Economic literacy and stock market participation:
An international comparison on the basis of country- specific social capital
Aron Joustra S2075571
University of Groningen Faculty of Economics & Business
MSC. International Financial Management
February 2017
Supervisor: dr. M. M. Kramer
Abstract
Traditional stock market participation research primarily focuses on individual characteristics that determine stock market participation. Later research has focused more on country- specific characteristics that aim at the disentanglement of traditionally assumed factors. This research applies the same rhetoric and discusses whether the traditionally positive influence of economic literacy on stock market participation is dependent on country- specific level of social capital. Social capital is measured using the country- average level of both trust and sociability. The findings of this research show that the relationship between economic literacy is dependent on country- specific social capital to the extent that the explanatory power of economic literacy weakens when country- specific social capital increases.
TABLE OF CONTENTS
I. Introduction... . 3
II. Literature review... 6
II. I Economic literacy and stock market participation... . 6
II. II Social capital and stock market participation... 9
III. Research design... 12
III.I Stock market participation ... 12
III. II Economic literacy... 13
III. III Social capital... 14
III. IV Control variables... 15
III. V Regression specification... 16
III. VI Sample splits... 17
IV. Data... 19
IV. I Sample... 19
IV. II Stock market participation... 20
IV. III Economic literacy... 22
IV. IV Social capital... 23
IV. V Control variables... 23
IV. VI. Sample validity...24
V. Results... 25
V. I Binary logistic regression requirements... 25
V. II Hypothesis 1: Binary logistic regression results... 26
V. III Hypothesis 2: Binary logistic regression results... 28
VI. Conclusion... 31
VI. I Conclusion of research... 31
VI. II Research limitations... 32
VI. III Future research and implications... 34
I. INTRODUCTION
Taking into consideration that the expected return on risky assets has consistently exceeded the return on risk- free assets, one might beg the question why investors are so reluctant to participate on the stock market. Therefore not surprising, stock market participation research has been a hot topic of interest in recent years. Creating additional insights in the behavioral determinants of stock market participation will potentially result in improved models regarding financial decision- making. As such, it can be of aid in the efforts to promote effective financial decision- making by households (Campbell, 2006).
Current literature brings forward two predominantly occurring phenomena. Firstly, stock market participation shows great heterogeneity across countries and secondly stock market participation rates are considered to be low. These low stock market participation rates are more commonly referred to as the stockholding puzzle. Figure 1 reports on the distribution of stock market participation across countries, derived from SHARE. The survey of Health, Ageing and Retirement in Europe (SHARE) is a multidisciplinary, cross- national panel database of which the latest data release (WAVE 5) is used for this research. 1
The average stock market participation equals only 15 per cent with countries like Sweden, Denmark and Switzerland showing participation rates exceeding 30 per cent, while more than halve of the countries show participation rates not exceeding 10 per cent. 2
FIGURE 1: Stock market participation rates across SHARE, WAVE 5 countries
Traditional research has primarily focused on individual characteristics that are assumed to determine the probability of stock market participation. An example of such an individual characteristic is the level of numeracy, often used as a proxy for economic literacy (Jappelli, 2010). According to Christelis, Jappelli and Padula (2010) those
1 Latest data release (WAVE 5) is conducted in 2013 and released in 2016
2The average stock market participation rate is based exclusively on the countries present in SHARE: WAVE 5. 0,0 0,1 0,2 0,3 0,4 0,5
AU BE CZ SW GE DE ES SP FR IT IS LU NL SE SL AV.
individuals with higher levels of numeracy are more prone to stock market participation. According to Mehra and Prescott (1985) individuals that are active on the stock market are able to accumulate more wealth. Additionally, according to Cocco, Gomes and Maenhout (2005) the welfare loss from not- participating is of such magnitude that it can result in almost 2 per cent of annual consumption.
In later research, the focus has shifted towards country- specific characteristics to shed light on the puzzle that surrounds stock market participation. Incorporating country- specific elements aims at disentangling traditionally assumed factors that influence the probability of stock market participation. For example, the traditionally assumed positive influence of wealth on the probability of stock market participation is questioned by Georgarakos and Pasini (2011) showing that investors on the lower halve of the wealth distribution in countries like Sweden and Denmark show participation rates twice as high as investors situated on the upper- halve of the wealth distribution in Austria, Spain and Italy. These differences in stock market participation rates are related to country- specific differences of both trust and sociability. Those countries having higher levels of both country- average trust and sociability are associated with higher stock market participation rates. Applying trust and sociability as country- specific characteristics implicates the incorporation of social capital. Traditional social capital research involves norms of trust, norms of reciprocity and the configuration and intensity of a network. With reference to both Georgarakos and Pasini (2011), and Christelis et al. (2010) this research aims at assessing whether country- specific levels of social capital also affect the relationship between economic literacy and stock market participation, resulting in the following:
Research question:
“To what extent is the traditionally assumed relationship between economic literacy and stock market participation dependent on country- specific levels of social capital?”
knowledge and skills effectively to manage resources for a long- term financial well- being (PACFL, 2008). To accurately measure economic literacy it is often proxied by the numeracy construct. Despite the fact that the numeracy construct measures only basic economic concepts, various authors have brought the strong and significant positive relationship between economic literacy and the numeracy construct forward (e.g. McArdle, Smith and Willis, 2009; Jappelli, 2010). The SHARE: WAVE 5 allows for such approximation of economic literacy and is therefore used in this research.
II. LITERATURE REVIEW
Despite a significant increase in stock market participation due to new technologies and an increasing amount of financial products (Thomas and Spataro, 2015) stock market participation rates across countries are still considered low and the heterogeneity it is accompanied by is as tenacious as ever before. The following section elaborates on the influence of both economic literacy and social capital on stock market participation.
II. I Economic literacy and stock market participation
Those investors making the conscious decision to participate on the stock market face a variety of costs. These costs are often unavoidable and investors are assumed to participate on the stock market when the expected return from a given investment exceeds these costs. Broadly spoken, these costs can be divided into entry and participation costs and although the amount of costs is impartial, the relativity of these costs is dependent on a variety of factors. Firstly, the relativity of these costs is based on the amount of resources readily available to the investor. According to Guiso, Haliassos and Jappelli (2003) the decision whether or not to invest in the stock market is dependent on a wealth- threshold. Those that exceed this explicit threshold are expected to participate on the stock market and those that do not exceed this wealth threshold are expected not to participate. Congruent with this hypothesis Alan (2006) shows that individuals that participate on the stock market are most likely to be situated on the “upper- halve” of the wealth- distribution. Those investors situated on the “lower- halve” of the wealth distribution face a higher relativity of these aforementioned costs and consequently face a lesser probability of stock market participation being worthwhile. Secondly, the relativity of the costs that accompany stock market participation is dependent on the cognitive ability of the investor. Stock market participation requires a significant investment in terms of both time and effort in order to get familiarized with the concepts related to the stock market. This familiarization process is of less significance for those investors having a higher level of cognitive ability (Christelis et al. 2010).3
Additionally, risk tolerance and the ability to process financial data are positively associated with cognitive ability, therewith increasing the probability of stock market participation (Christelis et al. 2010).
3 Christelis et al. (2010) investigate the distinct effects of three different elements of cognitive abilities, namely:
Cognitive ability is considered to be a coalescence of elements, of which the most important are fluency, numeracy and recall- skills. Numeracy, in literature often used as a proxy for economic literacy, shows to be an important determinant for the probability of stock market participation (Christelis et al. 2010). Investors with greater numerical ability are more probable of stock market participation, which is shown in figure 2. Using the numeracy construct (Dewey and Prince, 2005) allows to explicitly categorize individuals on the basis of their numerical abilities, with possible values range from 1 to 5. Figure 2 shows that individuals scoring a 3 or lower on the numeracy construct, have a probability of stock market participation somewhere in the range of 5 to 10 per cent. Those individuals exceeding a score of 3 show a stock market participation probability somewhere in the range of 10 to 30 per cent. The specific measurement of the numeracy construct is provided in section III.II.
FIGURE 2: Christelis et al. (2010) Numeracy and stockholding
Note: Figure 2 is derived from Christelis et al. (2010) showing exclusively the effect of direct stockholding across different values of numeracy. The effect of total stockholding, also depicted in Christelis et al. (201) is not relevant to this research and therefore not depicted in figure 2.
The usage of the numeracy construct as a proxy for economic literacy, despite measuring only basic economic concepts, is widely accepted in current finance literature. According to Delavande, Rohwedder, and Willis (2008) numeracy is strong and significantly related to financial- test scoring and this significance is dedicated to the fact that individuals with greater numerical skills are better adept to complex decision- making processes, e.g. financial decision- making. Jappelli (2010) confirms this line of reasoning by showing that the numeracy construct and the level of economic literacy have a significant positive correlation (0.79). It bears emphasis that the relationship between economic literacy and
the direct holding of stocks is more substantial than the relationship between economic literacy and other financial securities such as bonds. This is due to the fact that stocks are considered more “information- intensive” securities (Christelis et al. 2010). This difference is amplified in recent years due to the fact that deregulation and technological innovations have significantly increased the complexity of financial products (Jappelli, 2010). Given the abovementioned I propose the following first hypothesis:
H1: Ceteris paribus, economic literacy has a positive relationship with stock market
participation.
Much alike the distribution of stock market participation, the level of economic literacy is considered low and to show great heterogeneity across both individuals and countries (Jappelli, 2010). According to van Rooij, Lusardi, and Alessie (2011) a large fraction of the population knows very little about basic economic and financial concepts such as risk diversification, inflation and compounded interest. Jappelli (2010) presents an overview of the distribution of economic literacy across countries, of which the depiction in figure 3 is limited to the countries that are present in SHARE: WAVE 5.
FIGURE 3: Jappelli (2010) Economic literacy across countries
Note: Figure 3 shows the distribution of economic literacy across countries, derived from Jappelli (2010). The countries depicted in figure 3 are exclusively the countries present in the SHARE: WAVE 5 database used in this research. Other countries depicted in Jappelli (2010) are not included in figure 3.
3 4 5 6 7
AU BE CZ SW GE DE ES SP FR IT IS LU NL SE SL
According to Merton (1987) the choice of which stocks to purchase, in the absence of transaction costs, is solely dependent on which stocks are known to the investor. Guiso et al. (2003) have extended this notion by showing that if investors would be conscious of all risky assets, the stock market participation rate would increase significantly, potentially even double.
II. II Social capital and stock market participation
accordingly. As aforementioned, the relationship between stock market participation and elements of social capital in current research is also focused on the intensity of networks. The intensity of a network, more commonly referred to as structural social capital is explored by Hong, Kubik and Stein (2004) showing that the intensity of a network is of significant importance in determining the probability of stock market participation. 4
This is mainly due to the fact that information sharing through word- of- mouth has a diminishing effect on informational costs that accompany stock market participation. This implicates a higher propensity towards stock market participation for those social- households as opposed to non- social households. Countries that portray higher average sociability levels are associated with higher country- average stock market participation rates. Evidently, this effect is the strongest in countries where stock market participation rates are higher due to a higher probability of having someone in the network that is also active on the stock market. These findings are, to some extent, congruent with social capital research (Nolan et al. 2008) showing that people base their decision- making largely based on the doings of others. The influence of sociability is of such magnitude that it is assumed to potentially offset the discouragement effect of country- specific low average levels of trust (Georgarakos and Pasini, 2011).
The incorporation of social capital into a behavioral finance decision- making model aims at disentangling factors that are traditionally assumed to influence the stock market participation probability. An example of such an attempted disentanglement is provided by Georgarakos and Pasini (2011) showing that the traditionally assumed strong and significant relationship between wealth and stock market participation is not as straight forward as previously assumed. Rather, country- specific levels of both trust and sociability affect the relationship between wealth and stock market participation. Georgarakos and Pasini (2011) show that investors situated on the lower- halve of the wealth distribution residing in countries that have high average trust and sociability levels, such as Denmark, Sweden and Switzerland, show participation rates twice as high as investors that are situated on the upper- halve of the wealth distribution residing in countries with low country average trust and sociability levels such as Austria, Spain and Italy.
4 Hong et al. (2004) measure social interaction is through whether a household frequently interacts with
This research therefore incorporates a second hypothesis regarding the relationship between economic literacy and stock market participation and its dependency on country- specific levels of social capital. In other words: the magnitude of the relationship between economic literacy and stock market participation is compared on the basis of country- specific levels of social capital. Returning to the aforementioned results by Georgarakos and Pasini (2011) I propose the following: the relationship between economic literacy and stock market participation is of less significance in countries where social capital is greater. Vice versa, the relationship between economic literacy and stock market participation is of greater significance in countries with lower levels of social capital.
The informational constraint is of paramount importance in explaining the relationship between economic literacy and stock market participation. However, this informational constraint is potentially of differing significance resulting from differing country- specific social capital. This potential moderating effect of country- specific social capital on the relationship between economic literacy and stock market participation has not yet been established in current literature. Considering all aforementioned, hypothesis 2 is formulated as follows:
H2: The relationship between economic literacy and stock market participation weakens
when country- specific social capital increases.
III. RESEARCH DESIGN
The SHARE: WAVE 5 database, henceforth denoted as SHARE, is used in this research of which more specific details are provided in section IV. Using SHARE allows establishing an international comparative framework regarding economic literacy and stock market participation, and to what extent country- specific levels of social capital moderate this relationship. This section will provide description and measurement of the variables applied in this research.
III.I Stock market participation
SHARE provides various ways to determine the stock market participation rate across countries. Choosing the right measurement is solely dependent on the adequate distinction of stockholding, opposed to holding other financial securities such as bonds. This distinction is of paramount importance since, as aforementioned, stocks are considered to be more information- intensive securities (Christelis et al. 2010) and the holding of stocks is therefore assumed to have the strongest relationship with economic literacy. The following stock market participation measurements are present in SHARE:
1. “Do you have stocks”? (Yes/ no)
2. “Are the mutual funds you own, mostly stocks or bonds?” 3. “What is the amount of stocks currently held in stocks”?
With reference to Georgarakos and Pasini (2011) the second measure would potentially offer the most adequate distinction of stockholding. However, due to the framing of the answers in SHARE, this vital distinction is actually impaired. 5
Consequently, the second measure is not used for analysis. The stock market participation rates resulting from the first measure is greater than the stock market participation rates resulting from the third measure. I make the assumption that this is due to the fact that the third measure has a much higher degree of “intrusiveness” as opposed to the first measure. Simply stating whether you own stocks or not, is arguably less intrusive than providing the actual amount currently held in stocks. Given that this research is not concerned with the actual amount of money currently held in stocks, but is exclusively concerned with whether or not someone is active on the stock market, the first measure is applied in analysis. However, to account for the possibility of
differing results when using the first or third measure, the results of both measures are subjected to correlation analysis, showing a high (.987) and significant correlation. The possibility of the results being significantly different using the first or third measure is therewith assumed to be negligible.
Using the first measure: “Do you have stocks” implicates that stock market participation becomes a dichotomous variable, best suited for binary logistic regression analysis. Consequently, the assigned values in the measurement of stock market participation are 1 if the household reports having stocks, and 0 if the household reports not having stocks. With reference to hypothesis 1, stock market participation is denoted as 𝑆𝑀𝑃! in which i represents
the individual respondent and is denoted as:
𝑆𝑀𝑃!=
1 𝑖𝑓 𝑦𝑒𝑠
0 𝑖𝑓 𝑛𝑜 and
𝑣
! ~ ! (!,!)With reference to hypothesis 2 stock market participation is denoted as 𝑆𝑀𝑃!" where i
represent the individual respondent and c represents the specific country of residence.
𝑆𝑀𝑃!"=
1 𝑖𝑓 𝑦𝑒𝑠
0 𝑖𝑓 𝑛𝑜 and
𝑣
! ~ ! (!,!)III. II Economic literacy
The measurement of economic literacy is established through the usage of the numeracy construct as a proxy. The numeracy construct consists of the following 4 questions:
1. If the chance of getting a disease were 10 per cent, how many people out of one thousand would be expected to get the disease?
2. In a sale, a shop is selling all items at half a price. Before the sale a sofa costs €300,- How much will the sofa cost in the sale?
4. You currently have €2,000, - in a savings account. The account earns ten per cent interest each year. How much would you have in the account at the end of two years?
The score the respondent receives is a function of the amount of correctly answered questions. If a respondent answers question 1 correctly he/she is asked question 3. If they answer correctly again, he/she is asked question 4. Answering the first question correctly results in a score of 3, answering question 3 correctly results in a score of 4 and also answering question 4 correctly results in a score of 5. If the respondent answers question 1 incorrect, he/she is required to answer question 2. If they answer question 2 correctly he/ she receives a score of 2, while incorrectly answering both question 1 and 2 results in a score of 1 (Dewey and Prince, 2005).
Using the numeracy construct as a proxy for economic literacy, henceforth 𝐸𝐶𝐿!, results in the following, where irepresents the individual respondent.
𝐸𝐶𝐿! = 1 2 3 4 5
III. III Social capital
As was touched upon in the literature review, both trust and sociability are considered elements of social capital. As such, the country- specific level of social capital, henceforth 𝑆𝑂𝐶!, is derived from the country- average level of both trust and sociability. Both trust and sociability are given equal weighting in the determination of 𝑆𝑂𝐶! where c is the specific
country. 𝑆𝑂𝐶! is assigned to each respondent in each corresponding country. For example: every respondent in country K is assigned with the corresponding value of 𝑆𝑂𝐶!.
activities: voluntary or charity work, educational or training course; a sport, social or other kind of club; activities organized by political or community organization.
Given that country- average levels of trust and sociability are derived using different scales, I multiply the country- average level of sociability by 10, allowing both trust and sociability to have the same scaling.
III. IV Control variables
To determine the effect 𝐸𝐶𝐿! and 𝑆𝑂𝐶! on 𝑆𝑀𝑃!" the following demographic variables are controlled for: age, gender, health, depression, risk aversion and income, denoted as: 𝐴𝐺𝐸!, 𝐺𝐸𝑁!, 𝐻𝐸𝐴!, 𝐷𝐸𝑃!, 𝑅𝐼𝐴!, 𝐼𝑁𝐶!. These variables are assumed to either directly or indirectly influence 𝑆𝑀𝑃!, and are accordingly controlled for.
𝐴𝐺𝐸! is measured as years of age, of which the smallest value of age is 50 resulting from the fact that SHARE only includes respondents aged 50+. 𝐺𝐸𝑁! is measured as 1 for male respondents, and 0 for female respondents. 𝐻𝐸𝐴! is measured as 1 for respondents reporting “fair” or “poor” health, and 0 for respondents reporting “good” health. 𝐷𝐸𝑃! is measured as 1 when the respondent reports depression and 0 for respondents reporting no depression. 𝑅𝐼𝐴! is measured as 1 for respondents willing to take more than average risk with the prospect of higher expected average returns, and 0 for respondents not willing to take more than average risk. 𝐼𝑁𝐶! is measured as the income of an self- reported average month in the preceding year.
(2015) show that gender differences in financial literacy can explain a large part of the gender gap in stock market participation. In other words: men are assumed to have higher levels of financial literacy and are therefore more likely to participate on the stock market. As such, gender can be denoted as a both direct and indirect influence the stock market participation. With regards to age, according to Christelis et al. (2010) age is negatively associated with numerical abilities. In other words, when age increases the general level of numeracy decreases and therewith the probability of stock market participation. It must however be noted that the results regarding age denoted by Christelis et al. (2010) is only found to be statistically significant when considering the “total participation” on financial markets, and not statistically significant when considering direct stock market participation6
. Finally, wealth is assumed to have an influence on stock market participation. According to (Briggs et
al. 2015) the general conclusion of available literature, regarding the relationship between
wealth and stock market participation is that changes in wealth are associated with a higher probability of stock market participation. In this research the measure of wealth is proxied by self- reported income. It must be emphasized that income is not necessarily an equivalent of wealth but using SHARE it most adequately, without harming the sample, approximates wealth.7
Income is measured through the self- reported income of an average month of the preceding year.
III. V. Regression specification
Multiple regression analyses are performed to determine whether hypothesis 1 and hypothesis 2 hold true. With reference to hypothesis 1, the regression analysis is performed to establish the potential positive and significant relationship between economic literacy and stock market participation. As such hypothesis 1 aims at the confirmation of the results provided by e.g. Christelis et al. (2010) showing that numerical abilities are positively related to stock market participation. Abovementioned results in the following regression equation:
𝑆𝑀𝑃! = 𝛽1𝐸𝐶𝐿! + 𝛽2𝐺𝐸𝑁! + 𝛽3𝐴𝐺𝐸! + 𝛽4𝑅𝐼𝐴! + 𝛽5𝐷𝐸𝑃!+ 𝛽6𝐻𝐸𝐴! + 𝛽7𝐼𝑁𝐶! + 𝜀! (1)
6 The difference denoted by Christelis et al. (2010) regarding direct and total stockholding is that total
stockholding includes the holding of equity mutual funds and investment accounts.
7 SHARE also allows approximation of wealth, through the amount available for bequest. However, using this
Regression equation 1 does not incorporate 𝑆𝑂𝐶! since it is merely concerned with whether 𝐸𝐶𝐿! is positively related to 𝑆𝑀𝑃!, irrespective of the residential country of the investor. An error term is included, denoted as 𝜀!. The measurement of 𝑆𝑀𝑃!, 𝐸𝐶𝐿! and the control variables is denoted in sections III.I, III.II and III.IV, respectively.
With reference to hypothesis 2, whether the relationship between 𝐸𝐶𝐿! and 𝑆𝑀𝑃!" is moderated by 𝑆𝑂𝐶!, the regression equation includes an interaction term of 𝐸𝐶𝐿! with 𝑆𝑂𝐶! to establish a potential significant interaction. An error term is included, denoted as 𝜀!. Abovementioned results in the following regression equation:
𝑆𝑀𝑃!" = 𝛽1𝐸𝐶𝐿!+ 𝛽2𝑆𝑂𝐶! + 𝛽3(𝐸𝐶𝐿! ∗ 𝑆𝑂𝐶!) + 𝛽4𝐺𝐸𝑁!+ 𝛽5𝐴𝐺𝐸! + 𝛽6𝑅𝐼𝐴!+ (2) 𝛽7𝐷𝐸𝑃!+ 𝛽8𝐻𝐸𝐴! + 𝛽9𝑁𝐶! + 𝜀!
Again, the measurement of the 𝑆𝑀𝑃!", 𝐸𝐶𝐿! and the control variables is denoted in sections III.I, III.II and III.IV, respectively. The addition of 𝑆𝑂𝐶! is denoted in section III.III.
The interpretation of the coefficients resulting from binary logistic regression requires some additional remarks. The change of odds in the dependent variable as a result of a 1-unit increase in the independent variable is denoted as the odds- ratio, or Exp(b). The relationship between 𝛽, the odds ratio and the probability is the following:
𝐸𝑥𝑝 (𝑏) = 𝑒! with, 𝑝 = 𝐸𝑥𝑝 𝑏 1+𝐸𝑥𝑝 (𝑏)
III. VI Sample splits
IV. DATA
IV. I Sample
The survey of Health, Ageing and Retirement in Europe (SHARE) is a multidisciplinary, cross- national panel database of which the latest data release is used for this research. SHARE provides data on respondents that are 50+ years of age and is therefore representative for the corresponding aged population. SHARE provides data on respondents residing in the following 15 countries: Austria, Belgium, Czech Republic, Switzerland, Estonia, Denmark, Germany, France, Italy, Luxembourg, Israel, the Netherlands, Sweden, Slovenia and Spain.
Table 1 reports on the sample selection and shows the number of observations omitted with regards to economic literacy, country- average trust and sociability, and control variables. As was mentioned in section III.III the measurement of 𝑆𝑂𝐶! is based on country- average levels of both trust and sociability, and are therefore explicitly reported on in table 1. Approximately 65 per cent (28,465) of the sample is omitted due to the fact that respondents do not provide data on both stock market participation and the numeracy construct. The total number of valid observations used in analysis equals 11,174 across the 15 countries mentioned above.
TABLE 1: SAMPLE SELECTION
Non- missing values for stock market participation 43,908
Less missing values economic literacy (28,465)
Less missing values trust (884)
Less missing values sociability (757)
Less missing values control variables (2,628)
Total number of observations used in analysis 11,174
Note: Table 1 provides the sample selection procedure in which is denoted how many observations are omitted for economic literacy, country- average trust and sociability, and control variables. Full sample with non- missing observations for stock market participation consists of 43,908 observations. Total number of observations used in analysis equals 11,174.
TABLE 2: SUMMARY STATISTICS ACROSS COUNTRIES
N SMP ECL SOC GEN AGE HEA DEP RIA INC
Austria Mean Standard Deviation 126 0.03 .176 3.63 1.33 5.59 0.00 0.48 0.50 67.3 9.32 0.34 0.47 0.36 0.48 0.04 0.20 5,921 8,973 Belgium Mean Standard Deviation 823 0.16 0.37 3.45 1.07 5.16 0.00 0.44 0.50 64.5 10.35 0.28 0.45 0.43 0.50 0.04 0.19 6,887 19,984 Czech Republic Mean Standard Deviation 687 0.05 0.19 3.50 1.04 4.58 0.00 0.34 0.47 66,7 9.38 0.46 0.50 0.48 0.50 0.04 0.19 1,459 3,554 Switzerland Mean Standard Deviation 60 0.30 0.46 3.67 0.97 6.58 0.00 0.50 0.50 64.9 8.47 0.12 0.324 0.35 0.48 0.02 0.14 16,127 23,261 Germany Mean Standard Deviation 2313 0.12 0.32 3.57 1.03 5.22 0.00 0.49 0.50 64.6 10.40 0.40 0.49 0.48 0.50 0.02 0.15 6,063 15,255 Denmark Mean Standard Deviation 1098 0.31 0.46 3.61 1.11 7.29 0.00 0.49 0.50 65.0 10.28 0.21 0.40 0.30 0.46 0.12 0.33 6,281 17,358 Estonia Mean Standard Deviation 134 0.03 0.17 3.44 1.13 4.74 0.00 0.47 0.50 65,6 9.04 0.62 0.49 0.40 0.49 0.06 0.24 1,267 1,915 Spain Mean Standard Deviation 1172 0.04 0.21 2.66 1.04 3.97 0.00 0.48 0.50 67,4 11.41 0.40 0.49 0.33 0.47 0.02 0.14 4,159 12,568 France Mean Standard Deviation 138 0.07 0.24 3.22 1.10 4.28 0.0 0.43 0.49 65,7 10.18 0.32 0.47 0.47 0.50 0.04 0.21 4,903 9,272 Italy Mean Standard Deviation 917 0.04 0.19 3.08 1.18 4.18 0.00 0.45 0.50 66,3 10.33 0.41 0.50 0.41 0.49 0.05 0.18 4,117 9,622 Israel Mean Standard Deviation 120 0.05 0.22 3.88 1.07 4.72 0.00 0.43 0.50 66.8 10.28 0.29 0.33 0.33 0.47 0.07 0.26 3,468 8,163 Luxembourg Mean Standard Deviation 678 0.10 0.30 3.32 1.17 5.30 0.00 0.52 0.50 63.9 9.45 0.34 0.47 0.45 0.49 0.04 0.19 14,775 30,104 The Netherlands Mean Standard Deviation 898 0.10 0.30 3.66 1.14 6.65 0.00 0.50 0.50 66.0 9.91 0.30 0.46 0.31 0.46 0.03 0.17 4,187 9,111 Sweden Mean Standard Deviation 1538 0.39 0.49 3.60 1.00 6.60 0.00 0.50 0.50 67.7 9.66 0.23 0.42 0.31 0.46 0.13 0.33 5,070 10,023 Slovenia Mean Standard Deviation 475 0.07 0.26 3.33 1.06 5.01 0.00 0.40 0.49 67.3 9.80 0.40 0.49 0.45 0.49 0.01 0.10 2,382 4,511 Note: Table 2 provides the sample- summary statistics for each variable, denoting number of observations,
mean and standard deviation, across all countries denoted in SHARE. The total number of observations across all 15 countries is 11,174. The measurement of the variables is denoted in section III.
IV.II Stock market participation
rates are considered to be low and to show great heterogeneity. With regards to the low stock market participation rates, table 2 shows that more than halve of the countries have an average stock market participation rate not exceeding 10 per cent. Taking into consideration that the average stock market participation rate across all countries denoted in SHARE is approximately 15 per cent, the sample used shows that the stock market participation rate across countries is indeed rather low.
With regards to the heterogeneity in stock market participation, table 2 shows that 4 of the 15 countries have average participation rates not exceeding 5 per cent while Sweden, Denmark and Switzerland show participation rates of 30 per cent and higher. Taking into consideration that the standard deviation of country- average stock market participation rates equals approximately 11 per cent, the sample used indeed shows substantial heterogeneity in stock market participation rates across countries. In order to, validate the findings presented in table 2, the country- average participation rates are compared to the findings of Christelis
et al. (2010) and Georgarakos and Pasini (2011). Table 3 provides this comparison and shows
that country- average stock market participation rates are fairly comparable, increasing to some extent the confidence in the sample used.
TABLE 3: STOCK MARKET PARTICIPATION ACROSS COUNTRIES. COMPARISON OF THE SAMPLE USED WITH CURRENT LITERATURE
Sample used in this research (2013)
Christelis et al. (2010)
Georgarakos and Pasini (2011) Sweden 0.39 0.388 0.39 Denmark 0.31 0.332 0.32 Switzerland 0.30 0.262 0.25 Belgium 0.16 0.177 0.16 Germany 0.12 0.172 0.12 The Netherlands 0.10 0.128 0.14
Luxembourg 0.10 N/A N/A
France Slovenia Israel Czech Republic 0.07 0.07 0.05 0.05 0.156 N/A N/A N/A 0.14 N/A N/A N/A Spain 0.04 0.042 0.03 Italy Austria Estonia 0.04 0.03 0.03 0.045 0.055 N/A 0.05 0.06 N/A
IV. III Economic literacy
Spain, Italy and France show the lowest country- average economic literacy, while Israel, Switzerland and the Netherlands show the highest country- average economic literacy. According to Jappelli (2010) the level of economic literacy varies substantially across countries. Despite significant differences between country- average levels of economic literacy for the countries present in SHARE, it must be noted that if economic literacy were considered from a more global perspective the variation would be more substantial. This is due to the fact that the countries denoted in SHARE are neither top nor worst performers on economic literacy from a global perspective (Jappelli, 2010). In order to validate the sample country- average economic literacy findings of this research are compared to the findings of Christelis et al. (2010). Table 2 shows this comparison, and despite the fact that country- averages of economic literacy are fairly comparable, it must be noted that all countries are found to have higher country- average economic literacy in the findings provided by Christelis et al. (2010) as opposed to this research.
TABLE 4: ECONOMIC LITERACY ACROSS COUNTRIES. COMPARISON OF SAMPLE USED WITH CURRENT LITERATURE
Sample used in this research Christelis et al. (2010) Israel Switzerland 3.88 3.67 N/A 4.0 The Netherlands 3.66 3.8 Austria 3.63 3.8 Denmark 3.61 3.7 Sweden 3.59 3.8 Germany 3.57 3.7 Czech Republic Belgium 3.50 3.45 N/A 3.6 Estonia Slovenia Luxembourg France 3.44 3.33 3.32 3.22 N/A N/A N/A 3.3 Italy 3.08 3.1 Spain 2.66 2.7
IV. IV Social capital
Table 2 shows the substantial differences in 𝑆𝑂𝐶! across countries. Denmark and the Netherlands have the highest 𝑆𝑂𝐶! and Italy, France and Spain show the lowest. Table 5 reports on the categorization of countries based on 𝑆𝑂𝐶!, of which the procedures used are denoted in section III. VI. Columns (1) and (2) represent the sample split based on the first procedure, columns (3), (4) and (5) represent the sample split based on the second procedure.
TABLE 5: SAMPLE SPLITS: 𝑺𝑶𝑪𝑪 BASED CATEGORIZATION OF COUNTRIES
(1) (2) (3) (4) (5) Below 𝑆𝑂𝐶 Above 𝑆𝑂𝐶 Low 𝑆𝑂𝐶! Average 𝑆𝑂𝐶! High 𝑆𝑂𝐶! Spain Italy France Czech Republic Israel Estonia Slovenia Belgium Germany Luxembourg Austria Switzerland Sweden The Netherlands Denmark France Italy Spain Belgium Austria Slovenia Luxembourg Estonia Israel Germany Czech Republic Denmark Sweden The Netherlands Switzerland N: 7,457 3,720 2,227 5,353 3,594
Note: Table 3 provides the categorization of countries subsequent to two sample splits. The first sample split is based on above- and below 𝑆𝑂𝐶! denoted in columns (1) and (2), respectively. The second sample
split is based on low- average and high 𝑆𝑂𝐶! denoted in columns (3), (4) and (5) respectively. The
sample- split procedures are denoted in section III. VI.
IV. V Control variables
There are some remarkable findings with regards to the country- average levels of the control variables. Firstly, Denmark and Sweden show country- average 𝑅𝐼𝐴 that is substantially higher than for all other countries. Denmark and Sweden show country- average 𝑅𝐼𝐴 of 0.12 and 0.13, respectively while the mean of 𝑅𝐼𝐴 in all other countries (Denmark and Sweden excluded) is only 0.037. Secondly, the average income across countries shows substantial differences, with some arguably unlikely results. For example, the average income in this sample (per month) in Luxembourg and Switzerland equals 14,775 and 16,127 respectively. According to OECD (2017), Luxembourg and Switzerland are the 2nd
and 3rd
countries regarding yearly disposable income.8
Still, it must be emphasized that it is undoubtedly erroneous that the yearly average income in both countries would exceed 175,000. I ascribe these erroneous results to the sample used.
IV. VI Sample validity
To establish the validity of the sample, two different procedures are used. Firstly, denoted in sections IV.II and IV.III, the country- averages of both stock market participation and economic literacy are compared to the findings of current literature. The results are fairly comparable with regards to both economic literacy and stock market participation, and the differences that are present could be ascribed to a number of reasons. One of these possible reasons is the fact that both Georgarakos and Pasini (2011) and Christelis et al. (2010) base their findings on data originating from 2004, while the findings of this research are based on data originating from 2013.
Secondly, the sample used is compared to the observations that are omitted from SHARE. Given that SHARE is said to be representative for the corresponding aged population, the observations in the sample used, are compared to the mean of the missing observations per each variable. As such, the variables are subjected to One- sample T tests to validate whether the sample significantly differs from the missing observations mean. Column (1) represents the mean and standard deviation of the non- missing sample used, consisting of 11,174 observations. Column (2) represents the mean and standard deviation of the missing observations, of which the number of observations are denoted in section IV.I.
TABLE 6: SAMPLE VALIDITY
(1) (2)
Non- Missing Missing T- test
Mean SD Mean SD
Stock market participation 0.15 0.359 N/A N/A N/A Economic literacy 3.41 1.12 3.46 1.00 0.000 Gender 0.47 0.50 0.46 0.49 0.043 Age 65.90 10.25 66.93 10.20 0.000 Health 0.34 0.47 0.37 0.48 0.000 Depression 0.39 0.49 0.39 0.48 0.006 Risk aversion 0.05 0.23 0.04 0.19 0.000 Income 5,579 14,889 4,792 16,105 0.000
Note: Table 6 provides the comparison between the used sample used the procedure used with regards to omitting variables is denoted in section IV. I.
8 Luxembourg and Switzerland show the 2nd and 3rd highest disposable income for all countries present in
V. RESULTS
Performing binary logistic regression analysis implicates the necessity of the validation of the sample with regards to a variety of requirements. The following section will first address these concerns and subsequently proceed to the binary logistic regression results.
V.I. Binary logistic regression requirements
Current literature brings forward various rules of thumb regarding the sufficiency of the sample size when performing regression analysis. According to traditional logistic regression research each variable should at least contain ten subjects to provide reasonable stable estimates of the regression analysis (Peduzzi et al, 1996). Recent research points towards a minimum of at least twenty subjects to provide reasonable stable estimates. Another important pitfall with regards to binary logistic regression analysis is the correlation between independent variables. If independent variables are too strongly correlated, the estimates of variance can increase and cause the variance to be of implausible size (O’Brien, 2013).
The variation inflation factor and the correlation matrix are both widely- accepted procedures to signal the presence of multicollinearity. With regards to the variation inflation factor (VIF), various rules of thumb are known. According to O’Brien (2007) the strictest rule of thumb allows the VIF to be no more than 4 to safely assume the absence of multicollinearity. The independent variables assessed in this research show variation inflation factors ranging from 1.006 up to 1.089.9
The square root of the variance inflation factor describes the increase of the standard error of an independent variable opposed to if the independent variables were uncorrelated (O’ Brien, 2007). If the variation inflation factor would exceed abovementioned threshold of 4 it would implicate the necessity of that specific variable to be removed from the model. Given that the variation factors for the independent variables present in this model are well below 4, it is unnecessary to remove any variable from the model. With regards to the correlation matrix the strictest rule of thumb allows the correlation between independent variables to be no more than 0.7 to safely assume the absence of multicollinearity. Table 4 provides the correlation matrix of all independent variables. The largest correlation amongst the independent variables is between 𝐴𝐺𝐸! and 𝐻𝐸𝐴! and equals 0.234, allowing for the confirmation that multicollinearity is absent in this research.
9 𝐸𝐶𝐿
! denotes the lowest inflation factor (1.006) and 𝐻𝐸𝐴! denotes the largest inflation factor (1.089) of all
TABLE 7: CORRELATION MATRIX Variable 1 2 3 4 5 6 7 8 1. ECL - 2. SOC -.180 - 3. GEN -.021 .080 ** - 4. AGE -.056 ** -.062 ** -.042 ** - 5. RIA .020 -.004 .076 ** -.062 ** - 6. DEP -.023 * -.053 ** -.160 ** .030 ** -.023 ** - 7. HEA -.029 * .176 ** -.066 ** .234 ** -.050 ** .232 ** - 8. INC .24 .040 ** .003 ** -.010 .003 -.011 -.022 ** -
Note: Table 4 provides the correlation matrix of all variables present in the model. The greatest correlation (𝐻𝐸𝐴!,
𝐴𝐺𝐸!) is denoted in bold. Significance levels are denoted in asterisks of which ** denotes significance at the 0.01 level and * denotes significance at the 0.05 level.
V.II Hypothesis 1: Binary Logistic Regression results
It bears reminding that hypothesis 1 denotes the relationship between economic literacy and stock market participation, not yet incorporating the potential moderating effect of country- specific social capital. The relationship between economic literacy and stock market participation is controlled for using various demographic control variables. The measurement of stock market participation, economic literacy, social capital and all control variables is denoted in section III.I to III. V. For readability purposes all variables will henceforth be written in full.
Column (1), table 5, denoted in section V.III reports on the binary logistic regression results with regards to hypothesis 1. The results show that economic literacy is positive and significantly related to stock market participation, therewith allowing for the confirmation of the results provided by e.g. Christelis et al. (2010) showing that cognitive abilities are positively related to stock market participation.10
The relationship between economic literacy and stock market participation is significant and positive, even when controlling for age, gender, health, depression, risk aversion and income. Using binary logistic regression implicates that Exp(b), the odds- ratio, can be interpreted to determine the effect of a 1- unit increase in the independent variable towards the change of odds in the dependent variable. Economic literacy shows a 𝛽 of 0.170 and the Exp(b) is therefore equal to 1.186. As such, a 1- unit increase of economic literacy is associated with an approximately 19 per cent increase in the odds of stock market participation. It must be emphasized that the Exp(b) of 1.186 ratio
10 The confirmation of the findings of Christelis et al. (2010) exclusively relate to the relationship between
of economic literacy does not explain the change in probability of stock market participation from a 1- unit increase in economic literacy, rather it denotes the increase of odds regarding stock market participation as a result of a 1- unit increase of economic literacy. The relationship between 𝛽, the odds- ratio and the probability is denoted in section III. V.
probable that income has an affect on stock market participation, the regression results show a relationship of 0.000. Unfortunately, it is very likely that this is the result of a measurement error.
V.III. Hypothesis 2: Binary Logistic Regression results
It bears reminding that hypothesis 2 concerns the relationship between economic literacy and stock market participation, and to the potential moderating effect of country- specific level of social capital. Section III. V denotes the regression equation for hypothesis 2 and also elaborates on the procedures used for the sample splits. To determine whether the relationship is significantly different along different levels of social capital, an interaction term is included. The results of the regression analysis, including the interaction term, is denoted in column (2), table 5, showing a weak, yet significant result for the interaction term between country- specific social capital and economic literacy. This implies that the magnitude of the relationship between economic literacy and stock market participation weakens when country- specific social capital increases. In other words: the predictive power of economic literacy as a determinant for stock market participation is significantly weaker in countries with higher social capital as opposed to countries with lower social capital. Another important result is the strong and significant impact of country- specific social capital on stock market participation.
indeed the relationship between economic literacy and stock market participation is dependent on country- specific level of social capital. As aforementioned, Christelis et al. (2010) shows that age is only significantly related to total stockholding and not significant for direct stockholding. I propose that this is in part, due to the fact that SHARE respondents are aged 50+ years of age and the influence of age on stock market participation is therewith arguably diminished. The findings in this research also do not show a significant relationship between age and stock market participation across the whole sample, but does in fact find a negative and significant effect in the 𝑆𝑂𝐶! countries.
TABLE 8:
STOCK MARKET PARTICIPATION, ECONOMIC LITERACY, AND SOCIAL CAPITAL; BINARY LOGISTIC REGRESSION RESULTS
STOCK MARKET PARTICIPATION
(1) (2) (3) (4)
Variables All All Below 𝑆𝑂𝐶! Above 𝑆𝑂𝐶!
Constant -2.618 (.215)*** -6.756 (0.577)*** -2.310 (.333)*** -1.599 (.302)*** ECL 0.170 (0.025)*** 0.346 (0.150)** 0.181 (0.039)*** 0.039 (0.035) ECL*SOC -0.045 (0.025)* SOC 0.810 (0.090)*** GEN 0.363 (0.056)*** 0.374 (0.057)*** 0.498 (0.086)*** 0.285 (0.076)*** AGE (0.003) 0.004 (0.003) 0.000 (0.004)*** -0.12 (0.004) 0.006 RIA 1.441 (0.090)*** 1.172 (0.095)*** 1.206 (0.160)*** 1.173 (0.116)*** DEP -0.100 (0.059) -0.008 (0.061) 0.090 (0.089) -0.036 (0.084) HEA (0.069)*** -0.811 (0.071)*** -0.646 (0.104)*** -0.807 (0.096)*** -0.516 INC 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) Observations 11,180 11,180 7,457 3,723
Cox & Snell R2 0.050 0.092 0.029 0.046
Note: Table 5 provides the binary logistic regression analysis results. Column (1) shows the results binary logistic regression results with regards to hypothesis 1; in which the relationship between economic literacy and stock market participation, using a variety of control variables is denoted. Columns (2), (3) and (4) show the binary logistic regression results with regards to hypothesis 2; in which in column (2) the interaction between ECL and SOC is denoted and columns (3) and (4) represent the first sample split. The measurement of variables, and the procedure used for the sample split is denoted in section III. Asterisks represent the significance level in which * stands for significance at a 0.10 level, ** for a 0.05 significance level and *** for a 0.01 significance level.
second sample split applied. The procedure used in this second sample split is denoted in section III. VI. Columns (1), (2) and (3) represent low, average, and high social capital countries, respectively. Given the aforementioned one would expect that the relationship between economic literacy and stock market participation is the strongest in low- social capital countries, moderate in the average- social capital countries and the weakest in low- social capital countries. Columns (1), (2) and (3) show that indeed the relationship weakens when social capital increases. However, the relationship in both the high- and low- social capital category is not statistically significant. Arguably this could be the result of the sample being, subsequent to the sample split, of insufficient size to establish significant results. With the acknowledgement of the results not being statistically significant, the results do show that the relationship weakens when social capital increases. The odds- ratio for economic literacy is 1.234 in low social capital countries and 1.038 for high- social capital countries. This implies that the odds of stock market participation given a 1- unit increase of economic literacy increase with an approximate factor of 6 when residing in a low- social capital country as opposed to a high- social capital country.
TABLE 9:
STOCK MARKET PARTICIPATION, ECONOMIC LITERACY AND SOCIAL CAPITAL; BINARY LOGISTIC REGRESSION RESULTS
STOCK MARKET PARTICIPATION
(1) (2) (3)
Variables Low Average High
Constant (0.826)*** -2.551 (0.365)*** -1.993 (0.306)*** -1.541 ECL 0.133 (0.097) 0.103 (0.044)** 0.037 (0.035) GEN 0.594 (0.223)*** 0.484 (0.094)*** 0.282 (0.077)*** AGE (0.011) -0.018 (0.005)** -0.010 (0.004) 0.005 RIA 1.613 (0.327)*** 1.156 (0.182)*** 1.145 (0.117)*** DEP (0.239) 0.130 (0.096) 0.012 (0.085) -0.031 HEA (0.291)*** -0.990 (0.111)*** -0.758 (0.097)*** -0.510 INC 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) Observations 2,228 5,355 3,597
Cox & Snell R2 0.027 0.028 0.045
VI. CONCLUSION
VI. I Conclusion of research
Stock market participation research has predominantly brought two occurring phenomena forward. Firstly, stock market participation shows great heterogeneity across both individuals and across countries. Secondly, stock market participation rates are considered to be low, more commonly referred to as the stockholding puzzle. This research attempts to shed some light on the inaudibility surrounding stock market participation. Traditional stock market participation research has focused primarily on individual characteristics to establish the probability of stock market participation. However, more recent research shows inclination towards country- specific characteristics to explain the phenomena surrounding stock market participation. Georgarakos and Pasini (2011) provide such a research, in which they show that the relationship between wealth and stock market participation is dependent on country- specific levels of both trust and sociability. The incorporation of country- specific social capital aims at the disentanglement of traditionally assumed factors in the determination of stock market participation. This research has applied the same rhetoric and has examined the relationship between economic literacy and stock market participation and the moderating effect of country- specific social capital.
The first analysis of this research, regarding hypothesis 1, is merely concerned with the validation of current literature regarding the relationship between economic literacy and stock market participation. The results show that economic literacy is indeed significant and positively related to stock market participation, even when controlling for age, gender, risk aversion, health, depression and income. These specific control variables are included since current literature has established the relatedness of these variables to stock market participation. The findings presented in this research can only confirm the significant relationship of gender, health and risk aversion with stock market participation. Age, depression and income are not found to be significant related to stock market participation.
social capital countries. Also, results show that country- specific social capital is strong and significantly related to stock market participation. All aforementioned does beg the question why country- specific social capital is of such importance in explaining the relationship between economic literacy and stock market participation. With reference to Christelis et al. (2010), Guiso et al. (2008) and Georgarakos and Pasini (2011) I propose the following line of reasoning. Investors base their financial decision- making, in part, on whether or not the expected return of investment exceeds the required costs for that specific investment. However, the expected return on investment needs to be adjusted for a variety of factors. The relativity of this adjustment is of great importance. For example, investors with lesser cognitive abilities require a greater investment in terms of time and effort when investing in the stock market and as such require a more significant adjustment of the expected return. Also, both sociability and trust, the elements of social capital, are determinants in the degree to which the expected return requires to be adjusted. With regards to the former, informational costs are diminished when information- sharing through word- of mouth communication occurs. With regards to the latter, the expected return requires adjustment for the probability of being cheated, implicating that a higher level of trust results in a lesser required adjustment of the expected return. The contribution of this research is that abovementioned does not only hold when considering it from an individual perspective, but also when considering it from a country- level perspective. The relationship between economic literacy and stock market participation shows to be dependent on country- specific level of social capital, in which the explanatory power of economic literacy to determine the odds of stock market participation is weakened when country- specific social capital increases.
VI. II Research limitations
Evidently, this research is not without it’s limitations. This section will discuss a variety of limitations and subsequently address potential areas of future research.
stocks. Additionally I want to bring forward an assumption that potentially amplifies the abovementioned. The current state of the stock market potentially affects the honesty of the respondents regarding the question whether or not they are active on the stock market. When the stock market plummets, as was the case in the recent global financial crisis, one might argue that people are hesitant to admit they own stocks.
Another important caveat in this research is that the usage of SHARE, to some extent, complicates the measurement of the relationship between economic literacy and stock market participation. As has been denoted in section IV. I a large fraction of the sample had to be omitted due to the fact that respondents did not provide data on both stock market participation as well as economic literacy. Despite the fact that SHARE provides an enormously rich context regarding economic and demographic variables, the combination of economic literacy and stock market participation appeared not ideal. Also, given that the decision whether or not to become active on the stock market is arguably dependent on an arguably endless amount of factors, I would have liked to incorporate more control variables. The regression analysis shows low Cox and Snell R2
implicating that the model has little explanatory power.
Additionally, the measurement of some variables requires some additional remarks. With regards to the measurement of stock market participation, I felt that it was of paramount importance to apply the measure that would most adequately distinguish the holding of stocks. The measurement used however, with reference to Georgarakos and Pasini (2011) is a measure that is non- congruent with current literature and might therefore have biased the comparison between country- average levels of stock market participation with current literature. Also, relationship between income and stock market participation shows a 𝛽 𝑎𝑛𝑑 𝑆. 𝐸. of 0.000 in each regression analysis performed. This is undoubtedly the result of a measurement error regarding this control variable.
VI. III Future research and implications
Referring to low stock market participation as the stockholding puzzle implies that behavioral determinants are still not yet fully understood. With reference to the introduction of this research, creating additional insights in the behavioral determinants of stock market participation could potentially be of aid in the promotion of effective financial decision- making of households (Campbell, 2006). As such, future research should extend the path towards understanding more and more behavioral determinants of the stock market participation decision. I propose that the possible avenues for future research with regards to this topic are twofold. On the one hand, future research could dig deeper into country- specific characteristics that moderate traditionally assumed factors that determine stock market participation probability. On the other hand, this research has applied an equal weighting to the country- average levels of both trust and sociability to derive country- specific social capital. Future research could potentially investigate to what extent the distinct effects of both country- average levels of trust and sociability has on the relationship between economic literacy and stock market participation. It must be emphasized that the usage of SHARE allows for a rich international comparative context. Stock market participation research would undoubtedly improve if more countries would be included into SHARE.
VII. REFERENCES
Alan, S. (2006). Entry costs and stock market participation over the life cycle. Review of
Economic Dynamics, 9(4), 588-611.
Almenberg, J., & Dreber, A. (2015). Gender, stock market participation and financial literacy. Economics Letters, 137, 140-142.
Briggs, J., Cesarini, D., Lindqvist, E., & Ostling, R. (2015). Wealth and stock market participation: estimating the causal effect from Swedish lotteries. Work. Pap., New York
Univ.
Campbell, J. Y. (2006). Household finance. The Journal of Finance, 61(4), 1553-1604.
Christelis, D., Jappelli, T., & Padula, M. (2010). Cognitive abilities and portfolio choice.
European Economic Review, 54(1), 18-38.
Cocco, J. F., Gomes, F. J., & Maenhout, P. J. (2005). Consumption and portfolio choice over the life cycle. Review of financial Studies, 18(2), 491-533.
Dewey, M. E., & Prince, M. J. (2005). Health, ageing and retirement in Europe: First results from SHARE. Manheim, Research Institute for the Economics of Ageing, 108-117.
Delavande, A., Rohwedder, S., & Willis, R. J. (2008). Preparation for retirement, financial literacy and cognitive resources.
Gambetta, D. (2000). Can we trust trust. Trust: Making and breaking cooperative
relations, 13, 213-237.
Georgarakos, D., & Pasini, G. (2011). Trust, sociability, and stock market participation.
Review of Finance, 15(4), 693-725.
Guiso, L., Sapienza, P., & Zingales, L. (2004). The role of social capital in financial development. The American Economic Review, 94(3), 526-556.
Guiso, L., Sapienza, P., & Zingales, L. (2008). Trusting the stock market. The Journal of
Finance, 63(6), 2557-2600.
Hong, H., Kubik, J. D., & Stein, J. C. (2004). Social interaction and stock‐market participation. The Journal of Finance, 59(1), 137-163.
Jappelli, T. (2010). Economic literacy: An international comparison. The Economic Journal,
120(548), 429-451.
Kapucu, N. (2011). Social capital and civic engagement. International Journal of Social
Inquiry, 4(1), 23-43.
Mandell, L., & Klein, L. S. (2007). Motivation and financial literacy. Financial Services
Review, 16(2), 105-116.
McArdle, J. J., Smith, J. P., & Willis, R. (2009). Cognition and economic outcomes in the
Health and Retirement Survey (No. w15266). National Bureau of Economic Research.
Mehra, R., & Prescott, E. C. (1985). The equity premium: A puzzle. Journal of monetary
Economics, 15(2), 145-161.
Merton, R. C. (1987). A simple model of capital market equilibrium with incomplete information. The Journal of Finance, 42(3), 483-510.
Nolan, J. M., Schultz, P. W., Cialdini, R. B., Goldstein, N. J., & Griskevicius, V. (2008). Normative social influence is underdetected. Personality and Social Psychology Bulletin,
34(7), 913-923.
