The influence of the Big Five personality traits on saving
preferences
Saskia Milsted*, supervised by Dr. V. Angelini
19th of January, 2018
Faculty of Economics and Business Economics, University of Groningen, The Netherlands
Abstract
Using longitudinal data from the Dutch LISS data archive, this paper is the first to analysis the longitudinal influence of personality traits on different household saving types. High levels of individual within variation and significant population mean differences for the majority of the personality traits were found supporting the theory that personality traits do vary over a life course. The analysis divides household saving into six different categories; total, liquid, investment, insurance, durable goods and debt household saving. Dividing household savings resulted in more direct, significant influences of personality traits. This paper provides
evidence that this is because the different saving types are closely associated with specific risk attitudes. Results suggest that conscientiousness and agreeableness are associated with a risk-averse attitude therefore they are significantly correlated with low-risk types of household saving. Emotional stability, extraversion and openness are associated with a risk-taking attitude therefore they are significantly correlated with relative riskier types of household saving.
Key Words: Big Five personality traits, Personality trait consistency, Normative trend,
Household saving, Saving attitude, Risk attitude
JEL classifications: C23, C24, D01, D14, D15, D91
The author of this thesis would like to greatly thank Viola Angelini for her flexibility, guidance, and valuable knowledge in this field of expertise.1.
Introduction
Household saving is a major impetus for long-term economic growth as it is the main domestic source of funds to finance capital investments (OECD, 2015). Governments and businesses constantly intervene with saving policies and schemes in order to optimize aggregate household saving and stimulate economic growth. It comes as no surprise, therefore, that the analysis of household saving is one of the most active areas of research in economics. Deslsen & Smits (2014) recently suggested that the success or failure of a saving scheme or policy depends on a wide range of factors including personality characteristics. Personality characteristics as a determinant of household saving decision have had limited attention in literature despite the growing consensus that personality traits are as predictive in economic outcomes as cognitive factors1(Almlund et al., 2011). To the authors
knowledge only four studies specifically investigate the influence of personality traits on household saving (Schmölders, 1966; Brandstätter,1996,2000; Wärneryd,1996; Nyhus & Webley, 2001). These four studies are out-dated therefore this paper sets out revitalize research done on the influence of personality traits on household saving decisions.
This paper’s analysis is based on data from the Longitudinal Internet Studies for the Social Sciences (LISS panel) from the CentER data, a unique household survey about the Dutch population collecting longitudinal information on respondents’ socio-economic situation, demographics, family composition, health and personality traits. Personality traits are measured using the ‘Big Five’ taxonomy founded by McCrae & Costa (1992) and Goldberg (1993). According to this taxonomy, personality can be defined by five traits, namely; Openness to Experience, Conscientiousness,
Extraversion, Agreeableness and Emotional Stability. Table A1 in the appendix defines each trait and their associated facets (narrow personality traits) in detail.
development path over a life course as individuals from the same social institution, experience, and similar social tasks at similar ages (e.g. marriage, parenthood). This paper finds new evidence supporting the latter theory. High levels of individual within variation was found from this paper’s sample of 1296 Dutch respondents of all ages, over a time span of six years. Furthermore, significant population mean differences (taken from years 2009 and 2015) were found for different age
categories, including ages above 30, of the five traits. This provides evidence supporting the common personality trait development path over a life course.
Previous studies on the influence of personality traits on household saving behaviour (Schmölders, 1966; Brandstätter,1996,2000,2005; Wärneryd,1996; Nyhus & Webley, 2001) have failed to
adequately incorporate the inconsistent nature of personality traits over time. Previous estimators found that the influence of personality traits on household saving may be biased, therefore their internal and external validity may be questionable (Almlund et al., 2011). This paper addresses this shortcoming in existing literature by using unique longitudinal personality trait data to account for personality trait changes over time, while controlling for panel effects.
In order to find direct empirical influences of personality traits on household saving this paper divides saving into six different categories namely; liquid, insurance, investment, durable goods, debt and total household saving. More direct correlations are found because each different category is associated with a certain risk level and thus risk attitude. Saving attitudes such as risk have been proven to mediate the indirect influences of personality traits on household saving in past studies (Brandstatter, 1996; Warneryd, 1996). For example, investment saving is relatively risky therefore closely related to the saving attitude “risk-taking” whilst liquid saving is relatively low-risk therefore closely related to the saving attitude “risk-aversion”.
This paper uses fixed effect models to analysis the direct influence of personality traits on the different household saving types in order to control for individual unobserved heterogeneity. It distinguishes between two decision moments, namely the decision to engage in a certain saving type and conditional on having engaged, how much to save. A binary choice model and a linear regression conditional on ownership are used respectively. Furthermore the relationship between the five
personality traits and risk attitudes are analysed. The results are used to reason why dividing
OLS regression as the risk attitude variable is measured using a one-time LISS questionnaire from the year 2010. All factors considered the research question is the following:
What is the longitudinal direct influence of the Big Five personality traits on different types of household saving decisions?
Correctly assuming change in personality traits over time, dividing saving into its different categories and investigating the mediating role of risk attitudes extends the work done on the
personality traits and household saving. The empirical results suggest that all five traits have unique significant influences on specific household saving types and that this is due to the different risk levels associated to each saving type.
The remainder of this paper is structured as follows: section two will provide a literature review over the relevant findings on the influence of personality traits on saving decisions and attitudes. Furthermore, it will explain and analyse the current debate on the nature of personality traits. The third section consists of the hypotheses this paper sets out to prove, which are derived from the literature review. The fourth section highlights the data and methodology used. The fifth section illustrates the results and analyses them. This is followed by a discussion on possible improvements for further research and a conclusion.
2. Literature review
2.1 Personality traits as determinants of saving decisions
the respondent’s levels of discipline, conscientiousness, introversion/extroversion and easygoingness. He found that conscientious and self-disciplined people were three times more likely to be a regular saver than “carefree” and “easy-going” people.
Only 30 years later did Brandstätter (1996) and Wärneryd (1996) continue the research on empirical associations between personality traits and saving decisions. They both used data from the VSB-CentER saving project in Tilburg. The project collected data by means of a computer
questionnaire which it sent to a sample that was representative of the Dutch households. Wärneryd’s (1996) sample consisted of financial household managers that filled-in the questionnaire in both years. This summed up to a total of 3339 participants. Cross-sectional structural equations for both waves were used to test the influence of personality traits on saving attitudes, motives and behaviours. The respondents attitudes toward saving were split into five categories. Namely; thrift (overall
positive attitude towards saving), no need to save, saving involvement (save to make money), shame of incurring debt and saving is a habit. The saving motives were altruism, short-term buffer, long-term financial security and saving goals (i.e. save to buy a house). Saving behaviour was defined as
whether or not the respondent had saved between the year 1993 and 1994, and how much s/he had saved. Wärneryd (1996) found that conscientiousness and openness had a significant indirect
influence on household saving. While personality traits failed to come out significant when related to saving behaviour, conscientiousness and openness proved to be significant for all saving attitudes. Therefore the two traits indirectly significantly influence saving behaviour. Conscientious was
positively, and openness negatively, related to saving attitudes and therefore saving behaviour. It must be noted that emotional stability did show some direct significant correlation with saving behaviour. So when analysing only the direct effect of personality traits on saving decisions, emotional stability is the most important according to Wärneryd (1996) results.
Brandstätter (1996) at the same time as Wärneryd (1996) conducted a similar study. Yet, his main unit of analysis was couples (husband and wife), as they tend to make household decisions together. His sample consisted of 734 couples. Personality traits, saving attitudes, intentions and behaviour were measured and analysed as averages of the husband and wife. His study concludes that emotional stability, introvert and conscientious couples are more favourable towards saving. Similar to Wärneryd (1996) the relationship is argued to be mediated by saving attitudes. In this case the saving attitude was specifically the ability to delay gratification. Furthermore, Brandstätter (1996) points out that personality trait “packages” are also interesting to investigate. He finds that
introvert. Brandstätter & Guth (2000) later define the empirical association found regarding
introversion and saving behaviour as evidence of the ‘Social Comparison Theory’ (Schor, 1998). The theory states that saving and consumption decisions are influenced by people that one meets with regularly. Extravert people like to spend time with people and are socially engaged. A social life brings extra expenditures with it (diners, parties etc.). Thus, stable extraverts are more exposed to higher consumption patterns than unstable introverts and so they will be less inclined to save.
Brandstätter (1996) and Wärneryd (1996) both concluded that personality traits had no direct effect on the amount saved in that specific year. They argued that the relationship was indirect and mediated by the significant correlation between the personality traits and saving attitudes. Nyhus & Webley (2001) extend their research by focusing solely on the direct effect of personality traits on household saving. They used the Dutch CentER saving survey (CSS), also from Tilburg, which distinguishes between household liquid, investment, insurance, durable goods and debt saving. This provides a more in-depth empirical analysis of the direct influence of personality traits on household saving compared to Brandstätter (1996) and Wärneryd (1996) who only used total amount saved.
Nyhus & Webley (2001) sample consisted of only heads of households and their
delay gratification. In other words that the influence of personality traits on saving decisions may be mediated by certain saving attitudes (i.e. risk preference and delay gratification). This clarifies why Brandstätter (1996) and Wärneryd (1996) found significant effects on saving attitudes and not on total saving. The saving types used by Nyhus & Webley (2001) are by definition associated to different saving attitudes. For example, investment saving is more risky than liquid saving. Risk preference (risk averse of loving) is a saving attitude. Therefore by definition investment saving is associated with a risk loving saving attitude and liquid saving risk averse saving attitude. Total saving is the dependent variable Brandstätter (1996) and Wärneryd (1996) used for household saving. Total saving is the sum of the five different saving types so by definition total saving is associated with many different saving attitudes. Therefore the direct influence of personality traits on total household saving is insignificant.
In order to appropriately hypothesis on the empirical relationship between personality traits and different household saving types, one needs to understand the influence of personality traits on different attitudes related to saving decisions, such as delay discounting and risk behaviour.
2.2 Personality traits as determinants of saving attitudes
Delay Discounting, or delaying gratification, is an important attitude in saving decision making. It describes the extent to which the value of a reward decreases as the delay to obtaining that reward increases, in other words getting 50 euros today is valued more than receiving 50 euros in one year. Hirsch et al., (2008) set out to predict individual differences in delay gratification with
personality traits and cognitive ability. They used 97 undergraduate students from McGill university. The participants completed several questionnaire where-out measures for personality traits, cognitive ability and delay discounting rates were defined. Their regressions confirm the predictive validity of cognitive ability and personality in delay discounting rates. Emotional stability (extraversion) at higher (lower) levels of cognitive ability predicted reduced (greater) discounting. Thus emotional stability predicts the ability to delay gratification, while extraversion will most likely lead to impromptu saving decisions, for individuals with lower cognitive ability.
Factor model influences risk-preferences in different decision domains. Their sample (n=358) consists of participants that were likely to differ in risk-behaviour, they included academics, firefighters, mountaineers, city traders and chess players. The linear regression results revealed two groups. The first group (n=51) resides from individuals that have consistent risk preference over all decision domains (work, health and personal finance). Only the group with consistent risk averse behaviour (n=37) showed significant results; agreeableness and conscientiousness were positively associated with risk averse behaviour. The second group (n=307) consists of individuals with inconsistent risk preference over different decision domains. These individuals showed high significant levels of neuroticism, openness and also conscientiousness. This contradicts Nicholson et al.,’s (2005) findings that conscientiousness is associated with only risk averse behaviour, yet Soane & Chmiel (2005) explain that weighing alternative choices, a facet of conscientiousness, is associated with more situational variability thus inconsistent risk preference across different decision domains. All results and groups considered Soane & Chmiel (2005) conclude that extraversion and openness are related to risk-taking behaviour and conscientiousness related to risk-averse behaviour.
2.3 Debate on Personality Trait Consistency
From the previous two sections this paper can conclude that the studies conducted on the relationship between personality traits and household saving behaviour and decisions is limited. It is argued that this is because personality has a consistent state and economists are mostly interested in change and the cause of change rather than an enduring state (Wärneryd, 1996). Yet, literature has always been divided, and still is today, on the consistency of personality.
Roberts BW. and Helson R. take on a contextualised perspective on the nature of Personality Traits and argue that “change is complex and ongoing, owing to the many factors that can affect personality traits” (Srivastava et al., 2003). This theory states that Personality Traits are multiply determined, and environmental influences have a strong effect on change. This means that personality traits can continue to change in adulthood and late life (Roberts & DelVecchio, 2000; Srivastava et al., 2003; Roberts et al., 2006; Roberts & Mcroczek, 2008). Roberts & Mcroczek (2008) argue that these changes, past the age of 30, can be quite substantial and consequential and thus ignoring them can lead to misleading results when investigating the predictive powers of personality traits. Furthermore, more recent longitudinal studies have found a “normative” trend in personality trait development (Roberts et al., 2006). It is argued that this pattern stems from common ageing processes and social tasks/events, which lead most personalities to change in the same way.
Plomin & Nesselroade (1990) and McGue et al. (1993) conducted important longitudinal studies in the context of the personality trait consistency debate. They analysed twins across their life course and concluded that while personality traits appeared largely genetic during childhood, in adulthood genetics had only a very small influence. These studies provide evidence that personality traits are “multiply determined”, and that environmental factors play a large role in personality trait change.
(BLSA) that is made primarily of individuals above the age of 40. They analysis these individuals over short time spans of 3 to 6 years. This attenuates the amount of change they found in adulthood; none. Helson (1992) collects her data from Mills graduates, thus focusing more on change in young adulthood than midlife (<40). She focuses on single cohorts and analyses them in specific life periods of a life course (first career, marriage etc.) and for a long-time span (longer than 6 years). All factors combined Roberts et al. (2006) argue that the position of these two teams is eminently understandable given their data and the nature of their studies. Thus this paper concludes from this section that in order to accomplish a sound empirical investigation on the consistency of personality traits an appropriate data set is needed (including individuals of all ages), a long enough time span and the relevant methodological approach must be used. This paper sets out to further examine the nature of personality traits and more specifically investigate the likely existence of a common personality trait trend for a population.
2.4 The normative pattern of the Big Five Personality Traits
The methodological approach of examining mean-level changes has brought to attention a possible normative trend in certain personality traits. This normative pattern shows that as people enter young adulthood all five personality traits increase, and as they get older openness to experience and extraversion tend to decrease while conscientiousness, agreeableness and emotional stability continue to increase (Roberts et al., 2006; Roberts & Mrocbek, 2008). What is the reason for this trend? Roberts et al. (2006) and Helson et al. (2002) explain that certain life experiences are most likely the reason for these patterns. The majority of individuals engage in common universal social tasks, such as parenthood or marriage, which affects certain personality traits.
Mehrotra & Jolly (2000) argues the most powerful universal experiences in adulthood are
and turns up on time every day, s/he has met the roles expectation and will be rewarded (i.e.
compliment from boss or a promotion). This in turn will lead to greater social reinforcement, which increases the personality traits conscientiousness and emotional stability. Role expectations serve as guides for how an individual should act and thus possibly change. These roles and expectations, promoted by our western social institution, are the catalysts for the common pattern of personality trait development. So Robert et al. (2006) explain that if we can define the average ages associated with the most salient life experiences, one can get a better understanding of personality trait
development, which in turn could explain differences in preferences during a life-cycle such as saving decisions, which is the focus of this paper.
3. Research Design and Hypotheses
The limited studies done on the influence of personality traits on household saving all assumed personality trait consistency. Brandstätter (1996), Wärneryd (1996) and Nyhus & Webley (2001) all use cross-sectional data. It must be noted that the fixed personality variables used may have been due to limited longitudinal data as the studies were conducted more than 15 years ago. Either way failing to incorporate the possibility that personality traits do change over time may lead to biased estimators. The internal and external validity of the previous studies are questionable, considering the growing evidence that personality traits are inconsistent.
To the authors knowledge this is the first paper that studies the influence of personality traits on household saving using longitudinal data, allowing for personality trait inconsistency. Furthermore, inspired by Nyhus & Webley’s (2001) findings this paper will divide saving into the same six categories; liquid (relatively low risk), investment (relatively high risk), insurance, durable goods, debt (negative saving) and total saving. A more detailed description of the 6 categories is given in the Data section.
3.1 Hypotheses on the Big Five personality traits and saving type
3.1.1 Total saving
Total saving is the sum of all saving types (were debt is negative). Schmolder (1966), Wärneryd (1996) and Brandstätter (2005) used the current monetary amount of saving of their respondents as their dependent variable, which in this paper comes closest to ‘total saving’. All three authors concluded that conscientiousness was the most salient personality in predicting saving decisions. Nyhus & Webley (2001) results showed a positive significant association between
emotional stability and total saving. Furthermore, conscientiousness has been found to be related with risk-averse behaviour (Soane & Chmiel, 2005) and emotional stability with low discounting rates (Hirsch et al., 2008), both behaviours are related to increased saving. Therefore, this paper predicts emotional stability and conscientiousness will have a positive direct effect on total saving.
It must be noted that this prediction is based on past association found using fixed personality traits. This paper considers the possibility that personality traits change over a life course.
Conscientiousness and emotional stability have been shown to be self-reinforcing during a lifetime (Wärneryd, 1996; Roberts et al., 2006; Robert & Mcrozek, 2008) . As aforementioned, Roberts et al. (2005) and Helson et al. (2002) state that socially determined life experiences (normative
commitments) develop psychological maturity. Conscientiousness and emotional stability are strongly associated with psychological maturity. Once an individual experiences an increase in one of these two traits, the trait will continue to increase during a life course. Thus, taking into account that
conscientiousness and emotional stability are statically correlated with total saving and that both traits increase during a lifetime this paper predicts the following:
H1: High levels of conscientiousness and emotional stability are significantly positively correlated with total saving over a life course.
3.1.2 Liquid saving
which will most likely lead to relatively low-risk saving decisions such as liquid saving.
Conscientiousness and emotional stability tend to increase stably during a lifetime, since they are self-reinforcing. Thus much like total saving this paper hypothesizes:
H2a: Emotional stability and conscientiousness are significantly positively correlated with liquid saving over a life course.
Extraversion has been found to be negatively associated with delay-gratification behaviour (Hirsch et al., 2008) which will most likely lead to lower levels of liquid saving. Indeed, Brandstätter (2000) found that extraversion was negatively associated with saving. He explains that this is because extravert people attend more diners, parties and so forth which leads to greater consumption paths than more introvert people. Consumption and saving are substitutes, if you consume more you save less. Liquid saving is the closest to being a substitute of consumption from all the saving types. Thus, one can predict that extraversion is negatively correlated with liquid saving, at one specific point in time. Alongside extraversion this paper also predicts openness to be negatively correlated with liquid saving. Wärneryd (1996) found that openness was indirectly negatively associated with household saving. The indirect effect was mediated by saving attitudes. Soane & Chmiel (2005) find that openness is positively related to the risk-loving attitude when it comes to financial decisions. Therefore this paper predicts that openness is positively related with taking behaviour and risk-taking behaviour is negatively correlated with liquid saving as liquid saving is low-risk.
It must be noted that these findings for both extraversion and openness were built upon the Five-Factor theory that personality traits are stable during a lifetime. Extraversion and Openness follow a more complicated path than conscientiousness and emotional stability, over time. Both traits follow a curvilinear path, which increases at the beginning of adulthood; this can be associated with the educational and occupational age-graded roles, and decreases around the age of 40 (Roberts & Mroczek, 2008) which can be associated with parenthood and retirement age-graded roles. Even though extraversion and openness both decrease around the age of 40, liquid saving decisions are most salient before the age of 40, and those decisions will have a monetary consequence for the rest of a life course, it is therefore hypothesized;
3.1.3 Investment saving
Investment saving requires more risk because one puts his/her money into stocks which have uncertain returns unlike saving accounts (liquid saving) which has a set interest rate. As
aforementioned, openness is positively associated with risk-taking behaviour (Soane & Chmiel, 2005). Furthermore, a salient facet of openness is “novelty seeking”. Therefore, it is likely that individuals with high scores in openness will be more inclined to participate in the stock market because they seek novel activities. Alongside openness, extraversion has also shown to be associated with investment saving, yet the direction of the relationship is more ambiguous. On the one hand, Nyhus & Webley (2001) found a direct negative association between extraversion and investment saving. They argue that extravert people tend to consume more decreasing the money left to invest. On the other hand, extraversion has been found to have a direct positive effect with stock market participation (Christelis et al., 2010; Conlin et al., 2015). The theory the Word-of-Mouth effect has been proven to have a large impact on stock market participation (Liu et al., 2014). Extravert people are more sensitive to the Word-of-Mouth effect because they are relatively more socially engaged. This paper predicts the latter argument to be stronger than the former thus this paper predicts a positive direct relationship between extraversion and openness with investment saving.
Once again, it must be noted that these associations are derived from studies using time-invariant personality trait variables. As aforementioned, openness and extraversion follow a similar curvilinear pattern during a life course, that increases during young/middle adulthood and then experiences a decrease. Anew, this paper argues that investment saving decisions are most salient during young/middle adulthood and these decisions will have a monetary impact for the rest of the life course, thus:
H3: High levels of extraversion and openness are significantly positively correlated with investment saving over a life course.
3.1.4 Durable Goods Saving
durable goods saving is actually an expenditure, because one buys a house or car, but afterwards it can be understood as a type of saving. Instead of putting your money into a saving accounts you can put it into “bricks”, and if these durable goods are well maintained they can increase in value, creating financial gain. This paper believes that high levels of extraversion will most likely be related to durable goods saving, based on the “Social Comparison Theory” (Schor, 1998). As previously explained, the theory argues that saving and consumption decisions are influenced by people one meets with regularly. If ones friend buys a new car, the chances will increase that you will buy a new car. Extravert individuals are more socially engaged and therefore more likely to experience such a situation. As aforementioned the normative path of extraversion is curvilinear (decreasing at the age of 40). The normative individual and/or household tends to engage in durable goods saving before the age of 40, in other words buy a house or car at a stage in life where extraversion is prominent. These saving decisions in early/middle life stages have an impact for the rest of an individual's life
course, this paper predicts the following:
H4: High levels of extraversion are significantly positively correlated with durable goods saving over a life course.
3.1.5 Debt Saving
Debt saving can be understood as negative saving, and high in risk. Openness is positively associated with risk-taking behaviour (Soane & Chmiel, 2005), which mediates negative saving decisions. Facets of openness are desire to accumulate information, intellect and to seek novelty. Therefore someone with high levels of openness will most probably have high aspirations to educate him/herself and to undertake different adventures such as travelling. Both activities increase the likelihood of taking on different types of debt, such as study grants or loans. As previously mentioned, the normative pattern of openness during a life course is curvilinear, it tends to drop near the end of middle-adulthood (Roberts & Mroczek, 2008). So a normative individual has relatively higher levels of openness during young/middle adulthood. Taking on debt can have monetary consequence for the rest of a life course, therefore:
Conscientiousness and emotional stability have been directly positively associated with saving decisions (Wärneryd , 1996; Brandstätter, 1996) thus they will most likely be negatively associated with debt saving decisions (negative saving). Indeed, Nyhus & Webley (2001) find that the likelihood of a household holding debt is lower for a household with an emotionally stable head. They also find that the likelihood of a household holding debt is higher with an agreeable head. Nyhus & Webley (2001) argue that agreeableness is related to being concerned about others and thus increases the amount of loans and gifts to charities, friends, family and decreases the amount of money left to save, and so take on debt. Yet, Soane & Chmiel (2005) find that high scores in agreeableness are related to risk-averse behaviour, which mediates to positive saving decisions. The relationship is ambiguous, yet this paper supports the latter argument that agreeableness decreases debt saving, mediated by risk-averse behaviour. Agreeableness, similar to conscientiousness and emotional stability, is self-reinforcing during a lifetime. Taking into account that conscientiousness, emotional stability and agreeableness increase with age, it is therefore hypothesized:
H5b: High levels of emotional stability, conscientiousness and agreeableness are significantly negatively correlated with debt saving over a life course.
3.1.6 Insurance Saving
Insurance saving is correlated with the saving attitude risk aversion. An individual shields him/herself from the unknown future when engaging in insurance saving. As aforementioned Soane & Chmiel (2005) and Nicholson et al., (2005) find that conscientiousness is most associated with risk averse behaviour. So this paper predicts:
H6a: High levels of conscientiousness is significantly positively correlated with insurance saving over a life course.
early/middle stages of life. Under the assumption that saving decisions in early/middle life stages have an impact for the rest of an individual's life course, this paper predicts the following:
H6b: High levels of extraversion is significantly negatively correlated with insurance saving over a life course.
4. Data and Methodology
4.1 Data
This paper collects its data from the LISS panel (Longitudinal Internet Studies for the Social sciences). The LISS panel has been in full operation since 2007 and is based on a true probability sample of households drawn from the Statistics Netherlands. The LISS panel consists of 4500 households, which sums up to 7000 individuals. The data used in this paper is part of the LISS Core Study, which is a longitudinal study that provides the same variables each year for the same
households and individuals. These collected variables consist of demographic, health, economic situation (assets, income and housing) and personality traits making the LISS Core Study an attractive data source for this paper. The study provides measures of change in people’s lives as a reaction to life events or policy and societal changes.
This paper’s working sample is constructed as follows. The sample pools the 2009, 2011, 2013 and 2015 waves because these were the waves with available information on household assets.
under. The mean value of these categories was taken and used to replace the “9999999999” value, assigning a reliable monetary value to the specific saving types the respondents did not know the exact answer too (7457 observations). Lastly, participants could answer specific monetary value questions with “9999999998” if they refused to reveal the monetary value of a specific saving type. Once again, in order to avoid biased results these observations were removed from the panel data. Note if a
respondent refuses to answer in one year but provides an answer another years the individual remains in the sample but the observation in which s/he refuses to answer is removed. The final sample is unbalanced and consists of 4488 observations (1296 individuals).
4.1.1 Personality traits
In order to measure the Big Five personality dimensions this paper used the 50-item IPIP version of the Big Five model, provided by the LISS panel data (cp020 - cp069). The 50-items are concise questions ideal for a large survey where participants usually desire to devote limited time to the survey. The items can be viewed in Table A3 in the appendix. They are rated on a 5-point Likert scale ranging from 1 “Very Inaccurate” to 5 “Very Accurate”. Some items are negatively formulated and have to be inverted. Once inverted one can take the average of the 10 items assigned to a specific personality trait. The Big Five scales have shown great validity and reliability in past studies (John & Srivastava, 1999; Nyhus & Webley, 2001; Srivastava et al., 2003; Cobb-Clark & Schurer, 2012), which reasons this paper to use the Big Five model. Nonetheless, the paper still tests for reliability. The most common method is the Cronbach’s coefficient alpha. The Cronbach's alpha assesses the internal consistency, in other words whether the measurements used are a consistent measure of the concept considered. The results can be found in Table A4 in the appendix. In order for the measure to be acceptable the Cronbach’s alpha must exceed the threshold of 0.7. Therefore all five traits are reliable as their Cronbach alpha ranges from 0.76 to 0.8857.
It is appropriate to mention here that cognitive ability may cofound the interpretation of the measures of the Big Five personality traits, because when respondents answer the 50 item questions they use their cognitive skills to some extent. The relationship between cognitive ability and
standard measures of intelligence”. This issue is beyond the scope of this paper, but it is relevant to highlight.
4.1.2 Household Saving
As aforementioned, this paper divides saving decisions into six specific categories. Nyhus & Webley (2001) show that dividing household saving leads to more direct empirical associations with personality traits. This is because the different saving types are defined by their relative level of risk, so they are associated with the specific saving attitudes; risk-averse of risk-loving.
1. Household liquid saving is the most common and readily accessible type of saving, therefore it is very low risk. This includes current accounts, savings accounts, term deposit accounts, savings bonds or saving certificates, savings schemes and loans to family and friends.
2. Investment saving involves more risk than liquid saving. It includes growth funds, share funds, bonds, debentures, stocks, options, warrants and any other form of investment not mentioned such as antiques, jewellery, collections and so on.
3. Insurance saving is the total value of guaranteed minimum pay out of single-premium or life annuity insurance and total savings of endowment insurance. It is precautionary saving, individuals shield themselves from the unknown future. Therefore it is a relatively low risk type of saving.
4. Durable goods saving consists of the total value of all owned real estate and total sales value of car(s), motorcycle(s), boat(s), caravan(s) and so on. This is a relatively more high risk type of saving than liquid saving for example as the interest on the durable goods is often unknown. 5. Sum of debt. This can be seen as negative saving, hence important to take into account. It
includes all mortgages, remaining study grants and total amount of loans, credits and debts. This is a risky type of saving.
6. Total saving is defined as household liquid saving + investment saving + insurance saving + durable goods saving - sum of debt. Therefore there is no obvious associated saving attitude.
log transformation. Thus an inverse hyperbolic sine transformation is used for liquid, investment and total saving.
4.2 Empirical Methodology
This paper considers 6 different dependent variables each representing a different type of ‘saving’. Due to the nature of each saving type, different regression models will be needed. The saving types; insurance, investment, debt and durable goods have limited observations because not all respondent engage in these types of saving. A solution for these limited observations is to transform the dependent variables (insurance, investment, debt and durable goods) into binary variables, such that yit = 1 or otherwise yit = 0 where yit represents one of the four saving types mentioned for a
specific individual for a specific year. This allows for a two-part model. Assuming the decision to engage in one of the four different types of saving is independent to the actual monetary amount eventually saved, two regressions can be run to depict the two different decision moments; (1) whether or not to engage in a specific saving type, and conditional on having engaged (2) how much to save.
A probit regression will be used for the first part of the model, depicting the binary choice: whether or not an individual will save. The probit model estimates the probability a value will fall into one of the two possible binary outcomes (yit = 1or yit = 0) . The probit regression models the
conditional probability of a success outcome. Furthermore, Mundlak fixed effects will be added to the probit model in order to remove the unobserved time-invariant individual effects. Taking the first difference or within estimator (fixed effect method) is not appropriate in this nonlinear setting, therefore Mundlak fixed effects method is a neat solution. Mundlak (1978) discovered that by adding a vector of covariates of the mean of all time-varying variables to a random effects model it generates identical estimates for M.E. obtained from the fixed effects model (in the case of a linear model). Lastly, year dummies are added to model to control for aggregate time trends. One time dummy is dropped to avoid perfect multicollinearity. The two-way probit model with Mundlak fixed effects will give insights on the probability of engaging in a specific saving type:
In equation (1) 𝑇𝑖𝑡 is the vector of the Big Five traits, 𝑋𝑖𝑡 is the vector of control variable, 𝑍̅𝑖𝑡
represents the mean of all time variant variable, 𝐷𝑁𝑡 represents the time dummies and 𝑢𝑖𝑡 represents the error term.
The second regression of the two-part model will be a linear model conditional on a ‘success’ outcome (yit = 1). This part of the model will give insights on the influence of personality trait
changes on the amount held once an individual engages in a specific type of saving. The second regression will also include Mundlak fixed effects in order to remove the unobserved time-invariant individual effects. The linear regression will also include time dummies in order to control for aggregate time trends. The year 2015 is dropped in order to avoid perfect multicollinearity. The second part of this papers two-part model is a two-way linear regression model including Mundlak fixed effects, conditional on ownership:
log (𝑦𝑖𝑡) = {1 log (𝑦𝑖𝑡 ∗) = 𝛼 + ∑ 𝛽 1𝑇𝑖𝑡+ 𝛽2𝑋𝑖𝑡+ 𝛽4𝑍̅𝑖𝑡+ 𝛽5𝐷1𝑡+ ⋯ + 𝛽𝑁𝐷𝑁𝑡+ 𝑢𝑖𝑡 > 0 5 𝑖=1 0 𝑒𝑙𝑠𝑒 (2)
The dependent binary variable log(𝑦𝑖𝑡) takes on the value 1 if the individual i holds any
investment, insurance, debt or durable goods saving, and otherwise 0, at time t. The two-way linear regression model is estimated conditional on already owning the specific saving type (yit = 1). 𝑇𝑖𝑡
depicts the vector of the Big Five traits, 𝑋𝑖𝑡 being the vector of control variables, 𝑍̅𝑖𝑡 represents the
mean of all time variant variables, 𝐷𝑁𝑡 depicts the year dummies and 𝑢𝑖𝑡represents the error term.
Liquid saving and total saving do not suffer from limited observations. Considering the non-consistent personality trait theory, a fixed effect model is most appropriate. Fixed-effect models are designed to study the causes of changes within a person in a linear model. It furthermore controls for unobserved time-invariant heterogeneity (i.e. gender, nationality and so forth). This is crucial in such a study where unobserved heterogeneity is evidently present. Yet this brings about a drawback, namely that it cannot accommodate time-invariant variables. A fixed effect model allows for correlation between omitted heterogeneity and the regressor but cannot accommodate dummy variables such as gender or nationality2. The model will take on the following form:
log(𝑦𝑖𝑡) = 𝛼 + ∑5𝑖=1𝛽1𝑇𝑖𝑡+ 𝛽2𝑋𝑖𝑡+ 𝛽3𝐷1𝑡+ ⋯ + 𝛽𝑁𝐷𝑁𝑡+ 𝑢𝑖𝑡, 𝑓𝑒 (3)
𝑇𝑖𝑡 is the vector of the Big Five personality traits, 𝑋𝑖𝑡 a vector of covariates of time variant
independent variables, 𝐷𝑁𝑡 represent the time dummies for each year except 2015 in order to avoid
perfect multicollinearity and 𝑢𝑖𝑡 the error term. The Hausman test was taken for an empirical robustness check. Fixed effects were strongly advised for the two-way linear regression models.
5. Results and Analysis
5.1 Descriptive StatisticsTable 1 depicts the control variables used in the regressions; age, marriage, separation with partner, type of employment, education, health and children have all been included due to their relevance in financial decision making. The mean age of this sample is approximately 56. This slightly skewed age is most likely due to the fact that the sample only consists of individuals that make financial decisions. It must be pointed out the minimum age is 17 which is a young age for someone to be in charge of financial decisions. This is a single individual, making him/her an exception. The next age is 19. Notice that only 39.73% of the respondents have one child or more which is low for the relatively old sample. The Netherlands, for nearly half a century, has shown a steady decrease in net births (CBS, 2016). This could possibly manifest back into the normative pattern of personality trait change as agreeableness, conscientiousness and emotional stability tends to increase after parenthood (Roberts et al., 2006; Roberts & Mroczek, 2008). It is worth observing the low levels of employment; 41.08% employed and 3.8% self-employed. The OECD 2017 found a 75.4% employment rate and 16.8% self-employment rate. Note that employment rate is derived from individuals willing and able to work. If we summarize employment and self-employment for the sample excluding pensioners (31.73%) employment and self-employment increase to 60.73% and 5.57%, respectively. Yet both are still below the OECD rates. The different levels of education is evenly distributed over the population, with higher education being slightly higher than the other two levels. Medium education will be excluded in the regression analysis in order to avoid
Table 1
Descriptive statistics control variables. Data is unbalanced from years 2009, 2011, 2013 and 2015. Only respondents that make financial decisions for household or him/her-self.
Mean Standard Deviation Min Max Within Variation Observations
Male 0.5406 0.4984 0 1 0 4484
Age 56.1462 15.5674 17 93 1.7832 4484
Child(ren) 0.3973 0.4894 0 1 0.3339 4484
Live with Partner 0.4931 0.5000 0 1 0.1193 4448
Separated 0.1849 0.3883 0 1 0.0777 4484 Low Education 0.3226 0.4675 0 1 0.0582 4484 Medium Education 0.3017 0.4590 0 1 0.0813 4484 High Education 0.3737 0.4838 0 1 0.0695 4484 Employed 0.4108 0.4920 0 1 0.1606 4484 Self-Employed 0.0379 0.1909 0 1 0.0730 4484 Pensioner 0.3204 0.4667 0 1 0.1442 4484 Poor Health 0.1952 0.3964 0 1 0.1818 4484 Good Health 0.8048 0.3964 0 1 0.1818 4484
Log Household Income 7.6766 0.5763 4.3820 12.4303 0.1813 4310
Table 2 shows the descriptive statistics of the six different types of saving. 96.46% of the sample holds some sort of saving. Liquid saving is the most favourable type of saving; 92.63% of the respondents have some sort of liquid saving. 79.38% of the sample population owns durable goods saving. Less respondents hold debt than own durable goods, namely 58.17% of the sample population holds some sort of debt and as one can anticipate mortgage is the most common type of debt people hold, 48.52% of the sample population holds a mortgage (see table A5 in appendix). Investment and insurance saving are the least common saving types.
Table 2
Descriptive statistics of the dependent variables; invers hyperbolic sin value of total, liquid and investment saving and logarithmic value of debt, durable goods and insurance saving. Data is unbalanced from years 2009, 2011, 2013 and 2015.
Mean Standard Deviation
Min Max Observations
Holding Savings 0.9649 0.1838 0 1 4484
Log Total Savings 8.7411 7.5499 -15.86876 17.1408 4327
Holding Liquid Saving 0.9262 0.2615 0 1 4484
Log Liquid Saving 9.0818 4.1453 -13.9465 16.6356 4153
Holding Investment Saving 0.2448 0.4305 0 1 4484
Log Investments Saving 10.1022 2.0564 -5.9915 15.4642 1098
Holding Insurance Saving 0.1382 0.3452 0 1 4484
Log Insurance Saving 9.8593 1.3473 0 13.1224 620
Holding Durable Goods 0.7938 0.4046 0 1 4484
Log Durable Goods 11.2496 2.2134 0 16.9467 3562
Holding Debt Saving 0.5805 0.4933 0 1 4484
Log Debt Saving 11.1274 1.7151 0 16.0772 2603
Table 3 gives the descriptive statistics of the Big Five personality traits. All 5 mean levels are relatively high considering the ‘theoretical’ mean level is 2.5. Furthermore the minimum value for 3 out of the 5 traits is higher than 1. The table thus overall depicts a psychologically mature population. What is of most interest in table 3 is the within variation which represents:
𝑋𝑖𝑡−𝑋𝑖+ 𝑋̅ (4)
where Xit represent one of the five personality traits, for i a specific participant and t the
specific year. Thus Xi is the mean value of a trait for an individual i and X is the overall mean of the
Table 3
Descriptive statistics of independent variables, namely the Big Five Personality Traits. Data is unbalanced from years 2009, 2011, 2013 and 2015. Number of Observations 4479.
Mean Standard
Deviation
Within Variation
% of mean Min Max
Openness 3.5100 0.5047 0.1654 4.7111% 1.7 5
Conscientiousness 3.7891 0.5138 0.1757 4.6364% 1.6 5
Extraversion 3.2235 0.6531 0.1863 5.7791% 1 5
Agreeableness 3.8843 0.4858 0.1729 4.4525% 1.5 5
Emotional Stability 3.5311 0.7018 0.2299 6.5116% 1 5
The within variation shows the individual level of change in personality traits. In order to analyse personality trait change at a population level, and to examine the normative personality trait trend, this paper will use a mean-level approach (Roberts & Mroczek, 2008; Cobb-Clark & Schurer 2012). The measures of change are constructed as follows:
∆T = IT
2015 - IT2009 (5)
where I represents a specific individual and T represents one of the 5 Big Five personality traits. The mean of the differences is taken and tested to see if it is significantly different to 0, revealing consistency or not.
Table 4
Mean difference of personality traits between 2009 and 2015. The panel data is balanced, using only individuals that participated in the 2009 and 2015 questionnaire. The subsample consists of 408 participants. *= p-value<0.1, **=p-value<0.05, ***=p-value<0.01.
2009 2015 Mean Difference Standard Deviation t-test Openness 3.5368 3.4758 -0.0618*** 0.3534 -3.6007 Conscientiousness 3.8455 3.8318 -0.0137 0.3723 -0.7565 Extraversion 3.1972 3.1616 -0.0370** 0.3803 -2.005 Agreeableness 3.9026 3.8948 -0.0066 0.3446 -0.3946 Emotional Stability 3.4993 3.6056 0.1073*** 0.5101 4.3322
Table 4 shows the results for the full sample. The mean of the differences for emotional stability is strongly significant and positive, which was expected since emotional stability increases with age (Wärneryd, 1996; Roberts et al., 2006; Robert & Mcrozek, 2008). Yet, interestingly
time span used here is 6 years. Roberts et al. (2006) find that there is a positive relationship between the effect sizes and the longitudinal time span. The magnitude of mean-level changes increases as longitudinal studies exceed 10 years. This would be interesting to check once the LISS panel data has matured some more. While indeed the magnitude of the mean differences are small (do not exceed 0.11), Roberts et al. (2006) emphasize that small changes in personality traits may have a profound effect on successful development across one's life course. In order to further investigate the normative trend of personality traits over a life course, table 5 divides the subsample into different age categories in order to depict the possible relationship of normative commitments (stemmed from universal life experiences) and personality trait change.
Table 5
Mean difference of personality traits between 2009 and 2015. The panel data is balanced, using only individuals that participated in the 2009 and 2015 questionnaire. The subsample consists of 408 participants. Age is split into the CBS (Statistics Netherlands) categories, except for the first age group 20-35. Normal CBS category is 15-25 and 25 to 20-35. But in 2015 there were 0 individuals that still remained in the 15-25 age group since the youngest participant in 2009 was 19 years old. Hence the age group 20-34. There are 11 participants in the age group 20-23, 32 in the age group 35-44, 71 in the age group 45-54, 109 in the age group 55-64 and 185 in the group 65+. *= p-value<0.1, **=p-value<0.05, ***=p-value<0.01.
Mean Standard Deviation
As mentioned early in this paper, life experiences or roles and expectations are the catalysts for the normative pattern. Social institutions are a complex of these roles, expectations and values
embedded in a specific social structure that organises the sequence of human activity with respect to fundamental problems of producing life-sustaining resources, reproducing and maintaining a viable societal structure within one's environment (Turner 1997: 6). Mehrotra & Jolly (2000) defines occupation, marriage, parenthood (child bearing), post-parenthood (launching of child), retirement, grandparenthood and widowhood as the most powerful universal experiences.
According to the CBS report on the trends in the Netherlands 2016 occupation, parenthood and marriage all occur in the age category 20 to 34, and this can be seen in the results. The population experiences approximately a third point increase in emotional stability (0.3538) on a 5 point scale between the ages of 20 to 34. This was expected since these three life experiences motivate
individuals to become more psychologically mature (if expectations are met), which is associated with emotional stability. Psychological maturity is also associated with conscientiousness and
agreeableness, so it is surprising to observe that conscientiousness only shows a significant decrease at
old age (65+) in the opposite direction than predicted, and the mean differences of agreeableness are not different from 0 at any age.
Extraversion shows a large decrease between the ages of 20 to 34 approximately a quarter point increase on a 5 point scale. The normative trend predicts an increase in extraversion during young adulthood. One would predict that individuals that begin their career and are seeking their life partner tend to be very outgoing and sociable. Yet our results show the opposite. This might be explained by age frequencies in this age category. Only 2 individuals in this age category are 25 or younger, so the life experience parenthood will most likely have a more salient effect in this age group as the average age to bear a first child in the Netherlands is 29.4. Parenthood tends to decrease
extraversion due to the new responsibilities that come along with it. Furthermore, extraversion shows
a negative change at older age (65+). In this age category, the average individual experiences
also in line with Roberts & Mroczek (2008) findings that openness tends to decrease around the age of 40.
Overall these results clearly contradict the Five-Factor theory which predicts little or no change in personality traits, and if there is change it occurs before adulthood and stems only from biological causes (Mcrae & Costa, 2006; Borghans et al., 2008). The within variation is strong and emotional stability, openness and extraversion show mean differences significantly different to 0 at all ages. This paper has bought new evidence on the normative trend of personality traits. The results show that the largest changes occur in young adulthood which is in line with Robert & Mcorzek (2008) conclusion that the greatest changes occur between 20 and 40.
5.2 Regression Analysis
The results of the regression analysis are presented in the following section. For liquid and total saving the results of the two-way fixed effect regression will be reported in order to determine the influence of personality trait change. For investment, insurance, debt and durable goods saving the average marginal effects of the two-way probit regression with Mundlak fixed effects will be reported in order to analyse the rate of change, followed by the conditional two-way linear regression model with Mundlak fixed effects in order to determine the influence of changes on the monetary amount held, conditional on ownership of one the four specific saving types.
5.2.1 Total saving
are both positively associated with total saving. This is in line with the theory that parents increase saving when they have children, for their children’s health and education in the future (Schultz, 2004)
5.2.2 Liquid saving
The results found for liquid saving, in table 6, do not corroborate with H2a. Both emotional stability and conscientiousness are not significantly associated with liquid saving which is surprising. However, the results partly corroborate with H2b as a unit change in openness will lead to an
approximate 53.2% decrease in liquid saving. Note that this is a very large decrease, but a unit change in openness on a scale from 1 to 5 is also a very large change. Openness is associated with risk-taking behaviour which is negatively associated with liquid saving decisions. Once again, children are positively associated with liquid saving.
5.2.3 Investment Saving
The results in table 7 do not corroborate with H3 which predicts that traits correlated with risk-loving behaviour would be associated with the risky saving type investment (i.e. openness and
extraversion). Yet, therefore it is not a surprise to find that conscientiousness is negatively correlated with likelihood of holding investment saving. As conscientiousness is associated with risk-averse behaviour (Soane & Chmiel, 2005). However, it is unexpected to see that emotional stability is significantly positively associated with the influence of the change in (inverse hyperbolic sin)
monetary value of investment saving. In other words, a unit increase in emotional stability will lead to a 35.33% increase in the monetary value of investment saving. The results from the regressions run on the component “stocks” of investment saving (table 8 in the appendix) support the positive
relationship between emotional stability and investment saving. Namely emotional stability is
positively related to the probability of holding stocks and the monetary value of stocks conditional on already participating on the stock market. A valid explanation could be that an emotionally stable head tends to plan and think thoroughly before making a decision (i.e. not impulsive), so if an
Table 6
Two-way fixed effect regressions of the inverse hyperbolic sin value of total and liquid saving. Data is unbalanced and is from the LISS data archive including waves 2009. 2011. 2013 and 2015. Average marginal effects reported throughout. *=p-value<0.1. **=p-value<0.05. ***=p-value<0.01. Note that the observations are lower than the actual observations found for total and liquid saving in table 2. Observations are lost because there are 5 less observations for personality traits (4479 obs), 36 less observations for live partner (4448 obs) and 174 less observations for household income (4310 obs).This adds to a total of 215 lost observations. 186 observations are lost with total saving and 174 with liquid saving.
Log Total Saving Log Liquid Saving
M.E. t-stat. M.E. t-stat.
Openness -0.2271 -0.53 -0.5318** -1.99 Conscientiousness 0.6392 1.59 0.1301 0.52 Extraversion 0.0913 0.25 -0.2611 -1.13 Agreeableness -0.4323 -1.06 -0.1398 -0.55 Emotional Stability 0.0057** 0.02 0.2926 1.56 Male Age 0.7936** 2.09 -0.0726 -0.31 Children 0.9010** 2.59 0.0082** 0.04 Live Partner 0.5709 1.01 0.1098 0.31 Separated 0.8129 0.95 0.1808 0.34 Low Education 0.3332 0.30 0.4783 0.7 Medium Education High Education 1.3222 1.27 0.5690 0.90 Employed 0.7653 1.57 0.4019 1.32 Self-Employed -1.6499* -1.76 0.3059 0.53 Pensioner 1.0430** 2.00 0.3633 1.08 Poor Health 0.1685 0.47 -0.0485 -0.22 Good Health
Log household Income -0.2626 -0.71 0.2663 1.18
Year 2009 6.1417** 2.46 -0.8387 -0.54
Year 2011 4.4490** 2.54 -0.5178 -0.48
Year 2013 2.0222** 2.08 -0.2723 -0.45
F-test (joint significance of fixed effects) 4.36*** 3.7***
χ 2 (time) 3.17** 0.21
R2 0.1297 0.1410
Table 7
The dependent variable used in this table is investment saving. The first column depicts the results for the two-way probit model including Mundlak fixed effects. This represents the probability of owning some type of investment saving. The second column depicts the results for the two-way conditional linear regression model including Mundlak fixed effects. This represents the influence of changes on the inverse hyperbolic sin monetary value of investment saving conditional on ownership. The Mundlak fixed effects were included for the following variables; the Big Five personality traits, age, children, live partner, employed, pensioner, poor health, good health and the logarithmic value of household income. These variables have a within variation larger than 0.1. The coefficients for the Mundlak fixed effects are not depicted in the table for space reasons as they are not of interest. Data is unbalanced and is from the LISS data archive including waves 2009, 2011, 2013 and 2015. Average marginal effects reported throughout.
*=p-value<0.1, **=p-value<0.05, ***=p-value<0.01.
As aforementioned under table 6 a of total 215 observations are lost due to the personality traits, live partner and household income variables. Notice that the amount of observations lost for the binary dependent variable is also 215.
5.2.4 Durable goods saving
As aforementioned the nature of durable goods saving may can be slightly ambiguous. If an individual engages in durable goods saving s/he is actually buying a durable good. The initial decision to buy a house or car is actually a consumption decision, therefore it can be understood as a negative saving decision to some extent. This is depicted in the first column in table 8. Yet once the durable good is owned it becomes a saving decisions. Take a house for example, if an individual owns a house s/he can choose to sell it or keep and maintain it in the hope that it will increase in value. This is
Holding
Investment Log Investment
M.E. t-stat M.E. t-stat
Openness 0.0307 0.80 -0.1025 -0.40 Conscientiousness -0.0025* -0.07 -0.3131 -1.34 Extraversion 0.0073 0.22 0.2904 1.28 Agreeableness 0.0133 0.37 -0.0998 -0.41 Emotional Stability 0.0280 1.02 0.3533* 1.96 Male 0.0348** 2.40 -0.1022 -0.64 Age 0.0008* 0.12 0.0223 0.39 Children -0.0575** -1.99 -0.0418 -0.21 Live Partner 0.0008** 0.02 -0.2334 -0.65 Separated -0.0363* -1.88 -0.1268 -0.59 Low Education -0.0576*** -3.20 -0.2259 -1.11 Medium Education High Education 0.0631*** 4.05 0.4166** 2.40 Employed 0.0041* 0.10 -0.0385 -0.15 Self-Employed 0.1272*** 3.73 0.6304** 2.06 Pensioner 0.0160 0.34 0.1253 0.45 Poor Health -0.0476 1.36 0.0247* 0.10 Good Health
Log household Income 0.0317 0.91 0.4845** 1.98
agreeableness will lead to 0,25 percentage point lower probability of owning a durable good. An agreeable person is modest, sympathetic and experiences high levels of concern for others, therefore is less likely to spend money on valuable goods.
This paper predicted that high scores in extraversion would be positively correlated with durable goods saving (H5). Yet, the results in table 8 show extraversion to have an insignificant effect. However, a significant effect is found on both components that make durable goods; a dwelling or other type of durable good (i.e. car, boat…), these results are shown in table A10 and A11 in the appendix. High scores in extraversion significantly increases the chance of owning other durable goods and significantly decreases the chance of owning a house. This supports the “social comparison theory” (Schor, 1998) that if one’s friend owns a car, the chances will increase that you will own a car. Extravert individuals tend to have more friends therefore are more likely to own a durable good
(excluding houses). Extraversion has a negative relationship with the probability of owning a house, this is probably because a dwelling is a more ‘serious’ durable good. Extravert people are excitement seekers thus less likely to be pinned down by owning a house. The opposite effects on the two components of durable goods saving may be why extraversion is insignificant, as they cancel each other out.
Furthermore, in table 8, conscientiousness positively significantly influences the change in logarithmic monetary value of durable goods savings. More specifically, table A10 in the appendix shows that the relationship between conscientiousness and the logarithmic value of durable goods savings resides from the durable good dwelling. An individual with high levels of conscientiousness will own a dwelling with relatively higher (logarithmic) value (see table A10 in the appendix). High scores in conscientiousness have been directly associated with job performance and wages (Nyhus and Pons, 2005; Hogan and Holland, 2003; Salgado, 1997). This may be why more conscientious
Table 8
The dependent variable used in this table is durable goods saving. The first column depicts the results for the two-way probit model including Mundlak fixed effects. This represents the probability of owning some type of durable goods saving. The second column depicts the results for the two-way conditional linear regression model including Mundlak fixed effects. This represents the influence of changes on the logarithmic monetary value of durable goods saving conditional on ownership. The Mundlak fixed effects were included for the following variables; the Big 5 personality traits, age, children, live partner, employed, pensioner, poor health, good health and the logarithmic value of household income. These variables have a within variation larger than 0.1. The coefficients for the Mundlak fixed effects are not depicted in the table for space reasons as they are not of interest. Data is unbalanced and is from the LISS data archive including waves 2009, 2011, 2013 and 2015. Average marginal effects reported throughout.
*=p-value<0.1, **=p-value<0.05, ***=p-value<0.01.
Holding Durable Goods
Log Durable Goods
M.E. t-stat M.E. t-stat
Openness 0.0129 0.41 -0.1795** -1.99 Conscientiousness -0.0046 -0.16 0.2318*** 2.68 Extraversion 0.0304 1.12 0.0908 1.14 Agreeableness -0.0025* -0.09 -0.0879 -1.02 Emotional Stability -0.0152 -0.70 0.0500 0.78 Male 0.0515*** 4.55 0.0217 0.21 Age 0.0039 0.75 0.0434 1.35 Children -0.0203 -0.85 -0.0903 -1.24 Live Partner 0.0695* 1.77 0.0594 0.47 Separated 0.0175 1.29 -0.2967*** -2.58 Low Education -0.0435*** -3.27 -0.3878*** 0.11 Medium Education High Education 0.0344** 2.45 0.1534 1.43 Employed -0.0006** -0.02 0.1117 1.07 Self-Employed 0.1247*** 3.37 0.2777* 1.75 Pensioner 0.0499 1.38 0.0796 0.73 Poor Health -0.0166 -0.63 0.0647 0.83 Good Health
Log household Income 0.0264 0.91 0.2318*** 2.90
Year 2009 0.0362 1.40 0.3958** 1.97 Year 2011 0.0397* 1.86 0.2469* 1.70 Year 2013 0.0210* 1.86 0.1174 1.34 χ 2 (likelihood ratio) 1169*** 546.46*** χ 2 (time) 4.56 4.67 χ 2 (Mundlak) 52.79*** 107.63*** Obs 4273 3433 5.2.5 Debt Saving
Surprisingly, emotional stability is positively associated with the monetary value of debt held. The relationship between emotional stability and debt derives from debt’s component mortgage (see table A12 in appendix). A unit increase in emotional stability leads to a 19,46% increase in the logarithmic amount of mortgage, conditional on already holding mortgage. An explanation for this relation could be that mortgage is a more strategic take up of debt than other debt types such as credit card debt and overdrafts therefore requires more planning and a more stable profile, which is
associated with emotional stability.
Furthermore, it is interesting to notice that low education is negatively related to the
Table 9
The dependent variable used in this table is debt saving. The first column depicts the results for the two-way probit model including Mundlak fixed effects. This represents the probability of owning some type of debt saving. The second column depicts the results for the two-way conditional linear regression model including Mundlak fixed effects. This represents the influence of changes on the monetary value of debt saving conditional on ownership. The Mundlak fixed effects were included for the following variables; the Big 5 personality traits, age, children, live partner, employed, pensioner, poor health, good health and the logarithmic value of household income. These variables have a within variation larger than 0.1. The coefficients for the Mundlak fixed effects are not depicted in the table for space reasons as they are not of interest. Data is unbalanced and is from the LISS data archive including waves 2009, 2011, 2013 and 2015. Average marginal effects reported throughout.
*=p-value<0.1, **=p-value<0.05, ***=p-value<0.01.
Holding Debt Log Debt
M.E. t-stat. M.E. t-stat
Openness 0.0253 0.61 -0.0331 -0.35 Conscientiousness 0.0169 0.44 0.0677 0.73 Extraversion -0.0072 -0.20 -0.1271 -1.52 Agreeableness -0.0120 -0.31 -0.0103* -0.12 Emotional Stability -0.0149 -0.52 0.1533** 2.25 Male 0.0200 1.29 0.0995 1.02 Age 0.0089 1.24 0.0238 0.82 Children 0.0056 0.18 -0.1339* -1.77 Live Partner 0.0274 0.51 0.1846 1.39 Separated 0.0291 1.48 0.0469 0.40 Low Education -0.0644*** -3.63 -0.0673 -0.60 Medium Education High Education -0.0038 -0.21 0.0568 0.59 Employed 0.0447 1.01 -0.0164 -0.15 Self-Employed 0.0264 0.66 0.2437 1.47 Pensioner -0.0141 -0.29 -0.3011** -2.47 Poor Health 0.0230 0.63 -0.0539 -0.63 Good Health
Log household Income 0.0666* 1.67 -0.0565 -0.68
