Interest rates on savings accounts close to zero : shift of Dutch individuals towards the stock and bond market?’ : an analysis of the developments and determinants of holding risky financial assets and savings by Dutch

University of Amsterdam (UvA)

Bachelor Thesis (6013B0345W)

BSc Economics & Finance

A.A.2017/2018

June 2018

“Interest rates on savings accounts close to

zero: shift of Dutch individuals towards the

stock and bond market?’’

An analysis of the developments and determinants of

holding risky financial assets and savings by Dutch

individuals

Student: Maarten Tummers 11006021

Supervisor: Robin Döttling

Statement of Originality

This document is written by Maarten Tummers who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

In this thesis an extensive analysis is performed concerning the portfolio choice of Dutch individuals. National account data retrieved from Statistics Netherlands was used as well as 25 years of micro panel data (1993-2017) provided by CentERdata. The main factors of interest are the ratio of risky financial assets and the ratio of savings as part of total financial assets. In this thesis a macro analysis regarding financial asset holding over time was conducted as well as a micro analysis using the panel data to determine the factors influencing risky financial asset holding and savings.

Strong indications were found for a positive relationship between the ratio of risky financial assets and the level of wealth, university education, age and income. Gender (female), a high level of risk aversion and community were found as negative influencers of the ratio of risky financial assets. Furthermore, the results indicated a positive relationship between the ratio of savings and age, gender (female) and a high level of risk aversion. A high level of financial wealth was a negative influencer of the ratio of savings. A higher level of interest rate might have a slight negative effect on the ratio of risky financial assets held by Dutch individuals. This effect, however, remains unobserved when adding the variables related to risk aversion, income uncertainty and community.

Table of contents

Chapter Page(s)

1. Introduction ... 1-2 2. Literature Review ... 3-7 2.1 Factors influencing portfolio decisions ... 3-5 2.2 Interest rates effects ... 5-7 3. Research Method ... 8-13 3.1 Macro level ... 8-10 3.2 Micro level ... 10-13 4. Data Analysis ... 14-20 4.1 Macro level ... 14-16 4.1 Micro level ... 16-20 5. Conclusions ... 21-23 6. Review ... 23-24 7. References ... 25-26 8. Data sources ... 27 Appendix ... I-XV Appendix A: Tables and figures macro level analysis ... I-VII Appendix B: Tables micro level analysis ... VIII-XIII Appendix C: Measurement of variables ... XIV-XV

1. Introduction

In this thesis an extensive analysis will be conducted regarding the portfolio choice of individuals in the Netherlands. The role played by the level of interest rate on this choice is discussed more carefully. The main factors of interest are the amount of risky financial assets held as part of total financial assets and the amount of savings held as part of total financial assets by Dutch individuals. Using national account data provided by CBS, the central statistics institute of the Netherlands, a macro analysis can be performed. Data regarding national accounts is available from 1995 until 2016. Using this data, trends and developments in financial assets accumulation can be analyzed.

Furthermore, survey data is available provided by CentERdata. For the past 25 years, economic data has been collected via the Dutch National Bank Household Survey. This data, ranging from 1993 to 2017, will be used to perform several analyses. First of all, a similar analysis as mentioned earlier using the national account data will be conducted. This way a comparison can be made between the population data and the survey data. Besides using the data for macro purposes, survey data offers a wide variety of other possibilities. Instead of looking at the amount of a certain financial asset, the number of individuals holding that certain asset can be determined. Using the 25 years of available data an overview over time will be created. Subsequently, a more detailed overview is made for the amount of people holding risky financial assets and savings over the years. Finally, a regression on micro level will be performed testing the relevance of several factors suggested by the academic literature. Using the answers of respondents on the questions in the surveys an ordinary least squares (OLS) regression will be performed to test the statistical relevance of each of the variables of interest. Besides, for each year of the survey the appropriate level of interest rate will be added as a variable. By doing so the statistical relevance of the level of interest rate regarding the ratio of risky financial assets and savings can be tested as well. All of this will be done to be able to answer the following research questions:

• RQ1: What are the factors influencing the ratio of risky financial assets and the ratio of savings held by Dutch individuals?

• RQ2: What is the effect of the level of interest rates on the ratio of risky financial assets and the ratio of savings held by Dutch individuals?

Although various factors will be considered, special attention will be paid to the effect of the level interest rate on the portfolio choice of individuals. With interest rates on savings

accounts structurally below one percent since 2016 in the Netherlands (De Nederlandsche Bank 2018), Dutch individuals may be interested in alternative ways to invest their money. Since the period of the dot-com bubble there has not been seen such a large increase in the number of households in the Netherlands holding some form of stocks and/or bonds (Van Rein, 2017). The number of households in the Netherlands holding some form of stocks and/or bonds in August 2017 was 1,385,000, an increase of no less than fourteen percent compared to the year before (Van Rein, 2017). This increasing number of households participating on the stock and bond market might be influenced by the earlier mentioned decline of the interest rates on savings accounts. Meanwhile the number of brokers in the Netherlands is increasing rapidly. These facts could imply Dutch individuals see holding stocks and/or bonds as a serious alternative for holding money on a savings account.

The further structure of this thesis will be as follows: first of all, the relevant academic literature will be provided. This literature will be focused on several topics related to the research question. In this section, the available literature regarding the factors related to the choice of keeping risky financial assets and savings in the portfolio will be supplied first. This literature will provide information regarding factors influencing the chance of participation in certain markets. Furthermore, the effects of the level of interest rate is discussed. Literature with respect to interest elasticity of savings will be summarized. In addition, some indications for the change of the ratio of risky financial assets caused by a change in the level of interest rate is provided. Using the academic literature, the data necessary to answer the research question will be indicated.

In the following section the research design will be discussed thoroughly. The different methods to answer the research question will be provided as well as the reasons behind choosing these particular methods. Given the available data, the different analyses are discussed extensively in the research design section. After collecting the academic literature and constructing a research design it is time to perform an analysis of the data. First the macro analysis based on the national account data will be performed, followed by the macro analyses based on the survey data. Finally, the analysis on the micro level based on the survey data will be conducted.

Based on the results of the above mentioned data analysis, conclusions can be drawn regarding the research question. After stating the conclusions, a review of the performed research will be provided. Some factors that will be discussed are for example the weaknesses of the research as well as the limitations and finally some suggestions for additional survey questions and potential further research are stated.

2. Literature Review

2.1 Factors influencing portfolio decisions

Before discussing the academic literature concerning the effect of the level of interest rates on savings accounts and risky financial assets held by individuals, a substantial amount of useful and interesting literature available on closely related and overlaying topics will be discussed. First of all, it is important to take a close look at factors that potentially influence the behavior of individuals in the Netherlands in relation to holding risky financial assets and savings. Hochguertel, Alessie, and Van Soest (1997) found that the allocation between risk-free assets (mainly savings accounts) and risky assets (stocks and bonds) was determined by the level of financial wealth and the marginal tax rate. A higher amount of financial wealth might reduce the barrier to participate in the stock and bond market created by transaction costs. Besides, people with a higher level of education are more likely to hold significantly more bonds and stocks relative to savings (Hochguertel et al., 1997).

This research adds to the existing literature mentioned above in several ways. The research of Hochguertel et al. (1997) is based on data of 1988. In this research however, 25 years of survey is used from 1993-2017 resulting in a higher number of observations and an analysis based on more recent information. Furthermore, additional variables were able to be included in the micro regression of this research thanks to the broad range of questions asked in the surveys provided by CentERdata.

Financial literacy is another important factor regarding financial decision-making. People with low financial literacy are less likely to invest in stocks (Van Rooij, Lusardi, & Alessie, 2011). People often tend to directly connect the term ‘risk aversion’ to whether or not stocks are held in the portfolio. Haliassos and Bertaut (1995) found however that risk aversion per se does not account for the fact that a high percentage of households in the United States do not hold stocks despite the predicted equity premium. They did however find some support for a departure from the expected utility maximization, short sale constraints, inertia and income risk as possible explanations why so few hold stocks. Furthermore, Barasinska, Schäfer, and Stephan (2012) found that self-declared risk averse households are likely to hold portfolios which are incomplete, consisting mainly of a few risk-free assets. Using a basic investment game Charness and Gneezy (2007) found strong evidence that women are less likely to invest. This could indicate that women are financially more risk averse than men (Charness and Gneezy, 2007).

Instead of using 21 different questions concerning financial literacy like Van Rooij et al. (2011), only 2 questions will be used to estimate financial literacy in this research. These questions are based on self-declared financial knowledge and interest in financial matters. The way risk aversion is measured is however quite similar compared to the method used by Barasinska et al. (2012), based on one question. The research of Haliassos and Bertaut (1995) and the research of Barasinska et al. (2012) was based on data representative for the United States of America and Germany respectively, this research adds to this this existing literature by using Dutch data. Instead of the setting of an investment game like Charness and Gneezy (2007), this research will simply use survey data and estimate the effect of gender based on the actual amounts of risky financial assets and savings held by respondents.

Heaton and Lucas (2000) showed that entrepreneurial income risk is negatively related to the amount of wealth held in stocks. Likewise, holding stocks in the company where a non-entrepreneur works will have a negative effect on the amount of other stocks (not from the own company) held in the portfolio. Income risk might have an opposite effect on the amount of savings. According to Ochmann (2010) the amount of savings of an average household increases by 4.4 percent when transitory income uncertainty is doubled. Brown, Ivković, Smith, and Weisbenner (2008) have argued that the average stock market participation of an individual’s community is causally related to the decision of the individual to hold stocks. This is in line with the theory that stock market participation is influenced by social interaction. Households who go to church or interact with their neighbors, so called ‘social households’, are more likely to participate on the stock market (Hong, Kubik, & Stein, 2004).

In this research a broader term of income risk is considered rather than only entrepreneurial risk like in the research of Heaton and Lucas (2000). The savings decision is modeled as a function of total financial assets instead of a function of current income, the consumption price level and net returns to savings as in Ochmann (2010). Since information regarding social interaction and stock market participation of the respondent’s community is not available in the survey data, the approach of estimating the effect of an individual’s community is different from the researches of Brown et al. (2008) and Hong et al. (2004). This effect is estimated by using a respondent’s main source of financial advice.

According to Constantinides, Donaldson, and Mehra (2002) young consumers would like to hold stocks in their portfolios but because of borrowing constraints they are not able to achieve this. This is in line with the findings of Kreinin (1959). Kreinin found that the frequency of stock ownership is significantly related to age. This corresponds with the the results of King and Leape (1987) who found evidence that for older age groups the asset ownership rates for

stocks, bonds and mutual funds is substantially higher. In addition, Kreinin (1959) discovered a relation between stock ownership and education, liquid assets, occupation and income. Income is not only an influence on stock ownership, saving may be influenced by this factor as well. Harris, Loundes, and Webster (2002) found that saving is determined most importantly by current income. They also found supporting evidence for the level of economic optimism and demographics as determinants of household saving.

In this research panel data is used to analyze the role of borrowing constraints on the portfolio decision. More specifically, the self-declared level of difficulty to borrow money is used. In addition to the researches of Kreinin (1959) and King and Leape (1987), this research provides a more recent analysis on the effect of age as well many other potentially important independent variables on the dependent variables. Similar to the approach of Harris et al. (2002) survey data was used to analyze the effect of current income. In this research a longer time spawn is used (25 years), compared to the 6 years of data used in the research of Harris et al. (2002).

The internet may also play an important role when it comes to stock market participation. According to Bogan (2008) households using the internet participate more often on the stock market compared to households who do not use the internet. This is especially interesting when looking at the number of people 65 years and older. In 2017 in the Netherlands, 88.3 percent of people 65 years or older have access to the internet (CBS, 2018). In 2012, this percentage was significantly lower, just 64.6 percent (CBS, 2018). Combining the theory of Bogan (2008) with these numbers could imply a rise of stock market participation with respect to people 65 years or older.

By combining these findings and conduct the implied further research, a well thought-out answer concerning the research question should be able to be formulated.

2.2 Interest rates effects

According to Elmendorf (1996) the short-run interest elasticity of saving is probably positive. This means, a 1 percent increase in interest rate will lead to an increase in the amount of saving. An important theoretical framework behind this research is called the lifecycle model. The lifecycle model assumes that people are forward looking. Not only current income and current desired spending will determine the saving and consumption decision. Future income and future desired spending will affect their decision in each stage in their lives (Elmendorf 1996). The two basic preferences of consumers are described by Elmendorf as follows: the willingness of consumers to change their consumption between different stages in their lives is called the

‘intertemporal elasticity of substitution’ (Elmendorf 1996). The second preference is called the ‘rate of time preference’, which is basically a measure of patience. There are three effects following from an increase in the level of interest rate according to the lifecycle theory. These effects are: the substitution effect, income effect and the wealth effect (Elmendorf 1996). Substitution effect

If there is an increase in the level of interest rate, current consumption becomes more expensive relative to future consumption. In order to obtain an extra unit of consumption in the future, less current consumption must be sacrificed because the return on saving is higher as a consequence of the higher interest rate (Mankiw, 2013).

Income effect

The income effect has an opposite effect on consumption compared to the substitution effect (Elmendorf, 1996). If the consumer is a net-saver, an increase in the level of interest rate will lead to less expensive planned future consumption. This will lead to more current current consumption and less saving (Elmendorf, 1996).

Wealth effect

The final effect of an increase in the level of interest rate is the decline in the expected present discounted value of future income (Elmendorf, 1996). There will be a decline in the present discounted value of both human wealth (coming from labor and pension earnings) and financial wealth (coming from certain kinds of assets) (Elmendorf, 1996). In a lifetime sense people are worse off resulting in an increase of current saving and a decrease in current consumption (Elmendorf, 1996).

The findings of Elmendorf are in line with the results found earlier by Boskin (1978), who also found a positive interest elasticity of saving. An interest elasticity of saving around 0.3-0.4 was found (Boskin, 1978). On the contrary, some other authors have found results suggesting interest elasticity of saving is in fact not positive. Beznoska and Ochmann (2013) used German consumption data to research interest elasticity of saving. They found a value of uncompensated interest elasticity of around zero.

Different from the researches of Elmendorf (1996), Boskin (1978), and Beznoska and Ochmann (2013) the main goal of this research is not to estimate the level of the interest elasticity of saving. The effect of a substantial amount of other potential influencers on the ratio

of risky financial assets and savings will be analyzed as well. Similar to Beznoska and Ochmann (2013), micro data is used in this research.

Beer, Gnan, and Ritzberger-Grünwald (2016) researched the effect of a low interest environment on the portfolio choice of Austrian households. They found only modest evidence for a shift of portfolio allocation towards riskier assets. An important side note to this finding is the relative low participations of Austrian households in the stock market. Only 5.4 percent of Austrian households held some form of stocks in 2014 (Fessler, Lindner, & Schürz, 2016). In the monthly report of the Deutsche Bundesbank from October 2015 the saving and investment behavior of German households in the low-interest-rate environment was discussed. Although direct share purchases have remained muted since the financial crises of 2008, investment fund shares increased in popularity since 2013 (Deutsche Bundesbank, 2015). Households were particularly interested in buying equity funds.

In 2014, the Bundesbank conducted a household survey in Germany called ‘Panel on Household Finances’ (PHF) where the question was asked whether households were changing their savings behavior in response to the low-interest-rate environment. On average, about 15 percent of the households taking part in the survey have reduced their amount of savings and approximately 7 percent declared to have altered their investment behavior due to the low interest rates (Deutsche Bundesbank, 2015). The percentage of households choosing to change their saving behavior in response to the low-interest-rate environment is dependent on the category of the household. Wealthier households will change their saving behavior more often (Deutsche Bundesbank, 2015). Besides, less risk-averse households as well as households with more diversified portfolios declare more often that the low interest rates caused a change in their saving behavior. Of the very wealthy households, which generally manage a portfolio consisting of a substantial percentage of securities, 14 percent stated to invest their money differently than before in reaction to the low-interest rate environment (Deutsche Bundesbank, 2015).

This research adds to the literature mentioned above by considering a long period of time instead of the short (recent) time period focused on low interest rates considered by Beer et al. (2016). This research adds to the knowledge regarding the influence on interest rates on risky financial asset holding on the longer term. Instead of focusing mainly on a low-interest environment only like the questions in the PHF survey of 2014 (Deutsche Bundesbank, 2015), other interest rate situations are considered as well in this research. Moreover, this research is based on Dutch individuals rather than Austrian and German individuals.

3. Research Method

National Account

The first part of the research will focus on developments in the macro level. In order to get a broad overview of the developments in the last 25 years, national account data will be used. This data will be obtained from ‘CBS’, the national statistics bureau of the Netherlands. Data is available starting from 1995 until 2016. Sometimes, in later years, adjustments are made concerning the data of previous years. When having the choice between the original and the renewed version of the data the most recent version is preferred and therefore chosen. The focus of this analysis will be on the financial balance sheet of the sector households, including non-profit institutions serving households.

Using this data, the composition of financial assets can be compared throughout the years. The composition of total financial assets consists of the following elements: currency; transferable deposits; savings deposits; other deposits; bills and short term bonds; long term bonds; financial derivatives; short term loans; long term loans; shares and other equities; net equity of households in life insurance and pension funds reserves; other insurance technical reserves; other accounts receivable and payable. Elements of special interest are the ratio of risky financial assets compared to total financial assets and the ratio of savings compared to total financial assets. The following elements are part of risky financial assets: short term bonds; long term bonds; financial derivatives; shares and other equities. The savings part of financial assets consists only of savings deposits. This definition is chosen consciously to be able to make a clear distinction between savings and risky financial assets. Throughout the years the composition of the elements regarding financial assets has developed. This can be caused, for instance, by simple unpopularity of the asset. Financial derivatives for example were not included in the calculation of financial assets until 2003.

Survey data

By using survey data, a similar analysis can be performed as stated above using the national account data. This analysis however is certainly not the only tool that can be used to help us in answering the research question. Survey data opens a broad range of possibilities to approach the research question. Now there is not only information available about the total amounts of the several elements of financial assets, but information concerning the number of people holding these assets as well. This way several analyses can be conducted. Data can be retrieved

from CentERdata. This institution collects economic data on a yearly basis. The first year data was collected was in 1993, the most recent version of the survey available is from 2017. The data is categorized in different modules, namely: the Household Information Module, the Work & Pension Module, the Accommodation Module, the Income Module, the Wealth Module and the Psychological Concepts Module. Furthermore, two modules with aggregated values are available called the Aggregated Income module and the Aggregated Wealth Module.

The Aggregated Wealth Module is ideal to perform the macro level analysis. In this module the financial assets of each individual are distributed in several aggregate variables. The first analysis that will be done is used to determine the level of the individual financial assets. A separate variable must be added for risky financial assets as well as a variable for savings as part of financial assets. Where possible, the suggested definitions of asset classes of Alessie, Hochgürtel, and van Soest (2000) are used. Naturally, since their research dates back to the year 2000 some changes have occurred and some modifications of the definitions need to be made over the years. Total financial assets consist of the following elements: transaction and saving accounts; certificates of deposit; bonds; stocks; derivatives; mutual funds and managed investment accounts; defined-contribution plans; cash value life insurance; employer sponsored saving plans; other financial assets. The following elements are used to calculate the total amount of risky financial assets: bonds; stocks; derivatives; mutual funds and managed investment accounts; The savings part of financial assets is determined by the level of saving accounts and saving accounts (Postbank). Since 2001 the element ‘savings arrangements linked to a Postbank account’ is not longer treated as a separate component. This macro analysis based on the survey data can be compared to the previously mentioned macro analysis based on the national accounts. Although the two analyses will not be comparable one-on-one due to differences in definitions, the (slightly) different time spawn (1995-2016 for the national account data, 1993-2017 for the survey data) and the way pension fund reserves are treated, it will be possible to compare certain trends and analyze developments over the years.

The second step is to take a close look at the amount of individual people from the sample who hold the different financial assets instead of looking at aggregate values. This is one of the advantages of using survey data compared to national account data. By calculating these holding percentages it is possible to analyze the trends of the percentage of people holding certain assets over the years. This analysis can again be conducted using the Aggregated Wealth Module. Likewise, a table can be made concerning the choices of individuals regarding risky financial assets and savings. A separate analysis will be made to determine the percentage of people holding: savings, risky financial assets, savings and risky financial assets, savings and

no risky financial assets, risky financial assets and no savings and finally no savings and no risky financial assets.

3.2 Micro level

After looking at developments on the macro level the survey data can be used to make an analysis on the micro level as well. Each observation in the survey data is linked to a household number and the number of the person in that specific household. Using this information, it is possible to create a personal number so that each individual can be identified. The way to calculate this personal number is suggested in the codebook attached to the survey. The calculation has the following structure: personal number= number of household *100 + number of member.

From the literature there are several indications of variables that will possibly influence the dependent variables, namely the ratio of risky financial assets and the ratio of savings as part of financial assets. All these factors will be tested using a regression model. Statistical software called Stata, developed by StataCorp, will be used for the regressions. The regression model that will be used for this research is an OLS regression model. The assumptions behind this model include linearity of the parameters, an expected value of the mean of errors of zero, a random sample of observations, no multi-collinearity between the independent variables and errors should be independently and identically distributed.

The potential influencers of the ratio of risky financial assets and savings held by Dutch individuals can be summarized in the following regression model:

!" = $%+ $'()*+,∗ ./012ℎ + $456789:)+"5;∗ <=>?@AB02C=D + $EF(G5:)+"5;)*789:)+"5;∗ HI/J=B02C=D01?@AB02C=D + $EF(K;"L(FM"+N789:)+"5;∗ HI/ODCP/IQC2R?@AB02C=D + $S(;"5FG5:)+"5;)*789:)+"5;∗ T/DC=IJ=B02C=D01?@AB02C=D + $G5:)+"5;)*U5**(V(M∗

J=B02C=D01W=11/X/Q + $_YV( ∗ ZX/ + $_[()FS9FL(N∗ !/0ITAIP/R + $_\;:5](∗ ^DB=_/ + $_`);∗ a0D + $_\;+(F(M+b)+(∗ ^D2/I/Q2c02/ + $_{456b"MeYL(FM"5;}∗ <=>cCQfZP/IQC=D +

$_{`(8"9]b"MeYL(FM"5;}∗ a/@CA_cCQfZP/IQC=D + $_{\;:5](K;:(F+)";+N}∗ ^DB=_/ODB/I20CD2R + $_U5]]9;"+N∗ W=__ADC2R + $_{456g";);:")*\;+(F(M+} ∗ <=>hCD0DBC01^D2/I/Q2 +

$_{`(8"9]g";);:")*\;+(F(M+}∗ a/@CA_hCD0DBC01^D2/I/Q2 + $_{456g";);:")*i;56*(8V(} ∗

<=>hCD0DBC01jD=>1/@X/ + $_{`(8"9]g";);:")*i;56*(8V(</sub>∗ a/@CA_hCD0DBC01jD=>1/@X/ + $<sub>456k5FF56";VU5;M+F)";+</sub>∗ <=>l=II=>CDXW=DQ2I0CD2 + $<sub>`(8"9]k5FF56";VU5;M+F)";+}∗ a/@CA_l=II=>CDXW=DQ2I0CD2 + m_"

Where !" = c02C= =n cCQfR hCD0DBC01 ZQQ/2Q op c02C= =n T0PCDXQ

The variables regarding education, risk aversion, financial interest, financial knowledge and borrowing constraints are dummy variables used to find the relative effect to university education, high risk aversion, high financial interest, high financial knowledge and high borrowing constraints respectively.

Using the answers to the survey questions the values of the independent variables suggested by the literature can be estimated. The specific terminology of several of these questions can be found in appendix C. Looking at the academic literature it becomes clear a wide variety of factors are potential influencers of the decision of individuals to hold risky financial assets or savings. In the following section the independent variables are provided including their expected influence on the dependent variables and measurement method. The following variables of interest can be obtained from the survey data:

Wealth

In the literature review the importance of wealth in relation to the portfolio decision of individuals became clear. A higher level of (financial) wealth is expected to lead to an increase in the ratio of risky financial assets. An individuals level of wealth (total net worth) can be estimated using the ‘Aggregated Wealth Module’. The level of wealth is determined using the following equation: wealth= total financial assets + total non-financial assets – total debt. Education

According to the findings of Hochguertel et al. (1997) an increase in the level of education is expected to lead to an increase in the amount of risky financial assets held in the portfolio. Since the level of education is a subjective definition the best option is to create separate education categories. Information regarding the level of education can be found in the ‘Household Information Module’. The categories of education are: low education: kindergarten, primary education and special education; pre-vocational education: VMBO; pre-university education: HAVO and VWO; junior/senior vocational training: MBO; vocational colleges: HBO; university education: WO. These categories will be used as dummy variables.

Gender and Age

According to Charness and Gneezy (2007) women are less likely to choose a risky investment than men. The variables ‘man’ and ‘woman’ can be obtained from the ‘Household Information Module’ to test if this statement is indeed correct. Kreinin (1959) and King and Leape (1987)

suggested that age is an influencer of the portfolio choice of individuals. The age of the respondent will be estimated as the year of the survey minus the respondent’s year of birth. Income

The effect of the level of income regarding the ratio of risky financial assets is, based on the literature review, expected to be positive. Information regarding the level of a respondent’s income is available via the ‘Aggregated Wealth Module’. In this module a direct calculation of net total income is provided.

Income uncertainty

Based on the literature review the level of income uncertainty is expected to influence the ratio of risky financial assets as well as the ratio of savings. Income uncertainty is expected to have a negative effect on the ratio of risky financial assets held and a positive effect on the ratio of savings as part of the portfolio. Income uncertainty will be estimated using the respondent’s answer on a question, which is part of the ‘Income Module’. This question concerns the respondent’s expectation about future income. This question was asked from 1995 until 2017 and can be found in appendix C.

Financial literacy

Recent studies have looked at the effect of financial literacy concerning stock market participation. Van Rooij et al. (2011) found that people with low financial literacy are less likely to invest in stocks. The expectation regarding the effect of financial literacy on the dependent variable ‘ratio risky financial assets’ is as follows: a decrease in the level of financial literacy will lead to a decrease in the ratio of risky financial assets. Financial literacy will be estimated using the answers of respondents to two different questions, both part of the ‘Psychological Concepts Module’. The first question is based on self-declared financial literacy. This question was asked in the waves of 2001 until 2017. The second question is based on financial interest of the respondent. Answers to this question were available from 1993 until 2003.

Risk aversion

A higher level of risk aversion is expected to have a positive effect on the ratio of savings and a negative effect on the ratio of risky financial assets as part of total financial assets. The strategy to estimate a respondent’s risk aversion is based on the method of Alessie et al. (2000). They used a simple question that represented a choice between a safe and a risky investment to

create three separate dummy variables regarding the level of risk aversion. This information is available from 1995 until the most recent wave of 2017.

Community

Based on the literature of Brown et al. (2008) and Hong et al. (2014) the ratio of risky financial assets in an individual’s portfolio is expected to be positively influenced by both the average stock market participation of their community and their degree of ‘socialness’. As a proxy for the influence of community the answer to a question concerning the respondent’s main source of financial advice is used. This question was added to the survey of 1995 and remained present until the most recent version.

Borrowing constraints

According to Constantinides et al. (2002) young consumers would like to hold stocks in their portfolios. However, due to borrowing constraints they are unable to acquire these stocks. Higher borrowing constraints are expected to have a negative effect on the ratio of risky assets held as part of the portfolio. The level of borrowing constraints was estimated using the answer to a question about the self-declared level of difficulty to obtain a loan. This question is part of the ‘Psychological Concepts Module’ and is available in the waves of 2004-2017.

Interest rates

Based on the findings concerning the interest elasticity of savings as discussed in the literature review, a higher level of interest rate is expected to have a (slight) positive effect on the ratio of savings. Furthermore, a decrease in the level of interest rate is expected to lead to an increase in the ratio of risky financial assets as part of the portfolio. Although a wide variety of information can be derived from the surveys, the level of interest rate is not available. As a proxy for the level of interest rate the ‘NL Euro-Guilder Deposit Rate’ is used, provided by the Financial Times and Thomson Reuters. The questions regarding asset holding are based on the amounts on December 31st of the previous year. As an example, the question concerning mutual fund ownership in the survey of 2017 can be found in appendix C. As a consequence, the proxy for the level of interest is also based on the 31st of December of the previous year. The level of interest rate used in the regression is constant within years.

4. Data Analysis

Considering the financial asset composition of households (see table 1A/1B and figure 1A/1B in appendix A), several interesting observations can be made. The numbers are based on the closing balance sheet values. Looking at the percentage of risky financial assets there is a clear increase from 1995 until 2000, the period before the dot-com bubble burst. In 1995 the share of risky financial assets was 22.7 percent, in 2000 this share had increased to 29.4 percent. At the same time, the percentage of savings deposits decreased from 14.1 percent in 1995 to 10.5 percent in 2000. This increase in the ratio of risky financial assets and the decrease in the ratio of savings might indicate a shift in this period from risk-free (savings) assets to riskier assets.

In the period from 2001 until 2003, the period after the dot-com bubble, the ratio of risky financial assets fell down to 20.4 percent. In the meantime, the ratio of savings increased to 14.6 percent in 2003, indicating a reverse process compared to the developments of 1995-2000. This shift might be a response to a feeling of financial uncertainty and pessimism in this period. After 2003 the ratio of risky financial assets continued to decrease slowly before dropping sharply, from 17.7 percent in 2007 to 15.0 percent in 2008, in the beginning period of the most recent financial crisis. On the contrary, the ratio of savings deposits increased from 14.5 percent in 2007 to 17.7 percent in 2008. These developments are similar to the changes that took place after the dot-com bubble burst. The increase in the ratio of savings and the decrease in the ratio of risky financial assets might be an indication for a higher level of average risk aversion among households in the Netherlands. The ratio of savings deposits peaked in 2013, followed by a decreasing development until the most recent available year of 2016. The ratio of risky financial assets continued to decline slowly in the years after 2009. Note that after 2011 a different method of measuring the composition of financial assets was utilized, so comparison between the periods 1995-2010 and 2011-present must be made with caution. Survey data

After conducting the above provided analysis of the composition of financial assets based on the financial account data, it is time to do a similar analysis based on the survey data (see table 2 and figure 2). Let’s start with looking if the developments displayed by the national account analysis can also be found when looking at the composition of financial assets based on the survey data. Before doing so, a few notes about the differences between the national account

data and the survey data. Instead of the period ranging from 1995-2016 as in the national account analysis, the survey period ranges from 1993-2017. In the national account data ‘Net equity of households in life insurance and pension fund reserves’ plays a dominant role. Based on the survey data however, less is known about this variable, leading to (on average) higher asset ratios. Questions about asset holding in the survey are based on numbers from the 31st _of December of the year before (for an example, see appendix C). So, the national account numbers (based on the closing balance sheet values) must be compared to the survey numbers of a year later. When a certain period is discussed, the year of the national account data is leading. These differences are however not necessarily problematic, since the main goal is to compare developments rather than specific numbers.

Similar to the findings based on the national account data, the period of 1995-2000 resulted in a sharp increase in the ratio of risky financial assets from 19.5 percent in 1995 to 37.6 percent in 2000. The holding ratios displayed in table 3A and figure 3A indicate an increase in the percentage of people holding risky financial assets. There has also been a decline in the ratio of savings, although this decline was smaller than the decline observed in the national account data. Based on the increasing holding percentages of savings in the period of 1995-2000 it seems people have decreased the average amount of money on their savings accounts. Looking at the complete period of 1993-2017, the percentage of people holding savings has increased drastically from 23.8 percent in 1993 to 76.2 percent in 2017

The developments of 2001-2003 are observable in the survey data as well, the ratio of risky financial assets dropped to 22.8 percent in 2003 and the ratio of savings increased to 38.3 percent, compared to 25.5 percent in 2000. These developments can be observed in the holding ratios as well, the percentage of people holding a positive amount of risky financial assets decreased in this period and there was a steeper increase in the ratio of people holding a positive amount of savings in this period. The change in the percentage of people holding certain assets asset in reaction to the burst of the dot-com bubble seems to be less direct than the reaction in asset value compared to total financial assets. A possible explanation for this phenomenon is the direct effect of the burst resulting in a decrease in the value of financial assets. After 2003 risky financial assets increased in popularity, based on the ratio compared to total financial assets, peaking in 2007. This increase is however not observed in the holding ratio of risky financial assets, which stayed stable in this period. Based on these findings there are indications for an increase in valuation of risky financial assets and/or people choose to hold (on average) a higher amount of risky financial assets. Similar as the development based on the national account data, the ratio of risky financial assets declined sharply from 31.5 percent in 2007 to

25.3 percent in 2008. Again, the ratio of saving moved in the opposite direction from 40.5 percent in 2007 to 47.3 percent in 2008. The holding ratios in this period remained relatively stable, the decline in the ratio of risky financial assets might be partially explained by the decline in the valuation of these assets. In the years after the financial crisis the ratio of savings continued to increase, peaking at a rate of 58.9 percent in 2012. This peak in the savings ratio was thus observed one year prior to the peak in savings according to the national account data. The ratio of risky financial assets continued to decline, reaching a low of 17.8 percent in 2012. After 2012 a slight increase was noticeable. This increase is however not represented by an increase in the percentage of people holding risky financial assets.

Another interesting development that can be derived from the survey data is the role of mutual funds. In the period of 1992-2000 the ratio of mutual funds compared to total financial assets increased from 4.5 percent to 16.6 percent. Before 2000 the ratio of stocks and shares was larger than the ratio of mutual funds. After 2000, until the most recent survey, mutual funds were a larger part of total financial assets than stocks and shares. A similar pattern as the development of the ratio of mutual funds compared to total financial assets is shown when looking at the holding percentages. When looking at the holding percentages of stocks and shares however, a less drastic decrease can be observed compared to the development in the ratio of asset value.

The changes in asset ownership are quite similar when comparing the national account data and the survey data. So, based on developments over the years, the survey data seems to be a reasonably good estimation of the Dutch population.

4.2 Micro level

Since not all desired variables are available in the complete period of 1993-2017 several regressions will be done, each including a different set of independent variables. First of all, a linear regression is done using the variables available in all of the 25 years of survey data. The results of this regression are provided in table 5. The independent variables tested in this regression are: the level of wealth, the level of education, age, the year of the survey, income, gender and the level of interest rate. Starting in 1995, three new variables can be derived from the survey data: risk aversion, income uncertainty and community. These new variables and the variables mentioned in table 5 were used to perform the regression which resulted in the values provided in table 6.

The level of interest in financial matters is available in the period 1995-2003. For this period of time a separate regression was made using the dummy variables for financial interest

and the variables used earlier in the regression of table 6. Starting from 2001 the (self-declared) level of financial knowledge became available. Using the associated dummy variables and the variables tested in table 6, a regression for the period of 2001-2017 is performed. Finally, for the period of 2004-2017 a regression is performed including the dummy variables related to borrowing constraints, the level of financial knowledge and the variables mentioned in table 6.

A Breusch-Pagan/Cook-Weisberg test was performed to test for heteroskedasticity. In all regressions heteroskedasticity was indeed present. Therefore, robust standard errors are used in the regressions. Since the level of interest rate is constant within years, its effect in the model can not be estimated when year fixed effects are included. Hence, year fixed effects are not included in the model. The independent variables and their influences on the dependent variables will be discussed in the following section.

Wealth

The level of wealth has a positive relationship with the ratio of risky financial assets. In all five regressions wealth is significant at the 1 percent level. This is corresponding to the findings in the existing literature. Although this effect is highly significant, the actual change of the ratio of risky financial assets in response to an increase in wealth is rather low. An increase in wealth of €10,000 is expected to increase the ratio of risky financial assets by 0.09-0.17 percentage-points, ceteris paribus. Wealth has a negative relation with the ratio of savings at the 1 percent significance level in four of the 5 performed regressions. The estimated effect of an increase in wealth of €10,000 is a decrease of around 0.07 percentage-points on the ratio of savings, ceteris paribus. This contradictive effect might implicate a shift of individuals from savings to risky financial assets when wealth increases.

Education

University education has, in general, a positive influence on the ratio of risky financial assets when compared to other levels of education. In all the regressions, except for the one displayed in table 7, university education was significant at a level of significance of 5 percent and most times at the 1 percent level. Different from expectations, the difference between low education and university education was not the most dominant in all situations. University education is expected to increase the ratio of risky financial assets by 3.7-7.6 percentage-points when compared to pre-vocational education, ceteris paribus. The relation between the level of education and the ratio of savings is less clear. In all regressions however, university education increases the ratio of savings between 7.1-9.0 percentage-points compared to low education

(ceteris paribus), significant at the 5 percent level. University education thus seems to have a substantial positive influence on both dependent variables.

Age

Age has a strictly positive relationship with the ratio of risky financial assets, significant at the 1 percent level in all cases. This is in line with the findings mentioned in the literature review. According to the regression displayed in table 6 an additional year of age leads to an average increase in the ratio of risky financial assets of approximately 0.17 percentage-points, ceteris paribus. Age generally influences the ratio of savings in a positive way as well, being significant at the 5 percent level for all regressions except the one displayed in table 7 (including financial interest).

Income

There seems to be a positive relation between the level of income and the ratio of risky financial assets. This relationship is significant at 5 percent for all regressions except the regressions of table 6 and 7. When the yearly income increases by €1000 the ratio of risky financial assets will, dependent on which regression is used, rise by approximately 0.04-0.05 percentage-points, ceteris paribus. Based on the less extensive regressions displayed in table 5 and 6 one might see indications for a positive relation between the level of income and the ratio of savings. This effect seems to be smaller than the effect on risky financial assets, around 0.025 percentage-points increase (ceteris paribus) when income rises by €1000.

Gender

Like suggested in the literature, being a female seems to have a negative effect on the ratio of risky financial assets. In all regressions, except for the final one including borrowing constraints, the effect of gender is significant at the 5 percent level. Being a female results in a decrease in the ratio of risky financial assets of around 1.1-2.1 percentage-points, ceteris paribus. Contrary to the effect on the ratio of risky financial assets, being a female generally has a positive effect on the ratio of savings. This effect seems to be larger than the effect on risky financial assets. The ratio of savings increases by 5.7-7.6 percentage-points, ceteris paribus, when comparing to men. The ratio of savings thus seems to be influenced considerably by gender.

Risk aversion

Like suggested by the findings in the literature, a high level of risk aversion has a negative relationship with the ratio of risky financial assets (1 percent significance). Especially the difference between high risk aversion and medium risk aversion seems to affect the ratio of risky financial assets. This effect is equal to approximately 6.5 percentage-points, ceteris paribus. Like expected, high risk aversion has a positive relation in respect to the ratio of savings with a level of significance of 1 percent when compared to a low level of risk aversion. A high level of risk aversion increases the ratio of savings by 8.1-14.3 percentage-points, ceteris paribus. Based on these results, risk aversion seems to be a major influencer of the portfolio choice of individuals.

Income uncertainty

Income uncertainty shows a (surprising) positive relationship with the ratio of risky financial assets in the regression including financial interest (1995-2003) at 5 percent significance. In the other regressions where income uncertainty was included, its effect was negative. Nothing can be said in respect to the relationship of income uncertainty and the ratio of savings

Community

Another surprising effect is related to the variable community. When parents, friends or acquaintances are the main source of financial advice, the ratio of risky financial assets decreases at 5 percent significance. This negative effect might be explained by a high proportion of respondents naming their parents as their main financial advisors. Children might be advised by their parents to hold relatively safe financial assets. The estimated negative effect of community on the ratio of risky financial assets is 1.1-2.8 percentage-points, ceteris paribus. The influence of community on the ratio of savings is unclear.

Financial interest

A high level of interest in financial matters has a positive effect on the ratio of risky financial assets when compared to both low and medium financial interest at 1 percent significance. This result must be interpreted with caution, since the level of financial interest is only included in the regression ranging from 1995-2003 (table 7). The magnitude of this effect may however not be underestimated. A high level of financial interest raises the ratio of risky financial assets by 14.4 percentage-points compared to a low level of financial interest, ceteris paribus.

Financial knowledge

A high level of financial knowledge will result in a higher ratio of risky financial assets (1 percent significance) when compared to a low level of financial knowledge. This is based on the regressions ranging from 2001-2017 and 2004-2017 (table 8 and 9 respectively). The effect of financial knowledge seems to be smaller than the effect of financial interest. These results can however not be compared with certainty due to the different time spawns. Compared to low financial knowledge a person with (self-declared) high financial knowledge is expected to hold a portfolio where the ratio of risky financial assets is 4.3-5.0 percentage-points higher, ceteris paribus. There are no indications of an influence of financial knowledge regarding the ratio of savings.

Borrowing constraints

When analyzing the dummy variables related to the level of borrowing constraints, the final regression (table 9) shows that a high level of borrowing constraints compared to a low level of borrowing constraints leads to a decrease in the ratio of risky financial assets significant at 1 percent. A high level of borrowing constraints leads to a decrease in the ratio of savings as well (1 percent significance), when compared to both low borrowing constraints and medium borrowing constraints. As an effect of high borrowing constraints, the ratio of risky financial assets and the ratio of savings are expected to be 2.1 and 1.8 percentage-points lower respectively compared to low borrowing constraints, ceteris paribus.

Interest rate

The results of the regression presented in table 5 indicate a lower ratio of risky financial assets when interest rates are higher at 5 percent significance. When the level of interest rate rises by 1 point the ratio of risky financial assets will be approximately 0.36 percentage-points lower, ceteris paribus. However, when other variables are added, like risk aversion, income uncertainty and community in the regression of table 6 (1995-2017), the level of interest rate looses its significance. The effect of the interest rate level on the ratio of savings is not significant in all regressions.

5. Conclusions

Based on all information gathered in the previous sections, the research questions can now be answered. To repeat, the research questions were structured in the following way: what are the factors influencing the ratio of risky financial assets and the ratio of savings held by Dutch individuals? And furthermore: what is the effect of the level of interest rates on the ratio of risky financial assets and the ratio of savings held by Dutch individuals?

In the first part of the data analysis a general overview of financial asset composition over time was provided using both the national account data and the survey data. From these tables several conclusions can be drawn. Firstly, there was an increase in the ratio of risky financial assets in the period before the burst of the dot-com bubble accompanied by a decline in the ratio of savings. After 2000, an opposite trend was observable resulting in a decline in the ratio of risky financial assets and an increase in the ratio of savings. In the survey data an increase in the ratio of risky financial assets after 2003 can be observed, peaking in 2007. During the beginning of the most recent financial crisis a similar trend as observable in 2001 was present: an increase in the ratio of savings and a decrease in the ratio of risky financial assets. The ratio of savings continued to increase until 2013, possibly indicating an increased level of risk aversion of people in the Netherlands in reaction to the crisis.

Another conclusion that can be drawn is the growing importance of mutual funds over the years. Mutual funds peaked at a level of 16.6 percent of total financial assets in 2000 and has played a more dominant role in financial asset composition than shares and stocks ever since. This corresponds with the findings of the Deutsche Bundesbank (2015) who provided a possible explanation for this phenomenon by stating that household prefer to hire professionals to manage their riskier investment decisions. By doing so they face a more indirect form of exposure, corresponding with a higher level of risk aversion (Deutsche Bundesbank, 2015).

Some interesting conclusions can be drawn from the results obtained by performing the regressions on the 25 years of survey data. First of all, the conclusions regarding the ratio of risky financial assets will be provided followed by the conclusions concerning the ratio of savings. Finally, the influence of the level of interest rate on both ratios will be discussed.

There are strong indications for a positive relationship between the ratio of risky financial assets and the level of wealth, university education, age and the level of income of the respondent. All of these results are in correspondence with the hypotheses. Some less strong indications are present of a positive relationship between the ratio of risky financial assets and a high level of financial interest as well as a high level of financial knowledge. Some other

variables are found which have a significantly negative influence on the ratio of risky financial assets. These variables are: gender (female), a high level of risk aversion and community. High borrowing constraints and income uncertainty might also be causes of a lower ratio of risky financial assets. Being a female, having a high level of risk aversion, income uncertainty and high borrowing constraints was hypothesized to have a negative influence on the ratio of risky financial assets. Community however, was not. A possible explanation for this result is that of the people naming friends, family and acquaintances as their most important source of information regarding financial decision making, many named their parents. Parents might advise their children to hold a safe, traditional portfolio consisting mainly of risk-free assets.

Variables which show to have a significantly positive relationship with the ratio of savings are age, gender (female) and a high level of risk aversion. Some evidence is found for a positive effect of university education and high financial knowledge on the ratio of savings. The hypotheses that being a female and having a high level of risk aversion has a positive influence on the ratio of savings can thus be confirmed. A negative influencer of the ratio of savings is the level of wealth. A higher level of wealth results in a lower level of savings divided by total financial assets. There is also some support for a negative relationship between a high level of borrowing constraints and the ratio of savings.

When comparing the variables influencing the ratio of risky financial assets and variables influencing the ratio of savings an interesting pattern can be spotted. Some variables positively influencing the ratio of risky financial assets negatively influence the ratio of savings and visa versa. The level of wealth has a positive relationship with the ratio of risky financial assets but a negative relationship with the ratio of savings. Furthermore, a high level of risk aversion leads to a lower level of the ratio of risky financial assets but simultaneously to a higher level of the savings ratio. Finally, being a female negatively influences the ratio of risky financial assets but positively influences the ratio of savings.

There are slight indications that a higher level of the interest rate leads to a lower level of the ratio of risky financial assets. This effect, however, is not significant when controlling for additional variables like the level of risk aversion, income uncertainty and community. A possible explanation for the effect being insignificant is the monetary policy strategy of the ECB. To stimulate investments, the ECB adjusts market rates to a level below their long-term rates (European Central Bank, 2016). This policy is mainly chosen when the economy is in a situation of recession. In the macro analysis in the first part of the data analysis, a negative trend in the ratio of risky financial assets was observed in reaction to both the burst of the dot-com

bubble and the beginning of the most recent financial crisis. So, when interest rates are declining, the ratio of risky financial assets might already be relatively low.

6. Review

This research is based on analyses using data over a relatively long period of time, 22 years of national account data and 25 years of survey data. This way, developments over a broad period of time were able to be analyzed. However, data of the current year (2018) is not yet available. This might lead to unobserved recent trends since this information is not yet absorbed in the data. The most recent wave of the survey data (2017) contains information on asset holding based on the values at the 31st of December 2016. The increase in the amount of Dutch households holding stocks and/or bonds suggested by Van Rein (2017) might be too recent to show in the available data. In the next wave of survey data (2018) this trend might be more clearly observable. The relevance of this research is primarily based on the people of the Netherlands since all data used concerned Dutch households. So, extrapolating results to the setting of another country must be done with caution. Another limitation of this research is the fact that some variables are available for a limited period of time. In an ideal situation all variables of interest would be available in all the years, from 1993 until 2017. The dummy variables related to financial interest, financial knowledge and borrowing constraints all showed significant results. In this research however, these results must be treated carefully since these variables are not available in the full period of 1993-2017.

To be able to truly know if the reason of individuals to hold riskier financial assets is based on the level of interest rates some additional questions must be made part of the survey in the future. This way the effect of the level of interest can be stated with more certainty. In this question the effect of community can be incorporated as well. An example of such a question could be:

‘What has been your main reason to participate on the stock/bond market?’ - Current interest rate on savings accounts

- Advise from relatives/friends to participate - Tax reasons

Another additional question for future surveys could be related to the question if people consider risky financial assets as a substitute for holding money via a savings account. These questions could be structured in the following way:

‘Do you agree with the following statements?’

1. I consider stocks as an alternative for holding money on a savings account 2. I consider bonds as an alternative for holding money on a savings account

3. I consider mutual funds as an alternative for holding money on a savings account 4. Other statement…

By incorporating these new variables in the data analysis a more detailed answer can be provided concerning the factors influencing the portfolio choice of (Dutch) individuals.

An additional analysis that can be performed is an analysis based on the change in the ratio of risky financial asset ownership and the ratio of savings of each individual. By conducting this analysis, it will become clear what type of shifts have occurred between risky financial asset holding and savings over the years.

7. References

Alessie, R. J. M., Hochgürtel, S., & van Soest, A. H. O. (2000). Household Portfolios

in the Netherlands. (CentER Discussion Paper; Vol. 2000-55). Tilburg: Econometrics. Barasinska, Schäfer, & Stephan. (2012). Individual risk attitudes and the composition of

financial portfolios: Evidence from German household portfolios. Quarterly Review of Economics and Finance, 52(1), 1-14.

Beer, C., Gnan, E., & Ritzberger-Grünwald, D. (2016). Saving, portfolio and loan decisions of households when interest rates are very low–survey evidence for Austrian

households. Monetary Policy & the Economy, (1), 14-32.

Beznoska M, Ochmann R (2010) Household savings decision and income uncertainty. DIW discussion papers no. 1046. DIW Berlin, German Institute for Economic Research, Berlin

Beznoska, M., & Ochmann, R. (2013). The interest elasticity of household savings: A structural approach with German micro data. Empirical Economics, 45(1), 371-399. Bogan, V. (2008). Stock Market Participation and the Internet. Journal of Financial and

Quantitative Analysis, 43(1), 191-211.

Boskin, M. (1978). Taxation, Saving, and the Rate of Interest. Journal of Political Economy, 86(2, Part 2), S3-S27.

Brown, J., Ivković, Z., Smith, P., & Weisbenner, S. (2008). Neighbors Matter: Causal Community Effects and Stock Market Participation. Journal of Finance,63(3), 1509-1531.

Charness, G.B., & Gneezy, U (2007). Strong Evidence for Gender Differences in Investment. Mimeo, Department of Economics, University of California at Santa Barbara.

Constantinides, G., Donaldson, J., & Mehra, R. (2002). Junior Can't Borrow: A New Perspective on the Equity Premium Puzzle. The Quarterly Journal of

Economics, 117(1), 269-296.

Deutsche Bundesbank. (2015, October 13). German households’ saving and investment behaviour in light of the low-interest-rate environment. Retrieved June 6, 2018. Elmendorf D (1996) The effect of interest-rate changes on household saving and

consumption: a survey. Finance and economics discussion series, working paper no. 1996-27. Federal Reserve Board, Washington, DC

European Central Bank (2016). Addressing the causes of low interest rates. [online] European Central Bank. Available at: https://www.ecb.europa.eu/press/key/date/2016/html/ sp160502.en.html [Accessed 7 Jun. 2018].

Fessler, P., Lindner, P., & Schürz, M. (2016). In focus: Eurosystem Household Finance and Consumption Survey 2014–first results for Austria (second wave). Monetary Policy & the Economy, (2), 34-95.

Haliassos, M., & Bertaut, C. (1995). Why Do So Few Hold Stocks? The Economic Journal: The Quarterly Journal of the Royal Economic Society, 105(432), 1110-1129.

Harris, Mark N., Loundes, Joanne, & Fox, Elizabeth Vassall. (2002). Determinants of household saving in Australia. (*). Economic Record, 78(241), 207-223. Heaton, J., & Lucas, D. (2000). Portfolio Choice and Asset Prices: The Importance of

Entrepreneurial Risk. Journal of Finance,55(3), 1163-1198.

Hochguertel, S., Alessie, R., & Van Soest, A. (1997). Saving Accounts versus Stocks and Bonds in Household Portfolio Allocation. Scandinavian Journal of Economics,99(1), 81-97.

Hong, H., Kubik, J., & Stein, J. (2004). Social Interaction and Stock-Market Participation. Journal of Finance, 59(1), 137-163.

King, M. A. and J. I. Leape (1987), “Asset Accumulation, Information, and the Life Cycle,” NBER Working Paper, No. 2392.

Kreinin, M. (1959). Factors associated with stock ownership. The Review of Economics and Statistics, 41(1), 12-23.

Leibenstein, H. (1950). Bandwagon, snob, and Veblen effects in the theory of consumersʹ demand. The Quarterly Journal of Economics, 64(2), 183-207.

Mankiw, G. (2013). Macroeconomics. 8th ed. Houndmills: Palgrave Macmillan.

Uhler, R., & Cragg, J. G. (1971). The structure of the asset portfolios of households. The Review of Economic Studies, 38(3), 341-357.

Van Rein, E. (2017, November 9) Beleggingskoorts van late jaren negentig lijkt terug. Het Financieele Dagblad. p. 17. Retrieved from https://fd.nl/beurs/1226163/aantal-particuliere-beleggers-stijgt-fors

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

8. Data sources

In this paper use is made of data from the CentERdata databank. More specifically, 25 years of data collected via the DNB Household Survey (1993-2017).

In this paper used is made of data from CBS containing national account data ranging from 1995-2016. As an example, the source of the most recent version is provided, previous versions can be found on the same location: CBS. (2017, July 17). National accounts 2016. Retrieved May 23, 2018, from https://www.cbs.nl/en-gb/publication/2017/28/ national-accounts-2016

CBS. (2018, March 08). CBS StatLine - Internet; toegang, gebruik en faciliteiten. Retrieved June 03, 2018, from http://statline.cbs.nl/Statweb/publication/?DM=SLNL&PA= 83429NED&D1=0,2- 5&D2=0,3-

Datastream. (2018) Financial times and Thomson Reuters Datastream. [Online]. Available at: Subscription Service (Accessed: May 2018)

De Nederlandsche Bank. (2018). Deposito's en leningen van MFI's aan huishoudens, rentepercentages, gecorrigeerd voor breuken (Kwartaal) (updated June 4th 2018) [Online]. Retrieved from https://statistiek.dnb.nl/downloads/index.aspx#/details/ deposito-s-en-leningen-van-mfi-s-aan-huishoudens-rentepercentages-gecorrigeerd-voor-breuken-kwartaal/dataset/e1bfa590-095f-4c3f-8268-a1dc632732d2/resource/ 5f55f699-ad48-46aa-a54f-e2c1df318b2f