1
Master Thesis
Optimism and household portfolio choices:
empirical evidence from Dutch household survey
by Sargis Adikyan
MSc. Finance
MSc. Finance and Risk Management
Faculty of Economics and Business
Faculty of Economics and Business
Administration
University of Groningen
Alexandru Ioan Cuza University of
Iasi
Author:
Sargis Adikyan
2
Optimism and household portfolio choices: empirical evidence
from Dutch household survey
Sargis Adikyan
Abstract
In this study, we analyse the relationship of optimism and household level financial decisions. Using data from the DNB Household Survey and Human Mortality Database, we create a sample of 1,657 Dutch households counting 5,290 person/year observations for the period 2002 to 2009. The results show that optimism negatively affects the probability of owning stocks and risky financial assets in general. In addition, it is shown that there is no significant relationship between optimism and asset allocation decisions. The effect of optimism on financial decisions is further investigated for different subgroups of the population by disaggregating data based on individual level and household level characteristics such as gender, education and household size.
Keywords:
Dispositional optimism; Asset ownership; Asset allocation; Household finance; DNB Household Survey.3 Table of contents
1. Introduction ………. 5
2. Literature review ………... 6
3. Research design and methodology ………... 12
3.1. Hypothesis development ………... 12
3.2. Model specification ………... 14
3.3.Variable selection ……… 16
4. Data and descriptive statistics ……….. 19
4.1.Sample selection ………... 20 4.2.Variable description ………... 20 4.3. Descriptive statistics ……… 22 5. Results ……….... 27 6. Conclusion ………. 34 6.1 Main discussion ………... 34
6.2 Limitations and suggestions for further research ……….. 36
7. References ……….. 38
4 List of Tables
Table 1: Summary statistics of the main variables ……… 23
Table 2: Dynamics of ownership and optimism indicators ……….. 25
Table 3: Summary statistics for optimistic households ………. 26
Table 4: Regression results: ownership of risky assets and stocks ……….. 28
Table 5 Regression results: allocation of assets, OLS estimator ... 31
Table 6: Regression results: allocation of assets, Fixed effects estimator ………... 33
5 1.
Introduction
The portfolio selection process is divided into two stages according to Markowitz (1952). The first stage is concerned with observation and experience and summarizes the beliefs regarding the future performance of available securities. Then, having relevant beliefs regarding the future performances of the securities, the second stage results in the choice of the portfolio. Even though the study of Markowitz (1952) and Modern Portfolio Theory are more concerned with the second stage of portfolio selection, the role of individual level beliefs should not be neglected. The main assumptions of modern portfolio theory relate to homogeneous expectations of investors regarding the future returns of securities and to the fact that investors are rational. What if the expectations of individuals are formed not only by the available market information that they possess, but also by rather subjective and psychological factors? It is reasonable to assume that beside the attitude towards the risk of investor which affects the distribution of his money between risky and risk-free assets (Tobin, 1958), there are also other individual subjective level factors which can affect the distribution of money. Factors such as happiness, overconfidence, mood, and optimism have been widely discussed in the economic literature in this context in recent years. The subjective characteristic that is directly linked to the expectations and is considered as a personality trait is optimism. Compared to overconfidence or happiness, optimism is a broader category, which presents a bias in general expectancies of individuals. The generalized bias in expectations may influence every aspect of human life, including health, work, and financial decisions. In addition, it is shown that optimistic bias affects not only mental and physical, but also the economic well-being of individuals. There are many studies, that analyse the effect of optimism on various economic and financial decisions for different subgroups of population. However, the relationship of optimism and household level financial decisions is not thoroughly discussed in the literature and we aim to address this problem in our study.
6 allocation, but also ownership decisions of households, which has not been analysed in previous studies. We find evidence that optimism decreases the probability of owning risky assets and stocks by the households. At the same time, we indicate that optimism is not related to asset allocation decisions. We disaggregate our sample into different subgroups and analyse the moderating effects of higher education, household size, gender and other household characteristics. We find that the relationship of optimism and risky asset ownership decisions is not affected in any of the subgroups. In contrast, subgroup effects are significant in the relationship of optimism and stock ownership decisions.
The remainder of our study is organised as follows: Chapter 2 discusses the relevant literature of optimism and its relation with financial decisions, Chapter 3 provides the design of hypotheses and methodology, which is used in empirical analysis, Chapter 4 presents the available data and discusses sample selection process, Chapter 5 introduces results of regression analysis. Chapter 6 concludes our work and discusses possible limitations and further research questions.
2. Literature review
The following section discusses the definition of optimism and the existing literature about the relation of optimism with different aspects of human life. Further, different measures of optimism are presented, including direct psychometric tests and other individual study level instruments. Finally, a brief overview of the existing literature on the relation of optimism and financial decisions is presented.
Definition of optimism: Optimism is one of the general types of expectancies of the
individual. It is a broader category as it reflects the person’s general attitude towards different situations of life. In general, optimism is a psychological concept, which is defined by the personal expectation of the individual that good things will happen to oneself. Carver and Scheier (1985) view optimism as a global expectancy, which is relatively stable over time and context. They label this characteristic as dispositional optimism1, which is widely used in the field of economics. Dispositional optimism is defined as a generalized positive expectation of the individual about future outcomes in life.
1 Dispositional optimism is the scientific name of the psychological concept of optimism, however for writing
7
The role of optimism: Many studies have examined the effect of dispositional optimism
on different aspects of human life. There is evidence that dispositional optimism affects the physical and mental health of the individuals. According to a meta-analysis of 84 studies (Rasmussen, Scheier and Greenhouse, 2009), optimism significantly influences physical health. Optimists are less distressed in difficult situations and easily recover after a short period of time. Dispositional optimism is considered a personality trait, which can influence and predict human behavior. Particularly, dispositional optimism affects the ways individuals cope with difficulties (Scheier and Carver, 1992). This study shows that optimists face the difficulties without denial of the problem. They are interested in the details of the problem concentrating on problem solving. In contrast, pessimists tend to avoid the acceptance of the problem, they use emotion-based coping and get easily distressed in adverse situations.
Most of the research, which investigates the behavioral differences between optimists and pessimists, is in the context of health difficulties. Out of this context, researchers are interested in whether optimism can affect everyday decisions in combination with different characteristics of the individuals. Having generalized positive expectations regarding the future outcomes, it is logical that optimists will act according to their perception of the reality, which will play a deterministic role in the construction of their behavior.
8 currently, but they also want to work more for the future years, as optimistic people prefer to retire later.
So far, we have shown that optimism can influence physical, mental, and economic well-being of an individual. At the same time, many studies view specific levels of optimism as a characteristic of a particular group or occupation. There is evidence that entrepreneurs (Fraser et al., 2006) and inventors (Astebro, Jeffrey and Adomdza, 2007) are more optimistic than employees and the general population. Based on consumer confidence surveys of USA and 17 European countries, men are more optimistic than women are (Jacobsen et al., 2014). In addition, it is argued that not only optimism itself, but also the level of optimism is crucial for the well-being of an individual (Puri and Robinson, 2007). While moderate levels of optimism are constructive for the individual, extreme levels of optimism can lead to financial and health problems. So the measurement and categorization of optimism is another important topic to be discussed.
The measurement of optimism: The measurement of optimism is discussed in the
literature from the following perspectives: general expectations and attributional style, where the latter is the habitual way of explaining the origins of undesirable outcomes. In the perspective of general expectations, dispositional optimism is defined as suggested by Scheier and Carver (1985) with the corresponding definition of dispositional pessimism as generalized negative expectations about the future events. In the perspective of attributional style, optimistic people are those who consider the negative effects and their causes as external, temporary and unique whereas pessimistic people consider them as internal, constant and common. The research instruments, which present the attributional measures of optimism, are the Attributional Style Questionnaire (ASQ) and Expanded Attributional Style Questionnaire (EASQ) suggested by Peterson et al. (1982) and Peterson et al. (1988). Both questionnaires present different events to the individuals and ask them to assess the causes of those events supposing that events were to happen to them. EASQ presents only negative events, but ASQ presents both negative and positive events. Respondents need to assess the causes of the events based on three scales that express internality, stability and globality. Life orientation test (LOT) (Scheier and Carver, 1985) and broadly used revised Life orientation test (LOT-R) 2 (Scheier et al., 1994) are the methods of measurement of dispositional optimism from the expectancy perspective. The LOT-R is a questionnaire type test, which
2 The LOT-R test is publicly available at the following web-address:
10
Optimism and Financial decisions: Another valuable insight in this topic that is the main
interest of our study is the relation of optimism and financial decisions. In their household level extensive study, Puri and Robinson (2007) examine the effect of dispositional optimism on financial decisions of individuals. The study uses the Survey of Consumer Finance data for the U.S. economy for the years 1985, 1988 and 1991. Using the OLS estimator on time-averaged variables, it is shown that dispositional optimism affects the distribution of equity wealth among various instruments, however, it is not associated with the portfolio distribution between equity and debt. The corresponding dependent variables in the analysis are the ratios of stock investments over total equity and total equity over financial wealth. Further, using probit model, the authors show that optimistic individuals are more likely to spend less than their income. In addition, they argue that optimism positively affects total savings of the individuals standardized by the income level. The overall study focuses on asset allocation and saving decisions, leaving a question about the effect of optimism on the asset ownership decisions. The study of Rao et al. (2014) captures this question. The authors investigate the relationship between happiness and stockholding. At the same time, the effect of optimism on asset ownership decision is indirectly analysed in the study as they include optimism as a control variable. The dataset contains China Household Finance Survey data for the year 2011 that cover more than 8,000 households. They apply OLS estimator with tobit model specification and show that optimism positively affects the ratio of wealth invested in stocks through direct holdings. This result is analogous to the results of Puri and Robinson (2007). Moreover, the relationship is significant for the indirect holdings as well. Further, the authors show that optimism has explanatory power over the decision of directly or indirectly owning a stock, hence optimism affects the decision of stock market participation. It is worth noting that the measures of optimism in the studies of Puri and Robinson (2007) and Rao et al. (2014) are significantly different, again indicating the variability of measurement techniques. The former uses Life Expectancy Miscalibration and the latter uses principal component analysis to measure optimism based on households’ predictions of Chinese economy and consumer price index.
11 Puri and Robinson (2007) suggest that the effect of optimism on economical choices can be explained by the nature of optimism. In particular, optimism can arise from overconfidence and self-attribution bias. In this case, it is reasonable to consider that optimism affects the cautiousness of the household, more specifically optimistic households make imprudent and risky portfolio choices. Another explanation for the relationship is presented in the study of Christensen et al. (2006). The authors show that optimism positively affects the self-reported level of financial expertise of individuals. Optimistic individuals are more knowledgeable about their financial situation and about the economic developments. Both measures of expertise are subjective and self-reported. Further, they argue that there is a relation between optimism and the perception of the economic growth. Optimistic people are more likely to believe that there will be a high economic growth. These results suggest an alternative channel through which optimism affects financial decisions. The subjective perception of the future events not directly relating to the individual and the feeling of awareness about them is another way by which optimism affects financial decisions. In this context, it is logical to argue that positive expectations about the economy will infer same attitude towards the stock market, especially towards the stock returns. Finally, we can argue herein that optimism will affect the riskiness of the financial decisions, as optimistic people have all reasons to prefer risky assets to less risky investments. This relationship is particularly the main subject of our analysis.
12 3. Research design and methodology
This section presents a discussion of the hypotheses we suggest to check in the scope of our research. Further, we present the model specification and statistical techniques that are used in the empirical analysis. It is then followed by a discussion of dependent and independent variables and the rationale behind their inclusion in the analysis.
3.1 Hypothesis development
Apart from the investigation of the effect of optimism on financial decisions, it is of particular importance to understand the nature of the relationship. The objective of our analysis is based on the understanding of channels through which optimism affects financial decisions. It is presented in the literature that overconfidence, self-attribution bias, and subjective perception of events not relating to the personality are the channels through which optimism affects financial decisions. All three rationales eventually suggest that optimism affects the riskiness of financial decisions (Christensen et al., 2006; Puri and Robinson, 2007). So the risk is the spectrum through which we will investigate the relationship of optimism and financial decisions.
In order to empirically examine the set of questions that we address in our study, we present our hypotheses. We formulate the hypotheses in three levels: ownership, allocation and subgroup levels. The first level is ownership hypotheses. Optimistic people have generalized positive expectations about future outcomes and they work longer and retire later. Optimistic people value their work product and prefer to work for companies that use advanced technologies. In addition, optimistic people have positive expectations about the economy and they are more likely to believe that the economy will grow. We assume that beliefs of economic growth will infer same attitude towards the financial markets, particularly the stock market. In the end, we argue that optimistic people have enough incentives to invest in risky assets. We present our argument in two ownership level hypotheses. The first hypothesis presents the association of optimism and risky asset ownership decision and the second hypothesis presents the association of optimism and particular risky asset, which is the stock. The latter one is designed to check directly the effect of optimism on the decision of the stock market participation.
Ownership Hypothesis 1: More optimistic households are more likely to invest in risky assets
13
Ownership Hypothesis 2: More optimistic households are more likely to invest in stocks than
less optimistic households are.
We assume that the qualities that optimistic people have during difficult situations and the coping style of optimists help them to sustain a stable position in financial markets as well. Optimistic people are not easily stressed in adverse situations and they prefer to find solutions to their problems without denial, concentrating on the thorough investigation of the problem (Scheier and Carver, 1992). In addition, it is shown that optimistic individuals tend to believe that they have a higher level of financial expertise and are knowledgeable about the financial developments of their household as well (Christensen et al., 2006). Following these notions and combining them with our findings of the risk profile of optimistic individuals, we present allocation hypotheses. First, we argue that optimistic individuals have a higher ratio of their total assets kept in financial assets. Second, we assume that optimistic individuals allocate higher ratio of financial assets to risky assets. Finally, we want to check whether there is any preference among traditional investments in stocks and other risky assets following the evidence that optimists prefer to be stock pickers that is they prefer to invest in individual stocks (Puri and Robinson, 2007).
Allocation Hypothesis 1: More optimistic households allocate a greater share of their total
assets into financial assets than less optimistic households do.
Allocation Hypothesis 2: More optimistic households allocate a greater share of their
financial assets into risky financial assets than less optimistic households do.
Allocation Hypothesis 3: More optimistic households allocate a greater share of their risky
financial assets into stocks than less optimistic households do.
14 decisions. We assume that marital status and higher education are characteristics that can affect both the confidence and the self-reported financial literacy of the individual. Larger household size can increase the confidence of the individual, but at the same time, it increases his responsibility towards the other members of the household making him more cautious and prudent. So we are uncertain about the effect of household size on the relationship of optimism and financial decisions.
Herein we design subgroup hypotheses to present the moderating effect of marital status, higher education, gender, household size and the fact of main wage earning on the relationship of optimism and financial decisions, addressed in previous hypotheses. We present the moderating effect of family size in separate hypothesis, as we are not certain about the direction of the effect.
Subgroup Hypothesis 1: Belonging to the subgroup of male/higher educated/married/main
wage earner individuals, positively affects the relationship between optimism and household financial decisions, presented in ownership and allocation hypotheses.
Subgroup Hypothesis 2: Household size is a moderator variable in the relationship between
optimism and household financial decisions, presented in ownership and allocation hypotheses.
3.2 Model specification
The panel structure of our dataset allows us to analyse the effect of optimism on financial decisions taking into consideration the time variation of the variables. We present our model specification, which is then estimated using linear regression techniques. The main specification of our model has the following form:
𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐷𝑒𝑐𝑖𝑠𝑖𝑜𝑛𝑖,𝑡 = 𝛼 + 𝛽1𝑂𝑝𝑡𝑖𝑚𝑖𝑠𝑚𝑖,𝑡+ 𝛽2 (𝐼𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙 𝑙𝑒𝑣𝑒𝑙 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠)𝑖,𝑡+
+ 𝛽3 (𝐻𝑜𝑢𝑠𝑒𝑑𝑜𝑙𝑑 𝑙𝑒𝑣𝑒𝑙 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠)𝑖,𝑡+ 𝜆𝑡+ 𝜇𝑖+ 𝜀𝑖,𝑡 ,
(1)
where Financial Decision corresponds to asset allocation and asset ownership decisions presented in the hypotheses, Individual level controls refer to all control variables except household net income, household net worth and household size, which are presented as
Household level controls. Consequently, 𝜇𝑖- is the entity-specific effect which controls for
15 We start the analysis by the Pooled regression model using Ordinary Least Squares (OLS) estimator. In the Pooled OLS analysis, the observations are combined in a one set where the time and entity dimensions are neglected. The autocorrelation in variables, which is highly probable for surveys, may cause biased standard errors of estimates. Therefore, we run the Pooled OLS regression with cluster robust standard errors where the cluster dimension is the survey respondent. In case of binary dependent variables of risky asset ownership and stock ownership, we conduct analysis using a linear probability model with OLS estimator with cluster robust standard errors as well. The linear probability model has its drawback, as it is possible to predict probabilities that are out of the range (0, 1), however, acknowledging the issue we prefer to use this model as a starting point as it is easy to estimate and we are not involved in predictions, but we are rather interested in significant associations. In both models, we control for time trends, including a full set of time dummies.
Next, we analyse the relationship between optimism and financial decisions using individual-specific effects model, in order to exploit the panel structure of our dataset. In this model, the slope coefficient is kept constant, but the model allows variations in the intercept that captures the unobserved heterogeneity of the data (Cameron and Trivedi, 2005). The model is either defined as random effects or as fixed effects based on the assumption regarding the correlation of intercept and explanatory variables. The drawback of the random effects model is that it is valid when the composite error term is uncorrelated with all of the explanatory variables. Another inference is that the random effects model is valid when any unobserved omitted variable is uncorrelated with explanatory variables (Brooks, 2008). In case of microeconomic dataset, this assumption is very strict. We perform Hausman test in order to choose between the models. The null hypothesis of the test is that the coefficients of the efficient random effects estimator are equal to the coefficients of consistent fixed effects estimator. The fixed effects model is used when the null hypothesis is rejected. As it was expected, the fixed effects model is an appropriate one for our analysis. We perform regressions with robust standard errors, introducing a full set of time dummies.
16 𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐷𝑒𝑐𝑖𝑠𝑖𝑜𝑛𝑖,𝑡 = 𝛼 + 𝛽1𝑂𝑝𝑡𝑖𝑚𝑖𝑠𝑚𝑖,𝑡+ 𝛽2 (𝐼𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙 𝑙𝑒𝑣𝑒𝑙 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠)𝑖,𝑡+
+ 𝛽3 (𝐻𝑜𝑢𝑠𝑒𝑑𝑜𝑙𝑑 𝑙𝑒𝑣𝑒𝑙 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠)𝑖,𝑡+ 𝛽4((𝐼𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙 𝑙𝑒𝑣𝑒𝑙 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠)𝑖,𝑡× 𝑂𝑝𝑡𝑖𝑚𝑖𝑠𝑚𝑖,𝑡) +
+𝛽5((𝐻𝑜𝑢𝑠𝑒𝑑𝑜𝑙𝑑 𝑙𝑒𝑣𝑒𝑙 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠)𝑖,𝑡× 𝑂𝑝𝑡𝑖𝑚𝑖𝑠𝑚𝑖,𝑡) + 𝜆𝑡+ 𝜇𝑖+ 𝜀𝑖,𝑡, (2)
where coefficients 𝛽4 and 𝛽5 summarize the effect of the Optimism on Financial Decisions
for different individual and household level subgroups. Their sign will indicate whether the relationship between optimism and a particular financial decision is amplified or weakened for different subgroups.
3.3 Variable selection
While analyzing the relationship between optimism and household financial decisions, it is important to identify the object of analysis. The study of Puri and Robinson (2007) analyses the relationship of optimism and financial decisions at the individual level. However, it is likely that in case of financially illiterate individual or when an individual is not responsible for financial decisions, the decision of asset ownership or allocation can be driven by the advice of the financial consultant or by any other interconnected people such as spouse, parents, children or other relatives. In the study of Rao et al. (2014), the relationship between optimism and financial decisions is presented at the household level. Particularly, those individuals who are the oldest and have the highest level of education among the other household members are selected as representatives of the household. This approach has its drawbacks as well. The highest level of education and age do not necessarily relate to financial decisions as important financial decisions require financial expertise and understanding of financial matters. In our analysis, we follow the approach of Rao et al. (2014), however, we admit that the criterion for the selection of the representative should be adjusted. Summarizing, the object of our study is the household, which is considered as a single decision-making unit where financial administrators of the household make financial decisions. Consequently, all variables that we include in our study are discussed in this perspective.
Dependent variables: Following the design of proposed hypotheses, the selection of
17 The variables are updated annually, so for simplicity, we will assume that the values of variables are constant during the year. The second set of dependent variables corresponds to the financial decisions of asset allocation presented in allocation hypotheses. This set of dependent variables includes three variables. First, the ratio of financial assets over total assets, which describes the level of involvement of the household in financial markets. Second, the ratio of risky financial assets over financial assets, which describes the risk level of financial assets of the household. Third, the ratio of stock investments over risky financial assets, which describes the investment preference of the household between stocks and other risky financial assets.
Independent Variable: The main independent variable is dispositional optimism. Scheier
and Carver (1985) introduce dispositional optimism as generalized positive expectations about future events. In our analysis, we will use the measure suggested by Puri and Robinson (2007). According to this study, the measure of the dispositional optimism is the Life Expectancy Miscalibration calculated as the difference between self-reported life expectancy and the life expectancy inferred from actuarial tables. We do not argue that Life Expectancy Miscallibration is the most appropriate and accurate measure of optimism, however, it is a unique tool that can be implemented for a large-scale study. In addition, it provides the necessary robustness for the inclusion in the analysis.
Control Variables: In our analysis, we include individual level control variables such as
18 The relative risk aversion defined by Pratt (1964) and Arrow (1965) is a measure of attitude towards risk. When the increase of wealth results in an allocation of a greater proportion of wealth in risky assets, then households present decreasing relative risk aversion. In the case when an increase of wealth results in an allocation of a smaller proportion of wealth in risky assets, then households present increasing relative risk aversion. Pratt (1964) and Arrow (1965) hypothesize that the relative risk aversion increases with wealth. Many studies address this question for different subgroups of the population, different countries, and different periods. It is shown in the study of Cohn, Lewellen et al. (1975) that relative risk aversion declines with wealth when the latter is measured as the value of total marketable assets and not as net worth. Friend and Blume (1975) provide evidence that households are described by constant relative risk aversion. In this study, household wealth is calculated as net worth. In addition, they show that when household wealth is measured excluding automobiles and homes, then households are described by decreasing risk aversion. The studies of Siegel et al. (1982),Morin and Suarez (1983), Riley and Chow (1992), Halek and Eisenhauer (2001) and Watson and McNaughton (2007) investigate the effects of demographic and financial variables on relative risk aversion. It is shown that for less wealthy households the relative risk aversion increases (Siegel et al., 1982;Morin & Suarez, 1983) and when the household is in the wealthy top 10 % of the population the relative risk aversion decreases (Riley and Chow, 1992). The relationship between relative risk aversion and age is negative up to the age 65 (Riley and Chow, 1992; Halek and Eisenhauer, 2001). After the age 65, households exhibit an increasing risk aversion. Relative risk aversion is lower for males and is higher for married and self-employed individuals. With an increase in education level households present decreasing relative risk aversion (Riley and Chow, 1992). Generalizing, it is straightforward that relative risk aversion exhibits various patterns in different studies, which indicates that demographic variables and household financial measures affect the relative risk aversion and hence asset allocation measures, however the sign and significance of the relationship depends on a particular sample.
19 more likely to participate in the stock market as they have more cumulative information about the stock market (Bertaut, 1995). In all three studies, marital status, particularly being married and family size are not significant factors of the stock market participation. Wealthier households are more likely to invest in stocks, as they are not concerned about illiquidity problems. The wealth and income are shown to positively affect the decision of the stock market participation (Bertaut, 1998; Guiso et al., 2003).
The control variables of the main wage earner and household size are not widely discussed in the literature and we could not find any significant results indicating their effect on financial decisions. Our rationale of the selection of these variables is that we assume that being the main wage earner the financial administrator is more confident and exhibits a riskier financial behavior. The effect of household size is dichotomous: larger household increases the confidence of the financial administrator and is a supportive factor for his financial decisions. On the other hand, the financial administrator feels higher responsibility for other household members and is more prudent and cautious.
The demand for financial assets of the households described by the ratio of financial assets over total assets is not broadly discussed in the literature. The unique study that includes the ratio of financial assets over total assets in regression analysis is the study of Milligan (2005). It is shown that the age of the household positively affects the share of financial assets in total assets. The financial assets are the sum of risky and risk-free financial assets. The demographics and financial variables affect the ratio of financial assets over total assets by the means of risky financial assets. More precisely, keeping constant the value of real assets, the increase in risky financial assets increases the ratio of financial assets over total assets. Thus, ceteris paribus, we expect to have the same direction of influences of demographics and financial variables on the ratio of financial assets over total assets as we observe for the relative risk aversion.
4. Data and descriptive statistics
20 about assets, liabilities and income. In addition, there are two data sets that include aggregated data on income and aggregated data on assets, liabilities and mortgages that are presented at the individual level. The survey was initiated for almost 3,000 households in the year 1993 and by the year 2014, there were 2,072 households that have participated in the 22th wave of the survey. The dataset initially is not presented in a panel format and consists of 168 separate data files that are further combined manually.
The second source of the data is the Human Mortality Database3, which provides
mortality data consisting of the period life tables and Cohort life tables. We use this database in order to obtain actuarial data for our respondents from the DNB household survey. The actuarial data is further used to calculate dispositional optimism, our main independent variable.
4.1 Sample selection
The DNB Household Survey provides annual data for the period 1993 to 2014. Financial data is presented in different currencies before and after the year 2002, as long as starting from January 1, 2002 the Dutch guilder was replaced by euro coins and banknotes4. In our research, we use data for the period 2002 to 2014 in order to avoid the exogenous effects of the currency adoption and change in the country. The second determinant of the sample selection is the fact that Human Mortality Database does not provide period life tables for the period 2010 to 2014. Thus, the sample, which combines the DNB Household Survey and Human Mortality Database contains data for the period 2002 to 2009 and counts 36,037 observations. After restricting the sample to only financial decision makers, we obtain 14,855 observations. Finally, after dropping missing data on dependent, independent variables, and keeping the sample constant for all estimations, we obtain a final sample of 5,290 observations.
4.2 Variable description
In this section, we present the formal questionnaire definitions and constructions of dependent and independent variables.
3 The Human Mortality Database provides mortality and population data for 37 countries. The database is a joint
project of the Department of Demography of the University of California, Berkeley, USA, and the Max Planck Institute for Demographic Research in Rostock, Germany. Available at www.mortality.org or www.humanmortality.de.
4 Dutch guilder was the currency of The Netherlands until the adoption of the euro. The Netherlands adopted
21 We consider the household as a single decision-making unit, where the financial administrator who is in charge of financial decisions is the representative of the household. Our sample is restricted to the financial administrators of the households who are defined by the questionnaire as follows: “Are you the person who is most involved with the financial administration of the household? By financial administration, we mean making the payments for rent or mortgage, taking out loans, taking care of tax declarations, etc.” The financial administrator allocates total household wealth and is responsible for asset ownership decisions even if he is not the owner of the asset. Therefore, net income, net worth, the value of financial and risky financial assets5 and the market value of stocks are calculated at the household level as a sum of individual level values for the same variables. Then the household level value for a specific year is attached to the financial administrator.
We construct the risky financial assets as the sum of the following assets: investments in growth funds, mutual funds, bonds and mortgage bonds, stocks and shares, options, stocks from substantial holdings and employer-sponsored savings plan. The market value of stocks is the sum of minority and majority holdings excluding the value of shares, which are controlled through funds. Using the definitions of financial assets and risky financial assets as well as the market value of stocks, we construct dependent variables corresponding to asset allocation decisions. The dependent variables are as follows: the ratio of financial assets over total assets (Fin/Tot), the ratio of risky financial assets over financial assets (RiskyFin/Fin), the ratio of the market value of stocks over risky financial assets (Stock/RiskyFin).
The dependent variables corresponding to asset ownership decisions are binary variables defining the ownership of risky financial assets (RiskyFin_i) and stocks (Stock_i). These variables are calculated at the household level, indicating the ownership of the asset at least by one member of the household.
Our main independent variable is dispositional optimism (Optimism). It is calculated as a life expectancy miscalibration suggested by Puri and Robinson (2007). Since the life expectancy of the survey respondents is presented in probabilities, our measure of optimism is defined in probabilities as well and has the following form:
5 The value of financial assets is the sum of the values of the following components: checking accounts,
22 𝑂𝑝𝑡𝑖𝑚𝑖𝑠𝑚𝑖 = 𝐸𝐿(𝑆𝑢𝑟𝑣𝑖𝑣𝑖𝑛𝑔 𝑡𝑜 𝑁 | 𝑋) − 𝐸𝐴(𝑆𝑢𝑟𝑣𝑖𝑣𝑖𝑛𝑔 𝑡𝑜 𝑁 | 𝑋), (3)
where 𝐸𝐿(𝑆𝑢𝑟𝑣𝑖𝑣𝑖𝑛𝑔 𝑡𝑜 𝑁 | 𝑋) is the subjective probability of the individual to survive to
age N conditional on the vector of personal characteristics X at the moment of the survey. 𝐸𝐴(𝑆𝑢𝑟𝑣𝑖𝑣𝑖𝑛𝑔 𝑡𝑜 𝑁 | 𝑋) is the probability of surviving to age N conditional on the vector of personal characteristics X at the moment of the survey taken from actuarial tables. X is a vector of personal characteristics such as gender, health, age that can affect the subjective probability of the individual about his mortality. In our study the self-reported life expectancy is calculated based on the following question from the questionnaire: “How likely is it that you will attain at least the age of 80” with the corresponding scale from 0 to 10. Then we rescale the answer to the interval from 0 to 1. The vector of personal characteristics is limited in our data to gender, year of birth and survey date. The actuarial life expectancy is calculated using the data provided by the Human Mortality Database for The Netherlands conditional on the year of birth, gender and survey date. The data presents the probability of surviving 1 year for different age, gender and survey dates. Then we calculate the probability of reaching a particular age N as a product of probabilities of surviving the preceding years. The definitions of control variables that we use in our analysis is presented in Table 1.
4.3 Descriptive statistics
We have longitudinal data ordered by the time dimension of year and the entity dimension of the unique identification code of each financial administrator. The panel dataset is unbalanced. The time dimension has a minimum value of 1 year and max value of 8 years. We keep the sample constant for all regressions. The variables contain 5,290 observations consisting of 1,657 distinct households that provide self-reported data for 3.2 years on average. The ratio of the stock over risky financial assets has fewer observations as it is not defined for households who do not own risky assets.
23 Table 1
Summary statistics of the main variables
This table presents the summary statistics for main variables. The table presents the means, standard deviations, minimum and maximum values as well as the number of observations for each variable in three levels: the overall statistics, the between statistics counting for differences trough different entities (respondents) and the within statistics counting for differences through time for each entity. Numbers are rounded to the third decimal, half is rounded up.
Variable Description Mean Std. Dev. Min Max Observations
Fin/Tot Ratio of value of Financial assets over Total assets
overall 0.26 2.62 -188.08 3.46 N = 5,290
between 2.34 -93.74 1.66 n = 1,657
within 1.84 -94.08 94.61 T-bar = 3.19
RiskyFin/Fin Ratio of value of Risky financial assets over Financial assets
overall 0.23 1.70 -20.74 101.31 N = 5,290
between 0.93 -6.21 34.07 n = 1,657
within 1.43 -33.84 67.47 T-bar = 3.19
Stock/RiskyFin Ratio of value of Stocks over Risky financial assets
overall 0.12 0.28 0.00 1.00 N = 3,344
between 0.25 0.00 1.00 n = 1,154
within 0.12 -0.63 0.98 T-bar = 2.90
RiskyFin_i
Risky Financial asset indicator. Describes whether the household has risky financial assets as of the beginning of the year. Binary variable: 0-No risky assets, 1- Ownership of any risky assets
overall 0.63 0.48 0.00 1.00 N = 5,290
between 0.45 0.00 1.00 n = 1,657
within 0.24 -0.24 1.51 T-bar = 3.19
Stock_i
Stock indicator. Describes whether the household has investments in stocks excluding any holdings through funds as of the beginning of the year. Binary variable: 0-No, 1-Yes
overall 0.15 0.36 0.00 1.00 N = 5,290
between 0.33 0.00 1.00 n = 1,657
within 0.15 -0.72 1.03 T-bar = 3.19
Bond_i
Bond indicator. Describes whether the household has investments in bonds as of the beginning of the year. Binary variable: 0-No, 1-Yes
overall 0.04 0.20 0.00 1.00 N = 5,290
between 0.17 0.00 1.00 n = 1,657
within 0.09 -0.81 0.92 T-bar = 3.19
Saving Account_i
Saving Account indicator. Describes whether the household has a saving account as of the beginning of the year. Binary variable: 0-No, 1-Yes
overall 0.88 0.32 0.00 1.00 N = 5,290
between 0.30 0.00 1.00 n = 1,657
within 0.15 0.01 1.76 T-bar = 3.19
Optimism Dispositional optimism as a life expectancy miscalibration measure
overall -0.04 0.23 -0.79 0.54 N = 5,290
between 0.21 -0.72 0.54 n = 1,657
within 0.10 -0.61 0.42 T-bar = 3.19
Age Age of the respondent
overall 48.58 12.12 22.00 69.00 N = 5,290
between 12.84 22.50 69.00 n = 1,657
24
Table 1 (continued)
Variable Description Mean Std. Dev. Min Max Observations
Male
Gender of the respondent binary variable: 0-female,1-male between overall 0.58 0.49 0.50 0.00 0.00 1.00 1.00 N = 5,290 n = 1,657
within 0.00 0.58 0.58 T-bar = 3.19
High_Education Educational level of the respondent. Binary variable: 0-Other education, 1- University and vocational college education.
overall 0.41 0.49 0.00 1.00 N = 5,290
between 0.49 0.00 1.00 n = 1,657
within 0.06 -0.39 1.28 T-bar = 3.19
Job_contract Occupation of the respondent: Binary variable: 0-Other occupation, 1-Contractual basis job
overall 0.60 0.49 0.00 1.00 N = 5,290
between 0.47 0.00 1.00 n = 1,657
within 0.17 -0.27 1.48 T-bar = 3.19
Good_health
Health of the respondent. Binary variable: 0-Poor or Fair health, 1- Good or Excellent health..
overall 0.80 0.40 0.00 1.00 N = 5,290
between 0.36 0.00 1.00 n = 1,657
within 0.21 -0.08 1.67 T-bar = 3.19
Risk_tolerance
Risk tolerance of the respondent. Binary variable: 0-Moderate and High risk loving, 1-Low risk loving. Variable is measured based on the following question: “I am prepared to take the risk to lose the money when there is also a chance to gain money”. “Low risk loving” category includes the first 3 answers on the Likert’s scale from 1 to 7.
overall 0.69 0.46 0.00 1.00 N = 5,290
between 0.40 0.00 1.00 n = 1,657
within 0.28 -0.18 1.57 T-bar = 3.19
Married
Marital status of the respondent. Binary variable: 0-Currently not married, 1-Currently married. “Currently married” includes the following: married, living together with a partner.
overall 0.70 0.46 0.00 1.00 N = 5,290
between 0.44 0.00 1.00 n = 1,657
within 0.13 -0.17 1.58 T-bar = 3.19
Wage_Earner Describes whether the respondent is the household member with the highest income. Binary variable: 0-No, 1-Yes.
overall 0.73 0.44 0.00 1.00 N = 5,290
between 0.44 0.00 1.00 n = 1,657
within 0.12 -0.14 1.61 T-bar = 3.19
House_size Number of household members.
overall 2.43 1.31 1.00 8.00 N = 5,290
between 1.29 1.00 8.00 n = 1,657
within 0.28 0.18 5.28 T-bar = 3.19
Net_worth (thousands)
The sum of total assets and total debt of the household as of the beginning of the year.
overall 200.23 302.10 -1,106.58 5,351.90 N = 5,290
between 291.88 -439.02 3,475.22 n = 1,657
within 130.14 -1,896.68 2,205.63 T-bar = 3.19
Net_income
(thousands) Net income of the household as of the beginning of the year
overall 34.09 29.08 -3.56 1,185.98 N = 5,290
between 24.77 -1.60 547.78 n = 1,657
within 21.04 -173.89 1,006.89 T-bar = 3.19
25 The demographic related binary variables also have higher between variation and low within variation which indicates that there is less transition in the category such as married or contractual job during the time. The ratio of risky financial assets to financial assets is the only variable that has higher within variation, indicating that respondents have fewer differences in the weights assigned to their risky financial holdings, but these weights have higher variation over time, which once again indicates the necessity of analyzing relationships within the time frame. Table 2 presents the dynamics of ownership and optimism indicators. Optimism_i is a dummy variable which is equal to 0 when the optimism measure for the respondent is less than or equal to 0 indicating that the person is not optimistic and is equal to 1 when the optimism measure of the respondent is positive, indicating that the person is optimistic. According to Table 2, each year lower percentage of respondents from the total number of respondents for a particular year is investing money into risky financial assets. The same trend exists concerning the investments in the stocks. At the same time, the percentage of people investing in bonds and saving/deposit accounts have increased over the period 2002 to 2009. This means that during the time households prefer less risky and riskless investments to stock investment.
Table 2
Dynamics of ownership and optimism indicators
This table presents the proportion of respondents in the total number of respondents for the specific year sorted by five categories: Ownership of risky financial asset, ownership of stock, ownership of bond, ownership of savings account and optimism of the respondent. All numbers except the totals are in percentages. All indicators represent only positive outcomes.
Year RiskyFin_i Stock_i Bond_i Saving
Account_i Optimism_i Total obs. by Year 2002 76.42 16.67 3.72 84.40 53.55 564 2003 68.85 18.54 4.21 88.79 55.14 642 2004 64.07 15.98 3.96 89.07 51.50 732 2005 63.84 13.45 3.66 87.86 46.21 766 2006 60.76 14.45 3.82 88.10 47.17 706 2007 58.64 15.24 4.79 89.55 47.90 689 2008 57.10 14.52 5.38 87.93 39.80 613 2009 57.09 12.80 4.50 91.87 37.02 578
Total number of observations 5,290 Source: Author’s calculations
26 This suggests a clue about a positive association of optimism and risky financial asset ownership.
Table 3 presents the dynamics of average values for all variables from 2002 to 2009 for optimistic households. Table 3 shows that not only the proportion, but also the optimism level of optimistic households has decreased. By the year 2009, 85 % of optimistic individuals are males, which is in line with the literature that males are more optimistic than females. The average net income of optimistic households and average net worth have decreased dramatically in the sample period. By the year 2009 the former is about the half of the value of the year 2002.
Table 3
Summary statistics for optimistic households
This table presents the distribution of mean values of variables for the subgroup of optimistic households. Year 2002 2003 2004 2005 2006 2007 2008 2009 Fin/Tot 0.33 0.28 0.30 0.29 0.29 0.27 0.27 0.28 RiskyFin/Fin 0.31 0.22 0.20 0.21 0.20 0.24 0.16 0.16 Stock/RiskyFin 0.10 0.13 0.13 0.10 0.12 0.13 0.13 0.11 RiskyFin_i 0.78 0.70 0.68 0.64 0.64 0.62 0.59 0.61 Stock_i 0.17 0.20 0.18 0.12 0.14 0.17 0.16 0.14 Bond_i 0.04 0.04 0.03 0.03 0.03 0.05 0.06 0.04 Saving Account_i 0.84 0.90 0.90 0.87 0.87 0.89 0.87 0.92 Optimism 0.18 0.17 0.15 0.15 0.14 0.14 0.14 0.16 Age 45.58 45.15 45.40 46.33 48.11 48.66 52.89 53.35 Male 0.68 0.69 0.70 0.62 0.67 0.72 0.78 0.85 High_Education 0.46 0.45 0.42 0.42 0.45 0.43 0.43 0.46 Job_contract 0.69 0.71 0.69 0.63 0.61 0.64 0.57 0.60 Good_health 0.87 0.89 0.90 0.89 0.90 0.91 0.89 0.85 Risk_tolerance 0.66 0.71 0.63 0.70 0.62 0.65 0.65 0.64 Married 0.74 0.78 0.68 0.68 0.69 0.75 0.75 0.70 Wage_Earner 0.80 0.76 0.81 0.76 0.78 0.82 0.84 0.90 House_size 2.51 2.66 2.49 2.42 2.42 2.56 2.41 2.34 Net_worth (k) 352.97 162.76 158.80 155.47 170.90 215.20 244.79 203.17 Net_income (k) 62.29 30.12 32.92 32.05 31.13 30.79 31.01 32.61 Source: Author’s calculations
27 relative risk aversion. At the same time, the ratio of stock investments over risky financial assets presents a small improvement, it increases from about 10% to 11% for the sample period. The average number of optimistic households who own risky financial assets decreased by almost 17 p.p. and the average number of optimistic households who own stocks decreased from 17% to 14%. Overall, we argue that tendencies that we observe in variables suggest a positive association between optimism level and financial decisions of asset allocation and ownership, which we will further investigate in regression analysis.
5. Results
In this section, we present the results of regression analysis and put them in the context of the proposed hypotheses. We formulate our research questions in three main groups of hypotheses: ownership, allocation and subgroup level hypotheses. First, we investigate the relationship of optimism and financial decisions by estimating our models with control variables and a full set of time dummies. Then we introduce interaction terms in order to address Subgroup hypotheses. The estimations with the Pooled OLS estimator are the starting point of our analysis. However, the Pooled OLS estimator does not take into account time and entity variation of panel data. Moreover, when the fixed effects model is estimated with the Pooled OLS estimator, the estimates are inconsistent, but when the Pooled model is estimated with the fixed effects estimator, the estimates are still consistent (Cameron and Trivedi, 2005). So the fixed effects estimator is the instrument that we mainly rely on during our analysis.
28 Table 4
Regression results: ownership of risk assets and stocks
This table presents the results of the Pooled OLS and the fixed effects regressions for binary dependent variables of Risky financial asset indicator (RiskyFin_i) and Stock indicator (Stock_i). Each model is first calculated on the set of control variables and time dummies and then the interaction terms are added. All regressions are performed with robust standard errors.
Risky financial assets indicator Stock indicator
OLS estimator Fixed effect estimator OLS estimator Fixed effect estimator
29
Table 4 (continued)
Risky financial assets indicator Stock indicator
OLS estimator Fixed effect estimator OLS estimator Fixed effect estimator
M.1.1 M.1.2 M.1.3 M.1.4 M.2.1 M.2.2 M.2.3 M.2.4 2005 -0.025 -0.026 -0.532*** -0.525*** 0.036* 0.036* -0.117** -0.120** (0.026) (0.026) (0.078) (0.079) (0.020) (0.020) (0.058) (0.059) 2006 -0.061** -0.062** -0.728*** -0.718*** 0.030 0.029 -0.164** -0.169** (0.027) (0.027) (0.098) (0.100) (0.020) (0.020) (0.074) (0.076) 2007 -0.084*** -0.085*** -0.923*** -0.911*** 0.040* 0.040* -0.203** -0.209** (0.029) (0.029) (0.119) (0.121) (0.022) (0.022) (0.089) (0.091) 2008 -0.096*** -0.097*** -1.092*** -1.078*** 0.022 0.021 -0.250** -0.258** (0.030) (0.029) (0.138) (0.141) (0.023) (0.023) (0.105) (0.107) 2009 -0.092*** -0.093*** -1.250*** -1.234*** 0.010 0.008 -0.292** -0.301** (0.030) (0.030) (0.159) (0.163) (0.023) (0.023) (0.119) (0.122) Constant 0.419*** 0.425*** -5.802*** -5.698*** 0.045 0.046 -1.284** -1.359** (0.067) (0.067) (0.856) (0.884) (0.055) (0.056) (0.634) (0.652) Observations 5,290 5,290 5,290 5,290 5,290 5,290 5,290 5,290 R-squared 0.162 0.163 0.061 0.062 0.144 0.145 0.024 0.027 Cluster Robust
st. errors Yes Yes Yes Yes Yes Yes Yes Yes
Number of
respondents 1,657 1,657 1,657 1,657
Note: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
We fail to reject the null hypothesis that the coefficients are jointly zero, so the interaction terms are not affecting the relationship of optimism and risky asset ownership decision. For the whole sample, which is not disaggregated into subgroups, the sign of the significant Optimism coefficient (M.1.3) indicates that more optimistic households are less likely to hold risky financial assets, so we reject Ownership Hypothesis 1, which proposes positive relationship. The coefficients of time dummies for model specification M.1.3 are negative and highly significant. In addition, while approaching the year 2009, the negative effect of time on the probability of owning risky financial assets increases. The effect of the financial crisis of 2008 can be clearly identified through the time dummies. In model M.1.3, the coefficients of time dummies for 2008 and 2009 are respectively -1.25 and -5.8. This coefficients can not be used for forecasting purposes as they suggest that the probability range of the dependent variable will be out of (0,1) interval, however, the sharp decline in the coefficients makes vivid the negative effect of the financial crisis.
30 consideration moderating effects of different subgroups of the sample, the coefficient of Optimism indicates a significant relationship. Particularly, when introducing interaction terms in model M.2.4, we find evidence that optimism decreases the probability of owning stock for those households, who do not belong to any of the subgroups presented in interactions. As long as the total effect of each subgroup is the sum of the interaction coefficient and the Optimism coefficient, we test for joint significance of Optimism and each of interaction coefficients. Except for the interaction of male households, each of the interaction coefficients and Optimism coefficient are jointly significantly different from zero. However, only the interactions for the subgroups of the main wage earner and married financial administrators are independently significant.
Summarizing, we reject Ownership Hypothesis 2, which proposes a positive relationship between optimism and stock ownership for those households who do not belong to any of the subgroups. In addition, being married and being the main wage earner of the household positively affects the relationship between optimism and the decision of stock ownership. These results are in line with Subgroup Hypothesis 1. Moreover, the total effect of optimism on the probability of owning stock equals the sum of interaction coefficient and Optimism coefficient and has a positive sign (M.2.4). This implies that for households where the financial administrators are married, optimism increases the probability of owning stocks, which is in line with Ownership Hypothesis 2. In addition, based on joint significance tests we see that household size and education level are significant moderating variables in the relationship of optimism and stock ownership decisions and have a negative effect on the relationship. The coefficients of time dummies are negative and significant at 5% significance level for the years from 2004 to 2009 indicating that times-specific effects decrease the probability of owning stocks (M.2.4). Net worth has a positive significant effect on the probability of owning risky assets and stock in all specifications. In models M.1.3 and M.2.4 where optimism significantly affects ownership decisions, age has a positive significant effect on both decisions. Older people are more likely to own risky assets and stocks.
31 Table 5
Regression results: allocation of assets, OLS estimator
This table presents the results of the Pooled OLS analysis for three dependent variables: Financial assets over total assets (Fin/Tot), Risky financial assets over financial assets (RiskyFin/Fin) and stock over risky financial assets (Stock/RiskyFin). Each model is first calculated on the set of control variables and time dummies and then the interaction terms are added. All regressions are performed with robust standard errors.
Fin/Tot RiskyFin/Fin Stock/RiskyFin
32 Table 5 (continued)
Fin/Tot RiskyFin/Fin Stock/RiskyFin
M. 5.1.1 M.5.1.2 M. 5.2.1 M. 5.2.2 M.5.3.1 M. 5.3.2 2005 -0.080*** -0.082*** 0.076 0.074 0.026* 0.026* (0.021) (0.021) (0.129) (0.129) (0.016) (0.016) 2006 -0.068*** -0.069*** -0.090*** -0.092*** 0.022 0.022 (0.020) (0.020) (0.023) (0.023) (0.017) (0.017) 2007 -0.066*** -0.065*** -0.071** -0.071** 0.037** 0.037* (0.021) (0.021) (0.030) (0.030) (0.019) (0.019) 2008 -0.079*** -0.079*** -0.097*** -0.098*** 0.027 0.026 (0.023) (0.023) (0.026) (0.026) (0.020) (0.020) 2009 -0.087*** -0.087*** -0.016 -0.018 0.008 0.007 (0.025) (0.025) (0.121) (0.120) (0.020) (0.020) Constant 0.293 0.296 0.175 0.175 0.073 0.074 (0.324) (0.323) (0.153) (0.155) (0.055) (0.056) Observations 5,290 5,290 5,290 5,290 3,344 3,344 R-squared 0.004 0.004 0.004 0.005 0.085 0.087
Robust st. errors Yes Yes Yes Yes Yes Yes
Note: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
First, we analyse the effect of optimism on asset allocation decisions by estimating our models for control variables and a full set of time dummies. Then we include the interaction terms in the model specification. The only case when optimism significantly affects asset allocation decision is in the model M.5.2.2 that is a Pooled OLS analysis with interaction terms. In all other cases, the coefficient of Optimism is not significant. Based on the results of the fixed effects estimation, we could not find any evidence that optimism is related to asset allocation decisions presented in the models. Further, we check for joint significance of Optimism and each of the interaction coefficients. However, in all fixed effects models we fail to reject the null hypothesis that the coefficients are jointly equal to zero. This implies that we do not have evidence that will allow us to check Allocation hypotheses and Subgroup Hypothesis 1 from the perspective of asset allocation decisions. In addition, we reject the hypothesis that household size is a moderating variable in the relationship of optimism and asset allocation decisions.
33 Table 6
Regression results: allocation of assets, Fixed effects estimator
This table presents the results of the fixed effects analysis for three dependent variables: Financial assets over total assets (Fin/Tot), Risky financial assets over financial assets (RiskyFin/Fin) and stock over risky financial assets (Stock/RiskyFin). Each model is first calculated on the set of control variables and time dummies and then the interaction terms are added. All regressions are performed with robust standard errors.
Fin/Tot RiskyFin/Fin Stock/RiskyFin
34 Table 6 (continued)
Fin/Tot RiskyFin/Fin Stock/RiskyFin
M. 6.1.1 M.6.1.2 M. 6.2.1 M. 6.2.2 M.6.3.1 M. 6.3.2 2006 -0.424*** -0.377*** -0.706*** -0.578** -0.003 -0.003 (0.083) (0.116) (0.257) (0.260) (0.012) (0.012) 2007 -0.539*** -0.479*** -0.823*** -0.659** 0.008 0.008 (0.093) (0.132) (0.306) (0.318) (0.013) (0.013) 2008 -0.648*** -0.578*** -1.032** -0.840* 0.012 0.012 (0.106) (0.152) (0.423) (0.436) (0.013) (0.013) 2009 -0.759*** -0.682*** -1.093** -0.867 (0.119) (0.165) (0.521) (0.557) Constant -4.161*** -3.703*** -5.221** -3.846* 0.233 0.233 (0.622) (0.866) (2.079) (2.291) (0.155) (0.154) Observations 5,290 5,290 5,290 5,290 3,344 3,344 R-squared 0.003 0.003 0.010 0.012 0.013 0.015 Number of unique 1,657 1,657 1,657 1,657 1,154 1,154 Cluster Robust st. errors
Yes Yes Yes Yes Yes Yes
Note: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Summarizing, we do not have any evidence that optimism affects the asset allocation decisions. This result is questionable, as at the same time we find that optimism affects the asset ownership decision. If we assume that the asset ownership is a marginal case of asset allocation then it is logical to expect that optimism will affect asset allocation as well. However, based on the results we can argue that ownership decisions and allocation decisions have different determinants, and what affects ownership decisions does not necessarily affect allocation decisions.
6.
Conclusion
6.1 Main discussion35 The analysis of the channels through which optimism affects financial decisions, reveals the fact that optimism eventually affects the riskiness of financial decisions. Thus, we analyse the relationship of optimism and financial decisions through the prism of risk. We present our research questions in three groups of hypotheses: ownership, allocation, and subgroup level hypotheses. In order to check the proposed hypotheses, we conduct regression analysis using the Pooled OLS and the Fixed effects estimators.
The first contribution of our study to the existing literature is the analysis of the relationship of optimism and financial decisions within the time frame while taking into consideration the time and entity variation of our sample. The second contribution of our analysis, is the investigation of the relationship of optimism and asset ownership decisions, which is not thoroughly discussed in the existing literature even though asset ownership decision is chronologically preceding asset allocation decisions.
Our analysis provides evidence that optimism is negatively related to the probability of owning a stock or a risky financial asset in general. Due to the scarcity of the studies on this topic, we are not able to find any rationale for this result in the existing literature. Combining this result with negative time-specific effects and with our measure of optimism, we suggest that more optimistic households, who have positive expectations about their longevity, prefer not to risk their life span in difficult times. The analysis of the relationship of optimism and risky asset ownership decision for different subgroups of the sample does not provide any significant results. At the same time, we find evidence that belonging to different subgroups affects the relationship of optimism and stock ownership decision. Specifically, the relationship of optimism and decision of stock ownership is positively affected when the financial administrator is married and is the main wage earner of the household. In addition, we find that the total effect of optimism on stock ownership is positive for married financial administrators.
36 In addition, we find evidence that older households are more likely to own risky assets and to participate in the stock markets. Net worth is positively associated with ownership decisions. Wealthier households are more likely to own stocks and risky financial assets. This can be explained by the fact that wealthy households do not have liquidity constraints and are more stable to adverse movements of the financial market. We find evidence that household size significantly affects the relationship of optimism and stock ownership decision but is not significant in the relation of optimism and risky asset ownership decision.
6.2 Limitation and suggestions for further research
Limitations: Our study encounters several limitations that are worth to be mentioned.
Optimism is a subjective psychological characteristic, which assumes variation in measurement instruments. We use the life expectancy miscalibration measure for the measurement of optimism, however, one can argue that this measure is not proper for financial expectations. We use this measure, as it allows us to conduct a large-scale analysis. In addition, the study of Puri and Robinson (2007) ascertains that life expectancy miscalibration has a positive correlation with economical expectations and with the common LOT-R test. However, we agree that for further analysis, another measure of optimism that will directly capture financial expectations is preferable.
