Mortgage interest rate type choice and financial literacy

(1)

1

Mortgage interest rate type choice

and financial literacy

Anthony Wennemers

S2557975

MSc Finance

Thesis supervisor: Sander Sovago

University of Groningen – Faculty of Economics and Business

June 2019

Key words: Mortgage choice, fixed/adjustable interest rate mortgage, financial literacy, household finance.

Abstract

This paper examines the effect of basic and self-assessed financial literacy on the choice between a fixed interest rate and an adjustable-rate mortgage, while controlling for socio-economic factors. I use data from the DHS survey, an annual

household survey, to conduct factor analysis on various questions concerning financial concepts to create a basic financial literacy score. I find a negative and significant association between self-assessed financial literacy and the tendency

to pick an adjustable rate mortgage. Furthermore, I find no evidence for a significant effect of basic financial literacy on mortgage rate choice. Interaction variables with basic and self-assessed financial literacy are examined but mostly

return insignificant results. To conclude, instrumental variable regression is conducted to account for possible endogeneity and reverse causality issues but

(2)

2

1. Introduction

The current economic environment in which individuals must make critical financial decisions for their household and future is becoming increasingly complex. Government institutions in western countries are reducing their public services and crucial markets once regulated by governments are now at the mercy of market forces. Especially in the

Netherlands, individuals are expected to make appropriate financial decisions with diminishing governmental support. In 2006 for example, a new health insurance law was enacted in the Netherlands. Although a basic level of health care for all remained,

implementation was left to competing private health care providers and insurers. Since then, Dutch citizens must annually choose their preferred partners from these competing

institutions. Likewise, retirement planning is becoming more challenging for individuals. Stakeholders are currently negotiating a new pension system that is supposed to increase the freedom of choice provided to employees when planning for retirement. Both examples illustrate that times are changing, and financial responsibility is being shifted to the

consumer.

Specifically taking out a mortgage can be a complex decision for a household. Nowadays mortgages are offered by various financial institutions in a multitude of renditions (Amromin, Huang, Sialm and Zhong, 2011). When you take out a mortgage there are

(4)

4

illustrated by the decision households must make between either a fixed or adjustable interest rate, when taking out a mortgage. Although there are a lot of mortgages characteristics to take into consideration, many of the mortgage options provided to

consumers come with some or fully predetermined specifications. For example, the interest rate level on the mortgage loan. There is some diversity due to teaser rates, but in general financial institutions base rates on default expectations, macroeconomic tendencies and competition (Saunders and Schumacher, 2000). Thus, although the interest rate level is important to the borrower, it is not something an individual has a lot of influence over. Mortgage interest rate type differs from other mortgage features because borrowers are completely free to choose between a fixed rate and an adjustable rate mortgage.

Furthermore, a priori it is not clear whether a fixed or an adjustable rate mortgage will result in the financially most favourable outcome for a household. Thus, choosing a mortgage interest rate type requires knowledge about an individual’s own debt holding preferences and future economic expectations.

To properly choose a mortgage interest rate type that is fitting to a household’s situation, it is important that individuals are well equipped to make complex financial decisions. The commonly used measurement today to asses an individual’s financial knowledge and capability is financial literacy. Financial literacy is a not a new concept, but it has gained in popularity since the turn of the century. An important contribution to the research field of financial literacy was done by Van Rooij, Lusardi, and Alessie in 2011.

(5)

5

This paper examines the effect of financial literacy on mortgage interest rate type choice in the Netherlands, while controlling for socio-economic factors. The research question is: How do basic and self-assessed financial literacy influence the choice between a fixed interest rate mortgage and an adjustable interest rate mortgage? My sample is created by combining data from the 2005 core of the DHS with the additional single wave financial literacy module created byVan Rooij et al. (2011a, b). In line with their research, I also conduct factor analysis of survey responses to create a standardized basic financial literacy score. Besides this score, self-assed financial literacy is also gauged and used to produce regression results.Furthermore, I create multiple interaction variables to examine potential changes in the effect of (self-assessed) financial literacy on mortgage interest rate type choice if certain conditions hold. In conclusion, I briefly conduct Instrumental variable (IV) regression to account for possible endogeneity and reverse causality. This research adds to the existing body of research by focusing on mortgage interest rate type, a feature of mortgages that has received less focus. Furthermore, research focussing on explaining mortgage choices through financial literacy measurements has been limited. Previous

research has mainly focussed on explaining mortgage interest rate choice through economic, social and demographic variables (Koijen, Van Hemert, and Van Nieuwerburgh, 2009).

Previous research involving financial literacy has mainly focussed on stock market

participation or retirement planning (Lusardi, and Mitchell, 2017). By attempting to explain mortgage interest rate choices through financial literacy, I will be able to determine if financial literacy is a relevant parameter that must be considered. Ultimately the goal is to add to the comprehension of complex mortgage decisions.

My main findings are as follows. I found no significant association between basic financial literacy and mortgage interest rate type choice. However, I found a strong and significant negative association between self-assessed financial literacy and the tendency to pick an adjustable rate mortgage. This effect seems to be magnified if you are close to retirement age. My results show that not an individual’s financial literacy, but self-perception of this financial literacy is correlated with the mortgage interest rate type choice.

(6)

6

2. Literature review

2.1 Defining and measuring financial literacy

Measuring the financial capability of individuals is not a new research subject. Assessing an individual’s financial knowledge has already been attempted by researchers since the early nineties, but conceptualisation and interest were still limited. two decades ago, interest started to grow, partly due to the growth of financial opportunity and responsibility of the average consumer. This also caused a shift of focus within the field. Focus shifted from managerial financial capability and behaviour to the financial behaviour and understanding of ordinary consumers. It was around this time that the term financial literacy became popular to describe an individual’s financial knowledge and capability.

Financial literacy is a complex term that is not easily explained. It is perhaps best defined by Mason and Wilson (2000): “Financial literacy is an individual’s ability to obtain, understand and evaluate the relevant information necessary to make decisions with an awareness of the likely financial consequences”. Crucial in understanding this definition is that relevant

information is not merely limited to objective knowledge but also encompasses personal tendencies and preferences like risk tolerance. Furthermore, Lusardi and Tufano (2015) add that an individual’s financial experience is also a key aspect of financial literacy and is crucial when making financial decisions. The definition presented by Mason and Wilson (2000) covers a lot of what financial literacy connotates but fails to properly incorporate

experience. This is a testament to the intricate nature of financial literacy. This intricate nature also led to researchers using different methods of evaluating and measuring financial literacy.

This lack of consensus on proper definition and determination changed with the seminal work of Van Rooij et al. (2011a, b). The two financial (il)literacy survey modules they created are successful in accurately measuring financial literacy and score high when testing for internal consistency and reliability according to Hung et al. (2009). The first module designed to measure basic financial literacy contains five questions. The survey questions are

(7)

7

last question assesses whether the survey participant suffers from money illusion. Money illusion is the tendency to think about money in nominal terms rather than real terms. The exact wording of the financial literacy questions can be found in appendix A. Responses to these five questions are used for variable creation through factor analysis, which will be explained in detail in the data section.

The fact that financial illiteracy is widespread was already briefly touched upon in the previous section. Work by Lusardi and Mitchell (2014) compiling results from various papers shows that financial illiteracy is widespread. Dissection across socio-economic groups by Xu and Zia (2012) leads to some interesting conclusions. Xu and Zia (2012) prove that

geographical and ethnical differences in financial literacy are present. Urban populations score better on financial literacy questions than rural populations, while some ethnic groups score significantly worse than others. These researchers also argue that financial literacy follows an inverted-U shape with respect to age. This would mean that financial literacy grows as you reach adulthood, then stagnates and start to gradually lower after becoming of elderly age. Yet, it must be noted that Lusardi and Mitchell (2011) argue that for this result cohort effects cannot be ruled out. Lusardi, and Mitchell (2008) show that women in their sample are less likely to answer financial literacy questions correctly and report lower levels of self-assessed financial literacy. Lastly, Xu and Zia (2012) show that, poorer individuals and individuals with lower educational attainment, score worse on financial literacy questions than richer, better educated individuals.

(8)

8

can be increased through various channels. Financial education can help alleviate the problem of low financial literacy, according to Xiao and O’Neill (2016). However, Mandell and Klein (2009) show that following a financial management course later in life did not alter financial literacy levels. Research by Calcagno and Monticone (2015) indicates that non-independent advisors are unable to prevent low financial literate consumers of making suboptimal choices.

2.2 Mortgage interest rate type choice risk

The choice between a fixed rate mortgage and an adjustable rate mortgage is basically a trade-off between several types of risk, according to Campbell and Cocco (2003). The risk of an adjustable rate mortgage is that monthly payments can hike due to nominal interest rate and expected inflation increases. This ‘income risk’, which is evidently greater for low income households, could make it difficult for individuals to pay monthly mortgage expenses. This difficulty of payment only occurs when it is assumed that households face borrowing constraints and cannot borrow against future income perfectly, which is a

reasonable assumption. Borrowers with a fixed rate mortgage pay a fixed amount of interest monthly, and thus are less exposed to income risk due to unforeseen circumstances.

However, decreases in household income might still cause payment difficulties for

individuals. Furthermore, borrowers with a fixed rate mortgage are exposed to wealth risk. Wealth risk in this situation is the risk that the homeowner is stuck to an expensive

mortgage due to downwards movements of the nominal interest rate or expected inflation. Thus, consumers need to have a clear understanding of their own risk tolerance and of the liability their household is exposed to when facing income and wealth risk, to properly choose their mortgage interest rate type.

The choice between a fixed rate mortgage and an adjustable rate mortgage is especially relevant in the Netherlands. Van Ooijen and van Rooij (2016) explain nicely why this is the case. They state that “mortgage loans are with recourse in the Netherlands, meaning that borrowers are liable for the deficiency in case of default”. This increases the risk of a mortgage contract for the borrower. Also, the Dutch mortgage market is well-developed and complex mortgage products are abundantly available. Besides these intrinsic factors, several external factors set the Dutch mortgage market apart. Due to tax benefits on high mortgage interest payments, interest only mortgages are very popular in the Netherlands. Furthermore, borrowers in the Netherlands were allowed until recently to take out

(9)

9

savings. These conditions led to Dutch borrowers taking out large mortgage loans, which increases the effect the mortgage interest rate type choice can have on a households’ financial situation (DNB, 2000). It must be noted however that some of these beneficial conditions for big, un-amortizing mortgages have changed. Since 2013 your mortgage must amortize fully within 30 years to be eligible for tax benefits, and these tax benefits are also gradually being reduced. Moreover, since 2018 the value of the mortgage you take out cannot be greater than the market value of your home. Fortunately, this will not be an issue in this research since my dataset originates from the year 2005.

2.3 Control and interaction variables

Several borrower characteristics that are interesting have already been addressed and will be included in my regressions. These characteristics are: gender, wealth, age, risk tolerance, household income, educational achievements and time preference. Social status and

household characteristics are also included as a control variable in my regressions because peers can influence financial behaviour (Brown, Ivković, Smith and Weisbenner, 2008). Focus on economics during education is not used in initial regressions but is included in initial analysis. This is done, because according to Van Rooij, et al (2011a), focus on economics during education is a good proxy for financial literacy, and thus is useful when conducting IV regression Details concerning variable creation can be found in appendix B.

I create interaction variables between measured and self-assessed financial literacy to analyses the effect of self-confidence, like Kramer (2016). Interaction variables are also created between both financial literacy measures and the following variables of interest: age, gender, time reference, risk tolerance, wealth and income. These interaction variables provide additional insight in the effect of (self-assessed) financial literacy on mortgage interest rate type choice. They allow me to examine the change in the effect if certain

conditions hold (for example: the change in effect of financial literacy on the choice between a fixed and an adjustable mortgage when the borrower has a high risk tolerance).

(10)

10

Coulibaly, and Li, (2009) confirm that risk averse borrowers prefer a fixed rate mortgage. Also, Cocco (2013) proves that higher wealth levels and risk tolerance contribute to a preference for adjustable rate mortgage products. Finally, it is reasonable to expect that borrowers with high levels of time preference prefer an adjustable rate mortgage (Atlas, Johnson, and Payne, 2017). This is due to the fact that an adjustable rate mortgage has a lower initial interest rate ,since the fixed rate mortgage incorporates a positive term

premium (Fornero, Monticone, & Trucchi, 2011). These results indicate that adjustable rate mortgages are at least perceived as being riskier than fixed rate mortgages by

(11)

11

3. Data and descriptive statistics

3.1 De Nederlandsche Bank Household Survey

I use data from the De Nederlandsche Bank Household Survey (DHS). The DHS is an annual household survey that covers a broad range of economical and psychological topics. The surveys are conducted since 1993 by CentERdata, a research institute in the Netherlands. Recruitment of survey participants is done randomly and by phone. The distribution of the surveys is done via internet (an alternative method of answering is provided if there is no data connection present in the household) to the CentERpanel, a panel consisting of 2000 households that represent the Dutch population. This method of conducting a survey is preferred in comparison with phone surveys because internet surveys result in less reporting biases according to Chang and Krosnick (2009). I combine data from an additional single wave study from 2005 with the 2005 core of the DHS to create my sample. The additional survey is designed by Van Rooij et al. (2011a, b) and consists of the financial literacy questions discussed in the previous section.

3.2 Sample creation

Because not every individual completed both the DHS and the financial literacy survey, Combining the datasets results in a drop in sample size. The additional survey was only distributed to the person in charge of household financial matters, causing survey

respondents not in charge of financial matters to drop out. If a participant did not respond to the question “Is the interest rate of your mortgage a fixed interest rate?”, he or she is also dropped from the sample because this is the dependent variable in my regression. This causes a loss of a little over 2000 observations. Incomplete information on wealth leads to a loss of 61 observations. Furthermore, if a participant failed to provide any answer to one of the five basic financial literacy questions, so he or she simply leaves the question blank, the participant is excluded (answering “don’t know” doesn’t exclude the participant). This is done as it allows me to properly conduct factor analysis on the financial literacy questions. This restriction results in a loss of 362 observations. Finally, two individuals are dropped because they refused to answer the above-mentioned questions (refusal is a separate

answer category). The final sample has 470 observations. Summary statistics are displayed in table 1.The subjects in my sample are of old age, with the mean aging being 65.33.

(12)

12

indicate that they want to safe as much as possible). To conclude, exactly half of my sample is risk averse.

Table 1: summary statistics

mean Standard deviation Min max Continuous variables Age 65.33 13.72 40 104 Wealth 280893.90 173277.30 -9000 1419485

Net monthly household income 2587.27 1104.21 0 10000

Continuous variables on a scale

Self-assessed financial literacy 4.94 1.08 2 7

Time preference 5.07 1.16 1 7

Social status 3.73 0.98 1 5

Focus on economics during education 2.63 .98 1 4

Dummy variables

Mortgage interest rate type 0.12 0.33 0 1

Risk aversion 0.50 0.50 0 1

Gender 0.27 0.45 0 1

3.3 Mortgage interest rate type

My dependent variable is whether an individual currently has a fixed or adjustable interest rate on his mortgage. This is coded as a dummy variable that generates value 0 if someone has a fixed interest rate mortgage and generates value 1 if someone has an adjustable

interest rate mortgage. For how long the current interest rate type chosen is set is irrelevant. Of the 470 respondents in my sample, 58 (12.34%) chose an adjustable interest rate

mortgage (see table 2). This percentage is only slightly higher than the percentage of adjustable interest rate mortgages in the original full sample, which was 11.96%. So, sampling bias appears to be limited.

Table 2: mortgage choice for final sample

N=470 frequency percent

Fixed interest rate 412 87.66

Adjustable interest rate 58 12.34

total 470 100

3.4 The measurement of financial literacy

(13)

13

not use the second component because mortgages fall more into the basic financial product category. A mortgage is a complex product which requires a good understanding of one’s current and future financial situation. However, taking out a mortgage is often a necessity when buying a house. Whereas investing in stocks and bonds on the other hand is a choice and something most individuals will never do. Thus, for examine the effect of financial literacy on mortgage interest rate choice using the basic financial literacy survey is most fitting. This basic financial literacy variable will be referred to as basic financial literacy, financial literacy score or measured financial literacy.

(Weighted) responses to the basic financial literacy questions are reported in table 3 and 4. It can be observed that the questions are basic, with percentage correct answers being high. However, scoring drops as questions get more complex with only 74.89% of my sample correctly answering the money illusion question. Knowing that questions gradually become harder, it is interesting to see that participants struggle with the interest compounding question (a supposedly easier question). Exactly 14.26% gives the wrong answer to this question and 1.06% indicates that they do not know the answer to the question. In general, participants tent to answer “I do not know” more often on the latter, more complex

questions. Overall, only 52.13% (245 participants) managed to answer all five questions correctly. The average number of correct answers is 4.29, as can be seen in the last column of table 4. Furthermore, over one third of the sample makes at least one mistake. These tables show that survey respondents possess basic knowledge of a few concepts integral to basic financial literacy. However, only a little over half of participants possesses enough knowledge of all basic concepts. These results are in line with earlier work. Participants in my sample score similar results on the basic financial literacy questions as the participants of Van Rooij et al. (2011a, b) and Kramer (2016) do.

Table 3: responses basic financial literacy questions

N=470 Numeracy Interest

compounding

Inflation Time value of money Money illusion Correct 96.81 84.68 89.79 82.77 74.89 Incorrect 1.91 14.23 6.81 15.10 22.98 I do not know 1.28 1.06 3.40 2.13 2.13 Note: no refusals

Table 4: weighted responses basic financial literacy questions

N=470 0 1 2 3 4 5 mean

(14)

14 3.4.1 Factor analysis

I conduct factor analysis on the basic financial literacy questions to create a factor score that can be used in regression analysis (for the output tables of factor analysis, see appendix C). To check if the data is suited for factor analysis, Bartlett’s test of sphericity and the Kaiser-Meyer-Olkin Measure of Sampling Adequacy (KMO) are performed. The Bartlett’s null hypothesis of the correlation matrix of the five basic financial literacy questions being an identity matrix is rejected, indicating that factor analysis is helpful. This conclusion is strengthened by the KMO test which returns a value of 0.758. This indicates that a relative high proportion of variance of the 5 variables is caused by underlying factors. Thus, factor analysis is helpful. Continuing with initial factor analysis gives us only one factor with a high enough eigenvalue to be deemed relevant, as is expected.

Now that it is established that factor analysis is useful, I create dummy variables for the financial literacy questions. For each question a set of two dummies is created. The first dummy generates the value 1 if a question is answered correctly and the second dummy does the same if “I do not know” (DK) is answered. This is done because wrong and DK answers indicate different levels of financial knowledge, according to Lusardi and Mitchell (2007). Factor analysis with a minimal eigen value of one on all ten dummies and oblique rotation (since correlation between questions is probable) sets up our factor. The correct answers dummies have the highest uniqueness, indicating that they have the lowest levels of variance that is also explained by the other variables. Scoring coefficients for the ten dummies are calculated by using the Bartlett scoring method. These coefficients allow me to calculate a basic financial literacy score for every respondent. I conduct a Cronbach’s alpha test to test the reliability and internal consistency of my created basic financial literacy score. A score of 0.6680 is achieved, indicating that the internal consistency of my

constructed scale is sufficient but limited. This can be partly due to the intricate nature of the latent variable basic financial literacy. Basic financial literacy, as opposed to advanced financial literacy which involves more integrated topics, encompasses a broad range of abilities and knowledge required for day to day financial decision.

3.4.2 Self-assessed financial literacy and combination groups

(15)

15

low perceived-high actual, high perceived-low actual and high perceived-high actual. I do this so that I can check if overconfidence is a factor when choosing mortgage interest rate type.

3.5 Univariate analysis

Table 4 allows us to examine my sample more closely along socio-economic groups. Participants in my sample are well educated. Of the participants in my sample, 35.11% completed the higher vocational level and 15.11% obtained a university degree, as can be seen in the last column. The expectation that higher educational attainment participants score higher on financial literacy questions is confirmed, with participants that completed (pre) vocational level scoring lower than participants that completed (pre) university(see financial literacy column). The category consisting of people that only finished primary education is too small, it represents only 2.13% of my sample, to draw any conclusions from. It is interesting to note that the percentage of individuals that opt for an adjustable interest rate mortgage is stable. Around 90% of participants, not including those that completed pre-vocational level and university (of which 80.90% and 84.51% chose a fixed interest rate mortgage respectively), of every educational attainment group opted for a fixed interest rate mortgage.

Because the lowest and highest age groups score among the highest groups with respect to financial literacy, it can be concluded, that financial literacy does not follow the inverted-U shape with respect to age, in my sample. Assumptions about female financial literacy are confirmed, with only 36.43% correctly answering all financial literacy questions, which is less than their male counterparts. Female participants also chose less for an adjustable interest rate mortgage, only 7.75% did so. Wealthier and higher income participants have higher financial literacy scores, in line with expectations. These wealthier individuals also chose more often for an adjustable rate mortgage, as expected. Only a little over 30% of

participants in my sample considers their own financial literacy low (indicating a score of 4 or lower) and a mere 4% thinks they are extremely literate. Participants with high self-assessed financial literacy opted more often for a fixed rate mortgage than participants with low self-assessed financial literacy. Participants that have low measured financial literacy chose more often for an adjustable rate mortgage, which was not expected(although it must be noted that this is only the univariate analysis, so no concrete conclusions should be drawn). The combined financial literacy groups do not display any patterns. A slightly bigger preference for an adjustable rate mortgage is displayed by participants with high levels of time

(16)

16

Table 5: mortgage choice and financial literacy of socio-economic groups

N=470 % Fixed interest rate % high financial literacy (5 out of 5 correct) % of total sample Educational attainment Primary education 90.00 60.00 2.13 VMBO (pre-vocational) 80.90 32.58 18.94

HAVO, VWO (pre-university) 89.09 63.64 11.7

MBO (vocational) 91.25 37.50 17.02

HBO (higher vocational) 90.30 56.36 35.11

WO (university) 84.51 73.24 15.11 Age 25-34 97.22 56.94 15.32 35-44 91.95 57.47 18.51 45-54 88.37 50.39 27.45 55-64 84.27 43.82 18.94 65+ 78.49 53.76 19.79 Gender Male 85.92 58.06 72.55 Female 92.25 36.43 27.45 Household Characteristics Single living 86.96 43.48 24.47

cohabiting, without children 83.62 58.19 37.66 cohabiting, with children 92.22 52.10 35.53

Single, with children 80.00 40.00 1.06

Different, unspecified 100.00 50.00 1.28

Net monthly household income

€1150 or less 81.25 31.25 3.40

€1151 - €1800 86.52 34.83 18.94

€1801 - €2600 88.82 51.32 32.34

More than €2600 87.79 61.50 45.32

Household wealth

First quartile (poorest) 89.83 40.68 25.11

Second quartile 88.03 49.57 24.89

Third quartile 88.14 55.93 25.11

Fourth quartile (richest) 84.62 62.39 24.89

Social status 1 (low) 100.00 25.00 0.85 2 78.18 34.55 11.70 3 90.00 45.83 25.53 4 89.59 54.49 37.87 5 (high) 85.84 64.60 24.04

Self-assessed financial literacy

(17)

17 Measured financial literacy

Low (less than 5 correct) 87.11 0.00 47.87

High (all 5 correct) 88.16 100 52.13

Combined self-assessed and measured financial literacy

Low perceived-low actual 85.54 0.00 34.68 Low perceived-high actual 83.56 100 32.76 High perceived-low actual 88.03 0.00 13.19 High perceived-high actual 90.12 100 19.36

Time preference 1 (high) 100 100 0.43 2 88.89 44.44 1.91 3 89.66 48.28 6.17 4 87.67 46.58 15.53 5 89.61 58.33 33.19 6 83.90 50.85 25.11 7 (low) 92.50 40.00 8.51 I do not know 86.05 55.81 9.15 Risk aversion low 88.12 48.09 50.00 high 84.85 56.17 50.00

Focus on economics during education

1 (low) 81.08 52.70 15.74

2 86.67 46.67 25.53

3 88.51 48.28 37.02

4 (high) 91.67 66.67 20.43

I do not know 100 33.33 1.28

Note: Fixed and adjustable sum up to 100% (no refusals). Sample percentages may not sum up to 100% due to rounding. High financial literacy indicates in this table that the participants correctly answered all 5 financial literacy questions. The first two columns display percentages per category of

(18)

18

4. Methodology

To test if financial literacy has a predictive significance towards mortgage interest rate type choice, I conduct ordinary least squares (OLS) regression. Because it is possible that the error term in our estimation suffers from heteroskedasticity, I use heteroskedastic robust

standard errors. First, I conduct an OLS with solely my dependent variable mortgage interest rate choice and the independent variable basic financial literacy, to examine the sign of the independent variable. The second model includes self-assessed financial literacy. To examine the combined effect of various variables with the financial literacy scores, interactions and ultimately controls are included in the subsequent regressions. My model can be explained via the following equation:

𝑦_𝑖 = 𝐶 + 𝛽₁∗ 𝐵𝐹𝐿_𝑖+ 𝛽₂ ∗ 𝑆𝐴𝐹𝐿_𝑖+ 𝛽₃∗ 𝐶𝑉_𝑖 + 𝜀_𝑖 (1) With 𝑦𝑖 being the dependent variable mortgage interest rate choice(taking either value 0,

fixed, or value 1, adjustable)., 𝐵𝐹𝐿𝑖 being the basic financial literacy score created, 𝑆𝐴𝐹𝐿𝑖 is

the self-assessed financial literacy of a participant, CV consists of the multiple control variables in my regression, and 𝜀𝑖 is the error term. For the models with interactions

variables, the following equation holds:

𝑦𝑖 = 𝐶 + 𝛽1∗ 𝐵𝐹𝐿𝑖 + 𝛽2∗ 𝑆𝐴𝐹𝐿𝑖+ 𝛽3∗ 𝑉 ∗ 𝐹𝐿𝑖 + 𝛽4∗ 𝑉 ∗ 𝑆𝐴𝐹𝐿𝑖+ 𝛽5∗ 𝐶𝑉𝑖 + 𝜀𝑖 (2)

Where the variable interacting with self-assessed financial literacy (𝑆𝐴𝐹𝐿𝑖) and measured

financial literacy (𝐵𝐹𝐿𝑖) is a control variable of interest, for example gender (than 𝑉 would

(19)

19

5. Results

5.1 Initial results

In my OLS regression, basic financial literacy has a positive significant effect when it is included solely (OLS 1 in table 5) or together with self assed financial literacy (OLS 2 in table 5). For details on the Self-assessed financial literacy “I do not know” dummy, see appendix B. Without control variables, both measured and self-assessed financial literacy are significant, at the 5% and 1% level respectively. The coefficient of basic financial literacy is 0.0117 (0.0137 in the second regression). This coefficient indicates that participants with higher levels of basic financial literacy are more likely to opt for an adjustable rate mortgage. The coefficient of self-assessed financial literacy is -0.0350, indicating that every point increase in self-assessed financial literacy, propensity of a fixed rate mortgage increases with 3.5%.

Table 6: initial model regression results

VARIABLES OLS 1 OLS 2

financial literacy score 0.0117*** 0.0137**

(0.00387) (0.00624)

Self-assessed financial literacy -0.0350***

(0.0135) Self-assessed financial literacy “I do not know” dummy 3.313***

(1.270) Constant 0.123*** 0.296*** (0.0152) (0.0724) Observations 470 470 R-squared 0.001 0.014 Control variables NO NO

This table present the results of OLS regressions of (self- assessed) financial literacy on mortgage interest rate choice.

Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

(20)

20 5.2 Final results

The final model is displayed in table 7. To check for multicollinearity in this model, I conduct a variance Inflation factor test (see appendix E). I also perform a link test to check if there are any omitted variable or misspecification problems in this model. No causes for concern are found. Thus, it is reasonable to start analysing this model.

The significant positive correlation between financial literacy and choosing an adjustable rate mortgage does not survive when control variables are added. However, self-assessed financial literacy remains significant at the 5% level and has a coefficient of -0.0304. This means that in my model, every point increase in self-assessed financial literacy, leads to 3.04% more chance of picking a fixed interest rate mortgage. Furthermore, being between 55 and 64 years old further increases the effect self-assessed financial literacy has on

mortgage choice with -0.303%, through the implemented interaction variable (self-assessed financial literacy * age group 6 (55-64)). Participants in this age group thus have a bigger tendency to pick a fixed rate mortgage per point increase in self-assessed financial literacy. Interestingly, the two oldest age groups correlate positively and significantly with an adjustable rate mortgage. The coefficient of the age group 55-64 is 0.110, significant at the 5% level, and the coefficient of the 65+ group is 0.135, significant at the 10% level. Thus, the interaction between age and assessed financial literacy can also be interpreted as self-assessed financial literacy mitigating the effect age has on mortgage choice. Being of old age, on itself not in its interaction with other factors, thus increases the tendency of

choosing an adjustable mortgage rate. This is possibly due to the fact that older participants have accumulated more wealth during their lifetime. This wealth enables them to easier mitigate the potential adverse effects of an adjustable rate mortgages. Female participants are more biased towards a fixed rate mortgage. Being female increases the probability that you pick a fixed rate mortgage with 7.44% (significant at the 5% level). This is in line with expectations, considering that female participants scored worse than male participants on the financial literacy questions, adjustable interest rate mortgages are perceived as being riskier, and more financially sophisticated consumers opt more often for complex, riskier mortgages. Having an unspecified household composition greatly reduces your propensity of choosing an adjustable rate mortgage. Unfortunately, no extra information from which extra conclusions can be drawn is available on these participants.

(21)

21

Table 7: final model regression results

VARIABLES OLS 19

financial literacy score 0.00563

(0.00954)

self-assessed financial literacy -0.0304**

(0.0151) Self-assessed financial literacy “I do not know” dummy 2.861** (1.408) Age groups (base: 25-34)

35-44 (group 4) 0.0419 (0.0403) 45-54 (group 5) 0.0676 (0.0460) 55-64 (group 6) 0.110** (0.0527) 65+ (group 7) 0.135* (0.0701) (group 4) * financial literacy score -0.000373

(0.00134) (group 5) * financial literacy score -0.00169

(0.00283) (group 6) * financial literacy score -0.00303***

(0.00104) (group 7) * financial literacy score 0.000362 (0.00256)

Time preference -0.00184

(0.0135)

Time preference missing dummy 0.0325

(0.0934) Gender (base: male)

female -0.0744**

(0.0337)

Wealth -2.55e-09

(1.01e-07) Risk aversion (base: low)

High risk aversion 0.00210

(0.0370) Education level (base: primary)

VMBO (pre-vocational) 0.115

(0.136)

HAVO, VWO (pre-university) 0.0867

(0.136)

MBO (vocational) 0.0774

(0.137)

HBO (higher vocational) 0.0926

(0.134)

WO (university) 0.0845

(0.145) Household characteristics (base: single living)

cohabiting, without children -0.0132

(22)

22

cohabiting, with children -0.0429

(0.0488)

Single, with children 0.117

(0.240)

different, unspecified -0.187***

(0.0601) Net monthly household income (base: less than €1150)

€1151 t/m €1800 -0.0355 (0.105) €1801 t/m €2600 -0.0539 (0.101) More than €2600 -0.0382 (0.103) Social status (base: 1 (low))

2 0.159 (0.116) 3 0.0348 (0.113) 4 0.0333 (0.113) 5 (high) 0.0651 (0.125) Constant 0.149 (0.139) Observations 470 R-squared 0.074

Control variables YES

This table presents the results of OLS regressions of (self- assessed) financial literacy on mortgage interest rate choice, while controlling for socio-economic factors.

(23)

23

6. Conclusion

This paper has examined the effect of basic and self-assessed financial literacy on the choice between a fixed interest rate and an adjustable-rate mortgage, while controlling for socio-economic factors. Using data from the DHS survey, factor analysis was conducted on various questions concerning financial literacy to create a financial literacy score. Several

interactions between and with the financial literacy measures were explored but were ultimately not useful. In the final model, which includes abovementioned financial literacy measures, control variables and includes an interaction between age and self-assessed financial literacy, no significant effect of basic financial literacy on mortgage choice was found. However, self-assessed financial literacy has a significant, negative effect on the propensity of choosing an adjustable rate mortgage. This effect of self-assessed financial literacy is magnified if the participant is between 55 and 64 years old. Furthermore, in my sample, being of old age (55+) and being male increased the likeliness of an adjustable rate mortgage.

A limitation of this research is the possibility of endogeneity. The nature of the independent variable basic financial literacy makes it very likely that the variable is endogenous. As is set out by Xu and Zia (2012), many of the variables I control for can influence a participant’s financial literacy score. It should also be noted that, even though my list of control variables is extensive, it can still be the case that the coefficient is affected due to omitted variable bias. Furthermore, the possibility of reversed causality cannot be ruled out. Although it is unlikely that the actual choice of mortgage interest rate influences basic financial literacy, it is likely that the time and effort that is required to learn about the various options and implications of interest rate decisions increases basic financial literacy. To conclude, the dependent variable in my regression, mortgage interest rate type choice, can only take two values (fixed or adjustable rate mortgage). Thus, a probit model can potentially be more suited than my OLS model. To address endogeneity and other limitations of my model, I use IV regression in appendix F. Although my IV regression model is weak and results are limited and insignificant, they might induce to further research on the subject of endogeneity when examining mortgage choices and financial literacy.

In the introduction of this paper, I stated that my goal was to add to the comprehension of complex mortgage decisions. This goal has been achieved. Although results are not as clear-cut as anticipated, a little piece of the puzzle has been exposed. My results show that not an individual’s financial literacy, but his or her self-perception of this financial literacy is

(24)

24

(25)

25

References

Allgood, S., & Walstad, W. B. (2016). The effects of perceived and actual financial literacy on financial behaviors. Economic inquiry, 54(1), 675-697.

Amromin, G., Huang, J., Sialm, C., & Zhong, E. (2011). Complex mortgages. National Bureau of Economic Research no. w17315.

Atlas, S. A., Johnson, E. J., & Payne, J. W. (2017). Time preferences and mortgage choice. Journal of Marketing Research, 54(3), 415-429.

Brown, J. R., Ivković, Z., Smith, P. A., & Weisbenner, S. (2008). Neighbors matter: Causal community effects and stock market participation. The Journal of Finance, 63(3), 1509-1531. Calcagno, R., & Monticone, C. (2015). Financial literacy and the demand for financial

advice. Journal of Banking & Finance, 50, 363-380.

Caldera, A., & Johansson, Å. (2013). The price responsiveness of housing supply in OECD countries. Journal of Housing Economics, 22(3), 231-249

Campbell, J., Cocco, J., & National Bureau of Economic Research. (2003). Household risk management and optimal mortgage choice(Nber working paper series, no. 9759). Cambridge, Mass.: National Bureau of Economic Research. (2003).

Chang, L., & Krosnick, J. A. (2009). National surveys via RDD telephone interviewing versus the Internet: Comparing sample representativeness and response quality. Public Opinion Quarterly, 73(4), 641-678.

Cocco, J. F. (2013). Evidence on the benefits of alternative mortgage products. The Journal of Finance, 68(4), 1663-1690.

Coulibaly, B., & Li, G. (2009). Choice of mortgage contracts: evidence from the survey of consumer finances. Real Estate Economics, 37(4), 659-673.

Cox, R., Brounen, D., & Neuteboom, P. (2015). Financial literacy, risk aversion and choice of mortgage type by households. Journal of Real Estate Finance and Economics, 50(1), 74-112. DNB (2000). Het bancaire hypotheekbedrijf onder de loep; Rapport over de ontwikkelingen op de hypotheekmarkt in de periode 1994-1999. De Nederlansche Bank.

Fornero, E., Monticone, C., & Trucchi, S. (2011). The effect of financial literacy on mortgage choices.

(26)

26

Koijen, R., Van Hemert, O., & Van Nieuwerburgh, S. (2009). Mortgage timing. Journal of Financial Economics, 93(2), 292–324.

Kramer, M. M. (2016). Financial literacy, confidence and financial advice seeking. Journal of Economic Behavior & Organization, 131, 198-217.

Lusardi, A., & Mitchell, O. S. (2007). Financial literacy and retirement planning: New evidence from the Rand American Life Panel. Michigan Retirement Research Center Research Paper No. WP, 157.

Lusardi, A., & Mitchell, O. S. (2008). Planning and financial literacy: how do women fare?. American Economic Review, 98(2), 413-17.

Lusardi, A., & Mitchell, O. S. (2011). Financial literacy and retirement planning in the United States. Journal of Pension Economics & Finance, 10(4), 509-525.

Lusardi, A., & Mitchell, O. S. (2014). The economic importance of financial literacy: Theory and evidence. Journal of economic literature, 52(1), 5-44.

Lusardi, A., & Tufano, P. (2015). Debt literacy, financial experiences, and overindebtedness. Journal of Pension Economics & Finance, 14(4), 332-368.

Lusardi, A., & Mitchell, O. S. (2017). How ordinary consumers make complex economic decisions: Financial literacy and retirement readiness. Quarterly Journal of Finance, 7(03), 1750008.

Markow, D., & Bagnaschi, K. (2005). What American teens & adults know about economics. National Council on Economic Education.

Mandell, L., & Klein, L. (2009). The impact of financial literacy education on subsequent financial behavior. Journal of Financial Counseling and Planning, 20(1), 15–23.

Mason, C. L. J., & Wilson, R. M. (2000). Conceptualizing Financial Literacy (Business School Research Series. Paper 2000: 7). UK: Loughborough University.

Moore, D., (2003). Survey of Financial Literacy in Washington State: Knowledge, Behavior, Attitudes, and Experiences. Social and Economic Sciences Research Center, Washington State University, technical Report n. 03-39.

(27)

27

Parker, A. M., De Bruin, W. B., Yoong, J., & Willis, R. (2012). Inappropriate confidence and retirement planning: Four studies with a national sample. Journal of Behavioral Decision Making, 25(4), 382-389.

Saunders, A., & Schumacher, L. (2000). The determinants of bank interest rate margins: an international study. Journal of international Money and Finance, 19(6), 813-832.

Stafford, F. P., Hurst, E., & Chen, B. (2012). Diminishing margins: Housing market declines and family financial responses. Michigan Retirement Research Center Research Paper, (2012-276).

Van Ooijen, R., & van Rooij, M. C. (2016). Mortgage risks, debt literacy and financial advice. Journal of Banking & Finance, 72, 201-217.

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

Van Rooij, M. C., Lusardi, A., & Alessie, R. J. (2011b). Financial literacy and retirement planning in the Netherlands. Journal of Economic Psychology, 32(4), 593-608.

Xiao, J., & O'Neill, B. (2016). Consumer financial education and financial capability. International Journal of Consumer Studies, 40(6), 712-72

(28)

28

Appendix

Appendix A: financial literacy questions

Appendix table 1: basic financial literacy questions

Financial concept question Possible answers

1. Numeracy Suppose you had €100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow?

1. More than €102 2. Exactly €102 3. Less than €102 4. I refuse to answer 5. I Do not know

2. Interest compounding Suppose you had €100 in a savings account and the interest rate is 20% per year and you never withdraw money or interest payments. After 5 years, how much would you have on this account in total?

1. More than €200 2. Exactly €200 3. Less than €200 4. I refuse to answer 5. I Do not know

3. Inflation Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After 1 year, how much would you be able to buy with the money in this account?

1. More than today 2. Exactly the same 3. Less than today 4. I refuse to answer 5. I Do not know

4. Time value of money Assume a friend inherits €10,000 today and his sibling inherits €10,000 3 years from now. Who is richer because of the

inheritance?

1. My friend 2. His sibling

3. They are equally rich 4. I refuse to answer 5. I Do not know

5. Money illusion Suppose that in the year 2010, your income has doubled, and prices of all goods have doubled too. In 2010, how much will you be able to buy with your income?

1. More than today 2. Exactly the same 3. Less than today 4. I refuse to answer 5. I Do not know

(29)

29

Appendix B: (control) variable definition and determination

Likert scales of 5 or smaller are treated as categorical variables(with the exception of time devoted to economics during education, which is a continuous variable because it is used as instrumental variable). Likert scales of 7 or higher are continuous. Participants are allowed to answer “I do not know” to the survey questions. Some variables included have the following names: variablenameDK or variablenamemisssing. These dummy variables are created to mitigate misinterpretation of missing values and “I do not know” answers in continuous variables. Variables not included in this appendix are deemed self-explanatory.

Appendix table 2: (control) variable creation

variable type methodology

self-assessed financial literacy Continuous A score ranging from 1 (very bad) to 7 (very good) on the following survey question: How would you rate your own understanding of financial matters ? Variable name (for example

self-assessed financial literacy) “I do not know” dummy

Dummy The dummy takes value one if a participant answered “I do not know” to the question( in this case the self-assessed financial literacy question). This variable negates any outlier effects that can occur when “I do not know” is answered for a continuous variable.

Variable name (for example wealth) missing dummy

Dummy The dummy takes value one if there is no

information for a variable( in this case wealth). This dummy negates any outlier effects that can occur when no answer is given.

Time preference Continuous A score ranging from 1 (I like to spend all my money immediately) to 7 (I want to save as much as possible) on the following survey question: Some people spend all their income immediately. Others save some money in order to have

something to fall back on. Please indicate what you do with money that is left over after having paid for food, rent and other necessities.

Wealth Continuous Sum of amount invested in stocks, bonds and mutual funds, savings accounts and house value. Risk aversion Categorical A dummy variable that takes value 1 if someone

indicates that he or she would “rather pay more premium for a guaranteed pension” and takes value 0 otherwise.

Social status Categorical A self-assessment of social status, with a score ranging from 1 (low) to 5 (high).

Time devoted to economics during education

(30)

30 Appendix C: factor analysis

Appendix table 3: factor analysis

factor Variance Proportion

Factor 1 3.25631 0.8320

principal factor analysis of basic financial literacy questions from the DHS survey. Factors with an eigenvalue of 1 or higher are selected.

N=470

Number of retained factors=1 Oblique promax rotation

Appendix table 4: factor loadings

Question number Loading towards Factor 1 Uniqueness

Question 1 correct -0.6127 0.6246 Question 1 DK 0.8738 0.2365 Question 2 correct -0.2864 0.9180 Question 2 DK 0.8996 0.1908 Question 3 correct -0.4864 0.7634 Question 3 DK 0.6896 0.5245 Question 4 correct -0.2548 0.9351 Question 4 DK 0.6034 0.6359 Question 5 correct -0.2299 0.9471 Question 5 DK 0.6871 0.5280

Rotated factor loadings, used to create basic financial literacy score via Bartlett’s scoring method

Appendix table 5: factor scores

variable Factor score

Question 1 correct -0.08942 Question 1 DK 0.33689 Question 2 correct -0.02844 Question 2 DK 0.42982 Question 3 correct -0.05809 Question 3 DK 0.11987 Question 4 correct -0.02484 Question 4 DK 0.08651 Question 5 correct -0.02213 Question 5 DK 0.11865

(31)

31 Appendix D: regressions with interactions

I implement an interaction variable for measured and self-assessed financial literacy in OLS 3 and 4, to assess their combined effect. OLS 3 includes an interaction between the continuous (normal) variable versions of measured and self-assessed financial literacy. The coefficient of this interaction variable (self-assessed financial literacy * financial literacy score) is very small, positive, and insignificant. This means that this result is not statistically significantly different from 0, thus I cannot prove that self-assessed financial literacy influences the effect of financial literacy on mortgage interest rate type choice. Using dummy versions of

measured and self-assessed financial literacy, I create interactions variables that allow me to examine the effect of under- and overconfidence. This is in line with Kramer (2016) and shown in OLS 4, but results are insignificant. Both the interaction variable and the financial literacy dummies are insignificant in this regression. Moreover, the coefficient of the interaction variable switched signs and is negative in this regression, but this is most likely due to the information loss that is involved with dummy creation.

Appendix table 6: interactions included regression result

VARIABLES OLS 3 OLS 4

financial literacy score 0.0133*

(0.00683) self-assessed financial literacy -0.0350***

(0.0135) self-assessed financial literacy * financial literacy score 1.12e-05 (0.000158) Self-assessed financial literacy “I do not know” dummy 3.313***

(1.272) financial literacy dummy (base: low)

financial literacy high 0.0216

(0.0412) Self-assessed financial literacy dummy (base: low)

Self-assessed financial literacy high -0.0666 (0.0445) Self-assessed financial literacy high * financial literacy high -0.0693 (0.0571) Constant 0.296*** 0.147*** (0.0726) (0.0279) Observations 470 470 R-squared 0.014 0.025 Control variables NO NO

This table presents the results of OLS regressions of interactions between self-assessed and basic financial literacy on mortgage interest rate choice.

(32)

32

I conduct regressions, with variables interacting with both measured and self-assessed financial literacy, using the same approach as in OLS 3 and 4, in the following regressions. I let basic financial literacy interact with age (OLS 5 and 6), time preference (OLS 7), gender (OLS 8), risk tolerance(OLS 9), wealth (OLS 10) and income (OLS 11). Self-assessed financial literacy interacts with the same variables in OLS 12 through 18. No significant results were found for any of the interactions, except for the interaction between age groups and self-assessed financial literacy (which has a coefficient of-0.00170, significant at the 5% level). I exclude interactions that are insignificant from my final regression.

VARIABLES OLS 5 OLS 6 OLS 7 OLS 8

financial literacy score 0.00443 -0.0118 0.00662 0.0104 (0.00353) (0.0447) (0.00900) (0.00790) Age groups (base: 25-34)

35-44 (group 4) 0.0546 (0.0361) 45-54 (group 5) 0.0879** (0.0347) 55-64 (group 6) 0.142*** (0.0483) 65+ (group 7) 0.183*** (0.0467) (group 4) * financial literacy score 0.00615 (0.00536) (group 5) * financial literacy score 0.000500 (0.00951) (group 6) * financial literacy score -0.139

(0.179) (group 7) * financial literacy score 0.0468 (0.0363)

Age 0.00452***

(0.00109) Age* financial literacy score 0.000342 (0.000861)

Time preference 0.00561

(0.0128) Time preference* financial literacy

score

0.000891

(0.00132)

Time preference missing -0.00701

Female -0.0613**

(0.0305)

Female* financial literacy score -0.00197

(0.00858)

Constant 0.0281 -0.172*** 0.0931 0.140***

(33)

33

Observations 470 470 470 470

R-squared 0.038 0.037 0.002 0.008

Control variables NO NO NO NO

This table presents the results of OLS regressions of interactions between variables of interest and (self-assessed) financial literacy on mortgage interest rate choice.

Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

VARIABLES OLS 9 OLS 10 OLS 11

financial literacy score 0.0161*** 0.0154** 0.0131* (0.00423) (0.00610) (0.00772) Risk aversion (base: low risk aversion)

High risk aversion -0.0636* (0.0352) High risk aversion * financial literacy score -0.00455 (0.121)

wealth 6.24e-08

(8.66e-08) wealth* financial literacy score -2.20e-08

(2.63e-08)

Net monthly household income 6.55e-06

(1.75e-05) Net monthly household income * financial

literacy score -1.06e-06 Constant 0.156*** 0.106*** 0.107** (0.0239) (0.0277) (0.0474) Observations 470 470 470 R-squared 0.011 0.003 0.002 Control variables NO NO NO

(34)

34

VARIABLES OLS 12 OLS 13 OLS 14 OLS 15

Self-assessed financial literacy -0.000385 -0.00169 0.00785 8.62e-05 (0.000279) (0.00317) (0.00708) (0.00164) Age groups (base: 25-34)

35-44 (group 4) 0.0579 (0.0390) 45-54 (group 5) 0.0934** (0.0374) 55-64 (group 6) 0.139*** (0.0473) 65+ (group 7) 0.182*** (0.0520) (group 4) * Self-assessed financial

literacy

-0.000962 (0.000763) (group 5) * Self-assessed financial

literacy

-0.00102 (0.000619) (group 6) * Self-assessed financial

literacy

-0.00170** (0.000788) (group 7) * Self-assessed financial

literacy

0.000692 (0.00238)

Age 0.00447***

(0.00120) Age* Self-assessed financial literacy 1.51e-05

(5.77e-05)

Time preference 0.0141

(0.0147) Time preference* Self-assessed

financial literacy

-0.00151

(0.00114)

Time preference missing -0.0166

Female -0.0542

(0.0340) Female* Self-assessed financial

literacy -0.00134 (0.00170) Constant 0.0308 -0.164** 0.0504 0.140*** (0.0217) (0.0726) (0.0778) (0.0218) Observations 470 470 470 470 R-squared 0.035 0.037 0.003 0.008 Control variables NO NO NO NO

(35)

35

VARIABLES OLS 16 OLS 17 OLS 18

Self-assessed financial literacy -0.000641 -0.00255 -0.00495* (0.00115) (0.00165) (0.00293) Risk aversion (base: low risk aversion)

High risk aversion -0.0127 (0.0882) High risk aversion * self-assessed financial

literacy

-0.00966 (0.0157)

wealth 1.20e-08

(1.09e-07) wealth* self-assessed financial literacy 8.60e-09

(1.01e-08)

Net monthly household income -7.33e-06

(2.14e-05) Net monthly household income *

self-assessed financial literacy

2.36e-06 (1.90e-06) Constant 0.159*** 0.121*** 0.137** (0.0261) (0.0313) (0.0534) Observations 470 470 470 R-squared 0.009 0.003 0.006 Control variables NO NO NO

(36)

36 Appendix E: robustness checks

Multicollinearity issues are not uncommon in financial behaviour studies, since working with dummy variables for categorical responses leads to a lot of correlation between control variables. To check for multicollinearity in my final model (OLS 19), I conduct a variance Inflation factor test.

Appendix table 11: Variance inflation factor

VARIABLES VIF

financial literacy score 1.77

self-assessed financial literacy 190.25

Self-assessed financial literacy “I do not know” dummy 184.94

Age groups (base: 25-34)

35-44 2.56

45-54 3.26

55-64 3.01

65+ 3.62

Gender (base: male)

Female 1.29

Wealth 1.37

Risk aversion (base: low risk aversion)

High risk aversion 1.47

Educational attainment (base: primary education)

VMBO (pre-vocational) 10.91

HAVO, VWO (pre-university) 8.19

MBO (vocational) 10.96

HBO (higher vocational) 16.95

WO (university) 11.333

Household Characteristics (base: single living)

cohabiting, without children 2.43

cohabiting, with children 2.33

Single, with children 1.31

Different, unspecified 1.12

Net monthly household income (base: €1150 or less)

€1151 - €1800 5.85

€1801 - €2600 8.04

More than €2600 9.10

Social status (base: low)

2 16.79 3 31.65 4 39.53 5 (high) 32.58 4 (0.145) Time preference 2.96

Time preference missing 2.98

(37)

37

The social status and educational attainment dummies in appendix table 11 have high VIF values. However, these values are not problematic because both dummy groups measure one latent variable (social status or educational achievement). Self-assessed financial literacy and the dummy for a “I do not know” answer when asked to self asses financial literacy have high VIF values of 190.25 and 184.94 respectively. This might seem extreme but is due to the fact that both variables use the same underlying survey question. the self-assessed financial literacy “I do not know” dummy measure is included to solve mis estimation due to

statistical limitation.

I also perform a link test to check if there are any omitted variable or misspecification problems in this model. The results of this link test are displayed in appendix table 12.

Appendix table 12: link test

VARIABLES OLS LINKTEST

Hat 0.923** (0.405) Hat squared 0.293 (1.412) Constant 0.00269 (0.0281) Observations 470

(38)

38

Appendix F: addressing endogeneity through IV regression

The nature of the independent variable basic financial literacy makes it very likely that the variable is endogenous. To address endogeneity issues, I choose to conduct instrumental variable (IV) regression. Instrumental variable regression solves the reversed causality issue by using an instrumental variable that is associated with my independent variable basic financial literacy, but not with my dependent variable mortgage interest rate choice. IV regression also accounts for omitted variable bias. The case for IV regression is strengthened by Lusardi, Annamaria, and Olivia S. Mitchell. (2014), who show that IV regressions done previously by numerous authors lead to statistically significant results. Furthermore, the effect size of financial literacy is also larger than in normal OLS regression.

In line with Kramer (2016) and Lusardi and Mitchell (2007), respondents are asked how much time was devoted to economics during their education. I choose this as my

instrumental variable. Taking this variable as the instrumental variable has the advantage that reversed causality is ruled out, because the education was received earlier in their live, before mortgage decision would have to be made. Furthermore, two dummies are

constructed based on this variable to test for over identification. The method of IV regression is Two-Stage least squares.

It should be noted that IV regression might not be the appropriate method of combating endogeneity issues in my model. Although the assumption that a learning effect is present is reasonable enough that it merits further investigation, the direction of the effect is unclear. It is not clear if this learning effect is bigger for fixed or adjustable rate mortgages, or only present for one of the choices. Thus, choosing a mortgage interest rate type (the dependent variable in my regression) is not comparable to the choice to participate in the stock market (the dependent variable in various studies successfully attempting IV regression). Stock market participation is an opt in or out variable, you either invest or you don’t. This aspect of stock market participation, combined with the fact that financial literacy individuals participate more often in the stock market, makes the theory behind the learning effect stronger for stock market participation models than for my model.

(39)

39

the instrumental variable to test for over identification, in IV 2. IV 2 shows that the

coefficient of financial literacy becomes higher. the hypothesis that the instrument set used in IV 2 is valid, and the model is correctly specified cannot be rejected (checked with a test of overidentifying restrictions, which returned a p-value of 0.5976). The F statistic of my first stage regression is very low (1.70023), which indicates that time devoted to financial education is not a good instrumental variable. To mediate this problem, I re-regress the model with the same instrumental variable but this time I use the Limited information maximum likelihood (LIML) method. This method of IV regression is more appropriate when dealing with weak instruments. Using LIML in IV 3 alters standard errors and coefficients, and also lowers R-squared. To check if I ultimately should accept IV regression as the most appropriate model, I test for endogeneity. The null hypothesis that the variables in my model are exogenous cannot be rejected. Although the values of the Durbin and

Wu-Hausman scores are not high (0.3544 and 0.3687 respectively ), the original OLS model (OLS 19) used seems to be more appropriate. Retesting to account for the fact that I used robust standard errors didn’t alter the results from the endogeneity test much.

Table 7: final model regression results

VARIABLES IV 1 IV 2 IV 3

financial literacy score 0.0657 0.0695 0.0787 (0.0599) (0.0606) (0.0819) self-assessed financial literacy -0.0302** -0.0302** 0.000591

(0.0146) (0.0147) (0.00198) Self-assessed financial literacy “I do not

know” dummy 2.908** 2.912** 0 (1.376) (1.381) (0.00817) Age 0.00307** 0.00305** 0.00364** (0.00145) (0.00145) (0.00142) Time preference -0.000513 -0.000444 -0.00178 (0.0133) (0.0133) (0.0132) Time preference missing dummy 0.0405 0.0408 0.0566

(0.0895) (0.0897) (0.0893) Gender (base: male)

female -0.0602* -0.0596* -0.0431

(0.0340) (0.0342) (0.0363)

Wealth -9.52e-09 -9.72e-09 -3.34e-08

(9.95e-08) (9.97e-08) (1.01e-07) Risk aversion (base: low)

High risk aversion -0.0154 -0.0166 -0.0263

(0.0380) (0.0379) (0.0417) Education level (base: primary)

VMBO (pre-vocational) 0.0990 0.0987 0.0928 (0.129) (0.130) (0.125) HAVO, VWO (pre-university) 0.0830 0.0836 0.0766 (0.131) (0.131) (0.128)

MBO (vocational) 0.0813 0.0822 0.0936

(40)

40

HBO (higher vocational) 0.0833 0.0833 0.0870 (0.127) (0.127) (0.124)

WO (university) 0.0752 0.0754 0.0922

(0.138) (0.138) (0.135) Household characteristics (base: single living)

cohabiting, without children -0.0125 -0.0124 -0.0173 (0.0515) (0.0516) (0.0520) cohabiting, with children -0.0344 -0.0340 -0.0390 (0.0476) (0.0478) (0.0490)

Single, with children 0.183 0.189 0.229

(0.231) (0.232) (0.244) different, unspecified -0.173*** -0.173*** -0.176***

(0.0542) (0.0541) (0.0565) Net monthly household income (base: less

than €1150) €1151 t/m €1800 -0.0167 -0.0152 -0.0131 (0.104) (0.104) (0.108) €1801 t/m €2600 -0.0576 -0.0577 -0.0588 (0.0967) (0.0968) (0.0987) More than €2600 -0.0432 -0.0432 -0.0475 (0.0981) (0.0981) (0.0998) Social status (base: 1 (low))

2 0.165 0.165 0.152 (0.110) (0.110) (0.106) 3 0.0533 0.0539 0.0438 (0.103) (0.104) (0.101) 4 0.0481 0.0485 0.0327 (0.104) (0.104) (0.101) 5 (high) 0.0752 0.0751 0.0457 (0.114) (0.114) (0.112) Constant 0.00310 0.00345 -0.156 (0.162) (0.162) (0.139) Observations 470 470 470 R-squared 0.045 0.041 0.023

Control variables yes yes yes

This table presents the results of IV regressions of (self- assessed) financial literacy on mortgage interest rate choice, while controlling for socio-economic factors. Financial literacy is the

instrumented variable, the instrument is time devoted to economics during education. Robust standard errors in parentheses.

Mortgage interest rate type choice and financial literacy