

Appendix B

Factor analysis for all financial literacy questions, N=392.

Appendix table 1: Maximum likelihood factor analysis of all financial literacy questions in

the DHS survey.

Cutoff for factor selection: Minimum eigenvalue = 1.3 , factors have been rotated using a oblique varimax rotation.

Question No. Loading towards Factor 1 Loading towards Factor 2

Question 1 -0.17 0.92 Question 2 0.08 0.42 Question 3 0.41 0.29 Question 4 0.35 0.14 Question 5 0.15 0.15 Question 6 0.57 0.10 Question 7 0.45 0.10 Question 8 0.55 0.25 Question 9 0.58 0.12 Question 10 0.40 0.02 Question 11 0.53 0.03 Question 12 0.62 0.08 Question 13 0.56 0.08 Question 14 0.65 0.09 Question 15 0.54 0.19 Question 16 0.49 0.06

(33)

33 We use an oblique rotation since this allows for correlated factors, as seen in the table the correlation between the 2 factors is r=0.32. This makes sense because we expect basic financial literacy and advanced financial literacy to be correlated with each other.

Appendix C

Factor analysis of advanced literacy questions. In the table below all factor loadings for the advanced literacy questions and their ‘don’t know’ dummies are listed.

Appendix table 2: Maximum likelihood factor analysis of all advanced literacy

questions. Cutoff for factor selection: Minimum eigenvalue = 2.

Factor Factor loading

Advanced literacy question 1 0.58

Don't know -0.76

Don't know -0.65

Don't know -0.78

Don't know -0.69

Don't know -0.75

Don't know -0.77

Don't know -0.81

Don't know -0.75

Don't know -0.56

Appendix D

(34)

34 Appendix table 3: Factor scores, calculated using Bartlett's method and using

exact coefficients.

Factor Factor score

Don't know -0.10

Don't know -0.06

Don't know -0.10

Don't know -0.07

Don't know -0.09

Don't know -0.10

Don't know -0.13

Don't know -0.09

Don't know -0.04

Appendix E

Robustness check for follow-up questions: Factor analysis with and without follow up questions.

(35)

35

Appendix table 4: Maximum likelihood factor analysis of all advanced literacy questions. Cutoff for factor

selection: Minimum eigenvalue = 2.

Factor Factor loading, without follow up

questions

Factor loading, with follow up questions

Advanced literacy question 1 0.58 0.58

Don't know -0.76 -0.77

Don't know -0.65 -0.65

Don't know -0.78 -0.78

Don't know -0.69 -0.70

Don't know -0.75 -0.76

Don't know -0.77 -0.76

Don't know -0.81 -0.80

Don't know -0.75 -0.74

Don't know -0.56 -0.56

Total factor variance explained 9.26 8.91

Appendix F

Robustness check for the parental influence variable

In the table below the comparison between the two indices of parental influence can be found. The results following the specification of Model III (OLS-regression, including interaction effect between parental influence and financial literacy). Using all financial literacy questions reduces the parental influence coefficient and gives an interaction effect coefficient which is insignificant. We see this as support for our theory that the parental influence variable needs to be properly defined and that a revised questions set for parental influence should be introduced.

Appendix Table 5: Regression results of the effect of parental influence and financial literacy on stock

(36)

36

Model III (excluding Q2 and Q4 of parental influence questions)

Model III (including all parental influence

questions)

Estimation method OLS OLS

Advanced financial literacy 0.190*** 0.107**

(0.054) (0.053)

Parental influence -0.048 -0.038

(0.049) (0.050)

Parental influence * Advanced financial literacy -0.052 0.002

(0.034) (0.032)

Control variables Yes Yes

Pseudo R-squared 0.155 0.153

N 334 334

Note: Control variables include controls for gender, age, income, wealth, education, daily use of economics, marital status, occupation, basic financial literacy score and social status. Upper threshold for p-values: *** 0.01, ** 0.05, *0.10.

Appendix G

Variance inflation factors for Model III

In the table below, we can see the variance inflation factors for our final regression (Model III). We see that advanced financial literacy and the interaction effect have high VIF’s, this is logical because the same variables are used in computing these variables. We can see that multicollinearity is present in our model, although no variables have VIF’s that are extremely problematic. High VIF’s of dummy variable that measure a categorical variable (like

education) are not a problem, since they are all measure of the underlying variable (education).

Appendix Table 6: Centered variance inflation factors in Model III

(OLS-regression, interaction effect included, control variables included)

Variable Centrered VIF

Advanced financial literacy 14.01

Parental influence 1.820

Parental influence* Advanced financial literacy index 13.79

Education (Base group: University)

Primary 2.343

Preparatory intermediate vocational 7.489

Intermediate vocational 2.499

Secondary pre-university 3.292

Higher vocational 3.473

(37)

37

Second quartile 1.695

Third quartile 1.547

Fourth quartile (richest) 1.580

Marital status 1.354

Age (Base group: 20 - 30 years old)

Age 30-40 2.863 Age 40-50 3.402 Age 50-60 4.595 Age 60-70 4.734 Age 70-80 2.557 Gender 1.639

Occupation dummy (Base group: Employee)

Other 1.966

Retired 3.220

Self-employed 1.315

Log net household income 1.938

Social status 3.067

Daily use of economics 1.241

Basic financial literacy score 1.369

Appendix H

Wald test, null hypothesis is that the control variable has a coefficient of zero. When multiple dummy variables are grouped (for example with education), the null hypothesis is that all of these variables have a coefficient of zero. We find that only one control variable has a high enough p-value to possibly exclude it from the model (social status). However, this was not done because coefficients of other variables where practically unaffected by the exclusion of this variable.

Appendix Table 7: Wald tests of control variables (dummy variables are

grouped based on their underlying factors)

Model III Model IV

Control variable p-value Wald test p-value Wald test

Basic financial literacy score 0.538 0.558

(38)

38

Social status 0.896 0.864

Daily use of economics 0.597 0.598

Appendix I

Robustness checks for parental influence and age

We check the results of a correlogram between parental influence and age as continuous variable to get a general idea of the effect between the two variables:

Appendix Table 8: Correlogram between age and

parental influence

Parental influence Age

Parental influence 1 0.283

Age 0.283 1

We find a positive correlation between the two variables. This is in line with what we expected, since this indicates that if age increases, respondents were less likely to have had high parental influence.

To check the effects this correlation in our regression, we add interaction terms between parental influence and the dummy variables for age. The results are in the table below. We find that besides lowering coefficients of parental influence, the adjusted R-squared of the model goes down by adding the interaction effects between parental influence and age. Therefore, we do not use these interaction effects in our model.

Appendix Table 9 : Regression results of the effect of parental influence and financial literacy on

stock market participation

Model III (no interaction effects between parental influence and

age)

Model III (including interaction effects between parental influence and age)

Estimation method OLS OLS

Advanced financial literacy 0.190** 0.184**

(0.054) (0.056)

Parental influence -0.048 0.007

(0.049) (0.191)

Parental influence * Advanced

financial literacy -0.052 -0.043

(0.034) (0.034)

Control variables Yes Yes

Parental influence, financial literacy and stock market participation for households