Master Thesis Finance Habits in Investing

(1)

Master Thesis Finance

Habits in Investing

PIETER ODINK*

University of Groningen Faculty of Economics and Business

Abstract

Using data from the Dutch National Bank from 1999 to 2009, this paper examines the link between investing habitually and the optimality of the resulting portfolio. Results indicate that habitual behaviour is negatively related to the optimality of the portfolio, through information ignorance and cognitive dissonance. Main control variables in the study include herding behaviour and status quo bias, two other concepts related to behavioural finance.

JEL classification: C13, D03, D14, G11.

Key Words: Habitual behaviour, portfolio management, herding behaviour, status quo bias

*

(2)

1. Introduction

In the supermarket we buy the same brand of detergent every time we need it. In the lunchroom we buy the same type of coffee every time we want to have one. We also have the tendency to buy the same brand of car over time. Much of this behaviour takes place with a minimum of thinking. Behaviour as such, where actions are repeatedly being performed without deliberating too much, can be grouped under the concept of habits (Jager, 2003).

It is unclear, though, whether this type of habitual behaviour is also shown in financial decision making. When choosing to invest in different types of assets, we also have to make a decision on which ones to spend money on. Could this mean that investing in assets can also be done with a minimum of thinking? Can our investing behaviour also be classified as habitual behaviour?

When individual investors make their first decision to invest their money in different types of assets, they will take their time to decide on their personal optimal portfolio. The question arises whether this is also the case when new wealth is acquired and ready to be invested. In general, as Hindy and Huang (1993) argue, consumption is periodic when past consumption is sufficient to sustain the investor’s needs for some time. This indicates that in circumstances in which the individual investor is satisfied with his first investment portfolio, he will invest newly wealth in the same portfolio. However, when this investor is spending these new resources at some later period in time, this initial portfolio may not be his optimal portfolio. Hence, when the investor is still investing his resources in this same portfolio, it will result in a sub-optimal portfolio.

Research on habit formation might provide a reason why this irrational investing behaviour is shown. The presence of habits causes that new information is neither taken into account when performing the behaviour, nor actively searched for. Hence, whereas the habit may originate from a process of finding out the optimal behaviour given the prevailing circumstances, the circumstances may since then have changed such that alternative behaviour would yield better outcomes (Jager, 2003). Habitual behaviour then causes the individual investor to stay with his current portfolio, without deliberating new information and causing his portfolio to be sub-optimal.

The question arises whether habitual behaviour is theoretically a possible driver of suboptimal portfolios. For that to be true, the context in which investing behaviour takes place should be in line with the context of other habitual buying behaviour.

(4)

possibility to have habitual behaviour as antecedent of sub-optimal portfolios. Even in high-involvement decisions like choosing a car-brand, your habit to choose the same brand over time may win it from the rationale to search information related to other brands (Journal of Accountancy, 1993; Caballero, 1993).

In addition, as argued in the previous paragraph, habitual behaviour may cause the buyer/investor to ignore newly available information, even when changing circumstances may need an alternative behaviour to reach best outcomes (Jager, 2003). The result of ignoring new information is argued to be sub-optimal decision making. Since sub-optimal portfolios are deeply existent, but there is no clear understanding of why they are existent, the theory of habitual behaviour might provide new insights.

Thirdly, as Wood, Quinn and Kashy (2002) claim, habitual behaviour is likely to be developed in stable contexts. In this argumentation, a stable market is a market with a fairly stable number of alternatives, with a fairly stable price and with fairly stable quality. In the investing situation, individual investors have the choice to invest in a certain number of types of assets which do not change that much over time. The data used in this thesis comes from the Household Surveys of the Dutch National Bank in the previous 16 years. From this data, it can indeed be concluded that the composition of the portfolio does not change that much over time. In addition, risk and return provide alternatives for price and quality measurements in the investing context. In general, there is a separation between risky and risky assets. Certain assets are risky, and certain assets are non-risky. This classification does not change over time, indicating that also the price- and quality- characteristics of the stable market context are fairly stable over time.

At last, an additional feature of the context under which habitual behaviour is likely to be developed is given by Verplanken and Orbell (2003). They argue that making use of habits is appealing in the sense that they open mental capacity to perform other activities in the same time. This is very likely to be existent in cases of information overload. For inexperienced individual investors, the investing context indeed provides an overload of information. Again, this provides an argument that habitual behaviour is existent in the context of this study.

To summarize, it can be concluded that habitual behaviour might provide a new explanation for the question of why sub-optimal portfolios exist. The context in which investing behaviour takes place meets the requirements to provide a theoretical explanation for the relationship: Investing is related to buying, the investing context is fairly stable, decision making is complex due to information overload and sub-optimal decision making is existent in investing behaviour.

(5)

statement ‘Can habitual behaviour explain why individual investors have sub-optimal portfolios of assets?’ What exactly is meant with habitual behaviour will be made clear in section 2, the literature review. Together with this problem statement, several sub-questions will be answered to gain further insights. First, ‘Is habitual behaviour causing sub-optimality of portfolios?’ Second, ‘What other behavioural antecedents are driving portfolio choice? ’And thirdly, ‘Which other behavioural antecedents are causing sub-optimality of portfolios?’

It is important to note that this study tries to explain the investing behaviour of individual (non-professional) investors. Therefore, this problem will empirically be tested by making use of data from the Dutch National Bank. More specifically, in this paper use is made of data of the DNB Household Survey. Survey results from the survey of 1999, 2004, 2008 and 2009 are used to come up with indicators for the variables used in this thesis. The household survey is very well applicable since it is being administered to individual (non-professional) investors. The variables computed based on this household survey will be discussed more specifically in section 3.

Currently, two types of behavioural approaches are put forward to explain why sub-optimal portfolios exist. The first behavioural approach is related to herding behaviour, the second is related to status quo bias. Herding behaviour accounts for situations in which individual investors tend to follow the actions of other investors (Chang, Li and Tan, 2010). Chang, Li and Tan (2010) explain that in the circumstance of unstable markets, investors tend to suppress their own beliefs and their investment decisions are then more likely to be based on collective actions in the market. According to Rubaltelli et al. (2005), status quo bias is existent when individuals tend to postpone their decisions or to maintain their current situation unchanged instead of choosing an uncertain outcome.

Both approaches thus deviate from the habitual behaviour approach. The herding behaviour approach relates the behaviour of the investor to behaviour of other investors. The status quo bias approach relates the behaviour of the investor to a specific type of risk aversion, i.e. a resistance to change. The habitual behaviour approach relates the behaviour of the investor to repeated performance and the ignorance of new information, causing the behaviour to be unchanged (Jager, 2003).

(6)

one day to the consumption on the days before and try to conclude whether there is a pattern in the amount of consumption.

This study tries to explain another type of habitual behaviour related to portfolios of investors. This research will actually try to explain that habitual behaviour causes the investor to invest in the same types of assets, not considering the amount of consumption. This thesis is thus the first study to introduce this type of habitual behaviour in the financing context, which makes this thesis unique.

By introducing the new type of habitual behaviour in the financing context, this research is innovative in the way it will introduce a measurement for this type of habitual behaviour. A detailed overview on how this measurement is compounded, is given in section 3.3.

For practitioners who are investing their wealth, it is interesting to understand why their behaviour is not optimal, and how they could improve their performance. Results from this thesis will provide insights in directions of interest for practitioners as well, since it examines possible antecedents of sub-optimality of portfolios. A focus (e.g. reflecting their behaviour) of these practitioners on these antecedents might give a direction to improve their performance and might optimize the investments of their wealth.

(7)

2. Literature review

This section will examine the current literature in the field of habitual behaviour, the sub-optimality of portfolios and their relationship. Besides, it will describe the main control variables in this research, namely herding behaviour and status quo bias. General theoretical models, as well as empirical findings will be given. However, at first, the concept of habitual behaviour in this study is made clear.

2.1. Defining habitual behaviour

The concept of habitual behaviour will be explained in this subsection, making use of four elements of habitual behaviour theory; internal versus external habit specification, habit formation and development, the functioning of scripts and the advantages of habitual behaviour. The disadvantages related to habitual behaviour form the basis of the main hypothesis in this study and will be discussed in section 2.2.2. Before we start with outlining characteristics of habitual behaviour, remember the definition by Jager (2003) for habits, used in the introduction: ‘Much behaviour and decision making takes place with a minimum of thinking. Behaviour as such, where actions are repeatedly being performed without deliberating too much, can be grouped under the concept of habits (Jager, 2003).’

(8)

Secondly, it is interesting to see how habitual behaviour is originated, and how it develops. Fitts and Posner (1967) are the most influential researchers in this field, with their three stage learning process of human performance model. Fitts and Posner (1967) suggest that the learning process moves through three specific stages, in order to learn a new skill. Anderson (1982) used this three stage model and fitted it to the theory of habitual behaviour, distinguishing three stages in the development of a (new) habit. The declarative stage is the first stage in which people rehearse their information and actions, so that it is available to guide behaviour. The second stage, the knowledge compilation stage, describes the compilation processes by which the cognitive system goes from interpretive application of declarative knowledge to procedures that directly apply the knowledge. Shortly, in this stage the habit is thus defined. In the last, procedural stage, the habit is tuned. With tuning, Anderson (1982) means that with experience, the search becomes more selective and more likely to lead to rapid success. These steps can also be applied to the setting in which investors need to invest their wealth in assets. When an investor is investing for the first time, he/she will rehearse information on for instance risk, return and correlation measures. When the same investor is investing new wealth at a later stadium, he/she might pass over this step and will directly invest in the same portfolio as he/she did the first time. When these steps are repeated over time, the investor will need less and less rehearsing of information, and the habit is tuned.

Now that the development of the habit is made clear, naturally the scripts are included in defining habitual behaviour. A script reflects a specific rule stating that in a certain type of situation a specific response is adequate (Jager, 2003). Though, as f.i. Abelson (1981) points out, the script thus represents the knowledge structure behind the habit and is not equal to the habit itself. Fiske and Tayler (1991) explain that if a certain situation is recognised, the script will be activated and the person will not evaluate all available alternatives again. Frequent repetition results in the tuning in the procedural stage, as described by Anderson (1982).

(9)

2.2. Habitual behaviour and portfolio management

This section will review the current literature relating habitual behaviour to portfolio management. In addition, empirical studies are analysed which support the main question in this thesis, stating that habitual behaviour negatively impacts the optimality of the portfolio.

2.2.1. Current literature in the field of this study

Habit formation models have been increasingly successful in describing portfolio allocation puzzles (Brunnermeier and Nagel, 2008; Flavin and Nagakawa, 2008; Munk, 2008), stock market participation (Gomes and Michaelides, 2003; Kuznitz, Kandel and Fos, 2008), equity premium puzzles (Constantinidis, 1990; Meyer and Meyer, 2005; Otrok, Ravikumar and Whiteman, 2002) and consumption portfolio problems (Detemple and Karatzas, 2003; Ingersoll Jr., 1992; Pollak, 1970; Yang, 2000).

In the theory relating habit formation models to portfolio allocation puzzles, researchers explain why a certain part of the wealth of the investor is invested in a certain type of portfolio. Flavin and Nagakawa (2008) present the theory that the habitual level of consumption influences the risk aversion of the investor. Changing risk aversion on its turn decides whether the investor is investing his/her wealth more in riskless or risky assets. Munk (2008), on his/her turn, presents evidence that the wealth of a risk averse investor is unlikely to be invested in risky assets, since it is not insuring that future consumption will not fall below the habit level of consumption. This evidence thus shows that high habit persistence leads to lower allocation to risky assets. At last, Brunnermeier and Nagel (2008) theoretically explain that habit persistence drives the positive relationship between the level of liquid wealth and the proportion a household invests in risky assets. They, however, also show that their analysis fails to find this relationship and that the major driver of portfolio allocation changes seems to be inertia. It does however not show that habit persistence could be driving inertia as well, which would still imply that habit formation models are able to describe portfolio allocation changes.

(10)

motive to invest in risky assets, portfolio allocation theory describes changing allocation in another time period of life. Related to this explanation, Kuznitz, Kandel and Fos (2008) analyze the situation of an investor who derives pleasure from the contemplation of future consumption. They show that deriving utility from anticipation of future consumption lowers the mean allocation to stocks when the stock market exhibits mean reversion.

In 1985, Mehra and Prescott introduced the term equity premium puzzle. The term is related to the observation that the high return demanded from the stock market, compared to the bond market, must be driven by implausibly high risk aversion of individual investors (Mehra and Prescott, 1985). Constantinidis (1990) explains that this puzzle is resolved once the assumption of time separable utility is relaxed, which equals including habit persistence in the analysis. Meyer and Meyer (2005) empirically find evidence that letting the utility of consumption depend on decreasing relative risk aversion successfully eliminates one version of the equity premium puzzle. Otrok, Ravikumar and Whiteman (2002) found similar results as Constantinidis (1990) and add to the debate that habitual investors are much more risk averse to high-frequency fluctuations than to low-frequency fluctuations. They thus show that the equity premium puzzle is even more existent in times of unstable markets, indicating another argument to include habit persistence into the analysis. Constantinidis (1990) argues that the success of habit persistence in resolving the equity premium puzzle is essentially caused by the block it forms between the risk aversion of the investor and the elasticity of consumption.

(11)

Above described theories show that habit formation models are able to capture effects related to consumption level (indicating the amount of wealth an investor is investing), to equity premiums (indicating the level of risk premium the investor needs to invest in risky assets), to stock market participation (indicating the risk aversion of the investor) and the allocation of wealth within the portfolio to risky and non-risky assets. The current study will go one level deeper into the behaviour of the investor, namely the degree of optimality of the portfolio. Although the above described findings are thus not completely on the same level, it does generate confidence that the theory related to habit formation is able to describe patterns in portfolio management. The next section will conclude on theoretical findings, supporting the main hypothesis in this study.

2.2.2. Habitual behaviour and sub-optimal decision making

In this section, theoretical robustness is sought for the relationship between habitual behaviour and the sub-optimality of decision making. It will describe the influence of information ignorance and cognitive dissonance on decision making. Afterwards, these relationships are applied to the conditions of this study, to support the main hypothesis of this study.

(12)

relying on habits and vice versa. They concluded that the switch towards active thinking is only turned on when novel and discrepant conditions are set. Corresponding, they conclude that cognitive errors and problems are not the result of relying on habits, it is the failure of recognizing the presence of new conditions that fail to turn on active thinking that causes them. In general, habits are thus found to have a negative relationship with the appreciation and acquisition of information, summarized in this study as information ignorance.

(13)

For optimal decision making, it speaks to the mind that information about possible alternatives is needed to come to an optimal decision. In the case of habitual behaviour, where decisions take place with a minimum of thinking, information is not fully employed. This leads to the hypothesis that habitual decision making could lead to sup-optimal decision making and thus sub-optimal portfolios. Above described influences of information ignorance and cognitive dissonance provide robust theoretical implication to empirically research whether this relationship is indeed existent.

2.3. Herding behaviour

This section will describe the relationship of herding behaviour with portfolio management. To understand this relationship, the origin of the behaviour is characterized, as well as an important result of this behaviour, namely market bubbles. First, a definition is given.

Park (2009) defines herding behaviour as ‘the shift of individuals’ portfolio into the same direction as others’ portfolios’. Furthermore, two conditions are given under which it is likely that herding is produced rationally; substitutability between own income and relative income, as well as declining utility of the relative income. However, this study investigates irrational habitual behaviour. It is thus also more interesting to see under which circumstances herding behaviour is shown irrational. Bikhchandani, Hirschleifer and Welch (1992) indicate that it is likely for an individual to follow the behaviour peers in the same situation, when private information is rejected. Rejection of private information is argued to be caused by four mechanisms which all relate to group pressure. Related, although not the same is the theory underlying the ‘catching up with the Joneses’ feature. Kawamoto (2009) signifies that one is behaving like their neighbours to hunt for a certain socioeconomic status. Chang, Li and Tan (2010) argue that investors hold back their own principles and thoughts when the market is unstable. In unstable markets, individual investors are more likely to follow collective actions. Kim and Pantzalis (2003) found that their results were in line with theory suggesting that herding behaviour increases with task difficulty. Individual investors which are not capable of interpreting possible investments, are thus more likely to herd since they see investing as a difficult task.

(14)

of the bubble is a fact. Greenwood and Nagel (2008) show that inexperienced investors have a significant role in the creation of bubbles. Their argumentation follows the theory signifying that task difficulty originates herding behaviour. Cooper, Gulen and Rau (2005) point out that mutual funds successfully attracted new investments when their name was changed and applied to the current bubble. Since the performance of these funds had no improvements, it again shows that herding behaviour is also occurring irrational.

It is explained by Banerjee (1992) that herding behaviour leads to inefficient portfolios. Decision rules that are based on ignoring private information and following the herd will is argued to lead to a sub-optimal equilibrium. Grinblatt, Titman and Wermers (1995) found that mutual funds can be characterized as ‘momentum investors’. They buy stocks that were past winners and thus show a tendency to herd. However, they do not sell past losers, indicating a failure to herd. These momentum investors, on average, outperform the other funds which are not herding at all. An opposite conclusion of the relationship between herding behaviour and portfolio performance is thus found relative to the findings of Banerjee (1992).

2.4. Status quo bias

In this section, theory related to the influence of status quo bias on portfolio management is provided. The source of the behaviour is sought to grasp the explanation why the relationship exists. As a starting point, a definition for status quo bias is given.

(15)

positively related to the degree of status quo bias. Kempf and Ruenzi (2006) provide empirical support for this relationship.

(16)

3. Research design

In this part, the hypotheses, the measurement of the variables, and the methodology to test the hypotheses are outlined. In addition, the data needed for the study as well as the statistical equation are given.

3.1. Hypotheses

Based on the results of the literature, it is possible to derive hypotheses for habitual behaviour, herding behaviour and status quo bias and their relationship with the optimality of the portfolio of an individual investor.

Regarding habitual behaviour, it is expected that it has a negative relationship with the optimality of the portfolio of the individual investor. This is expected since theoretical evidence is found that cognitive dissonance and information ignorance drive sub-optimal decision making, as shown in section 2.2.2. The hypothesis settles as follows:

H1: The degree of habitual behaviour in decision making has a negative relationship with the optimality of the portfolio of the individual investor.

Concerning herding behaviour, opposing results are found in the literature. However, in both studies, it was found that investors with a low tendency to herd differ from investors with a high tendency to herd. The difference in optimality, as argued in section 2.3, related to the following hypothesis:

H2: Individual investors with a low tendency to herd differ from investors with a high tendency to herd, when it comes to the optimality of their portfolio.

In relation to status quo bias, theoretical results show that a negative relationship with the optimality of the portfolio might be expected. Through risk aversion and sub-optimal decision making, the degree of status quo bias is negatively influencing the optimality of choice making. These findings, given in section 2.4, are translated into the following hypothesis:

H3: The degree of status quo bias has a negative relationship with the optimality of the portfolio of the individual investor.

3.2. Data

(17)

2000 households participating in the survey. Each yearly study has an overlap with the respondents of the previous study (the longitudinal sample), as well as new respondents in the survey (the refresher sample). The panel reflects the composition of the Dutch-speaking population of 16 years and older. The survey contains information on both economical and behavioural aspects of financial behaviour. The survey consists of six questionnaires, namely Work and Pensions, Housing and Mortgages, Income and Health, Assets and Debt, Economic concepts and finally Psychological concepts. Data is publicly, and freely available for scientific purposes.

This research uses data from the survey waves of 1999, 2004, 2008 and 2009 o come up with indicators for the variables used in this thesis. Since this survey is distributed amongst individual investors, it perfectly suits the target population of this study. The variables computed based on this household survey will now be discussed more specifically in section 3.3.

3.3. Measurement of variables

In this section, it is explained which variables are used and how these variables are defined, based on the Household Survey of the DNB. The assets and questions derived out of the survey and used for this study are displayed in Table AI and Table AII in Appendix A.

First, for measuring the degree of optimality of the portfolio, indicator variables for 11 types of assets are used. These assets are presented in Table AI in Appendix A. The more different types of assets are held by the individual investor, the more investors diversify, and the more their portfolio is assumed to be optimal. The 11 types of assets include: employer sponsored savings plan, savings account, deposit books, savings certificates, annuity insurance, endowment insurance, mutual funds, bonds, shares, options and real estate. Based on the indicators, indicating whether the asset is held (1) or not (0), a percentage is calculated representing the percentage of the total of 11 assets the individual investors included in their portfolio. In addition, a count value is included for the Poisson regression. This variable is computed by simply counting the number of assets the respondent holds. Furthermore, an indicator indicating whether risky assets are held is included for the Probit regression. For this indicator, mutual funds, shares, and options are identified as risky assets. It is assumed that for the portfolio to be optimal, the investor should invest in risky assets next to investing in non-risky assets. For the composition of these three variables related to the degree of optimality of the portfolio, the 2009-survey is used.

(18)

variable representing the habitual behaviour in the long run. Comparisons are made between the portfolios in 2004 and 2009 to come up with a variable representing the habitual behaviour in the middle-long run. Comparisons are made between the portfolios in 2008 and 2009 to come up with a variable representing the habitual behaviour in the short run. The degree to which the portfolio in the three different time settings are the same indicates the strength of the habit of the investor to invest in the same type of assets over time. All analyses will be executed using the three variables separately, resulting in three different regression- equations each time. By doing so, robustness tests can be performed over different time spans for the relationship between the habitual behaviour variable and the degree of optimality.

Thirdly, based on four statements, a variable for herding behaviour is made. Based on the definition for herding behaviour provided by Park (2009): ‘the shift of individuals’ portfolio into the same direction as others’ portfolios’, seven questions out of the Psychology part of the survey are selected. Out of these seven questions, four statements significantly correlated with each other. The four statements relate to the degree to which the respondent feels little concern for others, is interested in other, feels comfortable around people, and is not interested in others (see Table AII in Appendix A). When someone is concerned for others, interested in others and feels comfortable around others, it is assumed that this person has a higher likelihood for herding behaviour. The scaling of the first and the last statement are therefore reversed, to match the scaling for herding behaviour. Items are measured on a five-point scale, indicating to what extent they (dis)agree with the statement. Cronbach’s alpha is used as reliability check for the correlation statistic. Results indicated that the four different questions measure the same concept. Based on this result, an average score is made out of the four individual variables. This newly created variable will serve as variable for herding behaviour. For this variable, the 2009-survey is used.

(19)

have a high status quo bias. High risk averse people are not tempted to take risk and will thus stick to their current situation more often. Changing behaviour thus corresponds to taking risks and therefore risk aversion is assumed to be positively related to status quo bias. In addition, neat investors are assumed to have high status quo bias. A person who is very neat and organized, will put everything back where it belongs. In this way, the situation will not change that much. Therefore neat and organized persons are assumed to have a higher status quo bias. Furthermore, respondents with an external locus of control are assumed to have a high status quo bias and internal locus of control to be negatively related to status quo bias. When someone has an internal locus of control, this person belief his/her luck is in his/her own hands. Therefore, this person will not stick to his/her current situation when this is not completely desirable. This is contrary to having a high status quo bias, so therefore internal locus of control is assumed to be negatively related to status quo bias. Since people with a high external of control do not belief that their luck is in their own hands, the argumentation is the other way around. Therefore, external locus of control is assumed to be positively related to status quo bias. The scaling for the internal locus of control bias is reversed to match the scaling of the status quo bias. Now that the four different variables are defined, Cronbach’s alpha is used to conclude if the four different variables measure the same concept. Reliability checks did point out that this is not the case. Therefore, it is not possible to come up with one single indicator for status quo bias. In further analyses, all four variables will be included and serve as variables for status quo bias. For these four variables, the 2009-survey is used.

At last, the following nine demographic variables are included: Age (divided by 100), Age2, Gender (male vs. female), Household size (count), Marital Status (married vs. unmarried), Education (low vs. medium vs. high), Main occupation (Employee, Self employed, Retired, Unemployed), Income (in €, divided by 1000) and Wealth (in €, divided by 1000). The values for age, income and wealth are reshaped to make it easier to interpret the betas, resulting from the regression analyses. For these variables, the 2009-survey is used.

3.4. Methodology and tested equations

Now that the variables and the hypotheses are made clear, it is time to define the methodology and to come up with the statistical equations which are tested in this study.

3.4.1. Methodology

(20)

Poisson regressions are used to check whether the results, found in the linear regression, are robust. Furthermore, the indicator indicating whether the respondent holds risky assets or not, is used in Probit regressions. These results are checked on their robustness by performing the same analyses with Logit regressions.

Since theory provided reasons to assume there is a causal relationship between the independent variables and the optimality of the portfolio, regression analyses is justified. In addition, the nature of the variable also make regression analyses possible from a practical point of view. 3.4.2. Tested equations

All analyses are intended to explain the optimality of the portfolio based on the three main independent variables (habitual behaviour, herding behaviour, and status quo bias) and the demographic variables. The basic relationship used in this study is formulated in the following equation (1). This basic equation is not empirically tested, it serves as starting point for the different equations used in the analyses.

(1) OPi = Intercept + β1*HBi + β2*HeBi + β3*RAi + β4*Nei + β5*ILi + β6*ELi + β7*Agei + β8*Age 2

i +

β9*Sexi + β10*Inci + β11*Weai + β12*MaSi + β13*LEdi + β14*MEdi + β15*Empi + β16*SEmi +

β17*Reti + εi

A list of all abbreviations used in the different equations, is provided in Table BI in Appendix B. The linear regression models make use of a specific type of variable for the optimality of the portfolio, namely the degree of diversification. For the linear regression analyses, the following equations (2.1, 2.2, and 2.3) are used:

(2.1) DDi = Intercept + β1*HB1i + β2*HeBi + β3*RAi + β4*Nei + β5*ILi + β6*ELi + β7*Age + β8*Age 2

+ β9*Sex + β10*Inc + β11*Wea + β12*MaS + β13*LEd + β14*MEd + β15*Emp + β16*SEm + β17*Reti

+ εi

(2.2) DDi = Intercept + β1*HB5i + β2*HeBi + β3*RAi + β4*Nei + β5*ILi + β6*ELi + β7*Age + β8*Age2

+ εi

(21)

To check the results of equations (2.1, 2.2 and 2.3) on its robustness, Poisson regression is used. These analyses make use of the count variable for the number of assets held. This results in the following equations (3.1, 3.2 and 3.3) for the Poisson regressions:

(3.1) NAi = Intercept + β1*HB1i + β2*HeBi + β3*RAi + β4*Nei + β5*ILi + β6*ELi + β7*Age + β8*Age 2

+ εi

In addition, it is tested which variable drive the individual investor to include or exclude risky assets in their portfolio. To test this, an indicator for holding risky assets is used as dependent variable in the Probit and Logit regression analyses. The corresponding equations (4.1, 4.2, and 4.3) settle as follows:

(4.1) IRi = Intercept + β1*HB1i + β2*HeBi + β3*RAi + β4*Nei + β5*ILi + β6*ELi + β7*Age + β8*Age 2

+ β9*Sex + β10*Inc + β11*Wea + β12*MaS + β13*LEd + β14*MEd + β15*Emp + β16*SEm + β17*Reti +

εi

+ β9*Sex + β10*Inc + β11*Wea + β12*MaS + β13*LEd + β14*MEd + β15*Emp + β16*SEm + β17*Reti +

εi

+ εi

(22)

4. Descriptive statistics

Before the results are presented, descriptive statistics of the data at hand are given in this section. Summary statistics for the sub-samples are given. The sub-samples are determined by the three different analyses based on the time spans of the habitual behaviour variable. Summary statistics are included for all dependent, independent and demographic variables. The degree of diversification and the number of assets held are based on the same variable, differing by the way they are presented (percentage versus count). Therefore, from these two variables, only the number of assets variable is discussed in this section. In addition, Age2 is of course directly linked to age, and therefore only the age statistic is described in this section.

4.1. Summary statistics of the short run sample

The total number of valid observations, which contain data for all variables in the short run sample, is 1010. This is 48.33% of the total observations.

At first, the dependent variables are observed. Table I, displayed at the end of this paragraph, contains information related to the dependent variables. It shows that the median number of assets held is 1.0000. The mean number of assets is 0.7119 higher than the median, which indicates a right tailed distribution. The standard deviation from the mean is 1.3106. Furthermore, it can be seen that at least one respondent holds seven out of the total 11 types of assets, and that at least one respondent does not invest at all. 23.96% of the respondents is investing in risky assets, shown by the mean for the indicator for risky assets held. This mean has a standard deviation of 0.4271, which is relatively high.

Table I

Descriptive statistics for the dependent variables

This table contains information related to the number of observations, the minimum, the maximum, the mean, the standard deviation and the median of the dependent variables in the short run sample.

Variable Observations Minimum Maximum Mean Std. Deviation Median Number of Assets held 1010 0.0000 7.0000 1.7119 1.3106 1.0000 Indicator Risky Assets held 1010 0.0000 1.0000 0.2396 0.4271 0.0000

(23)

shown by the mean of 3.8661. With respect to the locus of control variables, there is no clear direction. Both the means of the internal and external locus of control variables are around the neutral response, with mean scores of respectively 3.6147 and 3.2030.

Table II

Descriptive statistics for the independent variables

This table contains information related to the number of observations, the minimum, the maximum, the mean, the standard deviation and the median of the independent variables in the short run sample.

Variable Observations Minimum Maximum Mean Std. Deviation Median Habitual behaviour in the last year 1010 0.5385 1.0000 0.9292 0.0765 0.9231 Herding behaviour 1010 1.0000 5.0000 3.8683 0.6670 4.0000 Risk Aversion 1010 2.5000 7.0000 5.4700 1.0126 5.5000 Neatness 1010 1.2500 5.0000 3.8661 0.7901 4.0000 Internal Locus of Control 1010 1.1250 7.0000 3.6147 0.7360 3.6250 External Locus of Control 1010 1.0000 7.0000 3.2030 0.9694 3.2000

At last, descriptive statistics for the demographic variables are given in Table III, presented at the end of this paragraph. The gender distribution shows that the short run sample contains 41.29% females and thus 58.71% males. The average age of the respondents is 57 years old. The average household size is 2.3782. This does not directly mean that a small margin of the households has children, since only 76.73% of the respondents is married and living together. The average income level per year is €22740.60, with a standard deviation of €15389.30. Hence, although the average monthly income is modal, the standard deviation of this income is relatively large. The same conclusion can be made regarding wealth, with an average wealth of €56691.70 and a corresponding standard deviation of €124838.00. The distribution related to education level shows that 4.46% enjoyed low education, 45.54% enjoyed medium education, and 50.00% enjoyed high education. Regarding primary occupation, 44.65% of the respondents is employee, 3.17% is self employed, 20.40% is unemployed and 31.19% is retired.

Table III

Descriptive statistics for the demographic variables

This table contains information related to the number of observations, the minimum, the maximum, the mean, the standard deviation and the median of the demographic variables in the short run sample.

(24)

Indicator Low Education 1010 0.0000 1.0000 0.0446 0.2064 0.0000 Indicator Mediate Education 1010 0.0000 1.0000 0.4554 0.4983 0.0000 Indicator High Education 1010 0.0000 1.0000 0.5000 0.5002 0.0000 Indicator Employee 1010 0.0000 1.0000 0.4465 0.4974 0.0000 Indicator Self Employed 1010 0.0000 1.0000 0.0317 0.1752 0.0000 Indicator Unemployed 1010 0.0000 1.0000 0.2040 0.4031 0.0000 Indicator Retired 1010 0.0000 1.0000 0.3119 0.4635 0.0000

4.2. Summary statistics of the middle-long run sample

The total number of valid observations, which contain data for all variables in the middle-long run sample, is 714. This is 34.16% of the total observations.

Firstly, the dependent variables are examined. Table IV, presented at the end of this paragraph, contains information related to the dependent variables. It shows that the median number of assets held is 1.0000. The mean number of assets held is 1.7269, which is 0.7269 higher than the median. This indicates a right tailed distribution. The standard deviation from the mean is 1.3260. In addition, it can be seen that at least one respondent holds seven out of the total 11 types of assets, and that at least one respondent does not invest at all. 25.21% of the respondents is investing in risky assets, shown by the mean for the indicator for risky assets held. This mean has a standard deviation of 0.4345, which is reasonably high.

Table IV

This table contains information related to the number of observations, the minimum, the maximum, the mean, the standard deviation and the median of the dependent variables in the middle-long run sample.

Variable Observations Minimum Maximum Mean Std. Deviation Median Number of Assets held 714 0.0000 7.0000 1.7269 1.3260 1.0000 Indicator Risky Assets 714 0.0000 1.0000 0.2521 0.4345 0.0000

Subsequently, the independent variables are presented at the end of this paragraph in Table V. This table encloses descriptive statistics of the independent variables. It can be seen that on average, the portfolio of respondents for 94.24% the same as five years before, with a standard deviation of 7.76%. Respondents also show a tendency to herd, indicated by the mean of 3.8613, which is above the neutral value of 3.000. It is also indicated by the respondents that they are risk averse, depicted by the mean of 5.4739. In addition, respondents regard themselves as neat persons, shown by the mean of 3.8890. With respect to the locus of control variables, there is no clear direction. Both the means of the internal and external locus of control variables are around the neutral response, with mean scores of respectively 3.6422 and 3.2319.

(25)

This table contains information related to the number of observations, the minimum, the maximum, the mean, the standard deviation and the median of the independent variables in the middle-long run sample.

Variable Observations Minimum Maximum Mean Std. Deviation Median Habitual behaviour in the last five years 714 0.6154 1.0000 0.9424 0.0776 1.0000 Herding behaviour 714 1.0000 5.0000 3.8613 0.6634 4.0000 Risk Aversion 714 2.6667 7.0000 5.4739 1.0141 5.5000 Neatness 714 1.2500 5.0000 3.8890 0.7812 4.0000 Internal Locus of Control 714 1.7500 7.0000 3.6422 0.7359 3.6250 External Locus of Control 714 1.0000 7.0000 3.2319 0.9727 3.2000

Lastly, descriptive statistics for the demographic variables are observed. Table VI, presented at the end of this paragraph, includes this information. The gender distribution shows that the middle-long run sample contains 39.78% females and thus 60.22% males. The average age of the respondents is 59 years old. The average household size is 2.3515. Again, this does not directly mean that a small margin of the households has children, since only 77.45% of the respondents is married and living together. The average income level per year is €23066.10, with a standard deviation of €15554.20. Consequently, although the average monthly income is modal, the standard deviation of this income is relatively large. The same conclusion can be made regarding wealth, with an average wealth of €58255.90 and a corresponding standard deviation of €129784.50. The distribution related to education level shows that 4.34% enjoyed low education, 46.78% enjoyed medium education, and 48.88% enjoyed high education. Regarding primary occupation, 42.86% of the respondents is employee, 2.94% is self employed, 20.59% is unemployed and 32.91% is retired.

Table VI

This table contains information related to the number of observations, the minimum, the maximum, the mean, the standard deviation and the median of the demographic variables in the middle-long run sample.

(26)

Indicator Unemployed 714 0.0000 1.0000 0.2059 0.4046 0.0000 Indicator Retired 714 0.0000 1.0000 0.3291 0.4702 0.0000

4.3. Summary statistics of the long run sample

The total number of valid observations, which contain data for all variables in the long run sample, is 262. This is 12.54% of the total observations.

Initially, the dependent variables are studied. Table VII, accessible at the end of this paragraph, comprises information related to the dependent variables. It shows that the median number of assets held is 1.0000. The mean number of assets held is 1.6221. This mean is higher than the median and thus indicates a right tailed distribution. The standard deviation from the mean is 1.2619. In addition, it can be seen that at least one respondent holds seven out of the total 11 types of assets, and that at least one respondent does not invest at all. 24.81% of the respondents is investing in risky assets, shown by the mean for the indicator for risky assets held. This mean has a standard deviation of 0.4327, which is rather high.

Table VII

This table contains information related to the number of observations, the minimum, the maximum, the mean, the standard deviation and the median of the dependent variables in the long run sample.

Variable Observations Minimum Maximum Mean Std. Deviation Median Number of Assets held 262 0.0000 7.0000 1.6221 1.2619 1.0000 Indicator Risky Assets 262 0.0000 1.0000 0.2481 0.4327 0.0000

Subsequently, the independent variables are presented at the end of this paragraph in Table VIII. This table encloses descriptive statistics of the independent variables. It can be seen that on average, the portfolio of respondents for 88.17% the same as ten years before, with a standard deviation of 9.82%. Respondents also show a tendency to herd, indicated by the mean of 3.8788, which is above the neutral value of 3.000. It is also indicated by the respondents that they are risk averse, depicted by the mean of 5.5153. Besides, respondents regard themselves as neat persons, shown by the mean of 3.9695. With respect to the locus of control variables, there is no clear direction. Both the means of the internal and external locus of control variables are around the neutral response, with mean scores of respectively 3.6584 and 3.2084.

Table VIII

This table contains information related to the number of observations, the minimum, the maximum, the mean, the standard deviation and the median of the independent variables in the long run sample.

(27)

Herding behaviour 262 1.5000 5.0000 3.8788 0.6758 4.0000 Risk Aversion 262 2.6667 7.0000 5.5153 1.0333 5.5000 Neatness 262 1.2500 5.0000 3.9695 0.7857 4.0000 Internal Locus of Control 262 1.7500 7.0000 3.6584 0.7633 3.6250 External Locus of Control 262 1.0000 6.2000 3.2084 0.9498 3.2000

Lastly, descriptive statistics for the demographic variables are observed. Table IX, presented on the next page, includes this information. The gender distribution shows that the long run sample contains 35.11% females and thus 65.89% males. The average age of the respondents is 65 years old. The average household size is 2.0725. The lower household size probably relates to the higher average age. In addition, only 74.81% of the respondents is married and living together. The average income level per year is €23126.50, with a standard deviation of €15862.30. Consequently, although the average monthly income is modal, the standard deviation of this income is relatively large. The same conclusion can be made regarding wealth, with an average wealth of €60075.60 and a corresponding standard deviation of €110609.20. The distribution related to education level shows that 7.63% enjoyed low education, 43.89% enjoyed medium education, and 48.47% enjoyed high education. Regarding primary occupation, 32.44% of the respondents is employee, 2.29% is self employed, 20.23% is unemployed and 45.04% is retired.

Table IX

This table contains information related to the number of observations, the minimum, the maximum, the mean, the standard deviation and the median of the demographic variables in the long run sample.

(28)

5. Results

This section will elaborate on the results of the analyses in this study. Results are divided into two subsections; the first subsection contains the results related to explaining the degree of diversity of the portfolio of the individual investor, the second subsection contains the results related to explaining which variables drive the individual investor to invest in risky assets.

5.1. Results explaining the degree of diversity

The influence of the independent variables on the degree of diversity is presented in this section. At first, the equation including the habitual behaviour of the last year is estimated. Afterwards, the results of this equation will be checked on the robustness by using different time spans for the habitual behaviour measure. This is done both by estimating the degree of diversity and the number of assets held.

5.1.1. Results in the short run sample

First, assumptions related to linear regression are tested for equation 2.1. Afterwards, results are presented and it will be concluded whether results support the hypotheses. In addition, it is tested whether results are stable, by using the Poisson regression as well to estimate the equation. The equation estimated in this section is the following equation 2.1:

+ εi

(29)

be that these respondents have a low ability of self reflection. Examination of the 13 outliers with a positive value show that they result from respondents who have a high degree of diversification, and at the same time have high values for the independent variables (causing the predicted value of the degree of diversification to be low). They already had diversified portfolios in the past as well, which means that their habitual behaviour is high and their degree of diversification is as well.

Now that the assumptions are tested, results related to equation 2.1 can be given. The overall equation is able to account for 35% of the variance in the degree of diversity, indicated by a R2 of 0.350. The standard error of the predicted value is 0.0889, which roughly translates to holding one asset more or less than predicted. ANOVA statistics indicate that the equation as a whole has significant predictive power (F(18,1009)=29.597, p=0.0000).

Table X, displayed on the next page, provides the estimated coefficients of the explanatory variables in equation 2.1. In addition, the t-statistics corresponding to the results described in this paragraph can also be found in Table X. When these effects are examined, it can be seen that the habitual behaviour in the last year has a significant negative relationship with the degree of diversity. This result shows support for hypotheses H1, indicating that the more respondents are making their investing decisions habitually, the less optimal their portfolio is. In addition, it is shown that herding behaviour does not have a significant relationship with the degree of diversity. Hypotheses H2 is thus not supported by this result, which means that there is no support to assume that there is a difference in the optimality of the portfolio of respondents which are herding relative to respondents who do not herd. Related to the four indicators for status quo bias, it is observed that all signs are negative. On the other hand, only the risk aversion and the external locus of control variable have a significant negative relationship with the degree of diversity. With 90% confidence, it can be concluded that there is a significant negative relationship between the internal locus of control variable and the degree of diversity. The neatness does not show a significant relationship with the degree of diversity. Related to hypotheses H3, it is unclear whether the relationship exists, results are not fully in line with each other. On the other hand, the signs of the relationships are all in line with the hypothesized relationship, so no opposing results are found. This implies that the more respondents show a status quo bias, the less optimal their portfolio is.

Table X

Linear regression results for equation 2.1

This table presents the estimated coefficients (unstandardized and standardized) and the corresponding t-statistic for equation 2.1.

(30)

Intercept 0.3915 6.7013*** Habitual behaviour in the last year -0.2456 -0.1720 -6.3898***

Herding behaviour 0.0028 0.0172 0.6316

Risk Aversion -0.0137 -0.1272 -4.6500***

Neatness -0.0027 -0.0194 -0.7144

Internal Locus of Control -0.0052 -0.0349 -1.3019* External Locus of Control -0.0132 -0.1171 -4.2336*** Gender (1=man, 2=woman) -0.0170 -0.0769 -2.5653**

Age (/100) 0.4918 0.6782 3.7172***

(Age/100)2 -0.4511 -0.6738 -3.5191***

Household Size -0.0065 -0.0700 -2.0099**

Marital Status (0=single, 1=married) -0.0039 -0.0153 -0.4900

Income (/€1000) 0.0010 0.1406 4.2612***

Wealth (/€1000) 0.0002 0.2507 9.0632***

Indicator Low Education -0.0351 -0.0663 -2.4413** Indicator Mediate Education -0.0247 -0.1126 -3.9318***

Indicator Employee 0.0369 0.1681 4.2006***

Indicator Self Employed -0.0298 -0.0478 -1.6942*

Indicator Retired -0.0132 -0.0562 -1.3294

*** Significant at less than the 0.01 level. ** Significant at less than the 0.05 level.

* Significant at less than the 0.10 level.

In addition, it can also be seen in Table X that out of the main independent variables, the habitual behaviour in the short run has the largest impact on the degree of diversity. This is indicated by the standardized β of -0.1720. Subsequently, risk aversion and the external locus of control variable have roughly the same impact. At last, the impact of the internal locus of control variable, the neatness and the herding behaviour is quite minimalistic compared to the other variables.

(31)

portfolio. Highly educated respondents also show significantly more diversity in their portfolio relative to mediate educated and low educated respondents. In addition, differences are observed between unemployed respondents and employees in favour of the employees. Employees probably have more options to diversify their portfolio, since a number of assets are only assessable when someone is an employee. In addition, self employed respondents diversify significantly less than unemployed respondents. At last, no significant differences are found between retired and unemployed respondents.

Subsequently, it is tested whether the results of equation 2.1 are stable, by estimating the number of assets held with a Poisson regression. For this analysis, the following equation (3.1) is estimated:

+ εi

Related to the assumptions, no issues arise. Linearity is assumed between the independent and dependent variables. In addition, observations are independent.

(32)

variable, this result is not matching the result of equation 2.1. On the other hand, in equation 2.1 the result was not very strong so the contradicting in between the analyses is not large. The result related to the neatness variable indicates that there is no difference in the number of assets held between tidy and messy respondents.

5.1.2. Robustness checks

By making use of three different time spans for the habitual behaviour variable, the results of the short run sample can be tested on their robustness over different periods of habitual behaviour. Results will thus indicate whether habitual behaviour in the middle-long and in the long run has the same effect as the habitual behaviour in the short run. In addition, it will be indicated whether the results for herding behaviour and status quo bias are stable. For the robustness check related to the middle-long run sample, the following equation (2.2) is estimated:

+ εi

The robustness check, making use of the long run sample, is performed with the following equation (2.3):

+ εi

Table DII, presented in Appendix D, shows the estimated coefficients and the corresponding t-statistics related to equation 2.2. It can be concluded that the result for habitual behaviour in the short run sample does match the result for habitual behaviour in the middle-long run sample. The results show a significant negative relationship between the habitual behaviour in the middle-long run and the degree of diversity. The relationship is significant with 90% confidence and thus not as strong as in the short run sample. Stable results are also found for herding behaviour, since in this equation, the result is also not significant. Related to the four indicators for the status quo bias, results are also stable. The risk aversion and the external locus of control variable indicate a significant negative relationship with the degree of diversity. The neatness and the internal locus of control variable do not show a significant relationship with the degree of diversity. All these findings are in line with the results of the short run sample.

(33)

the short run sample. A significant negative relationship is found between the habitual behaviour in the long run and the degree of diversity. Results related to the relationship of herding behaviour and the degree of diversity show that there is no significant relationship. This matches the result of the short run sample. The risk aversion shows a significant negative relationship with the degree of diversity. The other three indicators for status quo bias do not show a significant relationship with the degree of diversity. The insignificant result for the external locus of control strikes the result of the short run sample. The results for neatness and the internal locus of control variable match the results in the short run sample.

In addition, the analyses in the middle-long and the long run sample are also performed with the Poisson regression. In the middle-long run sample, this relates to the following regression equation 3.2:

+ εi

In the long run sample, the analysis is performed by making use of the following equation (3.3): (3.3) NAi = Intercept + β1*HB10i + β2*HeBi + β3*RAi + β4*Nei + β5*ILi + β6*ELi + β7*Age + β8*Age

2

+ εi

Table DIV, displayed in appendix D, shows the results for equation 3.2. In this equation, there is no significant relationship found between the habitual behaviour in the middle-long run and the number of assets held. This is contradicting the results of equation 2.2, though in equation 2.2 the result was not very strong. Herding behaviour still does not show a significant relationship with the number of assets held. Also the results of the status quo bias indicators are in line with previous findings. Risk aversion and the external locus of control variable show a significant negative relationship with the number of assets held. The neatness and the internal locus of control variable do not show a significant relationship with the number of assets held.

(34)

control variable all do not show a significant relationship with the number of assets held. The result for risk aversion, neatness and internal locus of control match the result of previous outcomes, the result for external locus of control is not in line.

5.2. Results explaining holding risky assets

The relationship of the explanatory variables with whether the respondents holds risky assets is examined in this section. At first, the equation including the habitual behaviour of the last year is estimated. Afterwards, the results of this equation will be checked on the robustness by using different time spans for the habitual behaviour measure. This is done both by using the Probit and the Logit regression estimation.

5.2.1. Results in the short run sample

First, assumptions related to linear regression are tested for equation 4.1. Afterwards, results are presented and it will be concluded whether results support the hypotheses. In addition, it is tested whether results are stable, by using the Logit regression as well to estimate the equation. The equation estimated in this section is the following equation 4.1:

(4.1) IRi = Intercept + β1*HB1i + β2*HeBi + β3*RAi + β4*Nei + β5*ILi + β6*ELi + β7*Age + β8*Age2 +

β9*Sex + β10*Inc + β11*Wea + β12*MaS + β13*LEd + β14*MEd + β15*Emp + β16*SEm + β17*Reti +

εi

Before the results of the analyses are presented, the assumptions related to the Probit model are tested. Results related to the assumptions can be found in Table CII, in Appendix C. Firstly, no problems are observed related to the heteroscedasticity assumptions. In addition, when the Z-scores for each individual respondent is calculated, it can be seen that the Z-scores follow the normal distribution. The residuals, on the other hand, do not follow a normal distribution. However, in general the estimations related to the Probit model are quite robust, which eases the problem. In addition, it is checked whether results are stable, by estimating the same equations with the Logit model as well.

(35)

Table XI, shown on the next page, presents the results of the estimated coefficients and the Z-scores for equation 4.1. The result related to the habitual behaviour in the short run indicates that there is a significant negative relationship observed. It can be concluded that performing investment decisions habitually significantly decreases the probability of investing in risky assets, which is in line with hypothesis H1. On the other hand, hypothesis H2 is not confirmed by the results related to herding behaviour. Results show that an increased tendency to herd does not significantly lower the probability of investing in risky assets. Results related to the four indicators of status quo bias in most cases show support for hypothesis H3. Risk aversion, internal locus of control and external locus of control all show a negative relationship with the probability of investing in risky assets. It thus indicates that the more averse the respondent is to taking risk, the lower the probability of investing risky assets. In addition, the locus of control variables both conclude that the more someone feels his /her luck is not in his/her own hands, the less the probability of investing in risky assets. The only status quo bias variable which is not significantly related to the probability of investing in risky assets, is the neatness of the respondent.

Table XI

Probit regression results for equation 4.1

This table presents the estimated coefficients and the corresponding Z-score for equation 4.1.

Variable β Z-Score

Intercept 4.0855 3.4984***

Habitual behaviour in the last year -3.0420 -4.4353*** Herding behaviour -0.0759 -0.9527

Risk Aversion -0.3392 -5.3574***

Neatness -0.0376 -0.5679

Internal Locus of Control -0.0952 -1.3317* External Locus of Control -0.2487 -3.9630*** Gender (1=man, 2=woman) -0.0345 -0.2929

Age (/100) 3.9703 1.5552

(Age /100)2 -2.3913 -1.0008

Household Size -0.0816 -1.4104

Marital Status (0=single, 1=married) 0.1113 0.7830

Income (/€1000) 0.0064 1.7609*

Wealth (/€1000) 0.0027 6.2083***

Indicator Low Education -0.3627 -1.3456 Indicator Mediate Education -0.4562 -3.5308***

Indicator Employee 0.0645 0.3940

Master Thesis Finance Habits in Investing

Master Thesis Finance

Habits in Investing

Contents

1.

Introduction

2.

Literature review

3.

Research design

4.

Descriptive statistics

5.

Results