Consumer Behavior

Class 2: Research Methods

Prof. Ania Jaroszewicz
UC San Diego
Announcements
Survey Week 1 link will be posted on Canvas this evening
Due this Friday by 11:59pm
Make sure you hit “submit” at the end

#FinAid survey due 1/17

Office hours: Wednesdays 11-12, Wells Fargo Hall 3W123

2
Last time

3
 Why do consumers
prefer this product
over that one?

Do children
Will I get more
consume
sales if I use
Flintstone
Ad A or Ad B?
gummies
safely?
4
Last time

Relying on intuition may lead you astray.

Better to use data,
i.e. evaluate or conduct research.

5
 Today: research methods!
One of the most important classes. Why?

1 One of the goals of this course is not only understanding

what CB findings are
But also how researchers got there
→ Interpret and critique existing CB research, decide how
much to trust it, etc.

6
2 Sometimes, there is no existing information on your
specific question, e.g. Your Ad A vs. B
Learning research methods and skills allows you to
produce the knowledge yourself
• Non-profit or for-profit firm
• Marketing consulting agency
• Government agency
• Academia

7
Agenda
Correlation & causation

Experiments

Measuring variables
8
Agenda
Correlation & causation

Experiments

Measuring variables
9
Example 1 Suppose you get data on
UCSD students’
(1) exercise; and (2) GPA.
GPA
You find there’s a positive
relationship.

What are possible

explanations?
Take 1 min
to talk to a partner
Exercise
11
What can you conclude?

Exercising Good grades Something else

improves grades increase exercise increases both (e.g.,
(e.g., need less time wealthier students
(e.g., increased can afford tutors
to study and spend
oxygen flow) and gym
leisure time
exercising) membership)

12
Seems reasonable…
Let’s look at some other examples

13
Example 2

Giving out
Master’s
degrees
increases
box office
revenue?

Source: Statology
 Example 3
Kids who get tutoring have worse grades than those who
don’t get tutoring. Tutoring → worse grades?

15
Example 4 Staying married in Maine
decreases interest in margarine??
What’s going on here?
These data are showing correlations – not necessarily
causation!

Ex. 2: An alternative explanation is that the population is

increasing over time
Ex. 3: Kids who are struggling (have worse grades) seek
out tutoring
Ex. 4: One thing is likely decreasing divorce rates, and
something else is decreasing margarine consumption 17
Correlation ≠ Causation
Correlation = relationship between two variables
Causation = one variable producing an effect on another variable

18
Correlation Causation Cause or
“independent
variable”
How is How do changes
1 variable in 1 variable
associated with change
another variable? another variable?

Effect or
“dependent
variable”
19
Returning to exercise and grades…

How can you tell if

exercise increases grades
grades increase exercise
or something else increases both?

I.e., how can you tell if it’s a causal relationship (and if so,
the direction of causality)?

20
Let’s suppose we think exercise improves grades

Exercise = the cause or “independent variable” (IV)

→ the feature you think affects something else

Grades = the effect or “dependent variable” (DV)

→ the feature you think depends on the IV

Hypothesis: exercising more improves grades

(Note the directionality: “improves” rather than “changes”)
21
The information you’d ideally want…

You’d take a person

You’d prevent her

from exercising

Then measure her

grades at the end
of the quarter

22
 You’d take a person

You would go back in time…

You’d prevent her
from exercising

Then measure her

grades at the end
of the quarter

23
You’d take a person You’d take the same person

You’d prevent her You’d make sure she

from exercising exercised regularly during
the same quarter
Then measure her
grades at the end Then measure her grades at
of the quarter the end of the quarter

24
You’d take a person You’d take the same person

You’d prevent her You’d make sure she

from exercising exercised regularly during
the same quarter
Then measure her
grades at the end Then measure her grades at
of the quarter the end of the quarter

Compare grades 25
Sure, but…

Time
machines
are hard to
come by.

26
 What’s the next best thing?
An experiment!
aka. “A/B test” or “randomized controlled trial”

Instead of going back in time and

having the SAME person exercise and
not exercise…
Take DIFFERENT people and, over
the same time period, either have them
exercise or not exercise
27
Agenda
Correlational data

Experiments

Measuring variables
28
Experiments
Purpose of an experiment: establish causality.
How to do this:
1. Randomly assign people* to “control” or “treatment” group
2. Give each group a different level/value of the “independent
variable” (IV).
• Typically: Control group gets nothing or the status quo; treatment group
gets the thing you think will change the dependent variable
• Don’t change anything except for the IV
3. Measure the “dependent variable” (DV)
4. Compare the DV values of the 2 groups

*It could also be locations, days, etc. – see cookie example 29

The logic behind experiments
Random assignment to different groups (which get
different IV levels/values) ensures that, on average, the
groups are similar in all ways except for the IV
• E.g., similar gender breakdown, health, wealth…

So if you see differences in the DV, you know that those

differences must be because of the IV

30
1. Take a 2. For each 3. All heads are the 4. Give the 5. Measure 6. Compare
group of person, flip a “Control” group; all Control nothing / the DV for the DV (e.g.
people coin to tails are the status quo; give the both groups average) for
randomize “Treatment” group Treatment the in the same the 2
change way groups
 B A ➔ Exercise increases grades

B C ➔ Exercise decreases grades

B B ➔ No relationship

(Glossing over the entire field of statistics…) 32

Squirrel
example
• Suppose you are a manager at
a nut company.
• You want to know whether
sending a free gift of a squirrel
plushy to your customers
makes them more likely to
leave a positive review.
33
IV? Whether you send a squirrel plushy gift

Whether the customer leaves a positive

DV? review

Sending customers free squirrel plushies

Hypothesis? will increase the likelihood that they leave
positive reviews
34
1. Take a 2. For each 3. All heads are the 4. Give the 5. Measure 6. Compare
group of person, flip a “Control” group; all Control nothing / the DV for the DV (e.g.
people coin to tails are the status quo; give the both groups average) for
randomize “Treatment” group Treatment the in the same the 2
change way groups
Cookie smell example
Imagine you just opened a
café.

Would customers like your

café more if it smelled like
fresh cookies?

36
IV? Presence/absence of cookie smell

DV? How much customers like your café

Hypothesis? Maybe: more cookie smell makes

customers like your café more

37
1. Take all 2. For each 3. All days with heads 4. On Control 5. Measure 6. Compare
the days in day, flip a are the “Control” days, do nothing; the DV for the DV (e.g.
a month coin to group; all tails are the on Treatment days, both groups average) for
randomize “Treatment” group spray the scent in the same the 2
way groups
 Does playing scary background music in shark
Shark videos (vs. silence) make people more afraid of
sharks?
example
Take 2 mins to discuss with a partner: identify the
IV, DV, hypothesis, experimental design

39 39
 IV? DV?

Hypothesis? Experiment?

40
 IV? DV?
Video background music Video viewers’ fear
(scary or none) of sharks

Hypothesis? Experiment?
Randomize music in video
Scary music → more fear, (scary or none), measure
relative to silence fear, compare across groups
41
Just measure
the DV

Okay… how?
43
Agenda
Correlational data

Experiments

Measuring variables
44
Measuring the DV
Main categories:
1) Open-ended questions Qualitative (words)
2) Survey scales
3) Observing behavior Quantitative (numbers)
4) Biological markers

45
Measuring the DV
• All can be used in
Main categories: experiments (though
1) Open-ended questions quant. approaches are
better)
2) Survey scales • Each has strengths and
3) Observing behavior weaknesses
4) Biological markers • Best to combine
categories in
complementary ways

46
Measuring the DV
• All can be used in
Main categories: experiments (though
1) Open-ended questions quant. approaches are
better)
2) Survey scales • Each has strengths and
3) Observing behavior weaknesses
4) Biological markers • Best to combine
categories in
complementary ways

47
Open-ended
questions
Participants (consumers) use
words to describe experiences,
feelings, and reactions
Often face-to-face
Types:
• “In-depth interview” (1 consumer, 1+ researchers)
• “Focus group” (many consumers, 1+ researchers)
48
Open-ended
questions
The objective: explore the
subject matter in detail

Sometimes very broad

(“What thoughts come to
mind as you look at this
bottle of vitamins?”)
49
 Open-ended questions
Advantages: Disadvantages:
• Can go deep • Hard to analyze statistically; less
reliable for causal inference
• Can adapt questions based on
answers • Expensive; small # of participants
• Good for generating ideas / • Presence of researcher can
influence responses
new hypotheses
• Focus groups: influence of other
• Good for nuanced topics group members
• Focus groups: Can see how • Question wording may influence
people respond in social responses*
context • Responses may not reflect reality*

*See survey scale slides 51

Measuring the DV
Main categories:
1) Open-ended questions
2) Survey scales
3) Observing behavior
4) Biological markers

52
Survey
scales

53
Survey scales
Questions with a fixed set of response options the
consumer can pick from
Can be asked:
• On paper
• Online
• Phone
• Face to face
• Physical screen
54
 Survey scales
Advantages:
• More controlled
• Quick
• Inexpensive; can get many
participants
• Easy to quantify outcomes /
analyze

55
 Challenges of survey scales
1 How you ask the question affects the response
• Goal: neutral, easy-to-understand, unambiguous questions
Bad Better
Neutrality How awful is our competitor’s product? How would you rate our competitor’s
product?
Understandability Kindly expound upon and quantitatively Please rate your customer service
appraise the multifaceted dimensions of experience.
your subjective experiential interaction
with our client-oriented facilitation
personnel.
Ambiguity How satisfied are you with the food, How satisfied are you with each of the
service, and ambience? (1-5) following:
Food (1-5)
Service (1-5)
Ambience (1-5)
56
 Challenges of survey scales
2 Even if the question is asked well, people’s
responses may not reflect reality

57
Did they
tell the
truth?

58
How often do you go to the gym?

59
5≠2
Why did Gequan say “5”?

Take 1 minute to brainstorm some reasons with a partner

60
5≠2
Why did Gequan say “5”?

1) Misremembered?
2) Reported the goal, not the reality?
3) Wanted to sound impressive?

61
Let’s scale it up
Imagine you ask 100
people how often they
go to the gym and take
the average.

How would the self-

reports of gym
attendance compare to
actual attendance?
62
Some possibilities…
If people are only misremembering, individual people
would be wrong, but as a group they’d be right
Some would say 6/week when it’s 4
Some would say 2/week when it’s 4
→ high “variance,” but correct on average
If people are reporting goals or want to sound impressive,
they would generally report higher numbers than the truth
E.g., they’d say 6/week when it’s 4
→ high “bias” (not correct on average) 63
 What we 2nd best
want

Not good,
unless you
know the What we
direction fear
and amount
of bias
64
Survey questions often suffer from bias, variance, or both
Sometimes we can predict the direction of the bias
Examples:
• For gym attendance: bias upward
• Doctor asks about alcohol consumption: bias downward
• Professor asks about studying time: bias upward

65
Sometimes we don’t
even know the direction
of the bias
E.g., “how much do you
typically sleep?”
• May bias upward
because you know you
should sleep 8+ hrs
• May bias downward to
highlight busyness or
partying
66
 Survey scales
Advantages: Disadvantages:
• More controlled • Question wording may
• Quick influence responses
• Inexpensive • Responses may not reflect
reality due to variance, bias,
• Large sample or both
• Easy to quantify outcomes /
analyze Note: these are also
disadvantages of open-
ended questions!
67
Measuring the DV
Main categories:
1) Open-ended questions
2) Survey scales
3) Observing behavior
4) Biological markers

68
Observing behavior
Watching people’s behavior and recording it in an
objective way
People may or may not know that they are
being observed
Observation can be passive (e.g., gym card swipes)
Setting can be natural (e.g., store, home)
or artificial (e.g., in a lab)

69
Observing behavior in a natural setting
Company that markets Lee and
Wrangler jeans noticed women
trying on many pairs of jeans
The women felt they couldn’t
trust the size labels
Led to a new campaign to help
women choose best jean size
and style
70
Can be planted in a lab setting

71
Observing behavior
Advantages: Disadvantages:
• Real settings • Can be expensive /
• Less likely to be distorted logistically challenging
than survey responses [ex:
gym attendance]
• Captures the outcome that
arguably matters the most

72
Measuring the DV
Main categories:
1) Open-ended questions
2) Survey scales
3) Observing behavior
4) Biological markers

73
Biological markers
Biological measures of a biological state
Examples:
• Eye tracking
• Facial expressions
• Cortisol levels (stress)
• Sweat on your palms
• Heart rate
• Neurological activity
74
 Eye
tracking

75
Eye-tracking and facial coding

76
Biological markers
Advantages: Disadvantages:
• Very low possibility of • Super expensive
distortion • Can typically only be done in
• Can uncover insights labs (artificial settings)
consumers have no conscious • Can typically only be done
knowledge of with participants’ knowledge

77
Returning to the cookie smell example

How would you measure

the DV of “how much
customers like your café”?

78
Returning to the cookie smell example
Open-ended questions: Tell me what
thoughts came to mind as you walked
into the café.
Survey scale: How do you feel about
this café: ☺ or  ?
Behavioral observation: time in café
Biological marker: smiles

79
Returning to the How would you
shark example measure the DV of
“fear of sharks”?

Take 1-2 minutes to talk

to a partner:
• Come up with 2
measurements (different
categories)
• Identify 1 strength and 1
limitation of each
80
Open-ended question: asking Dig into a Hard to
participants in a conversation to nuanced topic analyze
introspect about how they feel about
sharks
Survey scale: How afraid of sharks are Easy to People may
you on a scale of 1 (not at all) to 7 compare bias downward
(very)?
across people
Behavioral observation: whether leave Captures Difficult to
the room, whether donate to a shark
conservation fund, whether swim in the outcomes that track
ocean afterwards matter

Biological marker: heart rate, palm Measures Cannot be

sweat, cortisol unconscious subtly
responses measured
Recap
Seeing that 2 variables move together doesn’t necessarily
mean 1 causes the other: “correlation is not causation”
To determine causality, use an experiment
• Randomize to Control or Treatment
• Give each a different level/value of independent variable
• Measure dependent variable
You can measure variables in many different ways
• Each has strengths and weaknesses
• Best to combine them
82
 Reminders
First weekly survey due
Friday at 11:59pm
• Link will be posted in
the “Survey – Week
1” assignment on
Canvas this evening
#FinAid
Monday: money!
Hidden Brain link on
syllabus
83