Journal of Behavioral and Experimental Finance 29 (2021) 100450

Contents lists available at ScienceDirect

Journal of Behavioral and Experimental Finance

journal homepage: www.elsevier.com/locate/jbef

Full length article

The impact of financial literacy and financial interest on risk tolerance

∗
Cecilia Hermansson a , ,1 , Sara Jonsson b ,1
a
Division of Real Estate Economics and Finance, Department of Real Estate and Construction Management, ABE School, KTH, Royal Institute of
Technology, Box 100 44, Stockholm, Sweden
b
Department of Finance, Stockholm Business School, Stockholm University, 106 91 Stockholm, Sweden

article info a b s t r a c t

Article history: We investigate and compare the effects of financial literacy and financial interest on risk tolerance,
Received 3 June 2020 evaluating not only at the means, but also the whole distribution. We use a unique sample of 12,156
Accepted 17 December 2020 Swedish bank customers combining bank-register data with survey data. Results show that both
Available online 31 December 2020
financial literacy and financial interest are associated with higher risk tolerance. They further show that
JEL classification: the impact of financial interest is significantly higher than the impact of financial literacy. Differences
D91 are also observed across the distribution. Quantile regressions show that financial interest has its
D83 greatest association at the medium-to-high range of risk tolerance, whereas financial literacy shows
its greatest association at the lower range of risk tolerance. Findings contribute to the literature on
Keywords:
risk tolerance, specifically pointing to the relevance of the noncognitive trait; interest, to individuals’
Financial-risk tolerance
Financial literacy risk tolerance.
Financial interest © 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction 2008). Conlin et al. (2015) conclude that personality traits affect
whether individuals invest is risky assets (stocks).
Financial-risk tolerance is commonly defined as the maximum The literature also suggests cognitive explanations for risk-
amount of variability in return that someone is willing to accept tolerance heterogeneity: Differences in risk tolerance could be
when making a financial decision. Because financial risk plays a due to differences in familiarity and understanding of the very
role in almost every important economic decision, understanding nature of financial risk (Sung and Hanna, 1996). In general, things
individuals’ tolerance toward risk links to the goal of understand- with which an individual is familiar and choice alternatives that
ing economic behavior. This paper investigates and compares the fall in the individual’s domain of competence will tend to be less
effects of financial literacy and financial interest on individuals’ risky. People prefer choices about which they can feel expert
risk tolerance. and competent (Heath and Tversky, 1991). Others argue that
Previous research has found various factors to explain risk understanding the nature of the risk concept decreases negative
tolerance. Included in this body of research are studies of how so- emotional reactions toward risk (cf. Croson and Gneezy, 2009;
cioeconomic factors such as age, income, and education affect risk Loewenstein et al., 2001; Slovic, 1987). In accordance, substan-
tolerance (Bucciol et al., 2017; Dohmen et al., 2011; Grable, 2000). tial evidence avers that financially literate individuals exhibit a
Further, a large number of studies investigate the relationship more risk-tolerant behavior and invest in risky assets to a larger
between gender and risk tolerance, concluding that men are more extent than individuals with lower financial literacy (e.g., Ban-
risk tolerant than women (Charness and Gneezy, 2012; Croson nier and Neubert, 2016; Dimmock et al., 2016; Van Rooij et al.,
and Gneezy, 2009; Dreber et al., 2011; Dreber and Hoffman, 2011). Financially literate individuals also hold more financial
2007; Eckel and Grossman, 2008; Gysler et al., 2002).2 Other assets in general (Feng et al., 2019). Considering the benefits
researchers find that genetic variation is an important source of of such behaviors, much effort has been put into improving
individual heterogeneity in financial risk-taking (Cesarini et al., consumers’ financial literacy, such as through work and high
school programs. However, it is unclear whether educational
programs actually affect risk-taking and financial behavior. Hung
∗ Corresponding author.
and Yoong (2013), for example, find that unsolicited advice has no
E-mail addresses: [email protected] (C. Hermansson),
[email protected] (S. Jonsson).
effect on investment behavior, whereas individuals who actively
1 The authors appear in alphabetical order and have contributed equally to solicit advice ultimately change behavior, in spite of relatively
this paper. low financial literacy. Their findings show that building financial
2 However, evidence exists that for lower risk-tolerance levels, the gender literacy can affect outcomes, but also that other (in their study
difference is insignificant (Säve-Söderbergh, 2012). unobserved) factors, such as inherent motivation, could be highly

https://doi.org/10.1016/j.jbef.2020.100450
2214-6350/© 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

relevant. Such motivational factors ‘‘do not appear to be perfectly (cf. (Wärneryd, 1996). We analyze unique data that combine
correlated with financial literacy’’ (p. 18). survey data (collected from 12,156 bank clients in 2013) with
Deci and Ryan (1991) suggest that ‘‘intrinsically motivated bank record data from each respondent. The analysis shows that
behaviors are those the person undertakes out of interest’’ (p. although financial literacy and financial interest are associated
241 italics as in the original). From this perspective, interest with higher risk tolerance, financial interest shows a signifi-
and intrinsic motivation are synonymous. Although knowledge of cantly higher association. We also find differences across the
financial issues, that is, financial literacy, is a cognitive construct, risk-tolerance distribution. Results show that financial interest
interest is motivational. In the education literature, researchers has its greatest association at the medium to high range of risk
discuss two forms of interest: Situational interest is best under- tolerance, whereas financial literacy shows its greatest associa-
stood as temporary interest associated with increased attention in tion at the lower range of risk tolerance. We also conclude that
a specific situation; Individual interest, in contrast, reflects a long- the interaction of financial literacy and financial interest on risk
term involvement in some activity or subject (e.g., Renninger, tolerance is insignificant. Hence, being financially literate and
2000; Hidi, 1990; Schiefele, 1991). Furthermore, individual inter- having a great interest in financial matters does not lead to addi-
est, or motivation, appears to be stable and long lasting toward tional risk tolerance. We find that the control variables (including
certain domains (Von Culin et al., 2014). Individual interest means gender, demographic, and socioeconomic data) explain approx-
a person values and likes a topic and looks for more information imately 13.6% of risk-tolerance heterogeneity. The model that
about it (Schiefele, 1991). We use the term financial interest to also includes financial literacy increases the explanatory power
describe a motivational state where the individual is interested to 18.4%, whereas the model that instead also includes financial
in economic issues and financial markets. Thus, financial interest interest yields an R2 of 24.5%. Hence, financial interest appear to
is here defined as an individual rather than a situational interest. offer additional explanation to risk-tolerance heterogeneity. Thus,
The association between financial literacy on financial risk tak- the findings contribute to the literature on financial-risk toler-
ing (i.e., stock market participation) has received considerable at- ance. Although much previous research concludes that financial
tention from researchers and policymakers. Specifically in the ed- literacy, a cognitive trait, affects risk taking (i.e., stock market
ucation literature, researchers give increased attention to noncog- participation), this study finds that non-cognitive traits (interest
nitive traits. Consistency of interest combined with perseverance or motivation) may have an even larger impact. The robustness
of effort significantly accounts for variances in success outcomes test confirms our findings.
(Duckworth et al., 2007). Research on the effects of interest is still In the educational literature, researchers often debate the
very scarce in the financial-literacy and behavioral-economics relationship between interest and knowledge, that is, how a
literature. Notable exceptions include early exploratory studies by person’s interest in a topic interacts with what he or she knows
Gunnarsson and Wahlund (1997) and Wärneryd (1999), who find about that topic (e.g., Tobias, 1994; Wlodkowski and Ginsberg,
that differences in the extent to which individuals express an in- 2017). Though the iterative process of interest and knowledge
terest in financial matters, that is, to what extent households have development is relevant to investigate, the cross-sectional nature
an interest and endeavor to acquire information about market of our data do not allow further investigation into this process.
developments and new forms of saving, correlates with differ- The remainder of the paper is organized as follows. In Sec-
ences in saving behavior. These studies conclude that financial tion 2, we describe the method and the data. Section 3 provides
interest appears to be the greatest among investors who invest in the results of the analysis where data were analyzed using linear
stocks and mutual funds, and own more complex products, such (ordinary least squares [OLS]) regressions and quantile regres-
as options. sions. Section 4 includes robustness checks, estimating a model
In this paper we propose a simple scale to measure financial with an objective risk-tolerance measure (the ratio of invest-
interest and investigate and compare the effect of financial liter- ments in stock to total amount invested through the bank) using
acy and financial interest on financial-risk tolerance. We evaluate fractional outcome regressions. Section 5 has our conclusions and
the effects not only at the means but also over the whole distribu- Section 6 provides limitations and suggestions for future research.
tion of risk tolerance. We further investigate the interactive effect
of these two constructs on risk tolerance. 2. Data and method
The literature proposes various means to measure risk toler-
ance with on-going discussion on the benefits of various mea- Data accrued from one of Sweden’s largest retail banks with a
sures. The discussion on suitable measures involves whether re- market share of approximately 20% of the Swedish retail market.
searchers should use subjective survey measures or objective We collected two types of data: anonymized data from the bank’s
measures from register data (Hermansson, 2018; Schooley and register of household customers (register data) and data from
Worden, 1996). Considerations when deciding on suitable sub- a survey sent to individual customers included in the register
jective risk-tolerance measures include whether context-related sample (survey data).
questions or more general questions should be posed (Dohmen In the spring of 2013, we drew a random sample of 90,528
et al., 2011). Another consideration is whether to use one- di- customers from the bank’s 2,254,420 Swedish customers. The
mensional or multi-dimensional constructs (MacCrimmon and conditions for including a customer in the sample was that the
Wehrung, 1986; Menkhoff and Sakha, 2017). Several studies use customer had an engagement with the bank and was 18 years or
data collected through the American Survey of Consumer Fi- older. The register data include individual-level demographic and
nances (SCF; e.g. Bonsang and Dohmen, 2015; Fisher and Yao, socioeconomic data (age, gender, geographical location, income,
2017). In that survey, respondents answered one question to rate financial assets, loans and mortgages).
their willingness to take risk when saving or investing. Others We sent a questionnaire by post in the spring of 2013 to
argued that multiple questions should be used when measuring all customers in the register sample. An academic institution
risk tolerance (Loomes and Pogrebna, 2014; Menkhoff and Sakha, was the sender – and also the receiver of the responses – in
2017). order to achieve independence from the bank. No reminders were
The present study uses a subjective, multi-item measure of sent. Returning the survey were 16,062 respondents, yielding
risk tolerance collected through a survey investigation. To test the a response rate of 17.7%. The survey data provided additional
robustness of our results, we also use an objective measure oper- demographic and socioeconomic data, such as marital and family
ationalized as the percentage of stocks to total financial wealth status, education, employment, and housing status. In addition,
2
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

we received data on customers’ risk tolerance, financial literacy, apartments score is lower (20.6% compared to 30%), and the
and financial interest. Of the returned surveys, 13,525 were com- share of houses scores higher (55.0% compared to 43%). Shares
pletely answered. We excluded surveys answered by respondents of tenant-owned apartments and farmhouses are in line with the
that stated they were clients in more than one bank. The reason national average.
for this was to obtain a fair approximation of the respondent’s We initially estimate the following relationship using OLS
wealth and proportion invested in risky assets. Wealth is oper- regression. The dependent variable is risk tolerance (RT) and
ationalized as the total amount invested through the bank, that the independent variables are financial literacy (FL) and financial
is, in savings accounts, mutual funds and stocks. The final sample interest (FI). The model also includes several control variables
size amounted to 12,156. See Table 1 for a descriptive summary. (CV), such as demographic and socioeconomic variables.
We acknowledge that the response rate is relatively low and
RTj = β0j + β1 FLj + β2 FIj + βi CVi j + εj (1)
that the sample used is a convenience sample. Still, the response
rate is in line with similar studies (Kramer, 2016; Lusardi et al., We also include an interaction variable in the equation above
2011). Comparing the survey data with the register data and the to investigate whether the interaction between financial interest
average Swedish population,3 the sample represents the Swedish and financial literacy is significant. We then use quintile regres-
population, except that customers responding to the survey are sion to evaluate if the effects of the financial literacy and financial
significantly older, wealthier, and better educated. We test for se- interest differ over the distribution of risk tolerance. We use
lection bias in the robustness test. The average age in the sample the interval from the 10th to the 90th percentiles. We run the
is 55.7 years (std. dev. 1.79 years), which is higher than the bank- equations simultaneously and obtain an estimate of the entire
register data (49.7 years) and the overall average age in Sweden, variance–covariance matrix of estimators by bootstrapping.
which is 41.2 years. The average age in Sweden, however, also
includes individuals younger than 18 whereas our sample only Q 0.90 (RT) = β0 0.90 + β0.90,1 FL + β0.90,2 FI + β0.90,i CVi (2)
includes individuals who are 18 years old or older. The sample
financial wealth is, on average, 512,289 SEK with a fairly large
standard deviation of 1,028,002 SEK. The survey financial wealth Q 0.10(RT) = β0 0.10 + β0.10,1 FL + β0.10,2 FI + β0.10,i CVi (3)
is higher than the average financial wealth for the Swedish popu- We then test if the difference in the quintiles is significant, for
lation, whose average financial wealth amounts to 305,000 SEK.4 example:
It is also higher than wealth reported in the bank’s register data
where the average wealth is 317,000 SEK. Education is measured Q 0.90 (RT) − Q 0.10 (RT) = (β0 0.90 − β0 0.10 )
according to five alternatives, from no finalized education to post- + (β0.90,1 FL − β0.10,1 FL)
gymnasial education, three years or longer. The most common
educational status is gymnasial education (equivalent to upper + (β0.90,2 FI − β0.10,2 FI) + (β0.90,i CVi β0.10,i CVi ) (4)
secondary school). Compared to the national average, this share
2.1. The dependent variable
is lower (27.0% compared to 45%), and the share of postgymnasial
education is higher (52.6% compared to 34%). Thus, the sample is
The measure of risk tolerance (RT) is assessed through sur-
better educated than Swedes in general.
vey questions about the trade-off between risk and return. Risk
In addition to age, wealth, and education, the models also
tolerance preferences are often measured using survey questions
include the variables gender, large city (i.e., Stockholm, Göteborg,
(see e.g., Dohmen et al., 2011; Grable and Lytton, 1999). The RT
Malmö, and Linköping/Norrköping), income, mortgage, work sta-
variable comprises three questions raised in the survey:
tus, family status, and housing as controls. The gender variable
(men = 1, women = 0), has a mean of 0.49, thus marginally 1. I can accept losing part of my saving if the chance of getting
implicating more women than men in the sample. A majority of a good return is great
the sample – and the overall population – lives in large cities 2. I think one has to take risks to gain something
(0.83, std. dev. 0.38). The mean monthly net income after tax 3. I would like to increase risk because the return is too low
averages 17,479 SEK (std. dev. 13,020 SEK). This is in line with
the income of the Swedish population, given the age structure The answers are indicated on a Likert-type scale ranging from
of the sample. The average mortgage is 337,520 SEK with the 1 (Totally disagree) to 7 (Totally agree). The confirmatory factor
standard deviation amounting to 963,325 SEK (other forms of analysis (CFA, presented in Table 2) provides a test for unidi-
loans and consumer credits are relatively small in Sweden). The mensionality (Gerbing and Anderson, 1988). The factor loadings
variable work-status variable includes six alternatives: working are above 0.5, as suggested by Bollen (1989). Item reliability
full or part time, being retired, long-term sick leave, student, (R2 ) measures show values above the recommended 0.5 (Bollen,
or unemployed. The most common work status is working full 1989), except for the third statement, where the R2 is 0.26.
time (44.2%). The share of retired persons is higher in the sample Composite reliability for the construct is 0.99, exceeding the rec-
than in the population at large, and is as expected, given the ommended level of 0.70 (Hair et al., 1988). The average variance
age structure of the sample. Family status includes the alterna- extracted is 0.98, and thus above 0.5 (as in Fornell and Larcker,
tives being single, married, or living in a couple relationships, 1981). The high reliability is due to small indicator measurement
and having or not having children. The most common family errors.
status is being married or living in a couple relationship with The subjective risk tolerance measure has content validity,
children (43.8%). Housing includes four alternatives: rental apart- because gains and losses are both included in the question (cf.
ment, tenant-owned apartment, house, or farmhouse. The most MacCrimmon and Wehrung, 1986).
common housing status is house ownership (55.0%). Compared
to the housing situation in Sweden at large, the share of rental 2.2. The independent variables

3 Data on the Swedish population is collected from Statistics Sweden (SCB) Financial literacy in this paper is defined as knowledge of
www.scb.se. financial concepts, that is, inflation and risk diversification (e.g.,
4 This information was reported in 2007, which is the most recent available (Anderson et al., 2017; Lusardi, 2008)). Financial literacy is mea-
statistics from Statistics Sweden. sured through a quiz including six questions (see Table 3). We
3
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

Table 1
Descriptive summary.
Variable Mean St.dev. Min Max Skewness Kurtosis
Risk tolerance 0.083 1.26 −1.98 2.80 0.13 2.10
Financial literacy 2.22 1.79 0 6 0.50 2.20
Financial interest 0.006 1.57 −2.33 3.09 0.15 1.97
Age 55.7 16.7 18 100 −0.33 2.33
Gender (Male) 0.49 0.50 0 1 0.06 1.00
Large city 0.83 0.38 0 1 −1.73 4.01
Income 17,479 13,020 0 366,957 3.72 60.0
Wealth 512,289 1,028,002 1 3.6e+07 10.5 225
Mortgage 337,520 963,323 0 5.4e+07 27.0 1389
Work status 2.27 1.39 1 6 0.94 3.33
Full time 0.442
Part time 0.093
Retired 0.343
Long-term sick leave 0.036
Student 0.040
Unemployed 0.048
Education 3.56 1.29 1 5 −0.46 2.16
No finalized education 0.089
Pregymnasial education 0.114
Gymnasial education 0.270
Postgymnasial ed. <3 yrs 0.204
Postgymnasial ed. ≥3 yrs 0.322
Family status 2.99 1.20 1 5 −0.69 2.07
Single w/o children 0.223
Single w children 0.037
Couple w/o children 0.284
Couple w children 0.438
Other 0.018
Housing 2.44 0.86 1 4 −0.54 2.16
Rental apartment 0.206
Tenant-owned apartment 0.199
House 0.550
Farmhouse 0.045

This table presents summary statistics on the sample of 12,156 survey respondents. The dependent variable, risk tolerance (RT), is a
construct captured through three survey questions. Financial literacy and financial interest are independent variables; the remaining
variables are controls. The variables’ means, standard deviation, min and max, skewness and kurtosis are reported. For the control
variables, we report work status, education, family status, and housing means and the shares that have stated a certain alternative.
Work status ranges from working full time to being far from employment (unemployed). Housing ranges from a low degree of
ownership (rental apartment) to a high degree of ownership (owning a farm house). Education ranges from a low level (no finalized
education) to a high level (postgymnasial ≥ 3 years) of education. Family status ranges from a low degree of relationship involvement
(single without children) to a high degree of involvement (couple with children). Considering the shares, in our sample 44.2% work
full time. Wealth is the respondent’s total amount invested through the bank in savings accounts, mutual funds, and stocks. Mortgage
is the respondent’s total amount of mortgage at the bank. Income, wealth, and mortgage are stated in Swedish crowns (SEK). On
June 24 2019, 1 SEK was equivalent to 0.11 USD. Income is the respondent’s monthly income.

Table 2
Measurement model for risk tolerance (RT).
Risk tolerance (RT) Mean Factor Measurement z-value Item Composite AVE
loading error Reliability Reliability
I can accept to lose part of my 2.99 0.83 0.007 116.27 0.69 0.99 0.98
saving capital if the chance of getting
a good return is great
I think one has to take risk in order 3.39 0.82 0.007 114.38 0.67
to gain something
I would like to increase risk since 4.98 0.51 0.008 63.83 0.26
return is too low

This table presents the confirmatory factor analysis conducted to assess the measurement model for the construct of RT, that is, the means, factor loadings,
measurement errors, z-values and item reliability for each survey question or statement included in the construct. The answers to the questions range from 1
to 7 on a Likert-type scale, where 1 corresponds to totally disagree and 7 corresponds to totally agree. The composite reliability for the construct exceeds the
recommended level of 0.70 (aligned with Hair et al., 1988). The high values of composite reliability and average variance extracted (AVE) are due to the small
indicator measurement errors.

developed the questions in accordance with the Swedish context it is common to save directly in stocks (about 32% of the pop-
and therefore they differ to some extent from questions used ulation) and in mutual funds (62% of the population), whereas
by, for example Anderson et al. (2017) and Lusardi (2008, 2012). direct bond investments are less common (8% of the popula-
In Sweden, a relatively large part of the population owns their tion; Swedish investment fund association, 2016).5 To know that
home or apartment (2/3 of the population according to Statistics mutual funds have different risk levels, and that saving in eq-
Sweden). Half of Swedish households have a mortgage, and a uity funds is riskier than saving in balanced or fixed-income
majority of these mortgages have variable interest rates. For funds, is therefore highly relevant. Other important concepts
the individual, the central bank (the Riksbank) has an inflation
target and to understand the relationship between inflation and 5 The shares above only include the direct private saving and not pension
nominal and real interest rates is relevant knowledge. In Sweden, savings managed by the pension system.

4
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

Table 3
Financial literacy (FL), questions and answers.
Question Answer Correct answer (%)
1. How high is the Riksbank’s inflation target? 2% 34.44
2. If there is a risk that the inflation will Raise the repo rate 37.05
exceed the inflation target, what should the
Riksbank do?
3. If the nominal interest rate is 5%, and the 3% 31.74
expected inflation is 2%, how high will the real
interest rate be (approx.)?
4. A savings product where you will receive a Equity-linked security 32.14
guaranteed amount at maturity, and the return
follows the equity market, is called:
5. Mutual funds have different risk levels; Equity fund 72.84
which of these mutual fund types is generally
viewed as having the highest risk?
6. The definition of the P/E-ratio is: Price per share divided 15.33
by earnings per share

This table presents the six financial literacy questions used to assess the variable Financial Literacy. We developed
the questions to fit the Swedish context (cf. Gunnarsson and Wahlund, 1997). The Table shows the question and
the correct answer. Respondents were instructed to select one of four alternatives, including the alternative ‘‘I don’t
know.’’ The table also shows the share of the respondents who selected the correct answer on each question.

Table 4 We conducted a confirmatory factor analysis (CFA, presented in

Financial literacy (FL) distribution of correct answers. Table 5. The factor loadings, the item and composite reliability, as
Number of correct answers Percentage Cumulative percentage well as AVE, are in line with recommended levels.
0 19.13 19.13
1 23.54 42.68
3. Results
2 17.18 59.86
3 14.34 74.20
4 11.87 86.07 3.1. Descriptive results
5 8.46 94.52
6 5.48 100.00
Table 6 shows the correlation matrix. The correlation between
This table presents the share of the sample with a certain number of correct the independent variables financial literacy and the financial-
answers. For example, the share of the sample with three correct answers is interest construct shows the highest correlation in the table
14.34%. The table also accumulates the correct answers. For example, close to
60% have less than 3 correct answers.
(0.5006 with a significance at the 1% level).

3.2. Empirical results

include price/earnings(P/E)-ratio and instruments such as equity-
linked securities. Gunnarsson and Wahlund (1997) use a similar The results from the OLS regression appear in Table 7. The
context-driven approach when measuring financial literacy. table reports standardized beta coefficients. In Column 1, we
Approximately 43% of respondents answered 0 or 1 question estimate the association of financial literacy and financial interest
correctly, and only 26% had more than 3 correct answers. The on risk tolerance. Results show that, on average, financial inter-
question about different risk levels of mutual funds (Q5) received est exhibits a higher association (0.309) than financial literacy
the highest percentage of correct answers (72.84%) and the ques- (0.179). The two variables have an explanatory power of 18.3%.
tion about the P/E-ratio (Q6) generated the lowest percentage Column 2 shows the results from estimating the association of
of correct answer (15.33%). The number of correct answers var- the control variables. Most control variables confirm previous
ied between 0 and 6 and the mean number of correct answers findings on explanations for risk tolerance. Specifically, men and
amounted to 2.24 (Std. dev. 1.79). Table 4 shows the distribution wealthy individuals are more risk tolerant. Risk tolerance is also
of the number of correct answers. higher among those who are well educated and those who work
Financial interest is measured by three survey questions: full time. The control variables yield a lower explanatory power
(13.6%) than the effects of financial literacy and financial interest
1. I am interested in economic matters and financial markets reported in column 1. In Column 3, we estimate the effect of
2. I follow the media about developments on the financial financial literacy together with controls, and in Column 4, we
markets estimate the effect of financial interest together with controls.
3. I follow the media about the developments of new saving The results illustrate that financial interest together with controls
products yields a higher explanatory power than financial literacy and
controls (24.5% vs. 18.4%).6 In Column 5, we estimate the full
Answers are indicated on a Likert-type scale ranging from 1
model. The results confirm the findings reported in Column 1.
(Totally disagree) to 7 (Totally agree). We intended the first
Financial interest shows a stronger association with risk tolerance
question, ‘‘I am interested in economic matters and financial
markets,’’ to capture the individual interest (rather than situa- (0.318) than financial literacy (0.119).
tional interest) and stable involvement in the subject (cf. Hidi, We further test these results by running the same regres-
1990; Schiefele, 1991). The other two measures intend to capture sion, this time operationalizing financial interest as a single-item
manifestations of claimed interest: The extent to which the re- construct. The question ‘‘I am interested in economic issues and
spondent follows media about developments of financial markets financial markets’’ has a strong congruence with the variable
and savings products. Interest means that the individual looks financial interest (cf. Bergkvist and Rossiter, 2007). We therefore
for more information about a subject. Thus, we intend the two
latter measures to capture the information-acquisition aspect of 6 Empirical models of financial risk taking report R2 –values at around 25%
interest. Together the three measures enhance content validity. (Cesarini et al., 2008).

5
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

Table 5
Measurement model for financial interest (FI).
Financial interest Mean Factor Measurement z-value Item Composite AVE
loading error reliability reliability
I am interested in economic issues 3.80 0.86 0.0035 246.65 0.74 1.00 0.99
and financial markets
I follow the developments in media 3.62 0.93 0.0031 302.96 0.86
about financial markets
I follow the development in media 3.08 0.76 0.0045 164.36 0.58
about saving products

This table presents the confirmatory factor analysis of the construct financial interest: the means, factor loadings, measurement errors, z-values, and item reliability
for each survey question or statement. The answers to the questions range from 1 to 7 on a Likert-type scale, where 1 corresponds to totally disagree and 7
corresponds to totally agree. The composite reliability for the construct exceeds the recommended level of 0.70 (Hair et al., 1988). The high values of composite
reliability and AVE are due to the small indicator measurement errors.

Table 6
Correlation matrix.
RT FL FI Age Gender Urban Income Wealth Mortgage Work status Housing Education Family status
RT 1.000
FL .3334* 1.000
FI .3987* .5006* 1.000
Age −.1392* .0780* .1800* 1.000
Gender .2256* .2939* .2514* .0611* 1.000
LargeCity −.0717* −.0861* −.0624* .1048* −.0193 1.000
Income .1667* .2151* .1286* −.0605* .1670* −.1140* 1.000
Wealth .1200* .2038* .2211* .2259* .0751* −.0262* .0756* 1.000
Mortgage .1483* .1649* .1331* −.0785* .1542* −.1574* .1810* .0007 1.000
Work status −.1392* −.1017* −.0429* .1722* −.1015* .0646* −.3470* .0355* −.1630* 1.000
Housing .0813* .1510* .1353* .1677* .0609* .1130* .0885* .0863* .1420* −.0907* 1.000
Education .1420* .2268* .0862* −.2545* −.0879* −.1487* .2243* .0263* .1267* −.1877* −.0078 1.000
Family status .0954* −.1079* .0463* −.1257* .0941* .0143 .1061* −.0377* .0993* −.1053* .3516* .0797* 1.000

This table presents the correlation between all the variables used in the regressions. Most correlations are significant at the 1% level (denoted * in this table). Work
status, housing, education, and family status are considered ordinal. Work status ranges from being far from employment (unemployed) to working full time. Housing
ranges from a low degree of ownership (rental apartment) to a high degree of ownership (owning a farmhouse). Education ranges from a low level (no finalized
education) to a high level (postgymnasial ≥ 3 years) of education. Family status ranges from a low degree of relationship involvement (single without children) to
a high degree of involvement (couple with children).

run the regression with solely this item as the operationalization For illustrative purposes, we also plotted coefficients for each
of financial interest. The results are similar as for the three item percentile in Figs. 1 and 2, with a 95% confidence interval, shown
construct, yielding a standardized beta value for financial literacy in the grey area. Quantile regressions are used to test if the effects
of 0.140 (t = 14.07) and 0.287 (t = 28.55) for financial interest. of the independent variables differ, depending on the distribution
Thus, the association between financial interest and risk tolerance of the dependent variable. Hence, the assumption is that the
remains higher than the association between financial literacy effects of financial interest may be greater on the xth percentile
and risk tolerance. The control variables in Column 5 show that than the yth percentile of the risk-tolerance distribution.
higher risk tolerance aligns with youth, male, wealth, and full- The results of the quantile regressions show that the associa-
time employment. Further, individuals who live in a house and tion between financial interest and risk tolerance is higher than
have finalized higher education are more risk tolerant. In Col- the association between financial literacy and risk tolerance over
umn 6, we estimate the interactive effect of financial literacy the whole risk-tolerance distribution. The results further show
and financial interest on risk tolerance. The results show that that the effects of financial literacy are greatest at the lower end
the interaction variable financial literacy × financial interest is of the distribution, at the 30th percentile (0.111), whereas the
insignificant; 0.005 (t = 0.36). Hence, being both financially lowest association is at the highest end of the distribution of risk
literate and having a strong financial interest is not associated
tolerance, that is, at the 90th percentile (0.035). In contrast, the
with additional risk tolerance.
effects of financial interest are greatest at the 60th percentile
To test whether standardized beta values of financial interest
of the RT distribution (0.327), and, contrary to financial liter-
and financial literacy are statistically significantly different from
acy, the smallest effect of financial interest on risk tolerance is
each other, we estimated their corresponding 95% confidence
noted at the 10th percentile (0.131). Hence, financial literacy has
intervals with bias corrected bootstrap (1000 replications). In
the event the confidence intervals overlapped by less than 50%, the greatest effect on risk tolerance for individuals with low-
the beta values would be considered statistically significantly range risk tolerance, whereas financial interest has the greatest
different from each other (Cumming, 2009). We standardize the association with risk tolerance for individuals with medium- to
variables in Model 1 and run the OLS regression. The result shows high-range financial interest. Further, financial literacy has the
that the confidence intervals of financial interest and financial lit- lowest association on risk tolerance for individuals with high risk
eracy do not overlap. The confidence interval of financial interest tolerance, whereas financial interest has the lowest association
ranges from 0.298 to 0.341, whereas the confidence interval for on risk tolerance for individuals with low risk tolerance.
financial literacy ranges from 0.099 to 0.1395. Hence, the associ- We further test if the differences between the percentiles are
ation between financial interest and risk tolerance is significantly statistically significant by using an interquantile range regression.
higher than the association between financial literacy and risk Table 9 refers to risk tolerance as the dependent variable and
tolerance. shows that the differences between the 90th, 50th, and 10th
To examine and compare financial-literacy and financial- percentiles are statistically significant for financial interest. Re-
interest differences in risk tolerance over the risk-tolerance dis- garding financial literacy, the differences between the 90th and
tribution, we present a series of quintile regressions in Table 8. 10th percentiles and between the 90th and 50th percentiles are
6
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

Table 7
Results from estimating the OLS regression model.
(1) (2) (3) (4) (5) (6)
Variable

Financial literacy 0.179*** 0.249*** 0.119*** 0.118***

(0.007) (0.006) (0.007) (0.007)
Financial interest 0.309*** 0.363*** 0.318*** 0.315***
(0.008) (0.008) (0.008) (0.012)
Literacy × Interest 0.005
(0.004)
Age −0.093 −0.176*** −0.204*** −0.230*** −0.229***
(0.005) (0.004) (0.004) (0.004) (0.004)
2
Age −0.090 −0.008 −0.007 0.021 0.021
(0.000) (0.000) (0.000) (0.000) (0.000)
Gender 0.212*** 0.143*** 0.131*** 0.108*** 0.108***
(0.023) (0.024) (0.023) (0.023) (0.023)
Large city −0.033*** −0.020** −0.014 −0.011 −0.011
(0.029) (0.028) (0.027) (0.027) (0.027)
Income −0.024*** −0.020** −0.010** −0.009 −0.009
(0.004) (0.004) (0.003) (0.003) (0.003)
Wealth 0.180*** 0.131*** 0.104*** 0.091*** 0.091***
(0.007) (0.007) (0.007) (0.007) (0.007)
Mortgage 0.038*** 0.027*** 0.014 0.013 0.013
(0.002) (0.002) (0.002) (0.002) (0.002)
Work status (Full time)
Part time −0.026*** −0.023*** −0.026*** −0.025*** −0.024***
(0.039) (0.038) (0.036) (0.035) (0.036)
Retired −0.030* −0.046*** −0.057*** −0.062*** −0.062***
(0.041) (0.039) (0.038) (0.038) (0.038)
Long-term sick leave −0.031*** −0.031*** −0.031*** −0.032*** −0.032
(0.060) (0.058) (0.056) (0.056) (0.056)
Student −0.011 −0.015* −0.021** −0.022** −0.022**
(0.062) (0.054) (0.059) (0.058) (0.058)
Unemployed −0.014 −0.013 −0.013 −0.012 −0.012
(0.056) (0.060) (0.051) (0.051) (0.051)
Education (No finalized education)
Pregymnasial education 0.021 −0.004 −0.002 −0.011 −0.011
(0.051) (0.050) (0.048) (0.048) (0.048)
Gymnasium 0.063*** 0.014 0.014 −0.004 −0.003
(0.046) (0.045) (0.043) (0.043) (0.042)
Postgymnasial education, <3 yrs 0.087*** 0.031** 0.030** 0.010 0.010
(0.046) (0.046) (0.044) (0.044) (0.044)
Postgymnasial education, ≥3 yrs 0.122*** 0.027 0.049*** 0.013 0.014
(0.045) (0.045) (0.042) (0.043) (0.043)
Family status (single w/o children)
Single w. children 0.001 0.004 0.002 0.004 0.004
(0.060) (0.059) (0.057) (0.057) (0.057)
Couple w/o children 0.014 0.002 0.001 −0.003 −0.003
(0.032) (0.031) (0.029) (0.030) (0.030)
Couple w. children 0.029** 0.017 0.025** 0.019* 0.019*
(0.032) (0.031) (0.030) (0.030) (0.030)
Other −0.009 −0.010 −0.010 −0.011 −0.011
(0.088) (0.085) (0.081) (0.081) (0.081)
Housing (Rental apartment)
Tenant-owned apartment 0.040*** 0.025* 0.017 0.013 0.013
(0.036) (0.036) (0.034) (0.034) (0.034)
House 0.059*** 0.040*** 0.037*** 0.030** 0.030**
(0.034) (0.033) (0.031) (0.031) (0.031)
Farmhouse 0.018* 0.008 0.009 0.005 0.005
(0.058) (0.056) (0.055) (0.054) (0.054)
Constant –*** –*** –*** – – –
(0.019) (0.140) (0.138) (0.134) (0.133) (0.135)

Adj. R2 0.183 0.136 0.184 0.245 0.252 0.254

Observations 12,156 12,156 12,156 12,156 12,156 12,156

This table reports the OLS regression in which risk tolerance (RT) is the dependent variable. Sample size is N = 12,156. The table reports
standardized beta coefficients. Column 1 shows OLS regression results for financial literacy and financial interest. Column 2 shows OLS
regression results for the control variables. Column 3 shows OLS regression results for financial literacy together with controls and Column
4 shows regression results for financial interest and controls. Column 5 shows the OLS regression results for the full model, as expressed
in Eq. (1). Column 6 shows the results when including the interaction variable financial literacy × financial interest. Income, wealth, and
mortgage represents ln(income), ln(wealth) and ln(mortgage). Work status, education, and family status are treated at categorical variables.
The base levels are indicated in the parentheses. Robust standard errors appear in parenthesis, ***, **, and * denote significance at the 1%,
5%, and the 10% level respectively.

7
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

Table 8
Results from estimating the quantile regression model.
Quantile Financial literacy Financial interest
Beta Beta
Q (0.10) 0.093*** 0.131***
(0.013) (0.013)
Q (0.20) 0.102*** 0.211***
(0.012) (0.013)
Q (0.30) 0.111*** 0.266***
(0.011) (0.012)
Q (0.40) 0.106*** 0.304***
(0.010) (0.011)
Q (0.50) 0.093*** 0.311***
(0.010) (0.011)
Q (0.60) 0.079*** 0.327***
(0.010) (0.012)
Q (0.70) 0.072*** 0.306***
(0.010) (0.012)
Q (0.80) 0.062*** 0.292***
(0.009) (0.011)
Q (0.90) 0.035*** 0.280***
(0.009) (0.013)
Controls Yes Yes
Observations 12,156 12,156

This table presents simultaneous quintile regressions with percentiles ranging from 0.10 to 0.90.
Sample size is N = 12,156. The coefficients are standardized beta coefficients. The regressions
follow the model expressed by Eqs. (2) and (3) above. Control variables are included. Bootstrapped
standard errors are given in parentheses. ***, **, and * denote significance at the 1%, 5%, and the
10% level respectively.

Fig. 1. The financial literacy coefficient estimates on risk tolerance. Note: The Model uses quintile regression with control variables and bootstrap method, plotted
with 95 percent confidence interval (grey area).

significant, whereas the difference between the 50th and 10th causes a bias in the OLS regression estimates because the regres-
percentiles is not significant. sor correlates with the error term. We correlate the independent
variables (financial literacy and financial interest) one at a time
4. Robustness tests with the error term of the OLS regression (cf. Wooldridge, 2015),
and found we cannot reject the hypothesis of reverse causality.
Our main results imply that financial literacy and financial in- This is as what could be expected because the process of de-
terest positively affect risk tolerance. We then conduct a number veloping financial interest, financial literacy, and risk tolerance
of robustness tests. Checking for multicollinearity, the VIF test over time is the most likely iterative. A person’s interest in a
shows values less than 1.55. subject interacts with what the person knows about the subject
The measures could potentially be endogenous, pointing to (Wlodkowski and Ginsberg, 2017).
reverse causality. For example, a person with higher risk tolerance We check for selection bias by conducting a number of group
might be more motivated to acquire knowledge about financial analyses including interaction variables to the regressions. The
concepts as well as follow financial-market development, that is, results appear in Table 10. We check whether the effects of
exhibit greater financial interest. Endogeneity of the regressors financial literacy and financial interest on risk tolerance differ
8
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

Fig. 2. The financial interest coefficient estimates on risk tolerance. Note. The Model uses quintile regression with control variables and the bootstrap method, plotted
with 95% confidence interval (grey area).

Table 9 Table 10
Results from estimating the inter-quintile regression model with RT as Interaction variables.
dependent variable. (1) (2) (3) (4)
Inter-quantile comparison Financial literacy Financial interest Variable
Difference in beta Difference in beta
(0.90–0.10) −0.059*** 0.144*** Financial literacy (FL) .155*** .068** .153** .127***
(0.014) (0.016) (.024) (.021) (.051) (.010)
(0.50–0.10) 0.002 0.175*** Financial interest (FI) .314*** .365*** .233*** .290***
(0.011) (0.012) (.028) (.024) (.058) (.011)
(0.90–0.50) −0.061*** −0.031** FL * Age −0.039
(0.012) (0.014) (.000)
Controls Yes Yes FI * Age .004
Observations 12,156 12,156 (.000)
FL * Education .062*
This table presents the interquintile range regression with RT as the dependent
(.005)
variable and financial literacy and financial interest as independent variables.
FI * Education −.050*
The sample size is N = 12,156. Regressions follow the model, expressed by
(.006)
Eq. (4), focusing on the interquintile comparison. Control variables are included.
FL * Wealth −0.037
Bootstrapped standard errors are given in parentheses. ***, **, and * denote
(0.004)
significance at the 1%, 5%, and the 10% level, respectively.
FI * Wealth 0.086
(0.005)
FL * Gender −0.020
(.014)
with regards to age, education, and wealth, because our sample
FI * Gender .041***
consists of older, more educated, and wealthier individuals com- (.016)
pared to the Swedish population at large. Prior researchers found Age −0.219***
that gender affects the effect of financial literacy on financial risk (0.004)
taking (e.g., Almenberg and Dreber, 2015; Bannier and Neubert, Education −.003
(.015)
2016). We therefore also interact gender with financial literacy
Wealth .097***
and with financial interest. (0.012)
We can conclude that age does not affect the association Gender 0.120***
between financial literacy and risk tolerance. The results further (0.039)
show that education has a significant positive effect (0.062) on
Controls Yes Yes Yes Yes
the association of financial literacy on risk tolerance and a signif-
Observations 12,156 12,156 12,156 12,156
icant negative effect on the association of financial interest with Adj. R2 .2524 .2526 .2524 .2529
risk tolerance (−0.050). However, the effects are small, only sig-
This table presents the OLS regression with interaction variables. The sample
nificant at the 10% level. We conclude that the interactive effect
size is N = 12,156. We interact financial literacy and financial interest with
of wealth is insignificant for financial literacy and for financial age (Column 1), education (Column 2), wealth (Column 3), and gender (Column
interest. The findings endure also when testing an interaction 4). Age, education, and wealth are continuous variables. Standardized beta
with the categorical variables of age and financial wealth. We coefficients are reported. Robust standard errors appear in parenthesis, ***, **,
further conclude that for men, the association between financial and * denote significance at the 1%, 5%, and the 10% level respectively.
interest and risk tolerance is significantly larger than for women
(0.041). However, the difference on the effect of financial literacy
is not significantly different between genders (cf. Almenberg and
Dreber, 2015).
9
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

Fig. 3. Quintile graph of Objective Risk Tolerance (ORT).

4.1. An alternative risk-tolerance measure 0.8 percentage points, on average, over the whole distribution of
financial interest. Thus, the marginal effect is higher for financial
As an extended robustness test, we investigate the effects of interest than for financial literacy.
financial literacy and financial interest on objective risk toler- To conclude, the overall findings endure also when using an
ance. The objective risk tolerance (ORT), is operationalized as the objective measure of risk tolerance. Financial literacy and finan-
ratio of investments in stock (equity) to total financial wealth cial interest significantly align with risk tolerance, and financial
(cf. Wärneryd, 1996, 1999). Information on the amount invested interest shows a stronger association.
in stocks as well as total amount invested through the bank
(i.e., wealth) was obtained from the bank’s register data.
Because respondents are clients in this bank – and no other 5. Conclusion
banks – the total amount invested through the bank is defined as
the client’s financial wealth. Because mutual funds also include This study investigates and compares the effects of financial
less risky fixed-income funds and balanced funds, we chose to literacy and financial interest on individuals’ financial risk tol-
include only stock investments to secure the content validity of erance. We use a unique data set that allows us to use both
the measure. ORT thus ranges from 0 to 1. The mean is 0.067 subjective and objective measures of financial risk tolerance. The
and the standard deviation in 0.16. The correlation between ORT analyses show that both financial literacy and financial interest
and RT is 0.1361 (significant at the 1% level). VIF test values are are associated with higher risk tolerance and we conclude that
less than 1.55. ORT exhibits a high degree of skewness (3.10) financial interest has a significantly higher impact on risk tol-
because a majority of individuals in the sample do not own erance than financial literacy. We also conclude that specifically
equity, demonstrated in Fig. 3. financial interest offers an additional explanation of financial-risk
We first estimate the following relationship. Objective risk tolerance. We also find differences in the association between the
tolerance (ORT) is the dependent variable. Similar to Model 1, variables and risk tolerance across the risk-tolerance distribution.
financial literacy (FL) and financial interest (FI) are independent The effects of financial literacy are highest at the lower end of
variables. The model also includes several control variables (CV), the risk-tolerance distribution and smallest at the higher end.
such as demographic and socioeconomic variables. Financial interest shows a different pattern with the highest
ORTj = β0j + β1 FLj + β2 FIj + βi CVi j + εj (5) effect at the medium-to-high range, whereas the lowest effect
of financial interest on risk tolerance is at the lowest end of the
Because the distribution of ORT lies in the interval [0,1], we distribution. Thus, individuals’ financial literacy has the greatest
also use a fractional outcome regression (Papke and Wooldridge,
association with risk tolerance for individuals with lower risk tol-
1996). Using non-linear least squares (NLS) is recommended for
erance, while financial interest has the largest association when
continuously distributed outcomes on a bounded interval.
the individual already has a fairly high risk tolerance.
E(ORT|x) = G(β0j + β1 FIj + β2 FLj + βi CVi j ) (6) Study results add to the growing literature on heterogeneity
in risk tolerance. They also contribute to the field of financial
Here G (·) is the probit function. A probit function can be used
literacy. Specifically, results point to the relevance of financial
with robust standard errors computed as defaults, even if the true
interest, or internal motivation, on risk tolerance. Assuming that
model is not probabilistic.
The results are reported in Table 11. The estimation of Model financial literacy does lead to improved financial behavior, we
5 shows that financial literacy and financial interest have signif- discuss what measures should be taken to improve the general
icant effects on objective risk tolerance. The standardized beta public’s knowledge of financial issues. In a combined survey and
coefficients amount to 0.080 and 0.119 respectively. Estimating experimental study, Hung and Yoong (2013) found that partic-
Model 6, we compute the average marginal effects of the vari- ipants who themselves take initiative to seek financial advice
ables at their means. The results show that one score higher in perform better than those who receive advice automatically and
the financial literacy test increases objective risk tolerance by for free. The results endure regardless of the level of financial
0.40 percentage points, on average, over the whole distribution literacy. Their results point to the relevance of interest and mo-
of financial literacy. The marginal effects show that a 1% in- tivation in positive financial performance. In this study, we test
crease in financial interest increases objective risk tolerance by this relevance explicitly and confirm its relevance.
10
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

Table 11
Objective risk tolerance, financial literacy and financial interest.
OLS- regression (1) Fractional outcome regression (2)
Variable Coef. Marginal effect
Financial literacy 0.080*** 0.044*** 0.004***
(0.001) (0.008) (0.001)
Financial interest 0.119*** 0.090** 0.008***
(0.001) (0.009) (0.001)
Age −0.169*** 0.000 0.000
(0.000) (0.006) (0.001)
Age2 0.237*** 0.000 0.000
(0.000) (0.000) (0.000)
Gender 0.028*** 0.040 0.004
(0.003) (0.025) (0.002)
Large city −0.016* −0.038 −0.003
(0.001) (0.032) (0.003)
Income −0.037*** −0.010** −0.001**
(0.001) (0.004) (0.000)
Wealth 0.236*** 0.242*** 0.022***
(0.001) (0.002) (0.001)
Mortgage −0.016 0.003 0.000
(0.000) (0.002) (0.000)
Work status
Part time 0.006 −0.046 −0.004
(0.004) (0.045) (0.003)
Retired 0.067*** 0.103*** 0.009***
(0.005) (0.037) (0.003)
Long-term sick leave 0.045*** 0.261*** 0.027***
(0.008) (0.068) (0.008)
Student 0.014** 0.047 0.004
(0.006) (0.089) (0.008)
Unemployed 0.014** −0.032 −0.003
(0.006) (0.076) (0.005)
Education
Pregymnasial education 0.013 0.065 0.005
(0.006) (0.050) (0.004)
Gymnasium 0.035** 0.089* 0.007*
(0.005) (0.048) (0.004)
Postgymnasial education, <3 yrs 0.027* 0.085* 0.007*
(0.006) (0.048) (0.004)
Postgymnasial education, ≥3 yrs 0.027 0.066 0.006
(0.006) (0.046) (0.004)
Family status
Single w. children −0.005 −0.135* −0.011**
(0.006) (0.072) (0.005)
Couple w/o children −0.014 −0.024 −0.002
(0.004) (0.031) (0.003)
Couple w. children −0.017 −0.035 −0.003
(0.004) (0.033) (0.003)
Other −0.006 −0.118 −0.010
(0.144) (0.011)
Housing
Tenant-owned apartment −0.002 0.051 0.004
(0.041) (0.004)
House −0.005 0.029 0.003
(0.038) (0.003)
Farmhouse 0.012 0.113* 0.010*
(0.061) (0.010)
Constant -*** −5.00***
(0.007) (0.202)
Observations

Adj. R2 0.142
Observations 12,156 12,156

This table presents the results from (1) the OLS regression and (2) the fractional outcome regression with
ORT as the dependent variable. The OLS-regression reports standardized beta coefficients. The fractional
outcome regression reports coefficients and marginal effects. The sample size is N = 12,156. Marginal
effects are average effects at the means of the variables. Standard errors are given in parentheses. ***, **,
and * denote significance at the 1%, 5%, and the 10% level, respectively. The fractional outcome regression
yields a pseudo R2 of 0.131.

6. Limitations and suggestions for future research study uses cross-sectional data, iterative processes of financial
knowledge and financial literacy remain to be explored. Fur-
The present study suffers from a number of limitations. It is thermore, because the items measuring financial interest were
almost a tautology that we know more about a subject that inter- self-reported, we cannot rule out the possibility that observed
ests us. Previous researchers established a link between interest positive associations were a consequence of positive desirability
in a topic and knowledge about the same topic. Because this bias.
11
C. Hermansson and S. Jonsson Journal of Behavioral and Experimental Finance 29 (2021) 100450

This study points to the relevance of analyzing long-term Eckel, C.C., Grossman, P.J., 2008. Men, women and risk aversion: Experimental
individual interest (e.g., Renninger, 2000) in relation to individ- evidence. In: Handbook of Experimental Economics Results, Vol. 1. pp.
1061–1073.
uals’ risk tolerance. Future research could investigate the interest Feng, X., Lu, B., Song, X., Ma, S., 2019. Financial literacy and household finances:
construct further, that is, the effects of various forms of interest, A Bayesian two-part latent variable modeling approach. J. Empir. Financ. 51,
such as temporary situational interest vis-à-vis long-term indi- 119–137.
vidual interest. Further investigations on the interest construct Fisher, P.J., Yao, R., 2017. Gender differences in financial risk tolerance. J. Econ.
Psychol. 61, 191–202.
also include investigations of its relationship to other noncogni- Fornell, C., Larcker, D.G., 1981. Evaluating Structural Equation Models with
tive traits (i.e., personality) as well as joint effects of cognitive unobservable variables and measurement error. J. Mark. Res. 18 (1, February),
and noncognitive traits on individuals’ risk tolerance and financial 39–50.
Gerbing, D.W., Anderson, J.C., 1988. An updated paradigm for scale development
well-being. The research on grit — passion for long term goals, for
incorporating unidimensionality and its assessment. J. Mark. Res. 25 (2),
example, finds that a consistency of interest combined with per- 186–192.
severance of effort explain individual success in various domains Grable, J.E., 2000. Financial risk tolerance and additional factors that affect risk
(Duckworth et al., 2007). taking in everyday money matters. J. Bus. Psychol. 14 (4), 625–630.
Grable, J., Lytton, R.H., 1999. Financial risk tolerance revisited: the development
of a risk assessment instrument. Final. Serv. Rev. 8 (3), 163–181.
Declaration of competing interest Gunnarsson, J., Wahlund, R., 1997. Household financial strategies in Sweden: An
exploratory study. J. Econ. Psychol. 18 (2–3), 201–233.
Gysler, M., Brown Kruse, J., Schubert, R., 2002. Ambiguity and Gender Differences
The authors declare that they have no known competing finan- in Financial Decision Making: An Experimental Examination of Competence
cial interests or personal relationships that could have appeared and Confidence Effects, Vol. 2002. Working Papers/WIF, (23).
to influence the work reported in this paper. Hair, Jr., J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1988. Multivariate Data
Analysis. Prentice-Hall, Upper Saddle River, New Jersey.
Heath, C., Tversky, A., 1991. Preference and belief: Ambiguity and competence
Acknowledgments in choice under uncertainty. J. Risk Uncertain. 4 (1), 5–28.
Hermansson, C., 2018. Can self-assessed financial risk measures explain and
We acknowledge financial support from Jan Wallander and predict bank customers’ objective financial risk?. J. Econ. Behav. Organ. 148,
226–240.
Tom Hedelius Foundation (grant 2015-0523), and also thank Hidi, S., 1990. Interest and its contribution as a mental resource for learning.
Swedbank for providing data. Rev. Educ. Res. 60 (4), 549–571.
Hung, A., Yoong, J., 2013. Asking for help: Survey and experimental evidence on
financial advice and behavior change. Mark. Retire. Financ. Advice 182.
References Kramer, M.M., 2016. Financial literacy, confidence and financial advice seeking.
J. Econ. Behav. Organ. 131 (Part A), 198–217.
Almenberg, J., Dreber, A., 2015. Gender, stock market participation and financial Loewenstein, G.F., Weber, E.U., Hsee, C.K., Welch, N., 2001. Risk as feelings.
literacy. Econom. Lett. 137, 140–142. Psychol. Bull. 127 (2), 267.
Anderson, A., Baker, F., Robinson, D.T., 2017. Precautionary savings, retirement Loomes, G., Pogrebna, G., 2014. Measuring individual risk attitudes when
planning and misperceptions of financial literacy. J. Financ. Econ. 126 (2), preferences are imprecise. Econ. J. 124 (576), 569–593.
383–398. Lusardi, A., 2008. Financial Literacy: An Essential Tool for Informed Consumer
Bannier, C.E., Neubert, M., 2016. Gender differences in financial risk taking: The Choice? (No. W14084). National Bureau of Economic Research.
role of financial literacy and risk tolerance. Econom. Lett. 145, 130–135. Lusardi, A., 2012. Numeracy, financial literacy, and financial decision-making.
Bergkvist, L., Rossiter, J.R., 2007. The predictive validity of multiple-item versus (No. w17821). National Bureau of Economic Research.
Lusardi, A., Schneider, D., Tufano, P., Morse, A., Pence, K.M., 2011. Financially
single-item measures of the same constructs. J. Mark. Res. 44 (2), 175–184.
fragile households: Evidence and implications (with comments and discus-
Bollen, K.A., 1989. Structural equations with latent variables. In: Wiley Series in
sion). In: Brookings Papers on Economic Activity. pp. 83–150, http://www.
probability and mathematical statistics, Wiley, New York.
jstor.org/stable/41228524.
Bonsang, Dohmen, 2015. Risk attitude and cognitive aging. J. Econ. Behav. Organ.
MacCrimmon, K.R., Wehrung, D.A., 1986. Risk Management. The Free Press, New
112, 112–126.
York.
Bucciol, A., Miniaci, R., Pastorello, S., 2017. Return expectations and risk aversion Menkhoff, L., Sakha, S., 2017. Estimating risky behavior with multiple-item risk
heterogeneity in household portfolios. J. Empir. Financ. 40, 201–219. measures. J. Econ. Psychol. 59, 59–86.
Cesarini, D., Johannesson, M., Lichtenstein, P., Sandewall, Ö., Wallace, B., 2008. Papke, L.E., Wooldridge, J.M., 1996. Econometric methods for fractional response
Is Financial Risk-Taking Behavior Genetically Transmitted? (No. 765). IFN variables with an application to 401 (k) plan participation rates. J. Appl.
Working Paper. Econometrics 11 (6), 619–632.
Charness, G., Gneezy, U., 2012. Strong evidence for gender differences in risk Renninger, K.A., 2000. Individual interest and its implications for understanding
taking. J. Econ. Behav. Organ. 83 (1), 50–58. intrinsic motivation. In: Intrinsic and Extrinsic Motivation. Academic Press,
Conlin, A., Kyröläinen, P., Kaakinen, M., Järvelin, M.R., Perttunen, J., Svento, R., pp. 373–404.
2015. Personality traits and stock market participation. J. Empir. Financ. 33, Säve-Söderbergh, J., 2012. Self-Directed Pensions: Gender, risk, and Portfolio
34–50. Choices. Scand. J. Econ. 114 (3), 705–728.
Croson, R., Gneezy, U., 2009. Gender differences in preferences. J. Econ. Literature Schiefele, U., 1991. Interest, learning, and motivation. Educ. Psychol. 26 (3–4),
47 (2), 448–474. 299–323.
Cumming, G., 2009. Inference by eye: reading the overlap of independent Schooley, D.K., Worden, D., 1996. Risk aversion measures: Comparing attitudes
confidence intervals. Stat. Med. 28 (2), 205–220. and asset allocation.
Slovic, P., 1987. Perception of risk. Science 2364799, 280–285.
Deci, E.L., Ryan, R.M., 1991. A motivational approach to self: Integration in
Sung, J., Hanna, S., 1996. Factors related to risk tolerance. J. Financ. Couns. Plan.
personality. In: Dienstbier, R. (Ed.), Nebraska Symposium on Motivation: Vol.
7, 11.
38, Perspectives on Motivation. University of Nebraska Press, Lincoln, pp. Swedish investment fund association, 2016. Årsrapport Fondsparande/Annual
237–288. Report Mutual Fund Savings.
Dimmock, S.G., Kouwenberg, R., Mitchell, O.S., Peijnenburg, K., 2016. Ambiguity Tobias, S., 1994. Interest, prior knowledge, and learning. Rev. Educ. Res. 64 (1),
aversion and household portfolio choice puzzles: Empirical evidence. J. 37–54.
Financ. Econ. 119 (3), 559–577. Van Rooij, M., Lusardi, A., Alessie, R., 2011. Financial literacy and stock market
Dohmen, T., Falk, A., Huffman, D., Sunde, U., Schupp, J., Wagner, G.G., participation. J. Financ. Econ. 101 (2), 449–472.
2011. Individual risk attitudes: Measurement, determinants, and behavioral Von Culin, K.R., Tsukayama, E., Duckworth, A.L., 2014. Unpacking grit: Motiva-
consequences. J. Eur. Econom. Assoc. 9 (3), 522–550. tional correlates of perseverance and passion for long-term goals. J. Posit.
Dreber, A., Hoffman, M., 2007. 2D: 4D and Risk Aversion: Evidence that Psychol. 9 (4), 306–312.
the Gender Gap in Preferences Is Partly Biological. Stockholm School of Wärneryd, K.E., 1996. Risk attitudes and risky behavior. J. Econ. Psychol. 17 (6),
Economics Working Paper, Stockholm. 749–770.
Dreber, A., Rand, D.G., Wernerfelt, N., Garcia, J.R., Vilar, M.G., Lum, J.K., Zeck- Wärneryd, K.-E., 1999. The Psychology of Saving – a Study on Economic
hauser, R., 2011. Dopamine and risk choices in different domains: Findings Psychology. Edward Elgar, Cheltenham, UK.
Wlodkowski, R.J., Ginsberg, M.B., 2017. Enhancing Adult Motivation To Learn: A
among serious tournament bridge players. J. Risk Uncertain. 43 (1), 19–38.
Comprehensive Guide for Teaching All Adults. John Wiley & Sons.
Duckworth, A.L., Peterson, C., Matthews, M.D., Kelly, D.R., 2007. Grit: perse-
Wooldridge, J.M., 2015. Introductory Econometrics: A Modern Approach. Nelson
verance and passion for long-term goals. J. Personal. Soc. Psychol. 92 (6),
Education.
1087.

12