Financial literacy and nescience
A search for the presence of the Dunning-Kruger effect
Jan Christian Scheweer
University of Groningen
2
Introduction
“I know nothing, except the fact of my ignorance”
(Socrates)
After Socrates’ friend Chaerephon was told by the Oracle at Delphi that there was no one wiser than Socrates, Socrates set out to find someone who had true wisdom in an effort to proof the Oracle wrong. In his search for true wisdom Socrates conversed with people who were experts in a variety of fields, for example ship builders, shoemakers, poets and military strategists. He found that they erroneously believed themselves to be experts in other fields as well, like morality or politics. After many conversations he came to the conclusion that he was wiser than those men, because he knew what he did not know. Put differently he was aware of his own ignorance. This is known as “Socratic Ignorance”. The term “simple ignorance” is sometimes used instead of Socratic ignorance. Whereas simple ignorance refers to someone being aware of his own ignorance, double ignorance is when someone is unaware of his ignorance, but thinks he is are aware.
Are things different now than they were over 2.000 years ago? Has humanity ascended to a higher state of self-awareness? Or have we not learned to recognize, acknowledge and accept our ignorance? The topic of this paper would suggest the latter to be closer to the truth. In 1999 Kruger and Dunning hypothesized a behavioural phenomenon that later became known as the “Dunning-Kruger effect” This effect states that unskilled people are unaware of their incompetence. Kruger and Dunning claim this is because the skills required to accurately assess your level of competence are the same as the ones needed to be competent. Competent people exhibit the opposite tendency as they tend to underestimate their level of knowledge and abilities. The phenomenon can be seen as a form of double ignorance.
The most important papers on the topic have been written by, other than Kruger and Dunning themselves of course, Krueger & Mueller (2002), Krajč & Ortmann (2008 and 2012) and Burson (2006). These will be discussed in the literature review section of this paper.
3 To our knowledge there has never been any research done to test the presence of the Dunning-Kruger effect through a nationally representative sample. The research done in this paper is performed on the basis of a nationally representative sample. In this way we add to the literature already written about the Dunning-Kruger effect. Most of the time students are the subjects of the research, making generalization of the findings not very trustworthy. The Dunning-Kruger effect has also never been tested in the domain of financial literacy or any other domain that qualifies as a behavioural finance topic. Finally, the methodology used in this paper is slightly different from other papers testing for the effect.
4
1. Literature review
In the five sections of the literature review we will first discuss the two most important subjects of the paper, namely The Dunning-Kruger effect and financial literacy. Thirdly, we will discuss criticism of the original paper on the Dunning-Kruger effect. After that other research trying to find the Dunning-Kruger effect will be discussed and finally we discuss literature on the possibilities of countering the phenomena underlying the Dunning-Kruger effect.
The Dunning-Kruger effect
In 1999 the Journal of Personality and Social Psychology published a paper written by J. Kruger and D. Dunning called: “Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments”. In this paper Dunning and Kruger hypothesize that individuals that have low competence in a certain field are unable to realize their incompetence, because they suffer a dual curse, namely the skills and knowledge needed to be competent are also those needed to accurately assess one’s own competence. They put it as follows: “We propose that those with limited knowledge in a domain suffer a dual burden:
Not only do they reach mistaken conclusions and make regrettable errors, but their incompetence robs them of the ability to realize it”. They find evidence of this hypothesis in
the four studies across different domains performed for the paper. An unexpected finding Kruger and Dunning made was that not only do the incompetent overestimate their abilities and knowledge, but the highly competent tend to underestimate their abilities and knowledge. Most likely they believe that, because they possess the knowledge others will as well. This is known as the false consensus effect. The two types of miscalibration discussed here, together are called the Dunning-Kruger effect.
A similar topic has been researched a lot, namely overconfidence, or “The tendency for people
to overestimate their knowledge, abilities, and the precision of their information or to be overly sanguine about the future.” (Ackert & Deaves 2016). Though the Dunning-Kruger effect and
5 through a more comprehensive methodology using counterfactual regression. They found that indeed the incompetent’s miscalibration was caused by erroneous beliefs of their own performance. Contrary to what was expected the highly competent not only misestimated the performance of their peers, but also underestimated their own performance.
Financial literacy
The concept of financial literacy has had many definitions throughout the last decades. A consensus about the definition has, as of yet not been reached. David Remund (2010) made an attempt to create a clear and comprehensive definition. He found that many of the definitions used were imprecise. The definition Remund proposed is the following: “Financial
literacy is a measure of the degree to which one understands key financial concepts and possesses the ability and confidence to manage personal finances through appropriate, short-term decision-making and sound, long-range financial planning, while mindful of life events and changing economic conditions”. This is also the definition we will use in this paper.
In the last decade financial literacy has received more attention from scholars and countries started paying more attention to the financial literacy of its citizens after the 2008 global financial crisis (Xiao & O’Neill 2016). The topic became even more important, because of market liberalization and the redesigning of the pension systems and social security, giving individuals more responsibility over their finances (Van Rooij, Lusardi & Alessie 2012).
These papers often relate financial literacy to financial advice seeking (Stolper and Walter) , household wealth accumulation (Van Rooij, Lusardi and Alessie, 2012; M. Debbic, 2015) and stock market participation (Van Rooij, Lusardi and Alessie, 2011). Many of them find a positive correlation between financial literacy and financial wellbeing and thus advocate for better financial education of consumers. Van Rooij et al (2011) find in their paper on the relationship between financial literacy and stock market participation that people who are more financially literate are more likely to invest in stock markets and reap the benefits of the equity premium. J.J. Xiao & N. Porto (2016) write that when consumers overestimate the level or accuracy of their knowledge they might decide not to enlist the help of a financial advisor. This in turn causes the consumer to not reap the benefits of a professionals advice.
6 Laurance E Willis writes, in his 2011 paper titled: “The Financial Education Fallacy”, that increasing financial literacy is an impossible and undesirable tactic to increase consumer financial wellbeing. Willis also writes that since we go to a doctor when we are ill or to a lawyer for legal advice why then should we have to learn to make financial decision by ourselves instead of going to a financial advisor? We tend to agree with this view on the issue of financial literacy of consumers.
If we find evidence of the Dunning-Kruger effect, there might be other solutions then education to make consumers more self-sufficient with respect to financial matters or more prone to seek financial advice. We would need to make people aware of their own incompetence. There are however some strong emotional motives to overcome, since people want to feel good about themselves and reject information that might prove a more negative view to be true (Baumeister & Newman, 1994). We will discuss some possible solution later on.
Criticisms and counter explanations to the Dunning-Kruger effect
Since the original paper of Kruger and Dunning was published in 1999 there have been a number of critiques on their analysis. In this section we will discuss some of them.
Krueger and Mueller (2002) argue that the phenomena that are deemed the Dunning-Kruger effect can also be explained by a combination of the better-than-average heuristic and regression effects (also known as regression to the mean). Regression effects tell us that if we would let a group of people make a certain test twice, the participants with extreme results (either high or low) will have a result closer to the mean the second time around (or vice versa). The better-than-average heuristic is defined as: the tendency for a person to rate
himself as above average in knowledge or skills (Ackert & Deaves, 2016). In the original 1999
paper Kruger and Dunning do mention the possibility of regression effects, but they believe that if this would explain the magnitude of the miscalibration it should be the same for the bottom and top quartile, which was not the case in their study. The bottom quartile was a lot more miscalibrated then the top quartile was.
Another possible error Kruger and Dunning made according to Krueger and Mueller (2002) is to assume that metacognitive abilities, intelligence and accurate self-assessment are a “package deal”. Krueger and Mueller argue that these skills do not have a one on one relation. The arguments given by Krueger and Mueller were tested and their original standpoints were defended by Kruger and Dunning (2002).
7 students at Cornell university. Krajč and Ortmann argue that the sample is not representative of the population for many reasons. The first is the fact that Cornell University admits approximately 29% of applicants and thus can select the best of the best. This makes the distribution non-normal and as they argue probably more like a backwards J-distribution. Then there is also the grade inflation phenomenon. This phenomenon makes grades less and less differentiating between students making inferences more difficult. They also argue that since it were students in their first year they had no previous experience with the subject they were tested on and thus the students face a signal extraction problem and the unskilled face a greater inference problem than the skilled.
Krajč an Ortmann (2008) conclude in their paper that the previously mentioned arguments together with the absence of controlling for diagnosticity of feedback and incentives for giving correct (self-)assessment, the methodology in inherently flawed and the conclusions therefore unreliable.
Burson, Klayman and Larrick (2006) claim in their study that all skill levels suffer from a similar degree of error and that this error depends on task difficulty. They put it as follows: “we
propose that people at all performance levels are equally poor at estimating the relative performance (e.g., their judgements are noisy) and equally prone to overestimating their percentile on tasks that are perceived to be easy and underestimating it on task perceived to be hard”. Their findings support this notion as the test results show that most inaccuracy
dissipates when accounting for task difficulty. They however do not dismiss the possibility that there is a correlation between cognitive and metacognitive skill, but think that it is not the primary cause of who is miscalibrated. To be able to determine a relation between cognitive and metacognitive skills the methodology has to be more comprehensive and first adjust for noise and a certain up or downward bias depending on (perceived) task difficulty. The findings and subsequent conclusions are in line with those of Krueger & Mueller (2002).
8 Finally there are floor and ceiling effects to consider. These occur because the individuals at the bottom or top can only have miscalibration in one direction. For the ones at the bottom they can only be correct or overestimate themselves and the ones at the top can only be correct or underestimate themselves.
Other research on the Dunning-Kruger effect
Over the last twenty years a lot of research has been done to find evidence of the Dunning -Kruger effect. Most of this research has been done, like the original paper, with the use of a student sample (See for example; Lindsey and Nagel, 2015 and Karatjas, 2013).
Research on the presence of the Dunning-Kruger effect has also been done with samples in the field of aviation (Pavel, Robertson and Harrison, 2012), wine tasting (Aqueveque, 2017) and among bridge players (Simons, 2013).
All the aforementioned studies have found evidence of the existence of the Dunning -Kruger effect. Although it is difficult to imagine that broader samples would yield different results, none of the research is done with a representative sample of the population and thus is generalization of the findings debatable.
Countering the Dunning-Kruger effect
The most obvious solution to countering the Dunning-Kruger effect of course would be to educate the incompetent so they would gain the meta-cognitive skills to realize their incompetence. The education would eliminate or at least reduce their incompetence as well. However educating everyone in all fields in which they are incompetent and unaware, would be an impossible task.
We were only able to find one paper that (briefly) discusses the possibilities of countering the Dunning-Kruger effect. Ehrlinger et al (2008) suggest it might be possible to encourage a more positive mind-set about learning and, by extension more accurate self-reflection. This is of course a method that will take a least some generations to be fully incorporated, if possible at all. Karl Popper said it as follows; “Our knowledge can only be finite, while our ignorance must
9
2. Hypotheses
After philosophy and theory it is now time to discuss empirics. The research question we wish to answer is: “Is the Dunning-Kruger effect prevalent in the domain of personal financial literacy?” To answer this research question we formulate two hypotheses. The first one is about the overconfidence that plagues the incompetent people.
Hypothesis 1
H0: Low actual financial literacy = Self-perceived financial literacy H1: Self-perceived financial literacy > actual financial literacy
The second hypothesis is for the underconfidence that the competent display. Hypothesis 2
H0: High actual financial literacy = Self-perceived financial literacy H1: Self-perceived financial literacy < actual financial literacy
10
3. Methodology
Regression
To test the two hypotheses outlined in the previous section, we will first have to create a measure of overconfidence (O). We do that by subtracting actual performance (A) from the participants self-assessed score (S). See equation 1. We recoded the self-assessment score so that it would take a value between zero and four1, like the actual financial literacy score.
𝑂 = 𝑆 − 𝐴 (1)
Overconfidence can thus have a value between minus four and four. Equation 2 shows the regression. If the Dunning-Kruger effect is present, we would expect the incompetent to have the highest degree of overconfidence and the highly competent to be the least overconfident or even underconfident.
𝑂 = 𝛼 + 𝛽𝐴 + 𝜀 (2)
We will thus expect the beta to be negative (i.e. the higher the actual performance the less overconfident the person will be). We will also perform a couple of regressions with control variables. A widely documented phenomenon is that men tend to be more overconfident that women. See for example Lundeberg, Fox and Punćochaŕ (1994). We will thus put a dummy in the regression to account for gender (G), where 1 = male and 0 = female. We will also put a interaction term in the regression (see equation 3).
𝑂 = 𝛼 + 𝛽1𝐴 + 𝐷1𝐺 + 𝛾𝐴 ∗ 𝐺 + 𝜀 (3)
Other variables we think can influence overconfidence are education, field of education, age and whether the participant is an investor in financial markets . These will also be tested in the regressions, with and without interaction terms.
Age can take a value between 16 (youngest participants) to 89 (oldest participants). Research shows that age can have a negative or positive relationship with self-assessment. The direction of the effect is determined through the question if the subject being evaluated improves with age (Zell and Alicke, 2011). In this case we would expect a positive correlation, since people are confronted with personal finances throughout their lives . Previous research show that older males who are highly educated are more overconfident when it comes to investment knowledge (Bhandari and Deaves, 2006).
11 Level of education can take six values, where 1 = Primary school, 2 = Junior high school 3 = Senior high school, 4 = Junior college, 5 = College and 6 = University. Although education is thought of as a method of decreasing overconfidence, when the education has no relationship with the topic of the questions there will not be a decrease of overconfidence. In fact there is an increase in overconfidence (Bhandari and Deaves, 2006). We will test this by regressing level of education and by accounting for participants who graduated from an economic or finance related education. We would expect these to have less overconfidence.
Finally we will account for whether the participant has experience as an (private) investor on financial markets. We expect a negative sign as more experience should lead to greater knowledge which in turn should lead to less overconfidence.
Non-parametric test
To add some robustness to the findings and to account for potential normality issues accompanied with ordinal data, we will also perform a non-parametric rank test. The test chosen is the Spearman rank correlation coefficient. The formula for calculating Spearman’s rho is given in equation 4.
𝑟𝑠𝑝𝑒𝑎𝑟𝑚𝑎𝑛 = 𝑆𝑠𝑎
𝑆𝑠𝑆𝑎 (4)
Where s and a are the ranks of self-assessed financial literacy and actual financial literacy, respectively, Ssa is the covariance of the values s and a, Ss is the standard deviation of the
values of s and Sa is the standard deviation of the values of a.
12
4. Data
In this paper we make use of data of the LISS (Longitudinal Internet Studies for the Social sciences) panel administered by CentERdata (Tilburg University, The Netherlands). The LISS panel is a representative sample of Dutch individuals who participate in monthly Internet surveys. The panel is based on a true probability sample of households drawn from the population register. Households that could not otherwise participate are provided with a computer and Internet connection. A longitudinal survey is fielded in the panel every year, covering a large variety of domains including work, education, income, housing, time use, political views, values and personality.
We use a single wave of data from the august 2011 survey in which participants were asked how they would rate there knowledge on financial matters and subsequently, they were asked to answer four question on financial matters. The four questions are presented is table 1. The original intention of the survey was to gauge the level of financial literacy of the Dutch population.
Table 1. Questions to measure financial literacy asked to the participants
Question Topic
Question A How much do you think you will have on the savings account after five years, assuming that you leave all your money on this savings account: more than 102 euros, exactly 102 euros, less than 102 euros?
Compounding interest
Question B Suppose that the interest on your savings account is 1% per year and that inflation amounts to 2% per year. After 1 year, would you be able to buy more, exactly the same, or less than you could today with the money on that account?
Inflation
Question C A share in a company usually offers a more certain return than an investment fund that only invests in shares.
13
Descriptive statistics
The number of participants that filled out the questionnaire is 4,860. For four of these 4,860 participants data other than the questionnaire on financial literacy was not available. The survey was filled out by 2,229 males and 2,627 females. The distribution of level of education is listed in table 2. In total 831 participants finished their education in the field of economics, management, business administration or accountancy, regardless of the level of education.
Table 2. Distribution of the level of education in the sample
Level of education Percentage (Absolute number)
Primary school 5.8% (281)
Junior high school 25.7% (1251) Senior high school 11.5% (557)
Junior college 21.9% (1064)
College 22.6% (1096)
University 8.1% (396)
Other 3.3% (162)
Not (yet) completed any education 0.8& (37) Not (yet) started any education 0.2% (12)
Missing 0.1% (4)
The average self-assessment score on financial literacy is 63.9%. 3.224 (66.3%) rate themselves higher than this average. The lowest score of 0% appears 66 times in the sample. 284 people gave themselves a perfect score of 7/7. Of these 284 only 63 were able to answer all questions correctly.
The number of questions answered correct on average is 56.67% or 2.25 questions. Table 3 summarizes the number of times each question is answered correctly, wrong or when they don’t know the answer.
Table 3. Frequencies and absolute numbers of times each question is answered correctly
Answered correctly Don’t know Wrong Question A 88.6% (4,304) 4.4% (215) 7% (341) Question B 76.9% (3,739) 10.8% (526) 12.3% (595) Question C 42% (2,043) 40.8% (1,984) 17.2% (833) Question D 19.2% (931) 46.9% (2,278) 33.9% (1,651)
14 Table 4 shows the distribution of the overconfidence values as calculated with equation 1. The mean value is 0.5216 with a standard deviation of 1.3869 indicating that there are large variations in overconfidence.
Table 4. Distribution of the overconfidence values
Overconfidence Percentage (Absolute number)
-4.00 0.1% (6) -3.00 1.8% (88) -2.00 5.8% (280) -1.00 13.3% (644) 0.00 27.7% (1348) 1.00 27.9% (1357) 2.00 16.6% (807) 3.00 5.4% (264) 4.00 1.4% (66)
As mentioned in the literature overview, Krajc and Ortmann (2008) comment on the method of Kruger & Dunning saying their sample is biased and that it most likely will follow a backwards J-distribution. The fact that the dataset used in this paper is made up of respondents representative of the Dutch population makes it so that we expect no such issues. Figure 1 shows the distribution of the number of question answered correctly. The figure confirms our expectation.
15
5. Results and analysis
Below we discuss the results from the statistical tests outlined in the methodology section. We will first discuss the regression results and secondly the results of the non-parametric test will be discussed.
Regressions
The results of the regressions without the interaction terms are summarized in table 5. As one can immediately see from the first regression is that actual financial literacy has a statistically significant negative relation to overconfidence. This is what we expected to see in the presence of the Dunning-Kruger effect. In the next regressions the factor stays significant and the most import one.
As predicted the gender variable shows that women tend to be less overconfident than men. The result is statically significant on all conventional significance levels.
The third regression includes the age variable. We find that age has a positives and statistically significant impact on overconfidence. The impact, though significant statistically, is much less significant in absolute terms. It seems that the wisdom that comes with age does not include a better ability of assessing one’s own capabilities and knowledge. At least not with respect to financial literacy. Ehrlinger et al (2008) come to a similar conclusion.
We perform a regression that includes the education term without the participants that answered “other” and “not (yet) started / completed any education” on the education question. We exclude these cases, because they make a correct interpretation of the regression result difficult if not impossible. The education variable shows a statistically significant influence at the five and ten percent significance level on overconfidence according to the model. Like the age variable the absolute impact is negligible. When accounting for the field of education we find a negative effect on overconfidence, like expected. The result however, is not statistically significant.
Finally we put in a dummy variable to account for experience investing in financial markets. The regression shows small and insignificant coefficients. Experience on financial markets has no significant impact on a person’s self-assessment. When comparing the means of actual financial literacy between those that are and those that are not active on the financial market we find that the mean of those active on financial markets is slightly higher, though not significantly so (57.05% versus 56.61%).
16 sample test. The test shows that there is no difference in the mean self-assessment, t-statistic -7.3274 P < 0.0001. This shows that the worst performers do not believe their knowledge and abilities are worse than the average.
18 We will also analyse the results of the regressions that include interaction terms. The results are summarized in table 6. Only the gender interaction term is significant at the five and ten percent level. The interaction term has a negative sign meaning that when actual skills are higher the difference in overconfidence between men and women becomes smaller. The total effect of gender diminishes with the use of the interaction variable, but it is still a significant result.
When we include all variables and all interaction variables into one regression we find that three variables and two interaction terms are statically significant namely, actual performance, gender, age, gender*actual financial literacy and age*actual financial li teracy. The results stay the same as with the individual regressions except for age. The age term has reversed its sign, becoming negative while the interaction term has stayed positive. This suggests that contrary to the other regressions, age has a decreasing effect on overconfidence, but only for the people scoring lower than 2.25 on the actual financial literacy test. Thus overconfidence decreases for participants that have lower scores and increases with higher scores on the financial literacy test. A result that is contrary to the theory.
Non-parametric correlation tests
20
6. Conclusions
Our analysis confirms the presence of the Dunning-Kruger effect. The regressions show a large and significant negative relationship between overconfidence and the number of questions answered correctly. This means that when financial literacy increases the level of overconfidence decreases and can even become negative (meaning underconfidence). This result is in favour of both our hypotheses.
Other factors we thought could influence overconfidence, like gender, age and education show varying results. Gender gives a statistically significant result. Like in other research men tend to be more overconfident then women. We also found a small but significant effect for age and level of education without the interaction term. Age yielded a counterintuitive result in the last regression where a better actual performance increases overconfidence. Field of education never showed a significant result. Experience on financial markets also shows no significant impact on the level of overconfidence.
The Spearman rank test shows the significant negative relationship between overconfidence and actual performance that we expect to see if the Dunning-Kruger effect is present. In the introduction we asked the question: “Is the Dunning-Kruger effect prevalent in the domain of personal financial literacy?” On the basis of the analysis presented in this paper we conclude that the Dunning-Kruger effect is present in the domain of personal financial literacy. In the literature review we discussed the potential dangers of the Dunning-Kruger effect being present in the domain of financial literacy. We also briefly discussed options on countering the effect. It is our belief that more research should be done on this subject. Until better and more simple methods are developed to overcome such deficiencies in self-assessment, it will be the responsibility of the government to minimize the loss in financial well-being of those that are unskilled and unaware. In the Netherlands the government has taken this responsibility in the past, by mandatory pension plans for employees, compulsory advice when taking out a mortgage, stringent duty of care regulation for financial institutions and so forth. As mentioned before, Van Rooij et al. (2012) wrote that a trend towards more personal responsibility is happening when it comes to personal finance. As long as people do not realize the extent of their ignorance, this to us, seems like an unwise trend.
21
7. Limitations & suggestions for further research
This section is used to discuss some of the limitations of the research done for this paper. On the basis of these limitations we will also make some suggestions for further research. The original paper by Kruger and Dunning received a number of critiques after being published. These critiques were discussed in the literature review. Some of these have been addressed in the analysis, others we could not take into account. These limitations and some of our own will be discussed next.
Limitations
22
Suggestions for further research
A great addition to the already existing research and this paper would be a repetition of this research based on a financial literacy questionnaire designed specifically for the purpose of testing for the Dunning-Kruger effect. Some of the aspects of this questionnaire should be;
1. Clear questions on relative knowledge on financial matters (self-assessment) 2. A more comprehensive questionnaire to assess actual financial literacy
3. A question asking an assessment of own performance after answering the questionnaire.
4. There should be more than one questionnaire asking the same question, but phrased differently to account for framing effect
5. The questionnaire should be filled out by participants on at least two different occasions to account for regression to the mean.
Some of these aspects are more difficult to realize, but surely a majority of them should be possible.
Feld, Sauermann and De Grip (2017) propose an instrumental variable (IV) approach to properly take measurement error into account. They argue that researchers use the same measure to define skill and to calculate overestimation and therefore the same measurement error is present in both components. In their research they use previously acquired grades to account for skill. This approach, however, is difficult to do on a national population representative scale. It is our opinion that repetition of the survey can also create the necessary adjustment to account for this “luck” factor.
23
Acknowledgements
I would first like to thank my thesis advisor Prof. Dr. B.W. (Robert) Lensink of the Faculty of Economics and Business Economics at the University of Groningen for his guidance during the writing of this paper. Furthermore I would like to thank Jens van der Meer for his input in developing and discussing the methodology.
24
Literature list
Ackert, L.F., Deaves, R., 2016. Behavioral Finance: Psychology, Decision Making and Markets. South western, Mason USA.
Aqueveque, C. 2018. Ignorant experts and erudite novices: Explorng the Dunning-Kruger effect in wine consumers. Journal of Food quality and preference 65, 181 – 184.
Baumeister, R.F., Newman, L.S. 1994. Self-regulation of Cognitive Inference and Decision Process. Personality and Social Psychology Bulletin, 20, 3 – 19.
Bhandari, G., Deaves, R., 2006. The Demographics of overconfidence. Journal of Behavioral Finance, Vol. 7, No. 1, 5 – 11.
Burson, K.A., Klayman, J., Larrick, P.L. 2006. Skilled or Unskilled, but Still Unaware of It: How Perceptions of Difficulty Drive Miscalibration in Relative Comparisons. Journal of Personality and Social Psychology, Vol. 90, No. 1, 60 – 77.
Debbich, M., 2015. Why Financial Advice Cannot Substitute For Financial Literacy. Working Paper 534. Banque De France.
Ehrlinger, J., Johnson, K., Banner,M., Dunning, D., Kruger, J. 2008. Why the unskilled are unaware: Further explorations of (absent) self-insight among the incompetent. Organizational Behavior and Human Decision Processes 105, 98 – 121.
Feld, J., Sauermann, J., Grip de, A. 2017. Estimating the relationship between skill and overconfidence. Journal of Behavioral and Experimental Economics 68, 18 – 24.
Huston, S.J., 2010. Measuring Financial Literacy. Journal Of Consumer Affairs, Vol. 44, No.2, 296 – 316.
Karatjas, A.G. 2013. Comparing College Students’ Self-Assessment of Knowledge in Organic Chemistry to Their Actual Performance. Journal of Chemical Education, Vol. 90, 1096 – 1099. Krajc, M., Ortmann, A., 2008. Are the unskilled really that unaware? An alternative explanation. Journal of economic Psychology 29, 724 – 738.
Krueger, J., Mueller, R.A., 2002. Unskilled, Unaware, or Both? The better-than-average Heuristic and Statistical Regression Predict Errors in Estimates of O wn Performance. Journal of personality and Social Psychology, Vol. 82, No. 2, 180 – 188.
25 Kruger, J., Dunning, D., 2002. Unskilled and Unaware – But Why? A Reply to Krueger and Mueller. Journal of Personality an Social Psuchology, Vol. 82, Ni. 2, 189 – 192.
Lindsey, B.A., Nagel, M.L., 2015. Do students know what they know? Exploring the accuracy of students’ self-assessments. Physical review special topics – Physics education research 11. https://journals.aps.org/prper/abstract/10.1103/PhysRevSTPER.11.020103
Lundeberg, M.A., Fox, P.W., Punćochaŕ, J., 1994. Highly Confident but Wrong: Gender Differences and Similarities in Confidence Judgments. Journal of Educational Psychology, Vol. 86, No. 1, 114 – 121.
Pavel, S.R., Robertson, F., Harrison, B.T. 2012 The Dunning-Kruger Effect and SIUC University’s Aviation Students. Journal of Aviation Technology and Engineering, 2:1, 125 – 129.
Porto, P., Xiao, J.J., 2016. Financial Literacy Overconfidence and Financial Advice Seeking. Journal of Financial Services Professionals, Vol. 70, No.4, 78 - 88.
Remund, D.L., 2010. Financial Literacy Explicate: The Case For A Clearer Definition In An Increasingly Complex Economy. Journal Of Consumer Affairs, Vol. 44, No.2, 276 - 295.
Rooij van, M.C.J., Lusardi, A., Alessie, R.J.M., 2012. Financial Literacy and stock market participation. Journal Of Financial Economics 101, 449 - 472.
Rooij van, M.C.J., Lusardi, A., Alessie, R.J.M., 2011. Financial Literacy, Retirement Planning And Household Wealth. The Economic Journal 122, 449 - 478.
Schlösser, T., Dunning, D., Johnson, K.L., Kruger, J. 2013. How unaware are the unskilled? Empiral Tests of the “Signal Extraction” counterexplanation for the Dunning -Kruger effect in self-evaluation of performance.
Simons, D.j. 2013. Unskilled and Optimistic: Overconfident predictions despite calibrated knowledge of relative skill. Psychon Bull Rev. 20, 601–607
Stolper, O.A., Walter, A. J. 2017. Financial literacy, financial advice, and financial behaviour. Bus Econ (2017) 87: 581. https://doi.org/10.1007/s11573-017-0853-9
Willis, L.E., 2011. The Financial Education Fallacy. American Economic Review, 101:3, 429 - 434.
Xiao, J.J., O’Neill, B., 2016. Consumer Financial Education And Financial Capability. International Journal Of Consumer Studies 40, 712 - 721.
