Available via license: CC BY 4.0
Content may be subject to copyright.
International Journal of Economics and Finance; Vol. 11, No. 2; 2019
ISSN 1916-971X E-ISSN 1916-9728
Published by Canadian Center of Science and Education
50
Confirmation Bias in Investments
Chu Xin Cheng1
1 Cate School, Carpinteria, California, United States of America
Correspondence: Chu Xin Cheng, Cate School, Carpinteria, CA, 93013, US. Tel: 1-805-755-9772. E-mail:
cloris_cheng@outlook.com
Received: November 24, 2018 Accepted: December 25, 2018 Online Published: December 30, 2018
doi:10.5539/ijef.v11n2p50 URL: https://doi.org/10.5539/ijef.v11n2p50
Abstract
Investors exhibit some well documented mistakes, such as the disposition effect and excessive trading. One
potential explanation of these phenomena is confirmation bias. People are inclined to be attached to their
investment thesis and are unwilling to consider or accept evidence that they are wrong. Thus, they make
speculative bets and hold onto them even as they show a downward trend. Confirmation bias may result from
people selectively acquiring information that allows them to continue believing what they initially believe. I
investigated selective information acquisition among investors with an experiment that gave participants the
choice to read an article supporting an investment they previously made or one opposing it. I discovered that
investors are significantly more likely to read the article that is supportive of their decision rather than the article
that opposed the investment they had chosen. This suggests that investors exhibit selective information seeking,
which could be a source of confirmation bias and is thus a plausible explanation for the investor mistakes
previously discussed.
Keywords: investments, confirmation bias, selective information acquiring
1. Introduction
1.1 Common Investment Mistakes and Its Relevance to Confirmation Bias
In financial markets we often see paradoxes that are hard to explain with mathematical models and theories. One
simple yet interesting question is “why do trades happen?” According to the No Trade Theorem, because
information is shared as common knowledge with private signals and the concurrent updates of beliefs,
speculations should not occur (Milgrom & Stokey. 1982). That is, if an investor is willing to sell his stocks in the
markets due to unknown sources of information, his willingness is a signal that the stock has the potential to
drop. From the opposite perspective, if a buyer is willing to buy a certain stock, the seller might be reluctant to
let go, for this signal from the buyer is indicating that the stock might rise. However, trades are indeed happening
daily around the world; moreover, there seems to be an excessive flow of trades going on. In 1999, Terrence
Odean examined the situation of the high trading volume in the markets by testing the hypothesis that
overconfidence would cause trades to happen more often.
In an earlier thesis in 1998, Odean illustrated that investors realize their losing stocks‟ situation at a much lower
rate than profitable stocks in what has been called the disposition effect. This results in investors selling their
winning stocks significantly before it reaches its pinnacle and holding on to losing stocks long after they first
started to fall. Colossal financial losses are thus inevitable. How and why does this overconfidence emerge?
What causes investors to be overly confident, even ignorant of their current situation?
One possible explanation for both the disposition effect and speculation is the confirmation bias. Confirmation
bias is a term that describes the reluctance people show to change their initial beliefs. People are going to be
more willing to see and accept new information consistent with what they already believe. A variety of
phenomena fall under the category of confirmation bias. What one might have encountered in one‟s life, such as
subconsciously rejecting certain information that holds against personal opinions or spontaneously favoring a
candidate‟s statement because it resembles one‟s moral beliefs. Such behaviors are often subconscious and
whoever was to review the incident with no involvement in the case might identify it better than the subject.
Motivated confirmation bias happens when a person holds a certain belief and the desire to confirm that belief
with the current information is strong. For instance, in a lawsuit, a juror who has already formed his beliefs may
be inclined to overestimate the importance of a piece of evidence that supports his previous judgment. There are,
ijef.ccsenet.org International Journal of Economics and Finance Vol. 11, No. 2; 2019
51
however, allied theories and concepts, overlapping with confirmation bias, that could potentially explain this
kind of behavior. For example, belief consonance, the preference to have common beliefs with other individuals
and the consequent disturbance of the conflicting beliefs, is similar to motivated confirmation bias. (Golman,
Loewenstein, Moene, & Zarri, 2016.) In fact, belief consonance might be the stimuli to the display of
confirmation bias: subconsciously misinterpreting information. On the other hand, the incoming information
could be easily seen as a threat to the subject‟s current identity when considering believe-based utility. Thus, the
subject will choose to avoid the information by means such as inattention, physical avoidance, and most
importantly, biased interpretation. (Golman, Hagmann, & Loewenstein, 2017).
1.2 Importance of Research
People are indeed reluctant to accept information that counters their existing beliefs. In the context of financial
markets, investors could continue to believe in their original decision (e.g. the stock will rise up) after it has
shown signs of dropping (generating the disposition effect), or after they‟ve found someone willing to sell the
stocks (generating speculation). Vice versa, when investors are willing to update their beliefs(due to stocks rising)
because it coordinates with their beliefs, speculations happen; and since people are more willing to accept
positive information, the winning stocks are sold at a higher rate than the losing stocks. Should one acknowledge
one‟s belief and be cautious of how it might influence future decision making, one could prevent financial loss
and possibly gain benefits.
1.3 Hypothesis Development
The guiding question to my research is: are investors more willing to accept information favoring or criticizing
their investment decisions? If so, such skewed information preference may well help explain the disposition
effect and excess speculative trade. My hypothesis is that investors are more inclined to accept information that
aligns with their previous investment decisions.
As prior theories have noted, information is valuable as it leads to better decision making, prevents mistakes, and
increases benefits. Thus, it seems irrational to actively avoid valid information. (Golman, Hagmann, &
Loewenstein, 2017). In the context of financial markets and investments, one should be open to and equally
interested in a broad spectrum of information because of the benefits. More specifically, investors should value
information criticizing their investments and favoring others. In doing so, investors can obtain a comprehensive
understanding of their opponents, change plans when necessary, and be prepared for challenges ahead. On the
other hand, positive information supports investors‟ decisions but doesn‟t provide much impact (whether positive
or negative) on their future profits. However, the theories fail to represent an accurate model. Investors‟ personal
portfolios are essential to their knowledge of loss and gain; therefore, investors should be checking their
portfolio regularly. Yet researchers have discovered that the frequency of checking their portfolios went down
(i.e. they avoided the information) when investors were experiencing financial downfalls. The confirmation bias
theory suggests that people are prone to review information supporting their own beliefs and tend to avoid
information that isn‟t consistent with their opinions. The research and the hypothesis are meant to find a
relationship between confirmation bias and investment decision makings.
In addition to the first hypothesis, a secondary hypothesis was formed: “After reading the article of their choice,
investors shift their investment plans in the direction of what they initially chose to invest in”. This hypothesis
would be a more direct test toward whether the participants indeed displayed confirmation bias, which is
investigated by the reconsideration question they were given.
2. Method
2.1 Participant Characteristics
I recruited participants of the age of 18 or older. There were no specific requirements for degrees or experience,
but a high school education was recommended for a clear understanding of the materials presented in the survey.
2.2 Sampling Procedures
2.2.1 Sample Size
We aimed for a total of 125 participants. The survey was created on Survey Monkey and was later released on
Amazon Mechanical Turk. The platform enabled a broad range of people with different backgrounds to respond
and thus supported a more generalized result. Participants indirectly assign themselves into 2 groups by
answering the questions.
2.2.2 Research Design and Measures
The purpose of this survey is to investigate the guiding question: are investors more willing to accept
ijef.ccsenet.org International Journal of Economics and Finance Vol. 11, No. 2; 2019
52
information consistent with their previous decision? The layout of the survey consists of two sections:
experimental questions and personal questions. Participant is presented with background information regarding
two kinds of mutual funds – actively managed funds and index funds. Actively managed funds are described as
risky and more rewarding and index funds as having a lower expense ratio and a careful investment. The
information is from Vanguard.com and contained three sections - goal, strategy, other things to consider - that
give an overall description of each kind of fund. Then the participant decides between these two funds as if they
are imagining making an investment of their own.
Next, participants are shown two headlines of articles regarding active management fund. The articles are
selected from seekingalpha.com. Both articles are similar in lengths, though one article had graphs representing
data mentioned in its contents; the articles are clearly supportive or oppositional based on headlines. 1st
Headline: The Case for Active Management. 2nd Headline: I hate “Active Management”. As participants
continue, the article they chose is shown, and they are given a chance to change their decision to invest more, no
change, and withdraw.
In the personal question section, questions are asked concerning age, gender, degree, occupation, experience in
economics and self-assessment of willingness to take risks. The answer to those questions were on a voluntary
basis. To prevent responder fatigue and for the designated purpose, individual workers (Turkers) can only do the
assignment (Human Intelligence Task) once. In addition, the reward for each participant to facilitate the
experiment is one dollar per submitted assignment. The assignment takes roughly ten minutes to complete.
The phenomena of interest (i.e. the willingness to read positive information) are measured by the number of
people who chose to read a certain article and their final decisions after reviewing the article.
2.2.3 Experimental Manipulation, Limitations, and Data Analysis Method
Response editing was turned off to ensure that choices are based on first-instinct for this experiment. However,
one limitation of my study is that I could not randomly assign participants to a treatment condition. The
treatment group and the control group are indirectly assigned: people who chose the active management fund are
in the treatment group (since they received information regarding their own investment) and people who invested
in index fund will be the control group (since they received information about an alternative investment option).
The subjects sort themselves into groups, and this could present a selection effect. That is, different kinds of
people choose to invest in actively managed funds and in index funds. It‟s possible that one of these differences
in the types of people in each condition could lead to any observed difference in information preference
behavior.
The data analysis is primarily focused on the choice made between two headlines. A mean is calculated to
measure the percent of people who chose to read “The Case for Active Management Fund” in invested in people
who originally invested in active management funds. Likewise, another mean was obtained for the index funds.
A 2-proportion z-test was also conducted for evaluating statistical significance.
3. Results
3.1 Recruitment
A total of 136 survey responses was received. Five responses were incomplete as they were not in the submitted
status; one responder didn‟t correctly enter his Mechanical Turk ID number as required; one responder stated his
age as 10, which was smaller than the criteria – an age of 18 or older. Disregarding those seven responses, we are
left with 129 valid responses. Since this was an online survey, the knowledge of the accuracy of the information
regarding personal information is scarce.
The maximum age of the sample is ninety-year-old and the minimum age is twenty-two. Participants have had
education ranging from high school to masters. Average value of the self-evaluation regarding risk-taking is 6.1
on a 10 point scale. Ninety-two participants are identified as male and 34 females, with 2 exception of gender
non-binary (or “Others”).
3.2 Statistics and Data Analysis
The total number of people who chose to invest in actively managed funds is 74, and the number of investments
for index funds is 55. 65 people both invested in active management funds and read the article “The Case for
Active Management”; 35 people that invested in index funds read the article “The Case for Active
Management”.
ijef.ccsenet.org International Journal of Economics and Finance Vol. 11, No. 2; 2019
53
Table 1. Results of investment decisions and article preferences
People who choose to read “The
Case for Active Management”
People who choose to read “I
Hate „Active Management‟”
Total number of people
People who invested in Active
Management Funds
65
(88% of total)
9
74
People who invested in index
funds
35
(64% of total)
20
55
Total
100
29
129
Bar Graphs of table above:
Figure 1. People invested in active management funds
Figure 2. People invested in index funds
The data table is of one degree of freedom. Two independent proportions was calculated by using the formula:
ratio = people who choose to read X / people who choose to invest in X. Considering the proportions, I decided
to use a 2 proportion Z-test to measure the statistical significance. The alpha value set for p is 0.01(p<0.01).
ijef.ccsenet.org International Journal of Economics and Finance Vol. 11, No. 2; 2019
54
Table 2. 2-Proportion Z-test for information preference
Proportions
Total Number of Participants
P-hat
Q-hat
Z-value
p
Active Management Funds Investors
0.8783784*
74
0.775
0.225
3.256
0.00171
Index Funds Investors
0.6363636*
55
*1. The proportion is obtained by 65/74.
*2. The proportion is obtained by 35/55.
After reading the article, participants were given a choice as to reconsider their investment decisions. The topics
shown in the first row represents the options the participants could see. The table shows their reconsidered
investments after reading.
Table 3. Results for Re-consideration question
No Change
Invest(more) in Active
Invest(more) in Index
Withdraw from Active
Withdraw from Index
Total
Invest in Active + Read
“The Case for AM”
12 (18.75%)
33. (51.56%)
14 (22.87%)
5 (7.81%)
0
64
Invest in Active + Read
“I Hate AM”
5 (50%)
0
4 (40%)
0
1 (10%)
10
Invest in Index + Reda
“The Case for AM”
11 (33.33%)
14 (42.42%)
3 (9.09%)
2 (6.06%)
3 (9.09%)
33
Invest in Index + Read
“I Hate AM”
15 (75%)
1 (5%)
4 (29%)
0
0
20
Note. Active = Active Management Funds; The Case for AM: The Case for Active Management; Index = Index Funds; I hate AM = I Hate
“Active Management Funds”; Total = Total number of people who answered the follow-up question.
3. Discussion
The Z-test gave a result of a p-value smaller than 0.01. That is, people who invested in active management funds
are indeed more likely to accept information favoring active management funds. According to the confirmation
bias and the information preference theory, people are inclined to accept information or beliefs that corresponds
with one‟s own belief and avoid information that is inconsistent. By having participants review materials
regarding the mutual funds and deciding on their own, a belief is formed (consciously or subconsciously) that
their investment is the “right” decision (the decision that will gain the most utility). Later, participants did
display behaviors of seeking for information that could be reassuring and confirming their previous judgment.
However, limitations in this research do exist. Due to software functions, the experiment was originally designed
to have 2 sets of headlines: one set had headlines both regarding the active management funds and the other set
had headlines concerning the index funds. We were not able to randomize the different sets to present to
participants, thus preventing us from confirming a causal relationship between confirmation bias and information
avoidance.
In regards to the reconsideration question after reading the article, one can see from the results that reading tends
to move people toward active management (probably due to that most people read the article favoring active
management), but the move toward active management is not stronger for those who initially invested in active
management than for those who initially invested in an index fund. Therefore, the data fails to support the
secondary hypothesis. Evidence for information preference that would be consistent with confirmation bias is
present, but there is no direct evidence for confirmation bias itself.
4. Conclusion
There is a strong relationship between the investment participants made and their article of choice. Previous
studies showed that, like the information avoidance theory (especially the misinterpretation and inattention to
information), investors will evaluate an information as more reliable and authentic if the information conforms to
their prior beliefs. (Park, Konana, Gu, Kumar, & Raghunathan, 2013) My research presents an intrinsic
relationship between decision making and receiving information. Furthermore, the relationship serves as a
warning to those who try to stay objective. Future research could be dedicated to finding the causation for the
information avoidance phenomena displayed in investors and thus helping people prevent financial loss.
In retrospect, the result indirectly responds to the speculation paradox previously introduced. With the benefit of
ijef.ccsenet.org International Journal of Economics and Finance Vol. 11, No. 2; 2019
55
knowing that people do display this skewed information preference toward what they already believed, we can
see that traders are more likely to react to positive information (i.e. sell the stocks) and avoid accepting
information that suggests signs of downfall. While this imbalance between the information giver and the receiver
exists, speculation happen to an extent of excess trade.
Another motivation for this research is its policy implications. To ensure the openness of information and to
refrain from sudden market changes because of inside information, governments have regulations that require the
frequency and quantity of company disclosures. One would suspect that disclosures are regarded by people as
valuable information and that people would read it often enough, for information means profit in the markets.
However, evidence suggests otherwise. Fewer than 3% of people read privacy disclosures (Jensen, Potts, &
Jensen, 2005. Privacy practices of Internet users: self-reports versus observed behavior.) For all the effort to
ensure the openness of information, these policies are not as effective as we‟d hoped. One possible reason could
be that people are reluctant to review disclosures because they are afraid that the information might go against
their beliefs (e.g. A disclosure might imply that the company is experiencing downfalls and thus influence the
stock markets). If we were to find that people indeed display confirmation bias when dealing with disclosures,
we could provide an explanation for why the policy doesn‟t perform well.
Acknowledgements
I am grateful for Professor Russell Golman who provided expertise and clear guidance throughout the research
process. I would also like to show my gratitude to Eureka Research Program for the opportunity of this research
and the comments that greatly improved my paper.
References
Golman, R., Hagmann, D., & Loewenstein, G. (2017). Information avoidance. Journal of Economic Literature,
55(1), 96-135. https://doi.org/10.1257/jel.20151245
Golman, R., Loewenstein, G., Moene, K. O., & Zarri, L. (2016). The preference for belief consonance. The
Journal of Economic Perspectives, 30(3), 165-187. https://doi.org/10.1257/jep.30.3.165
Jensen, C., Potts, C., & Jensen, C. (2005). Privacy practices of internet users: Self-reports versus observed
behavior. Int. J. Hum. Comput. Stud., 63(1-2), 203-227. https://doi.org/10.1016/j.ijhcs.2005.04.019
Milgrom, P., & Stokey, N. (1982). Information, trade and common knowledge. Journal of Economic Theory,
26(1), 17-27. https://doi.org/10.1016/0022-0531(82)90046-1
Odean, T. (1998). Are investors reluctant to realize their losses? The Journal of Finance, 53(5), 1775-1798.
https://doi.org/10.1111/0022-1082.00072
Odean, T. (1999). Do investors trade too much? The American Economic Review, 89(5), 1279-1298.
https://doi.org/10.1257/aer.89.5.1279
Park, J., Konana, P., Gu, B., Kumar, A., & Raghunathan, R. (2013). Information valuation and confirmation bias
in virtual communities: Evidence from stock message boards. Information Systems Research, 24(4),
1050-1067. https://doi.org/10.1287/isre.2013.0492
Copyrights
Copyright for this article is retained by the author(s), with first publication rights granted to the journal.
This is an open-access article distributed under the terms and conditions of the Creative Commons Attribution
license (http://creativecommons.org/licenses/by/4.0/).
