Professional Documents
Culture Documents
Financial Market
▪ If you are taking someone out to a dinner to close a business deal, the
general etiquette is to avoid the MOST and LEAST expensive wines
on the list, choosing a bottle of wine in the middle of the list
This requires that people should be able to update their beliefs correctly
given a series of prior outcomes.
P( E H ) P( H )
Bayes’ Theorem: P( H E ) =
P( E )
Example:
Suppose that John and Wes are arguing about whether a coin brought to
class by a student has two heads or whether it is fair. Imagine that there are
no other possibilities. For whatever reason, the student will not let them
inspect the coin, but she will allow them to observe the outcome of coin flips.
Let H be the hypothesis that the coin has two heads, so that ¬ H means that
the coin is fair. Let us consider John first. He thinks the coin is unlikely to
have two heads: his prior probability, Pr (H), is only 0.01. Now suppose the
student flips the coin, and that it comes up heads. Let E mean “The coin
comes up heads.”
Q. What probability should John assign to H given that E is true?
Bayesian Updating
Example:
Pr (H) = 0.01
Q. What probability should John assign to H given that E is true?
➔ What is P(H|E), which is the posterior probability?
➔ P(not H) = 0.99
➔ P(E|H) = 1, P(E|not H) = 0.5
𝑃 𝐸 𝐻 ∗𝑃 𝐻 1∗0.01
➔𝑃 𝐻𝐸 = = =
𝑃 𝐸𝐻 ∗𝑃 𝐻 + 𝑃 𝐸 𝑛𝑜𝑡 𝐻 ∗𝑃 𝑛𝑜𝑡 𝐻 1∗0.01 + 0.5∗0.99
0.0198 ≅ 0.02
Bayesian Updating
Example:
You are conducting an exit poll at the 2016 U.S. presidential election.
Prior to voting, 50% of people say they would vote for Hillary Clinton, with
the rest 50% supporting Donald Trump.
However, Trump supporters tend not to reveal their preferences so openly,
so they lie 20% of the time at exit poll and say they voted for Clinton. Clinton
supporters, on the other hand, lie only 5% of the time, correctly saying they
voted for Clinton 95% of the time.
Q. If someone tells you at the exit poll that he/she voted for Clinton, what is
the probability that he actually voted for Clinton at polling booth?
Bayesian Updating
Example:
H: The person actually voted for Clinton at the polling booth.
E: The person says they voted for Clinton at the exit poll.
P(H) = Probability of actually voting for Clinton = 0.5, P(not H) = 0.5
P(E|not H) = 0.2, P(E|H) = 0.95
Q. If someone tells you at the exit poll that he/she voted for Clinton, what is
the probability that he actually voted for Clinton at polling booth?
➔ What is P(H|E), which is the posterior probability?
𝑃 𝐸 𝐻 ∗𝑃 𝐻 0.95∗0.5
➔𝑃 𝐻𝐸 = = =
𝑃 𝐸𝐻 ∗𝑃 𝐻 + 𝑃 𝐸 𝑛𝑜𝑡 𝐻 ∗𝑃 𝑛𝑜𝑡 𝐻 0.95∗0.5 + 0.2∗0.5
0.8261 ≅ 0.826
Deviations (1): Availability Heuristic
▪ But finding a word with K on the third letter takes more effort!
take, fake, make, acknowledge……….
Example:
Your optometrist tells you that your new contacts are so good that you can
wear them night and day for up to 30 days straight. She warns you that there
is some chance that you will develop serious problems but says that she has
seen many clients and that the probability is low. After a week, you
develop problems and see an ophthalmologist. The ophthalmologist tells
you that he is the doctor who removes people’s eyes when they have been
injured as a result of improper contact use. He tells you that the probability
of developing serious problems is high.
Q. Explain how they can report such different views about the likelihood of
developing serious problems as a result of wearing contacts.
Deviations (1): Availability Heuristic
Example:
Q. Explain how they can report such different views about the likelihood of
developing serious problems as a result of wearing contacts.
➔ Given that the optometrist mainly sees contact users without problems,
an image of a healthy user is most available to her. Given that the
ophthalmologist mainly sees users with problems, an image of an
unhealthy user is most available to him.
➔ The two are prone to the availability bias, the optometrist is likely to
underestimate, and the ophthalmologist to overestimate, the probability
of developing serious problems as a results of wearing contacts.
Deviations (1): Availability Heuristic
• Housing market bubble: Prior to the recent financial crisis, investors were
influenced by stories of others who had made significant profits by
flipping homes or investing in real estate, leading to a dramatic surge in
housing demand and overvalued properties
Deviations (2): Representativeness Heuristic
Yeji is 27 years old. Her father is a renowned church minister famous for
handing out free meal to homeless people in Yongsan area. Yeji has always
been deeply concerned about social injustice in Korea, so she took a degree
in social welfare at university, where she was also involved in the hunger
awareness movement and wrote more than 200 personal letters to the
president about it.
Q. Out of the two scenarios below, which is more likely?
(A) Yeji works for Samsung.
(B) Yeji works for Samsung and is an active charity volunteer.
Deviations (2): Representativeness Heuristic
▪ The examples deal with a situation where the evidence is observed with
a high probability when your initial hypothesis is right.
Example 1.
H: The woman has breast cancer.
E: The mammogram test is positive.
P(H) = Probability of actually having breast cancer = 0.01, P(¬H) = 0.99
P(E|H) = Probability of testing positive given that the woman has breast
cancer = 90% = 0.90
P(E|¬H) = Probability of testing positive given that the woman does not
have breast cancer = 10% = 0.10
𝑃 𝐸 𝐻 ∗𝑃 𝐻 0.9∗0.01
➔ 𝑃 𝐻𝐸 = = =
𝑃 𝐸𝐻 ∗𝑃 𝐻 + 𝑃 𝐸 𝑛𝑜𝑡 𝐻 ∗𝑃 𝑛𝑜𝑡 𝐻 0.9∗0.01 + 0.1∗0.99
0.08333 ≅ 8%
Deviations (3): Base Rate Fallacy (기저율 오류)
Example 2.
H: The patient has the MERS virus.
E: The test is positive.
P(H) = Probability of actually having MERS = 0.001, P(¬H) = 0.999
P(E|H) = Probability of testing positive given that you have the MERS = 1
P(E|¬H) = Probability of testing positive given that you do not have the
MERS = 5% = 0.05
𝑃 𝐸 𝐻 ∗𝑃 𝐻 1∗0.001
➔ 𝑃 𝐻𝐸 = = =
𝑃 𝐸𝐻 ∗𝑃 𝐻 + 𝑃 𝐸 𝑛𝑜𝑡 𝐻 ∗𝑃 𝑛𝑜𝑡 𝐻 1∗0.001 + 0.05∗0.999
0.0196 ≅ 1.96%
Deviations (3): Base Rate Fallacy (기저율 오류)
Example 3.
H: The cab involved in the accident was Blue.
E: The witness identified the cab as Blue.
P(H) = Probability that a cab involved in the accident is Blue = 0.15, P(¬H) =
0.85
P(E|H) = Probability that the witness correctly identifies a Blue cab as Blue =
0.80
P(E|¬H) = Probability that the witness incorrectly identifies a Green cab as
Blue = 0.20
𝑃 𝐸 𝐻 ∗𝑃 𝐻 0.80∗0.15
➔ 𝑃 𝐻𝐸 = = =
𝑃 𝐸𝐻 ∗𝑃 𝐻 + 𝑃 𝐸 𝑛𝑜𝑡 𝐻 ∗𝑃 𝑛𝑜𝑡 𝐻 0.80∗0.15 + 0.20∗0.85
0.414 ≅ 41.4%
Deviations (4): Gambler’s Fallacy
▪ Suppose you tossed a coin three times and you get Head, Head, Head.
What is the probability of getting a Head on the fourth toss?
Deviations (4): Gambler’s Fallacy
▪ Suppose you tossed a coin three times and you get Head, Head, Head.
What is the probability of getting a Head on the fourth toss?
▪ Still 0.5!
▪ Yet, human minds often think “it’s about time a Tail came up.”
▪ This is like drawing a ball from an urn of finite size (without replacing).
If an urn contains 5 “Heads” and 5 “Tails,” the probability of getting
a Head on the fourth toss becomes 2/7. or 28.6%.
→ Human mind is more comfortable with this idea.
Deviations (4): Gambler’s Fallacy
The “hot hand effect” is the seeming opposite of the gambler’s fallacy.
▪ If a player has made two hits in his first two at-bats, fans and colleagues
believe he is having a “hot” day and more likely to make a hit on his third.
▪ While the gambler’s fallacy and the hot hand effect appear difficult to
reconcile, Barberis, Shleifer and Vishny (1998) show they are both at
work in stock markets.
▪ But after a while, they begin to expect a “hot hand” regime, overinvesting
in the momentum stock and causing it to be overpriced (reversal).
Deviations (4 & 5): Synthesis
Q2. Do you think you are better than average when in a karaoke?
▪ People tend to attribute their successes in life to their own ability or skill,
but blame the situation or other people when they fail.
When nothing happened on Oct. 28, 1992, they didn’t change their beliefs.
They instead claimed their “prayer” has been answered.
Self-Evaluation Bias
Typical example:
8 (considered lucky) is third most popular as a last digit on the limit order,
but 4 (symbol of death) is least popular.
Causes of Irrationality