40
Financial Literacy and the Under-Insurance Puzzle
Martijn Vast S2738376
Faculty of Economics and Business Economics University of Groningen
Master Thesis Finance & Economics Supervisor: Dr. Carolina Laureti Date: 11-01-2021
ABSTRACT
This study examines whether households’ insurance choices are affected by their actual level of financial literacy and overconfidence regarding their financial literacy. This paper uses data from the De Nederlandsche Bank’s household survey conducted in 2005, which was accompanied by an exclusive financial literacy module. This dataset allows for analysing the level of insurance of endowment insurance, annuity insurance, and mortgage insurance using logistic regression. Households with a higher financial literacy or confidence in their financial literacy possess a significantly higher level of insurance. Furthermore, financial literacy overconfidence does not affect their level of insurance choices in an aggregate insurance environment, yet it does negatively impact the levels of insurance zooming in on the individual insurance products. This shows the importance of taking financial literacy confidence into account when constructing financial education programs, which thus far have been ineffective in tackling sub-optimal decision making. Lastly, the findings regarding mortgage insurance are in dissonance with these general observations.
1
I.
Introduction
After the 2008 global financial crisis, governments have been encouraging households to get more involved with their own personal finances. As a result, they got more financial responsibilities in planning their future needs. Simultaneously, the financial markets and its products have gotten more complex over the last decade. Combined, this highlights the increased importance of households their ability in managing their finances. Unsurprisingly, researchers have uncovered they are experiencing difficulties with governing their financial planning. The choice of what level of insurance to adopt is one of the areas households struggle with. Recent research1 related to the coronacrisis showed that households have underestimated the risk of job loss when taking on a mortgage, which has led to low levels of insurance. Yet, this crisis emphasizes how uncertain the future actually is and how impactful the consequences can be. Other empiric studies reported low levels of insurance among households as well. A mortgage insurance adoption study by Cox and Zwinkels (2019) reported individuals to under-insure against low-probability, high consequence risks. Moreover, Ameriks et al. (2018) observed individuals to insure little against late-in-life risks, such as long-term care. These findings contrast with predictions based on classical risk aversion theory (Schwarcz, 2010).
Considering people are extremely risk-averse in other economic domains, the phenomenon of under-insurance is quite striking. High risk-exposure to various uncertainties, such as natural disasters and negative home equity, is observed by many studies (Ameriks et al., 2018; Cox and Zwinkels, 2019; Bernheim et al., 2013; Von Peter, Von Dahlen, and Saxena, 2012). Standard economic theory predicts great value for risk-averse individuals, yet, private insurance markets remain small over the globe (Brown and Finkelstein, 2007; Mitchell et al., 2011). Several approaches to explain the under-insurance puzzle have been taken. However, researchers fail to fully understand what drives demand. For instance, product design flaws only partially explain the puzzle (Ameriks et al., 2018). Even subsidizing insurance against disaster risk in disaster-prone regions has little impact on demand (Viscusi, 2010).
Sub-optimal financial decision making in general has been studied extensively. Within this stream of research, financial literacy has received a lot of attention. Many studies report individuals to make poor financial decisions, due to the lack of the required knowledge and skills to adequately evaluate all factors influencing their decision. Van Rooij et al. (2012) showed that households who have little grasp of certain economic concepts, such as risk diversification and interest compounding, tend to plan less for retirement. Demand for insurance
1
2 also appears to be affected by financial literacy. Giné and Yang (2009) state that a lack of understanding index insurance negatively affected demand of the product in Malawi. Such financial illiteracy does not appear to be limited to countries with undeveloped markets only. Countries such as Germany and The Netherlands report very low levels of financial literacy as well (Lusardi and Mitchell, 2011).
Perceived literacy and the discrepancy between actual literacy and one’s perception might also affect financial decisions. Anderson, Baker, and Robinson (2017) studied whether actual or perceived financial literacy is a better predictor of precautionary savings. Their results showed that perceived literacy has more predictive power than actual financial knowledge. Moreover, experimental results documented by McCannon, Asaad, and Wilson (2016) indicate that individuals who are confident in their own financial competence make more trusting investments, indicating risk-loving behaviour. This shows the importance of taking into account perceived literacy, as well as actual literacy. However, this allows room for household’ misjudgements regarding their financial knowledge and abilities to impact behaviour as well. The degree of misjudgement, or miscalibration, is referred to as overconfidence. Overconfidence has been shown to lead to sub-optimal decision making and excessive risk-taking behaviour (Daniel and Hirshleifer, 2015; Malmendier and Tate, 2008; Chu et al., 2017).
This study attempts to get a deeper understanding of insurance choices made by households, by researching the influence of financial literacy and financial literacy overconfidence on such decisions. The methodological approach closely resembles the work of Kramer (2016) and Hogeboom (2019). First, actual financial literacy will be assessed by a series of questions regarding basic and more advanced economic concepts. Subsequently, perceived financial literacy will be evaluated based upon their own assessment of their financial knowledge. Based upon actual and perceived literacy a combined measure of financial literacy is constructed, aiming to represent overconfidence. This measure allows for identifying households that are overconfident as well as households that underestimate their own financial knowledge.
3 Although these do not present a complete representation of total insurance choices, it does allow for obtaining preliminary insights into insurance behaviour regarding late-in-life risks.
Although financial literacy has been a popular subject of interest among economic researchers, most have focussed on investor behaviour instead of household behaviour. This study adds to the literature of household behaviour by obtaining a deeper understanding of the way in which financial literacy and overconfidence influence decision making, especially for decisions related to insuring against low probability, high consequence scenarios. This should provide interesting insights for policy makers of financial education programs. Such programs have become the common conception of how to tackle sub-optimal decision making due to a lack of financial knowledge. However, there is some disagreement on its effectiveness. Martin (2007) documents that its benefits are indeed observed for a variety of household decisions, such as saving and retirement planning. Nevertheless, other studies have shown that its effectiveness is quite heterogeneous/limited (Kaiser and Menkhoff, 2017; Fernandes et al., 2014). Therefore, more insights in what drives household behaviour is required to provide well-tailored education programs or other tools that aim for sound financial decision-making by households.
The remainder of this paper follows the following structure. The next section will provide a synopsis of household behaviour, financial literacy, overconfidence, socio-economic variables and how they are interrelated. Section III and IV provide a thorough explanation of the methodological approach taken, as well as an in-dept description of the DHS dataset and financial literacy module. Hereafter, the main results will be discussed in detail. Section VI concludes and considers interesting directions for future research based upon the limitations and results of this study.
II.
Literature review
2.1 Under-Insurance Puzzle4 Gandolfi and Miners (1996). They observed that education of wives is negatively correlated with insurance on the husband. Earlier findings on life insurance suggest income, household size, age, and education to be affecting demand as well (Burnett and Palmer, 1984; Truett and Truett, 1990; Fischer, 1993; Guillemette et al., 2015). This supports the modern belief that insurance demand is not only driven by risk-aversion.
A recurring topic within the insurance literature is the observation of under-insurance by households. Ameriks et al. (2018) state that “only ten percent of elderly Americans hold long-term care insurance”, whereas one in three will eventually end-up in a nursing home with an average cost of $84,000 a year. Cox and Zwinkels (2019) report that mortgage insurance is merely adopted by 30 percent of households, despite mortgage debt being a substantial liability of households’ balance sheets. This contrasts with the idea of agents behaving risk-averse, observed for many other financial decisions; one of the main motives of saving, taking precautions, is driven by such risk-aversion (Skinner, 1988).
Many researchers have attempted to explain the under-insurance puzzle using various angles. Adverse selection is brought up by several studies as a possible explanation for little demand in annuity insurance markets (Brown, 2001; Brown et al., 2007). They provide some evidence that individuals with poor health are less fond of annuitizing as a result of reduced mortality. Ameriks et al. (2018) investigated whether product design flaws influence insurance decisions. Yet, product imperfections only provide a partial explanation, as demand is still lower than their model predictions when facing the ideal product. Other explanations range from heterogeneous preferences, asymmetric information, and framing (Cutler, 2008; Brown, 2008). Although these studies have some traction in terms of explanatory power, its statistical power remains limited or creates other puzzles. Therefore, this paper attempts to explain the under-insurance puzzle from another angle that has been able to explain several other sub-optimal patterns of financial decision-making: financial literacy.
2.2 Financial literacy and household decision-making
5 are low (Lusardi and Mitchell, 2014; Klapper et al., 2015). South-Asia is considered one of the most financially illiterate regions. However, financial illiteracy is not only observed for developing countries; Lusardi and Mitchell (2011) also reported low financial literacy levels for Germany and The Netherlands.
Numerous studies have provided more insights in the determinants of financial literacy levels. Lusardi and Mitchell (2010) reported significant gender related literacy differences; women are less financially literate than men are. Chen and Volpe (2002) obtained a similar assessment among college students. This relation between gender and financial literacy levels is even more profound for the elderly (Lusardi and Mitchell, 2008). Age appears to be negatively correlated with financial knowledge. Income, education, and employment have also been shown to positively affect financial literacy levels (Al-Tamimi and Bin Kalli, 2009; Lusardi et al., 2010; Guiso and Jappelli, 2008).
Several patterns of financial decision-making have been attributed to varying levels in financial literacy. Van Rooij et al. (2011) find that stock market participation is positively affected by having a greater understanding of financial concepts. Not only participation, but also stock market behaviour is positively affected. Financially literate households hold better portfolios in terms of diversification. However, negative effects of financial literacy are widely reported as well, especially concerning household behaviour. Moore (2003) and Miles (2004) showed that financial illiteracy increases the likelihood of taking out a mortgage loan with unfavourable terms. Unpreparedness for retirement is extensively documented as well (Lusardi and Mitchell, 2007a; Van Rooij et al., 2011).
Overall, higher literacy levels appear to go hand in hand with sound financial decision-making. Expectations are that this conclusion carries over to demand for insurance. For this reason, this study proposes that a higher actual financial literacy increases insurance levels held by households.
2.3 Financial literacy overconfidence
6 decisions. Allgood and Wallstad (2012), therefore, propose a combined financial literacy measure of actual and perceived financial competence. Others have already adopted such a combined measure and confirmed the relevance of perceived financial literacy, such as Van Rooij et al. (2011). Bannier and Neubert (2016) even observed perceived financial competence to be a stronger determinant of sophisticated financial decisions than actual financial competence. For that reason, this study expects to find the same relation between insurance demand and perceived financial literacy as with actual financial literacy: perceived financial literacy increases households’ demand for insurance products.
The interaction between objective and perceived financial knowledge have been shown to affect financial decision-making as well. Allgood and Walstad (2012) state that a combined measure of actual and perceived financial literacy provides more insights in behavioural tendencies compared to actual literacy on itself. Individuals overestimating their own financial knowledge and capabilities are defined as overconfident2 (Bhandari and Deaves, 2006). This behavioural trait manifests itself in several ways: miscalibration, above-average effect, and excessive optimism. Miscalibration refers to the overestimation of one’s beliefs regarding its own knowledge and information (Ackert and Deaves, 2011). The above-average effect describes the tendency to overestimate their personal capabilities compared to the population. Lastly, the excessive optimism captures the flawed estimation of likelihoods, overestimating the likelihood of the desired outcome compared to the unfavourable outcome (Weinstein, 1980).
Several explanations for the factors driving overconfidence have been brought up by the literature. One of them is the self-attribution bias, which captures the misattributing of good performance to one’s ability and blaming the environment or bad luck to bad performance (Kim, 2013). In addition, the illusion of control and knowledge fuels excessively optimistic behaviour; the vast quantity of data and information available leads to overestimating the accuracy of their forecasts (Barber and Odean, 2002). Another driver is the confirmation bias; the selective interpretation of new information in such a manner that it supports their beliefs (Duong, Pescetto, and Santamaria, 2014). All factors could drive individuals to misperceive actual probabilities. Considering insurance decisions are highly dependent on probability assessments, it is plausible that they affect precautionary financial decision making such as insurances.
Besides the extensive research on the cognitive side what is driving overconfidence, there has also been a stream of research attempting to tie demographics/personal characteristics to overconfident behaviour. One of the main findings is the presence of gender effects. Although
7 there is overconfidence present within men and women, men are more overconfident than women are (Barber and Odean, 2001; Van Rooij et al., 2011). This appears to be even more prevalent in male-dominated contexts, such as the realm of finance. Men are more confident in their financial competence than women (Prince, 1993). Besides gender effects, education level has also been shown to affect confidence. Bhandari and Deaves (2006) reported the intuitive result that overconfidence is more likely to be observed among highly educated individuals. Deaves et al. (2010) reported similar findings, as well as experience having a positive correlation with overconfidence. Other studies reported overconfidence increasing with income (Pirinsky, 2013; Bhandari and Deaves, 2006) and with age (Pirinsky, 2013; Pak and Chatterjee, 2016).
The consequences of overconfidence for investors are unmistakable. Excessive trading might be one of the most widely accepted results brought up by the field of behavioural finance. Many studies confirmed this investor tendency and showed that their trading performance was hurt by it (Grinblatt and Keloharju, 2006; Daniel and Hirshleifer, 2015; Barber and Odean, 2000). Malmendier and Tate (2008), Barros and Silveira (2007), and Ferris et al. (2013) showed that CEO’s decision-making is affected by overconfidence as well. They report overconfident CEOs to choose financing structures with higher leverage, higher volume of acquisition offers due to overestimating their value creation abilities, and that CEOs are more likely to partake in value destroying M&A.
However, little is known about how household behaviour is affected. The general conclusion that can be drawn from the scarce number of studies that have looked at this behaviour is that overconfident household tend to make more risky decisions. For instance, Hauff and Nilsson (2020) studied indebtedness among young adults and found that those who are overconfident about their financial knowledge borrow more. This contrasts with Hogeboom’s (2019) findings, who was unable to find a relation between overconfidence and saving and borrowing behaviour. Another study documented early retirement withdrawals to be more likely among overconfident individuals (Kim et al., 2019). Moreover, they state that the individuals do not fully understand the consequences of early withdrawal. This matches the negative relationship of objective knowledge with early withdrawal documented as well. Financial advice demand has also been shown to be negatively related with overconfidence (Kramer, 2016; Porto and Xiao, 2016; Gentile et al., 2016).
8 2.4 Financial literacy and insurance decisions
Despite numerous studies using financial literacy (overconfidence) to successfully explain behavioural patterns within household finances, research combining financial literacy and insurance decisions is scarce. Among the few studies that in fact explored this relation is the work of Allgood and Walstad (2016). They reported that financial literacy indeed positively affects insurance coverage. Respondents in their financial behavior survey had to specify whether they possessed one or more of three separate types of insurance: auto insurance, life insurance, and health insurance. Results showed that respondents with a high actual financial literacy and high-perceived financial literacy were four percentage points more likely to have any insurance, compared to those with low actual financial literacy and low-perceived financial literacy.
Similar results are reported by Lin et al. (2017). Their analysis also revealed a positive association between financial literacy and insurance demand. Unsurprisingly, the component of financial literacy concerning insurance and retirement planning has the strongest explanatory power. However, Mahdzan and Victorian (2013) could not establish this positive relation between financial literacy and life insurance demand. Their study into the determinants of life insurance demand yielded the unexpected conclusion of financial literacy having no significant impact on life insurance decisions. Due to the lack of research into the relation between financial literacy and insurance levels, in combination with mixed evidence, it is too premature to draw any sensible conclusions. Therefore, this study aims to contribute to this stream of research. The following section presents the econometric approach taken to quantify the insurance-financial literacy relation. Specifically, the following hypotheses are tested:
Hypothesis 1: Higher actual financial literacy increases households’ level of insurance. Hypothesis 2: Higher perceived financial literacy increases households’ level of insurance. Hypothesis 3: Overconfident households adopt a lower level of insurance.
III. Methodology
9 within the 0 to 1 domain, which is impossible. To get rid of this problem, models are required to take a non-linear approach. The conventional advice is the use of logit or probit models. The logit model and the probit model use a function that transforms the regression model to ensure the values are bounded within the (0, 1) interval. Visually, the fitted regression model appears as an S-shape rather than a straight line. Considering OLS can only estimate linear models, the logit model and the probit model are estimated using maximum likelihood. These two models produce similar results, making the choice between the two arbitrary (Brooks, 2008). This study adopts the logit model. Although the differences are trivial, the results will be replicated using a probit model as a robustness check.
The specification of the logit model is as follows:
Ii = α1+ β1AFLi+ β2PFLi + β3FLOi+ θ1IEi+ θ2CFi+ εi (1) where Ii is replaced by one of four dependent variables: annuity insurance, endowment insurance, mortgage insurance, or any insurance. AFLi represents the number of correctly answered financial literacy questions for respondent I; PFLi represents the score of perceived financial literacy assigned by respondent I; FLOi is a dummy variable with value one in case respondent i has been classified as overconfident and zero otherwise; IEi represents interaction effects between overconfidence and several socio-demographics for respondent I; CFi represents several socio-demographic control factors for respondent I; and εi is the error term. To avoid biased standard errors due to possible heteroscedasticity, this study uses heteroscedasticity-robust standard errors in all estimations.
10
IV.
Data
4.1 Dataset
This study uses the data obtained from the DNB Household Survey (DHS) conducted by CentERdata, a survey agency located at the Tilburg University. Approximately 2000 Dutch households participate every year, providing information on demographic characteristics and economic aspects, such as employment, income and insurance choices. This study merely uses the data of the survey conducted in 2005, as this year’s survey included an exclusive financial literacy module required to obtain a measure of the respondent’s financial literacy. This module is devised by Van Rooij et al. (2011) and includes several questions on financial concepts varying in difficulty. The questions were administered only to those who are in charge of the household’s finances, with a response rate of 74.4%. Selection bias is dealt for by providing the required technological equipment, such as a computer, to those who lacked possession.3
Of the 1508 observations the financial literacy module provided, 481 were deleted from the sample as those individuals did not provide information on the insurance products in their possession. Moreover, several observations are excluded from the sample as they did not answer the question on perceived financial literacy or failed to provide information about their net income. Similar to Van Rooij et al. (2011), the top and bottom 1% of the net monthly income distribution are trimmed to prevent skewing results due to outliers. The final sample, therefore, consists of 937 observations. As a robustness check, the analysis will be replicated without the income-related data trimming.
4.2 Key variable construction
This subsection describes the construction of all key variables included in the model. The dependent variables are described first, followed by the main explanatory variables and the socio-demographic control factors.
The dependent variables are constructed as follows. Respondents were asked to indicate whether they have single-premium annuity insurance policies and endowment insurance policies in their possession. Both insurance products are set up with a wealth accumulation phase, to be paid out after the maturity date. Moreover, to assess mortgage insurance possession hey were asked whether they have a mortgage backed by the National Mortgage Guarantee (NMG). This implies that homeowners are provided a safety by a government-backed foundation under specific circumstances, such as losing your job or the passing of your spouse. For each of the three insurance products, a dummy variable is constructed. A value of one is assigned in case the individual in fact possessed the insurance product in question, and zero
11 otherwise. Lastly, the fourth dependent variable indicating whether the individual possess any form of insurance at all is constructed in a similar fashion.
The main explanatory variables relate to the respondents’ financial literacy. Actual financial literacy is observed by examining 5 basic and 11 more advanced financial literacy questions included in the module designed by Van Rooij et al. (2011). The basic questions primarily assess financial numeracy and interest compounding, whereas the more advanced questions are related to the characteristics of financial instruments. The exact wording of the 16 questions is described in Appendix A. Corrects answers to these questions are counted to obtain a score out of 5 and 11, for the basic financial literacy and the advanced financial literacy test respectively. “Don’t know” answers are regarded as incorrect answers. To take into account the possible differences between answering incorrectly and “Don’t know” pointed out by Lusardi and Mitchell (2007b), this study will repeat the analysis using an alternative scoring method to check for robustness. The alternative score for financial literacy is constructed as a percentage of correctly answered questions out of the total questions answered, where “Don’t know” answers are excluded.
A measure for perceived financial literacy is obtained by inspecting the respondents answers on the following question: “How would you rate your knowledge regarding financial matters on a scale of 1 to 7?”. Based on these scores, respondents can also be classified in terms of financial literacy confidence. Most importantly, respondents perceiving themselves as above average while scoring below average on the advanced financial literacy test are classified as overconfident. Based on this classification a dummy variable is constructed with a value of one in case the individual is classified as overconfident and zero otherwise. Similarly, a dummy is constructed to indicate under confidence; respondents scoring above average on the financial literacy test, but having a below-average perception of their financial skills. Considering under confident behaviours might differ from those with the “right” perception and could, therefore, affect baseline behaviour. This procedure corresponds to the approach previous researchers have taken (Van Rooij et al., 2011; Hogeboom, 2019). The exact choice of the thresholds indicating performance and perception above or below average are discussed in section 4.4 and 4.5.
12 4.3 Descriptive statistics
The general characteristics of the final sample are summarized in table 1. The respondents are between 22 and 90 years old, with the average head of household aged 51 years. The sample is overrepresented by men, as almost 70% of the respondents is male. Apparently, within the household men are more often taking on the responsibility of financial planning. Education, however, seems to be more evenly represented across the sample. Moreover, the average respondent is most likely to be in employment or already retired, as this comprises approximately 75% of the sample.
The basic financial literacy test is scored quite adequately, with an average of 4.16 questions answered correctly over a total of 5. Whereas the average of correctly answered questions of the more advanced financial literacy test is only slightly more than half the questions; an average of 6 correctly answered questions out of 11. Consistent with the literature, most possess the knowledge required to grasp standard economic concepts. However, they lack the knowledge to optimally evaluate the situation as decisions get more complex. Considering the little variability in basic financial literacy scores, these will not be incorporated as an explanatory variable of insurance behaviour. Only the scores on the advanced questions will be taken into account, as these are more evenly distributed. This follows the approach taken by Hogeboom (2019).
13
Variable N Range Mean Std. Dev.
Insurance behaviour Endowment insurance 937 0-1 0.125 0.331 Annuity insurance 937 0-1 0.228 0.420 Mortgage insurance Any insurance Financial literacy
Basic financial literacy score Advanced financial literacy score Perceived financial competence Financial literacy overconfidence Low perceived – Low advanced Under confident
Overconfident
High perceived – High advanced Socio-demographic controls Age
Net monthly income (€) Male Female High education Middle education Low education Employed Self-employed Retired Other
One or more children No children 488 937 937 937 937 937 937 937 937 937 937 937 937 937 937 937 937 937 937 937 937 937 0-1 0-1 0-5 0-11 1-7 0-1 0-1 0-1 0-1 22-90 250-4600 0-1 0-1 0-1 0-1 0-1 0-1 0-1 0-1 0-1 0-1 0-1 0.336 0.403 4.157 6.575 4.767 0.205 0.158 0.219 0.418 51 1782 0.693 0.307 0.446 0.311 0.242 0.568 0.042 0.250 0.141 0.297 0.703 0.472 0.491 1.030 2.949 1.152 0.404 0.365 0.414 0.494 15.176 719.418 0.462 0.462 0.497 0.463 0.429 0.500 0.200 0.433 0.348 0.457 0.457 Table 1 Descriptive statistics
14 4.4 Actual financial literacy
The following two subsections discuss the content of table 2. This table provides a cross-sectional overview of insurance levels and the respondents’ financial literacy scores on the basic and the advanced literacy test. Moreover, a similar overview on insurance levels, perceived financial literacy, and financial literacy overconfidence is presented.
Table 2 provides some interesting preliminary insights into insurance choices. The results hint at a positive correlation between adoption of any type of insurance and performance on the basic financial literacy test. Of the respondents answering all questions correctly, 43% possess one or more of the insurance types of interest. This is significantly higher than the 13% observed for those answering zero questions correctly. When deconstructing any insurance adoption into the three insurance products, this relation is also observed for endowment insurance and annuity insurance. However, the inverse appears to hold for mortgage insurance. This could be a first indication of counterintuitive results that will be tested and discussed in section 5.
15 Insurance products
Endowment Annuity Mortgage Any
Basic financial literacy indicator
None correct 0.00 0.13 0.00 0.13 1 0.07 0.13 0.75 0.33 2 0.04 0.19 0.36 0.23 3 0.10 0.22 0.39 0.36 4 0.12 0.20 0.37 0.41 5 0.35 0.26 0.30 0.43
Advanced financial literacy indicator None correct 0.04 0.13 0.38 0.23 1 0.00 0.06 0.53 0.29 2 0.03 0.19 0.44 0.32 3 0.09 0.09 0.40 0.34 4 0.08 0.10 0.54 0.30 5 0.07 0.24 0.32 0.40 6 0.09 0.23 0.51 0.42 7 0.12 0.28 0.32 0.38 8 0.14 0.38 0.30 0.42 9 0.21 0.49 0.27 0.55 10 0.17 0.40 0.29 0.41 11 0.23 0.44 0.17 0.49
Perceived financial literacy
1 0.00 0.25 0.00 0.25 2 0.03 0.20 0.27 0.26 3 0.06 0.18 0.30 0.30 4 0.09 0.20 0.33 0.38 5 0.12 0.21 0.38 0.40 6 0.19 0.30 0.29 0.48 7 0.27 0.30 0.42 0.50
Financial literacy overconfidence
Overconfident 0.07 0.18 0.47 0.36
Other 0.14 0.24 0.31 0.42
Table 2 Distribution actual and perceived financial literacy
This table provides a cross-sectional overview of insurance levels and financial literacy scores obtained by the respondents of the financial literacy module designed by Van Rooij et al. (2011). The scores are separated in two parts, the basic and the advanced set of questions. Moreover, the table denotes the perceived financial literacy score distribution based upon the following question: “How knowledgeable do you consider yourself with respect to financial matters, on a scale ranging from 1 to 7? Total insurance levels and a breakdown of its composition is denoted per score group, with levels ranging from 0-1. Any insurance might state a lower percentage than mortgage insurance, as the latter percentage is constructed using only the homeowners sample. Data is obtained from the DNB household survey conducted in 2005.
16 4.5 Perceived financial literacy and overconfidence
In general, the respondents in the sample appear to perceive themselves as quite knowledgeable regarding financial matters. Table 2 shows that the rating they awarded themselves are heavily distributed around the higher end of the 1 to 7 scale. Moreover, it suggests a positive relation between insurance levels and one’s perception of financial literacy, as was hypothesized in section 2. Once more, the statistics on mortgage insurance are in dissonance with this general finding.
Considering the average perception of financial literacy is approximately 4.75 out of 7, the cut-off to determine above average perception is set at 5. In combination with the cut-cut-offs chosen that determine above-average performance on the advanced financial literacy test, this allows for classifying the respondents based on their financial literacy confidence levels. Table 1 reports how this splits the sample into four types. For instance, those who have a high perception of their financial literacy, but actually performed below-average on the advanced financial literacy test are classified as overconfident. Almost 22% of the sample can be characterized as such. As can be deduced from table 2, individuals with this particular trait appear to exhibit deviating behaviour in terms of insurance coverage. Overconfident individuals adopt substantially less endowment and annuity insurance, as was expected based on existing literature. However, they do possess more mortgage insurance. This is the opposite relation compared to what was hypothesized. Although unexpected, this finding is consistent throughout the whole dissection of insurance choices discussed in section 4.4 and 4.5.
V.
Results
This section provides the results of the logit estimations. First, the results regarding the relation between actual financial literacy and insurance choices are discussed. Thereafter, the impact of perceived financial literacy and the overconfidence trait will be analysed. Lastly, any interaction effects with overconfidence are discussed.
5.1 Actual financial literacy
17 Any insurance Endowment insurance Annuity insurance Mortgage insurance Advanced 0.014** (0.006) 0.021*** (0.005) 0.015*** (0.005) -0.014* (0.008) Age dummies
(base group: Age ≤ 30)
31-40 0.149** (0.061) 0.042 (0.050) 0.061 (0.058) 0.112 (0.090) 41-50 0.195*** (0.061) 0.067 (0.049) 0.103* (0.057) 0.105 (0.088) 51-60 0.237*** (0.060) 0.065 (0.049) 0.178*** (0.057) 0.070 (0.090) 61-70 0.146* (0.077) 0.052 (0.062) 0.072 (0.073) -0.064 (0.117) >70 -0.167* (0.094) -0.212* (0.112) -0.361*** (0.129) 0.034 (0.132) Male -0.023 (0.037) 0.004 (0.027) -0.010 (0.034) 0.003 (0.056) Education dummies
(base group: low education)
Intermediate education 0.045 (0.042) 0.014 (0.030) 0.009 (0.039) 0.151** (0.062) High education 0.080* (0.041) -0.022 (0.032) 0.013 (0.037) 0.155** (0.060) Ln(net income) 0.127*** (0.045) 0.042 (0.035) 0.088** (0.041) -0.183*** (0.066) Children 0.063* (0.035) 0.040* (0.023) 0.013 (0.030) -0.065 (0.049) Employment status
(base group: no work)
Employed 0.137*** (0.047) 0.060 (0.039) 0.066 (0.042) 0.167** (0.071) Self-employed -0.190** (0.092) -0.099 (0.082) -0.180* (0.098) 0.005 (0.134) Observations 937 937 937 488
Table 3 Insurance levels and financial literacy
18 Although the results provide evidence for age effects, a different relation than hypothesized has been brought to light. Lusardi and Mitchell (2008) reported a negative correlation between age and financial knowledge. Combining this with the idea that financial knowledge and level of insurance are positively correlated, age and insurance levels were expected to be negatively correlated as well. However, the logit estimation actually discovered a positive relation between insurance levels and age, similar to Hogeboom (2019). This positive relation dampens after their 60th life year, with households older than 70 adopting even less insurance than the base group. A possible explanation for this concave relation might be related to different considerations throughout the lifecycle. For instance, at the earlier stages of the lifecycle households might be more interested in insurance adoption for mortality or longevity risk reasons and insuring themselves against the many uncertainties ahead. However, for the older generations there is less need for insurance as a result of the lower expected life expectancy. Such a life-cycle stage dependent argument seems to have more sensible explanatory properties than the financial literacy argument. More research is required to make strong conclusions on the age-insurance relation.
Furthermore, gender does not significantly affect insurance choices. Earlier studies into socio-demographic determinants of insurance levels of Truett and Truett (1990) and Burnett and Palmer (1984) did not report gender to be significant either. However, they did report highly-educated individuals to have higher levels of insurance. This study observed high education to be a significant determinant as well (at a 10% significance level). Highly educated households are 8% more likely to have any insurance compared to households with a low education. They appear to be more aware of the benefits of insuring against future adversity. Net income has also been shown to positively influence insurance choices; earning 1000 euros more a month increases the probability of a household possessing any insurance with 3.7%. Being employed and having children also positively affect the level of insurance a household decides to adopt. This corresponds with the work of Truett and Truett (1990) and Burnett and Palmer (1984) who stated that for larger families the head of the household feels more responsibility for ensuring the wellbeing of their family and, therefore, chooses a higher level of insurance to protect against adversities. Quite surprisingly, self-employment has a significantly negative effect on insurance choices. Based on the additional risks involved with having your own business, such as the lack of employee benefits, one would actually expect a higher level of insurance to counter the heightened risk-profile concerned with their employment status. A possible explanation might be that those who are self-employed are less risk-averse in general, which translates to adopting a lower level of insurance as well.
19 policies. The endowment and the annuity insurance policy are financial products that guarantee an income stream after the individual has gone through an accumulation phase. Thus, these insurance types are not that different from a saving policy. However, mortgage insurance’s features differ substantially from a saving policy. Although payments are made upfront, there is no accumulation of funds as with the other policies. Therefore, households could perceive this as making investments without certainty of its benefits. Perhaps, more financially literate households are more confident in their ability to construct their own “safety net” through insurance-related savings policies. Therefore, they are less interested in insurance products without certain wealth accumulation, such as a mortgage insurance.
A second explanation could be related to the NHG-borrowing limit which determines whether a household is eligible for taking on an NHG-backed mortgage. The exact limit for 2005 could not be retrieved, however, in 2020 this limit to apply for an NHG-backed mortgage had been set at €310,000. Thus, households with a mortgage greater than this limit do not even have the option to adopt this insurance. Wealthier households are intuitively more likely to have a mortgage above this limit, as they are more inclined to buy houses of a greater value. Considering financial literacy is positively correlated with net-income, see table 10 in appendix D, it is more likely for households with a high financial literacy to not be eligible for a NHG-backed mortgage. This could skew the relation in such a manner that negative coefficients are of no surprise.
Furthermore, the influence of socio-demographic factors differs per insurance type as well. No significant age effects are observed for the endowment and mortgage insurance policy, whereas annuity insurance levels appear to vary significantly per age cohort. Moreover, education does not significantly impact endowment and annuity insurance levels, but does impact mortgage insurance levels. Seemingly socio-demographics influence insurance behaviour differently when looking at the global picture of insurance coverage than when looking at a specific insurance product. At a micro-insurance level, unambiguous conclusions cannot be drawn with respect to socio-demographics.
As a robustness check, the logit estimations are replicated without trimming the dataset based on income as described in section 4.1. Table 12 in appendix D presents these results and shows the results are robust against the method of data trimming based on income. Moreover, using a probit estimation instead of a logit estimation produces trivial differences as can be derived from table 17 in appendix E. This confirms the statement by Brooks (2008) that the choice between the two is arbitrary.
20 11 in appendix D show, the reported marginal effects are only deviating slightly from the original results and not impacting the sign of significance of the included variables. Considering previous studies showed that income significantly impacts insurance choices (Truett and Truett, 1990; Burnett and Palmer, 1984), I have elected to keep income as a control variable for all logit estimations. Although the accuracy of the results might be slightly affected by the moderate multicollinearity, I do not deem this to be a problem as the intent of this study is not to accurately predict insurance levels using financial literacy scores. The intent is to provide more insights into the nature of the financial literacy-insurance relation, which are merely slightly.
5.2 Perceived financial literacy
This section discusses the results of explaining insurance choices using a combined measure of actual and perceived financial literacy. Table 4 presents the marginal effects and table 8 in appendix C presents the original coefficients. Allgood and Walstad (2012) proposed such a combined measure, stating that it deepens one’s understanding of how financial literacy affects household behaviour. As hypothesized, both actual and perceived financial literacy significantly affect insurance choices, although the significance of actual financial literacy has dropped. The probability of possessing any insurance increases by 1.1% with a one-unit increase of the actual financial literacy score, and increases by 2.7% with a one-unit increase of perceived financial literacy (at a 10% significance level). This corresponds with the results of Bannier and Neubert (2016), who found perceived financial literacy to be a stronger determinant of financial decision-making than actual financial literacy. However, as can be derived from table 10 in appendix D, actual and perceived financial literacy are moderately correlated. Such moderate multicollinearity could affect the precision of the estimates and thus its significance. However, previous household behaviour studies have safely ignored similar magnitudes of correlation (Hogeboom, 2019). Comparing the extended model with the model discussed in the previous section, no differences are observed regarding the impact of socio-demographics on total insurance coverage. The results are robust against the method of income trimming, see table 13 in appendix E.
21 between the two policy structures are minimal. The reason for this difference in interest between the endowment and the annuity insurance product, might be of an emotional nature. The endowment insurance often has the unique property that the accumulated funds will be paid out when the policy holder passes away, with the intent to provide financial stability for the family after its death. Considering confident individuals are more likely to act on their own ideas (Heath and Tversky, 1991), it makes sense that they have a higher interest in the endowment insurance due to the emotional associations with it. The annuity insurance, however, is not accompanied by certain associations that appeal confident individuals. Therefore, they do not have a heightened interest in annuity insurance compared to those with a lower perceived literacy. Moreover, although insignificant, the coefficient reported for mortgage insurance is of a positive sign. The forces at work with actual financial literacy and perceived financial literacy regarding mortgage insurance adoption appear to counteract with each other.
Again, no unambiguous conclusions can be drawn regarding the control factors on a micro-insurance level. Age effects are observed for annuity micro-insurance and endowment micro-insurance levels for age cohorts 41-50, 51-60, and >70, whereas no age effects are observed for mortgage insurance levels. Similarly, a household’s education level significantly affects mortgage insurance adoption, but not annuity and endowment insurance adoption. Net monthly income’s impact on insurance choices seems to be more robust, yet still not reported to be a significant determinant of annuity insurance.
Any insurance Endowment insurance Annuity insurance Mortgage insurance Advanced 0.011* (0.006) 0.017*** (0.005) 0.014*** (0.006) -0.017** (0.008) Perceived 0.027* (0.014) 0.028*** (0.011) 0.008 (0.013) 0.028 (0.021) Age dummies
(base group: Age ≤ 30)
31-40 0.165*** (0.062) 0.059 (0.050) 0.066 (0.060) 0.134 (0.090) 41-50 0.212*** (0.062) 0.084* (0.049) 0.109* (0.058) 0.124 (0.088) 51-60 0.252*** (0.061) 0.081* (0.049) 0.183*** (0.057) 0.088 (0.091) 61-70 0.163** 0.069 0.078 -0.047
22 (0.078) (0.061) (0.073) (0.116) >70 -0.144 (0.095) -0.188* (0.111) -0.353*** (0.129) 0.064 (0.133) Male -0.026 (0.037) 0.001 (0.026) -0.011 (0.034) -0.001 (0.056) Education dummies
(base group: low education)
Intermediate education 0.048 (0.042) 0.015 (0.029) 0.009 (0.039) 0.154** (0.062) High education 0.087** (0.041) -0.015 (0.031) 0.016 (0.037) 0.162*** (0.060) Ln(net income) 0.121*** (0.045) 0.035 (0.034) 0.086** (0.041) -0.188*** (0.066) Children 0.058* (0.035) 0.035 (0.023) 0.011 (0.030) -0.069 (0.049) Employment status
(base group: no work)
Employed 0.136*** (0.047) 0.058 (0.038) 0.066 (0.041) 0.165** (0.070) Self-employed -0.188** (0.092) -0.094 (0.081) -0.179* 0.098 -0.002 (0.135) Observations 937 937 937 488 5.3 Overconfidence
The previous sections discussed how actual financial literacy and perceived financial literacy affect insurance levels independently. This section discusses the results using an overconfidence index that internalizes the effect of both actual and perceived financial literacy. This provides more information on the joint impact of actual financial knowledge and their perception. Table 5 displays the marginal effects of the logit estimations, whereas table 9 in appendix C displays the original coefficients.
23 Although overconfidence does not appear to significantly affect decision-making in a total insurance coverage environment, overconfident households’ behaviour differs significantly on a micro-insurance level. They hold 7.6% less endowment insurance and 5.8% less annuity insurance, significant at a 5% and 10% significance level, respectively. Moreover, overconfident households are 13% more likely to possess mortgage insurance, significant at a 5% significance level. At first this seems to be contradictory with the idea of overconfidence leading to excessive risk-taking behaviour. However, when broadening the idea of overconfidence leading to excessive risk-taking to sub-optimal decision making, the results are more consistent. The mortgage insurance has the least appealing properties, as there is no certainty of wealth accumulation whilst having to pay annual fees. Yet, they have a significantly higher interest in the mortgage insurance and a significantly lower interest in the two insurance products with certain wealth accumulation, compared to non-overconfident households. This shows that the results are actually in accordance with sub-optimal behaviour as a result of overconfidence.
The contradicting significance of overconfidence between the total insurance analysis and the micro-insurance analysis could be due to the absence of information regarding quantity differences. The analysis thus far evaluates whether the household has any insurance coverage, not taking into account any differences in the total quantity of insurance coverage. Yet, it is possible that overconfident households possess significantly less insurance types than non-overconfident households, which cannot be observed due to how the dependent variable is constructed. Combined with the notion that even overconfident households probably decide its best to hold at least some, it could explain why the effect of the overconfidence trait does materialize when looking at the specific insurance types individually, but not when looking at the choice to hold any insurance. Therefore, in a total insurance context the impact of the overconfidence trait cannot be fully analysed without quantity-related information. The results of a first attempt into quantity analysis are presented in table 19 in appendix F. The logit estimation is repeated using an alternative dependent variable with value 1 if the household has two or more of the three different insurance products, and 0 otherwise. This does in fact show financial literacy overconfidence to significantly decrease the level of insurance a household chooses to adopt.
24 levels (at a 10% significance level). It appears under confidence has some influence on decision-making, making it an interesting topic for further research. As a second robustness check, the analyses are repeated without the income-related data trimming. The results reported in table 14 in appendix E show that the conclusions are robust against the choice of income-related trimming. Any insurance Endowment insurance Annuity insurance Mortgage insurance Overconfidence -0.036 (0.036) -0.076** (0.031) -0.058* (0.035) 0.130** (0.051) Age dummies
(base group: Age ≤ 30)
31-40 0.143** (0.062) 0.033 (0.050) 0.053 (0.059) 0.123 (0.091) 41-50 0.188*** (0.062) 0.056 (0.050) 0.094 (0.058) 0.113 (0.089) 51-60 0.235*** (0.061) 0.058 (0.049) 0.173*** (0.057) 0.077 (0.091) 61-70 0.136* (0.078) 0.038 (0.064) 0.061 (0.073) -0.072 (0.118) >70 -0.179* (0.095) -0.235** (0.114) -0.377*** (0.130) 0.036 (0.132) Male -0.007 (0.036) 0.024 (0.026) 0.005 (0.033) -0.003 (0.058) Education dummies
(base group: low education)
Intermediate education 0.052 (0.042) 0.022 (0.031) 0.015 (0.039) 0.167*** (0.062) High education 0.093** (0.041) -0.010 (0.033) 0.023 (0.037) 0.171*** (0.061) Ln(net income) 0.151*** (0.044) 0.080** (0.036) 0.116*** (0.040) -0.200*** (0.064) Children 0.062* (0.035) 0.038 (0.023) 0.011 (0.030) -0.065 (0.049) Employment status
(base group: no work)
Employed 0.131*** (0.048) 0.052 (0.040) 0.061 (0.042) 0.155** (0.071) Self-employed -0.198** (0.094) -0.115 (0.085) -0.191* (0.099) -0.002 (0.134) Observations 937 937 937 488
Table 5 Insurance levels and overconfidence
25 5.4 Interaction effects
The literature review uncovered that the degree of overconfidence is related to socio-demographic characteristics. Barber and Odean (2001) and Van Rooij et al. (2011) stated that men are more overconfident than women, especially in male-dominated contexts such as financial planning. Furthermore, overconfidence is reported to be increasing with age (Pirinsky, 2013; Pak and Chatterjee, 2016), education (Bhandari and Deaves, 2006; Deaves et al., 2010), and income (Pirinsky, 2013; Bhandari and Deaves, 2006). Therefore, this study examined whether interaction effects are present between overconfidence, gender, age, education, and income, which could affect insurance related decision-making. This section discusses the impact of the interaction effects on the choice to hold any insurance.
The results of the interaction effect analysis regarding holding any insurance are presented in table 6. The first logit estimation includes an interaction term for overconfidence and being male. No significant interaction effect has been revealed. Thus, the influence of financial literacy overconfidence on holding any insurance is not dependent on gender. This contrasts with the work of Hogeboom (2019), who reported a significant dependence on gender within a savings behaviour setting. The results regarding age and net income are more in line with Hogeboom’s (2019) conclusions; financial literacy overconfidence’s impact on holding any insurance is not dependent on age and net income.
However, the analysis does show that having a high education alters the manner in which financial literacy overconfidence influences insurance behaviour. The interaction term between being highly educated and financial literacy overconfidence has a significant positive effect on insurance levels (at a 10% significance level). This implies that interpreting education and overconfidence individually leads to incomplete/misleading conclusions. Therefore, they should be interpreted conjointly to draw sensible conclusions. Secondly, it shows that the manner in which overconfidence affects insurance levels is dependent on the attained level of education. In fact, being highly educated changes the relation completely. Individual inference of the influence of overconfidence on insurance levels shows that overconfident households hold less insurance. However, the total effect of financial literacy overconfidence on insurance levels is actually positive when the household has attained a high level of education. This education-dependency could explain the reported insignificance of overconfidence as a determinant of insurance levels in section 5.3. Repeating the logit estimation using this interaction term, overconfidence indeed reveals to be of significance (see table 18 in appendix F).
26 (1) (2) (3) (4) Overconfidence -0.146 (0.268) -0.725 (0.532) -0.521** (0.235) -0.822 (0.500) Male 0.062 (0.183) Overconfidence*Male -0.158 (0.353) Age -0.002 (0.008) Overconfidence*Age 0.010 (0.011) High education 0.226* (0.213) Overconfidence*High education 0.689* (0.353) Net income 0.000** (0.000) Overconfidence*Net income 0.000 (0.000)
Control factors YES YES YES YES
Observations 937 937 937 937
McFadden’s R2 0.0940 0.0936 0.0964 0.0945
Table 6 Interaction effects: Any insurance
27
VI.
Conclusion
Many studies have reported sub-optimal financial behaviour of households, such as the observation of households under-insuring against late-in-life risks. This is often attributed to a lack of financial knowledge and capabilities. Financial education has been the common approach to tackle insufficient financial literacy. However, the effectiveness of financial education programs is unsatisfactory and shows the need for improvement of such programs or the development of other tools helping households with their financial planning. To aid in this process, this study has adopted a three-part measure of financial literacy to obtain better insight in what drives household behaviour, focussing on insurance decisions. The first part examines actual financial literacy and the second part examines perceived financial literacy. The third part combines actual and perceived financial literacy to assess financial literacy overconfidence’s impact. This study finds evidence for each of the three financial literacy components to affect insurance choices.
The first hypothesis predicted actual financial literacy to influence levels of insurance positively. The logit estimation confirmed this prediction; a higher score on the advanced financial literacy test has a positive influence on households’ total insurance coverage. A similar association is observed for endowment and annuity insurance when zooming in on the individual insurance products. However, the results for mortgage insurance are in dissonance with this general observation. They showed actual financial literacy to negatively influence mortgage insurance adoption, contrasting with the findings of Cox and Zwinkels (2019). An explanation for this counterintuitive result is that more financially literacy households are more skilled in constructing their own “safety nets”. Therefore, they are less interested in insurance types that do not have a certain wealth accumulation property, such as the annuity and endowment insurance products.
28 Lastly, a negative relation between financial literacy overconfidence and the levels of insurance has been hypothesized. Although the nature of the relation between financial literacy overconfidence and the choice to adopt any insurance is indeed reported to be negative, the analysis did not render overconfidence as a significant determinant. Interestingly, overconfidence does significantly affect insurance choices at an individual insurance product level. Overconfident households hold significantly less endowment and annuity insurance, whereas mortgage insurance adoption is significantly higher among overconfident households. This appears to be contradictory to the belief of overconfidence leading to excessive risk-taking behaviour, but does fit into the greater picture of sub-optimal financial decision-making as a result of overconfidence observed throughout the literature.
Of course, this study is subject to several limitations, which are mainly related to the characteristics of the DHS dataset. The sample size was relatively small as only the data for the year 2005 could be used, due to the restriction of the financial literacy module exclusively conducted in that year. Moreover, it is possible that the data is outdated. We could only analyse behaviour in 2005, while it is reasonable to assume households behave differently nowadays, especially after the financial crisis in 2008. Therefore, replicating the analysis using a larger, more recent dataset could prove insightful.
Moreover, expanding the questions included in the survey allows for more extensive analyses. The survey questions include only 1 question regarding perceived financial literacy. Increasing the number of questions regarding perceived financial literacy could provide more reliable results and aid in getting more insight in what is driving perception. Similarly, only three different insurance products are considered in the survey. Including more different types of insurance products, such as car insurance and health insurance, would provide a more complete representation of what drives insurance choices. Moreover, the survey merely asks whether the households has purchased a particular insurance product. The information obtained from these questions does not display anything about differences in the number of insurance policies a household has chosen to adopt, nor the total size of insurance coverage of a particular insurance. Looking into insurance behaviour using data on insurance coverage quantities could be an interesting avenue of further research. A first attempt at quantifying the effect of financial literacy overconfidence executed in section 5.3 did already provide some evidence of significance.
30
References
Ackert, L., Deaves, R., 2011. Behavioral finance: Psychology, decision-making, and markets. Cengage learning Inc, Boston, United States.
Allgood, S., Walstad, 2016. The effects of perceived and actual financial literacy on financial behaviors. Economic Inquiry 54, 675-697.
Al-Tamimi, H., Bin Kalli, A., 2009. Financial literacy and investment decisions of UAE investors. Journal of Risk and Finance 10, 500-516.
Ameriks, J., Briggs, J., Caplin, A., Shapiro, M., Tonetti, C., 2016. Late-in-life risks and the under-insurance puzzle. Unpublished working paper. National Bureau of Economic Research, Cambridge. Anderson, A., Baker, F., Robinson, D., 2017. Precautionary savings, retirement planning and misperceptions of financial literacy. Journal of Financial Economics 126, 383-398.
Bannier, C., Neubert, M., 2016. Gender differences in financial risk taking: the role of financial literacy and risk tolerance. Economic Letters 145, 130-135.
Barber, B., Odean, T., 2000. Trading is hazardous to your wealth: the common stock investment performance of individual investors. Journal of Finance 2, 773-806.
Barber, B., Odean, T., 2001. Boys will be boys: gender, overconfidence, and common stock investment. Quarterly Journal of Economics 116, 261-292.
Barber, B., Odean, T., 2002. Online investors: do the slow die first? The Review of Financial Studies 15, 455-488.
Barros, L., Silveira, A., 2007. Overconfidence, managerial optimism and the determinants of capital structure. Brazilian Review of Finance 6.
Bernheim, B., Forni, L., Gokhale, J., Kotlikof, L., 2003. The mismatch between life insurance holdings and financial vulnerabilities: evidence from the health and retirement study. American Economic Review 93, 354-365.
Bhandari, G., Deaves, R., 2006. The demographics of overconfidence. Journal of Behavioral Finance 7, 5-11.
Brown, J., 2001. Private pensions, mortality risk, and the decision to annuitize. Journal of Public Economics 82, 29-62.
Brown, J., 2008. Why don´t people insure late life consumption? A framing explanation of the under-annuitization puzzle.
Brown, J., Casey, M., Mitchell, O., 2007. Who values the social security annuity? Evidence from the health and retirement study. Unpublished working paper. National Bureau of Economic Research, Cambridge.
Brown, J., Finkelstein, A., 2007. Why is the market for long-term care insurance so small? Journal of Public Economics 91, 1967-1991. American Economic Review 98, 304-309.
Burnett, J., Palmer, B., 1984. Examining life insurance ownership through demographic and psychographic characteristics. Journal of Risk and Insurance 51, 453-467.
31 Chu, Z., Wang, Z., Xioa, J., Zhang, W., 2017. Financial literacy, portfolio choice and financial well-being. Social Indicators Research 132, 799-820.
Cox, R., Zwinkels, R., 2019. Mortgage insurance adoption in the Netherlands. Real Estate Economics 47, 977-1012.
Cutler, D., Finkelstein, A., McGarry, K., 2008. Preference heterogeneity and insurance markets: explaining a puzzle of insurance. American Economic Review 98, 157-162.
Daniel, K., Hirshleifer, D., 2015. Overconfident investors, predictable returns, and excessive trading. Journal of Economic Perspectives 29, 61-88.
Deaves, R., Luders, E., Schröder, M., 2010. The dynamics of overconfidence: evidence from stock market forecasters. Journal of Economic Behavior and Organization 75, 402-412.
Duong, C., Pescetto, G., Santamaria, D., 2014. How value-glamour investors use financial information: UK evidence of investor’s confirmation bias. The European Journal of Finance 20, 524-549.
Fernandes, D., Lynch, J., Netemeyer, R., 2014. Financial literacy, financial education, and downstream financial behaviors. Management Science 60, 1861-1883.
Ferris, S., Jayaraman, N., Sabherwal., 2013. CEO overconfidence and international merger and acquisition activity. Journal of Financial and Quantitative Analysis 48, 137-164.
Fisher, S., 1973. A life cycle model of life insurance purchases. International Economic Review 14, 132-152.
Gandolfi, A., Miners, L., 1996. Gender-based differences in life insurance ownership. Journal of Risk and Insurance 63, 683-693.
Gentile, M., Linciano, N., Soccorso, P., 2016. Financial advice seeking, financial knowledge and overconfidence: evidence from the Italian market. Unpublished working paper. CONSOB Statistics and analyses, Rome, Italy.
Giné, X., Yang, D., 2008. Insurance, credit, and technology adoption: Field experimental evidence from Malawi. Journal of Development Economics 89, 1-11.
Goldsmith, A., 1983. Household life cycle protection: human capital versus life insurance. Journal of Risk and Insurance 50, 473-486.
Grinblatt, M., Keloharju, M., 2006. Sensation seeking, overconfidence, and trading activity. Journal of Finance 64, 549-578.
Guillemette, M., Hussein, M., Philips, G., Martin, T., 2015. Minority household size and the life insurance purchase decision. Financial Services Review 24, 37-50.
Guiso, L., Jappelli, T., 2009. Financial literacy and portfolio diversification. Unpublished working paper. Centre for Studies in Economics and Finance, University of Naples, Italy.
Hauff, J., Nilsson, J., 2019. Determinants of indebtedness among young adults: impacts of lender guidelines, explicit information and financial (over)confidence. International Journal of Consumer Studies 44, 89-98.
Heath, C., Tversky, A., 1991. Preference and belief: ambiguity and competence in choice under uncertainty. Journal of Risk and Uncertainty 4, 5-28.
32 Kaiser, T., Menkhoff, L., 2017. Does financial education impact financial literacy and financial behavior, and if so, when? The World Bank Economic Review 31, 611-630.
Kim, K., Lee, J., Hanna, S., 2019. The effects of financial literacy overconfidence on the mortgage delinquency of US households. Journal of Consumer Affairs 54, 517-540.
Kim, Y., 2013. Self-attribution bias of the CEO: Evidence from CEO interviews on CNBC. Journal of Banking and Finance 37, 2472-2489.
Klapper, L., Lusardi, A., Panos, G., 2013. Journal of Banking and Finance 37, 3904-3923.
Kramer, M., 2016. Financial literacy, confidence and financial advice seeking. Journal of Economic Behavior and Organization 131, 198-217.
Lin, C., Hsiao, Y., Yeh, C., 2017. Financial literacy, financial advisors, and information sources on demand for life insurance. Pacific-Basin Finance Journal 43, 218,237.
Lusardi, A., Mitchell, O., 2007a. Financial literacy and retirement preparedness: evidence and implications for financial education. Business Economics 42, 35-44.
Lusardi, A., Mitchell, O., 2007b. Financial Literacy and retirement planning: new evidence from the Rand American Life Panel. Unpublished working paper. Michigan Retirement Research Centre, Ann Arbor.
Lusardi, A., Mitchell, O., 2008. Planning and financial literacy: how do women fare? American Economic Review 98, 413-417.
Lusardi, A., Mitchell, O., 2011. Financial literacy around the world: An overview. Journal of Pension Economics and Finance 10, 497-508.
Lusardi, A., Mitchell, O., 2014. The economic importance of financial literacy: theory and evidence. Journal of Economic Literature 52, 5-44.
Lusardi, A., Mitchell, O., Curto, V., 2010. Financial literacy among the young. Journal of Consumer Affairs 44, 358-380.
Mahdzan, N., Victorian, S., 2013. The determinants of life insurance demand: a focus on savings motives and financial literacy. Asian Social Sciences 9, 274-284.
Malmendier, U., Tate, G., 2008. Who makes acquisitions? CEO overconfidence and the market’s reaction. Journal of Financial Economics 89, 20-43.
Martin, M., 2007. A literature review on the effectiveness of financial education. Unpublished working paper. Federal Reserve Bank of Richmond, Richmond.
McCannon, B., Asaad, C., Wilson, M., 2016. Financial competence, overconfidence, and trusting investments: Results from an experiment. Journal of Economics and Finance 40, 590-606.
Miles, D., 2004. The UK mortgage market: taking a longer-term view. Unpublished working paper. UK treasury.
Mitchell, O., Piggott, J., Takayama, N., 2011. Revisiting retirement payouts: market developments and policy issues. Oxford University Press, Oxford.
Moore, D., 2013. Survey of financial literacy in Washington state: knowledge, behavior, attitudes, and experiences. Washington State Department of Financial Institutions.
