You are on page 1of 60

Behavioural Finance

Jindal School of Banking and Finance


April-May, 2024

Lecture 2: Heuristics and Bias


Amlan Das Gupta
Heuristics and Bias
 Often we have to take decisions based on an estimate.

 We usually use heuristics to make these estimates.

 Often these heuristics make us prone to systematic errors.


Problems that we struggle with
 Estimating with limited and unreliable information.

 What is the probability that a particular element belongs


to a particular category?

 What is the probability that A caused B?

 How can we predict something based on some other but


related information?
Limited Sample Size
Law of large numbers
 Bigger the sample size, closer is the sample mean to the population mean - - Law of
large numbers.

 Imagine trying to test firecrackers to find the proportion of duds. There are 10000
boxes.

 Works even if there are unlimited boxes.


Implication
 Variability is high from estimates derived from small samples.

 Or in other words, extreme predictions are more common.

 Consider a box where there are 10 red and 10 black balls and 4 are
drawn at random with replacement.
 A study of the incidence of kidney cancer in the 3,141 counties of the
United States reveals a remarkable pattern. The counties in which the
incidence of kidney cancer is lowest are mostly rural, sparsely populated,
and located in traditionally Republican states in the Midwest, the South,
and the West.

 What do you make of this?


Also consider:

 The counties in which the incidence of kidney cancer is highest. These ailing
counties tend to be mostly rural, sparsely populated, and located in
traditionally Republican states in the Midwest, the South, and the West.

 How do you reconcile the two?


Issues with experiments

 We tend to believe results from small samples:


In a telephonic poll involving 500 individuals 75% women said pink was their favourite colour

 Is this believable?
 What are the issues?
Law of small numbers

 Unfortunately even trained statisticians will make this mistake.


 The main problem lies with our system 2 which recognises doubt and has to
spend energy retaining it.
 So unless immediately falsified we may start believing these thing.
Cause vs Chance
 Consider the question: If in a coin toss game you have had 5 heads in a
row, is it more likely that you will get a tail in the next one?

 Another question: Consider the urn example from the last slide.
Suppose we don’t know the distribution inside. Following are three
draws:
RRBBB
BRRBB
RRRRR

Which one do you think is random?


Ideas about Randomness
 We often assume erroneously that certain patterns are incompatible with randomness.

 In such cases we invent stories to justify some sort of cause.

 May have had evolutionary advantages.

 Also, responsible for major biases. Belief that some people are superior.
Map of London showing V-1 rocket strikes
(Adapted from Gilovich [1991])

 People believed that areas to


escape from bombing were home
to German spies.
 Later it would be confirmed that
these patterns are consistent
with randomness.
Anchoring
Anchoring
 Suppose I tell you that the population of New York was 85 lakhs
in 2018

 Now guess the population of Washington D.C in 2019.

 Your answer will be influenced by the number above even if you


are not aware of it.
Anchoring
 The most robust and well documented psychological phenomenon in
behavioural economics.

 Anchoring is present even if the anchor is completely uninformative.

 Example: If a number is generated from a roulette wheel it still


influences an answer to the following:

What percentage of UN countries are from Africa?


Anchoring Index
 Anchoring is a phenomenon that is not just demonstrable but also
quantifiable.

 Consider the height of the Eiffel tower.

 Is it greater than 200 feet?


 Is it less than 1000 feet?

 What is the actual height?


Abuse of Anchors.
 Anchoring is powerful, so it is liable to be misused. – Judges, realtors.

 Arbitrary rationing in supermarkets, first offers in negotiations – are misuse.

 Suggested donations or tip amounts could be good.

 Trick is to argue against the anchor. “think opposite”.


Availability
Availability
 This is applicable when we are asked to estimate the size or
frequency of a category.

 Example: You judge a place by the first visit experience.

 We define the availability heuristic as the process of judging


frequency by “the ease with which instances come to mind.”
Ease vs actual number

 When we judge the ease of remembering cases is it


necessary to remember a certain number?
 The answer is NO

 Which set of letters would make the higher number of


words:
 XUZONLCJM
 TAPCERHOB
Availability and Bias
 Any factor which increases the chance of that category coming to
mind will lead to bias.

 A rude person you have met today will increase your estimate of
the proportion of rude people.

 Dramatic events.
 Personal experiences.
 We need constant awareness to fight such biases and it takes
effort.
Availability and the systems
 Availability is a result of a combination of system 1 and 2.

 System 1 searches for memories related to the category and makes an


impression of how easy that is.

 However, you may reverse its judgement if you are able to surprise it.
Experiment
List X ways in which you have been assertive in the last week.

Now give an estimate of how assertive you are.

A second set were told to remember X instances where they were not
assertive.

Then asked to give an estimate of how assertive they were.


 People who were asked to remember less, concluded that they were
assertive.

 Those who had to remember more tend to conclude the opposite.

 “unexplained unavailability”: The mechanism is that system 1’s


expectation is getting challenged.

 This does not work if you are able to explain the challenge.
Availability and risk perception

 The availability bias is very important for our perception of risk and
the business of insurance.
 After any disaster we rush to take precautions, but our zeal wanes
with time.
 Also, we find it difficult to envisage disasters that are unprecedented
– e.g. climate change.
Availability and the Affect Heuristic

Paul Slovic’s finding: Ideas that affect us more emotionally will be perceived as
more risky.

 Strokes cause almost twice as many deaths as all accidents combined, but
80% of respondents judged accidental death to be more likely.
 Tornadoes were seen as more frequent killers than asthma, although the
latter cause 20 times more deaths.
Risk perception: Public vs Experts

 Slovic argues that there is no such thing as objective measure of risk.


 Using the expert’s opinions on risk perception ignore the emotional cues
general people associate with different outcomes.
 As such description of risk depends on power!
Sunstein’s view

 Sunstein believes that often public policy is dictated by biases caused due to
availability.
 Experts can help prevent loss of resources and lives by being objective.
 Describes a phenomenon where availability can lead us to a cascade.
Representativeness
Sirohan is a student. Rank the following schools in
order of your estimated probability that Sirohan is a
student there.

 JGU
 Delhi University
 JNU
 IIT Delhi
 AIIMS
Next redo the ranking after reading the description below
of Sirohan written after observing him

He is of high intelligence, although lacking in true creativity.


He has a need for order and clarity, and for neat and tidy systems
in which every detail finds its appropriate place. His writing is
rather dull and mechanical, occasionally enlivened by somewhat
corny puns and flashes of imagination of the sci-fi type. He has a
strong drive for competence. He seems to have little feel and little
sympathy for other people, and does not enjoy interacting with
others. Self-centered, he nonetheless has a deep moral sense.
Predicting by representativeness

 This process usually ignores base rates and the reliability of the information.
 Mostly we look at how well the description is representative of the
stereotypes we carry for the different universities.
Predicting by representativeness

 Our systems 1 and 2 combine to answer probabilistic questions

 The prospect is first visualized by system 1, and then system 2 steps in to see
how well the vision matches with our memories.

 Usually this process yields good results.


Problems with representative
predictions
 We tend to neglect base rates and predict extreme outcome more often.
 Often we use information with less predictive power.
 Sometimes the validity of the information we use is also questionable.

 Classic case: Movie Moneyball.


Predictive power and validity
Imagine you are traveling by train and your co-passenger is reading
thinking fast and slow. Which of the following are you likely to predict:

1. The person has a graduate degree.


2. The person has no graduate degree.

Which would you choose?


Disciplining your belief

 We need to make sure our beliefs follow the logic of probability.


 If think P(we are right)=0.4 then P(we are wrong) needs to be 0.6.
 We cannot think that P(we are right in the afternoon) will be 0.5.
 If we are looking at our belief about two things happening together then it
has to be less than either.
 If we think of our belief about two things and either can happen then the sum
rule has to be used.
Regular six-sided die with four green faces and two
red faces, which would be rolled 20 times. Which
combination would you choose to bet on.
 RGRRR
 GRGRRR
 GRRRRR
Base rate neglect.

 Consider the question: What is the probability that X, is an


Indian, plays basket ball?

 Next consider: What is the probability that X is an Indian


male, who is tall, interested in sports and plays basket
ball?
All Indians All basket ball players who are Indians

Male, athletic …etc.


 When we narrow our definition we are actually evaluating two events
happening jointly.
 Being male and a basketball player.
 Different from the conditional probability of being a basketball player
given you are male.
 Even in the second case the base rates are important
Visualization helps
How do we change a belief?
Application of Bayesian updating
 Consider the following:
 A cab was involved in a hit-and-run accident at night.
 85% of the cabs in the city are Green and 15% are Blue.
 A witness identified the cab as Blue.
 The witness identifies the colour of the cab correctly 80% of the time.

What is the probability it was a blue cab?


Calculating the true probability
 Without any witness our base rates are our best guess.
 Given that the witness identified a blue cab. We revise the estimate.
 Probability that it was a blue cab
If we change it a bit
 Consider :
 A cab was involved in a hit-and-run accident at night.
 85% of the accidents in the city are caused by Green cabs and 15% by Blue.
 A witness identified the cab as Blue.
 The witness identifies the colour of the cab correctly 80% of the time.
Statistical base rate vs Causal base rates

 Statistical base rates are usually ignored.

 Causal base rates are like stereotypes and they are often used as data.

 Stereotypes are statements about a group that is taken as true for all
members.
Regression to the mean
Regression
Consider the following story from an experienced flight instructor:

“On many occasions I have praised flight cadets for clean execution of
some aerobatic manoeuvre. The next time they try the same manoeuvre
they usually do worse. On the other hand, I have often screamed into a
cadet’s earphone for bad execution, and in general he does better on his
next try. So please don’t tell us that reward works and punishment does
not, because the opposite is the case.”
Regression to the mean
 Francis Galton observed that offspring of taller parents were not as tall and tended to revert to
some mean.

 This was the origin of the linear regression technique.


Galton’s observation about seeds from
plants:
They yielded results that seemed very noteworthy, and I used
them as the basis of a lecture before the Royal Institution on
February 9th, 1877. It appeared from these experiments that the
offspring did not tend to resemble their parent seeds in size, but
to be always more mediocre than they—to be smaller than the
parents, if the parents were large; to be larger than the parents, if
the parents were very small…The experiments showed further
that the mean filial regression towards mediocrity was directly
proportional to the parental deviation from it.
 Any outcome will be a combination of genuine factors and luck

 Any exceptional good performance is likely to be followed by worse and vice versa.

 Eg: Commentators curse, jinxing someone, hot streak in basket ball, debut
centurions in cricket.
Regression with two variables

 Suppose we are interested in predicting some variable on the basis of another.


(Son’s height by father’s height)
 Then mean reversion will happen at every level of father’s height.
 Connecting the average of all son’s heights for a particular value of father’s
height helps us get to the regression line between the two.
When does mean reversion not happen?
 Whenever there is randomness in the outcome mean reversion will happen.

 Correlation coefficient between two measures, which varies between 0 and 1, is


a measure of the relative weight of the factors they share.

 So if there is perfect correlation between two variables there is no mean


reversion.
Example:

The bigger is the role of the common factor, lesser is the reversion to mean.
Intuitive prediction and regression based
predictions
 Consider the following questions:
 If my father’s height is 1.85 m how tall am I?
 If a party has won 65% votes in an election how much are they going to win in
the next?
 If Pilu as 4 year old could ride a bicycle will she become an Olympic athlete at
the age of 25?
How we predict?

 We evaluate how connected the two variables are.


 We make a ranking of how precocious the first observation is.
 We translate the rank to a scale of the second variable.
 We give them the same rank in terms of the second variable as we have in the first.
Using regression

 Start from your baseline prediction towards your intuitive prediction.


 Adjust only to the extent that you think there is correlation between the two
outcomes.

You might also like