Chapter 11: Multivariate Methods of Marketing Reseach II: Conjoint Analysis and
Multidimensional Scaling
TRUE/FALSE
1) In interdependence methods, like conjoint and factor analyses, one or more variables are
designated as being predicted by (dependent on) a set of independent variables.
Answer: False
The description is of dependence methods, which include conjoint analysis, but not factor
analysis (an interdependence method).
2) Conjoint analysis is a set of methods designed to understand and quantify trade-offs.
Answer: True
Conjoint analysis quantifies trade-offs made in consumers’ minds while evaluating products.
3) Conjoint allows researchers to determine how a product should be priced.
Answer: True
The nature of economic transactions means that everything is ultimately traded off against
price, which is almost always an element of conjoint studies. One helpful feature of conjoint
is that it quantifies how any attribute is traded off against price and, therefore, how each
possible product should be priced.
4) To determine the optimal levels of attributes for a product, researchers should use
multidimensional scaling.
Answer: False
Determining the optimal levels of attributes for a product is a use of conjoint analysis.
5) An advantage of test markets is that products can be adjusted on the fly if targets are not
being met.
Answer: False
In a test market, making changes to the product is difficult or impossible, whereas in a
conjoint study, they can be adjusted on the fly if targets are not being met or the results
indicate that the tested configurations need to be fixed.
6) The overall goal of conjoint analysis is to measure the “part worths” or “importance
utilities” using information provided by respondents, such as rankings, ratings, or choices.
Answer: True
Conjoint analysis attempts to measure trade-offs using a consumer’s evaluations.
NOTE: Refer to this output regarding customer preferences for light bulbs relative to
brightness and bulb life to answer the following questions.
7) The preferences expressed in the output above are consistent and do not show the
customers making any trade-offs between brightness and bulb life.
Answer: False
Buyer A generally prefers bulb life, but trades off some bulb life for brightness in ranking #3.
Buyer B makes the opposite trade-off in ranking #3.
8) The output above shows that Buyer A generally values brightness more than bulb life,
whereas Buyer B generally has the opposite preference.
Answer: False
The output shows the reverse; Buyer A generally prefers longer bulb life, whereas Buyer B
generally prefers a brighter bulb.
9) The output above includes the relative importances of the attributes of brightness and bulb
life and the relative importances of the various levels of these attributes.
Answer: False
The output above does not include the relative importances. These can be estimated by
spanning each variable from 0 to 100 and assigning intermediate values for the rankings.
10) The following set of part worths reproduces the rankings of Buyer A above:
100 watts: 100
75 watts: 60
60 watts: 50
2500 hours: 50
2000 hours: 25
15 hours: 0
Answer: True
This set of part worths results in the rankings of Buyer A above.
11) The set of utility values for Buyer A given in question #10 is the unique solution to the
rankings given.
Answer: False
The rankings can be reproduced by assigning part worths to the attribute levels that fit the
buyer’s rankings, but these values will not be unique for this set of rankings.
12) An orthogonal conjunct design means that what is observed for one variable is unrelated
to what is observed for any of the others.
Answer: True
An orthogonal design is one in which, when several quantities have to be measured in an
experiment, manipulations of one quantity will not affect measurements of another.
13) One technique that can be used with an ordinal level dependent variable and nominallyscaled independent variables is conjoint measurement.
Answer: True
Conjoint can employ these data types.
14) In adaptive conjoint, the research objectives are adapted on the fly to market changes.
Answer: False
In adaptive conjoint, a respondent’s answers to prior questions are used to determine
subsequent questions, to help best use respondents’ time and obtain superior statistical
estimates.
15) In real life, consumers generally decide their preferred brands by mentally ranking their
options.
Answer: False
In real life, consumers generally decide their preferred brands by choosing one option from
an available set based on perceived trade-offs.
16) It is important to use a “none” or “no choice” option in all choice-based conjoint analysis.
Answer: True
Respondents must have the opportunity to choose none of the choices in case all of the
choices are below the person’s minimum threshold.
17) Conjoint is very useful for predicting market share of an optimal product.
Answer: False
Conjoint predicts preference, not sales or market share.
18) Multidimensional scaling summarizes data about associations between a fixed set of
objects by determining the minimum “dimensionality” required to represent the objects’
interrelationships and the position of each object on each dimension.
Answer: True
Multidimensional scaling summarizes data about associations between a fixed set of objects
to reveal relationships between them. It does this by determining two things: (1) the
minimum “dimensionality” (e.g., 2-D, 3-D) required to represent the objects’
interrelationships well, and (2) the position of each object on each dimension (i.e., its location
in the map).
19) The virtual of the perceptual map produced by MDS is that it is always a 2-dimensional
image that can be quickly understood by researchers and that is not burdened with
quantitative information.
Answer: False
MDS can also produce a 3-dimensional image. It preserves and illustrates the quantitative
information about the relationships. If humans were capable of visually processing more than
3 dimensions, it could produce those as well.
20) In order to interpret the dimensions created by MDS, researchers must overlay external
information.
Answer: True
MDS will always provide as faithful a representation of relative similarity or distance as
possible, but researchers can only attach meaning to the dimensions in the derived graph by
tying in other kinds of data.
21) MDS starts being more reliable, in terms of unambiguous interpretation of the resulting
spatial dimensions, for large numbers of object comparisons.
Answer: True
For n objects, data on n(n – 1)/2 possible pair-wise comparisons will be used to estimate 2n
coordinates in two-dimensional space. The resultant MDS starts being more reliable for large
numbers of object comparisons.
22) Stress in MDS is a goodness-of-fit measure between the input rank order and the output.
Answer: True
In multidimensional scaling, stress is a measure of how well the derived MDS map actually
represents the original pairwise similarity data.
23) A stress level of over 0.2 may indicate that a higher-dimensional solution may be called
for.
Answer: True
A stress level of over 0.2 is considered quite poor, indicating that there is a good deal of
disparity between the derived map and the original distance data, and therefore that a higherdimensional solution may be called for.
24) Fully metric MDS methods require interval- or ratio-scaled input measures and generate a
set of relationships among objects that is also interval or ratio.
Answer: True
This is the definition of fully metric MDS methods.
25) Nonmetric multidimensional scaling takes rank-ordered input measures and generates a
set of relationships among the objects which is ordinally-scaled.
Answer: False
Nonmetric multidimensional scaling takes rank-ordered input measures and generates a set of
relationships among the objects that is interval.
MULTIPLE CHOICE
1) The research method that is used to understand and quantify trade-offs is
a. factor analysis
b. discriminant analysis
c. conjoint analysis
d. multidimensional scaling
Answer: C
Conjoint analysis is useful for examining why consumers make the choices they do by
understanding and quantifying trade-offs.
2) Conjoint analysis can be used to answer all of the following questions except:
a. How much are specific attributes of a product or service valued?
b. What levels of each attribute should the product or service have?
c. How should a product or service be priced?
d. What brands do consumers tend to view as being similar?
Answer: D
In looking at what brands consumers see as being similar, cluster analysis will work the best.
3) Which research method is considered excellent as a premarket testing methodology?
a. discriminant analysis
b. cluster analysis
c. conjoint analysis
d. multidimensional scaling
Answer: C
Because of the cost of test marketing, conjoint analysis can be used prior to test marketing to
discover consumer trade-offs.
4) A valenced attribute is
a. an attribute with many, usually greater than 7, levels
b. an attribute where some levels are universally preferred
c. an attribute with one or more undetermined levels
d. a constructed attribute that mixes one or more levels of other attributes
Answer: B
With attributes that are valenced, everyone will prefer some levels of the attribute over
others. Everyone, for example, will prefer a camera that is less prone to breaking,
NOTE: Refer to this output regarding part worths for brightness, bulb life, and price in light
bulbs to answer the following questions.
5) Given the part worths above, which lightbulb will be preferred in the market, a “super
bright” bulb that is 100 watts but burns out in 1500 hours and sells for $2.50, or a “long life”
bulb that lasts 2500 hours, is 75 watts and sells for $3.00?
a. the super bright bulb
b. the long life bulb
c. The two bulbs will be equally preferred.
d. There is no way of determining this without more data.
Answer: B
The sum of the part worths for the super bright bulb is 105, whereas the sum of the part
worths for the long life bulb is 110; so the long life bulb will be slightly preferred.
6) Given the part worths above, what is the utility of a 60-watt bulb that lasts 2000 hours and
costs $2.00?
a. 0
b. 20
c. 45
d. More information is needed to determine this.
Answer: C
The utility of the bulb is the sum of the part worths, 0 + 25 + 20 = 45.
7) Given the part worths above, should the company produce the long life bulb, the super
bright bulb (both of which are described in question #5), or both of the bulbs?
a. the long life bulb
b. the super bright bulb
c. both of the bulbs
d. The answer cannot be determined from this information.
Answer: D
The question of whether both bulbs, or any other configurations, should be produced is
known as the product line design problem and has been studied by academic researchers for
decades. To solve it requires a great deal of information over and above conjoint part worths,
such as production costs, retailing operations, distribution, market size, and relative demand.
8) The set of conjoint analysis methods used by marketing professionals include of all of the
following except
a. k-means conjoint
b. orthogonal designs
c. adaptive conjoint
d. choice-based conjoint
Answer: A
The k-means method is a nonhierarchical clustering technique.
9) In conjoint analysis, _______________ allows researchers to offer a subset rather than all
of the possible combinations when looking a trade-offs on a number of product attributes.
a. the k-means method
b. orthogonal designs
c. adaptive conjoint
d. choice-based conjoint
Answer: B
To consider all possible trade-off combinations would be too cumbersome for respondents,
therefore, researchers can use an orthogonal design.
10) For k attributes, k-1 trade-off matrices are produce in what type of conjoint study?
a. k-means conjoint
b. adaptive conjoint
c. ranking data
d. choice-based conjunct
Answer: C
To use ranking data for k attributes requires k – 1 trade-off matrices.
11) By using _______________, a large conjoint design with 15,000 possible profiles can be
reduced down to just 25.
a. correlations
b. factors from factor analysis
c. orthogonal designs
d. clusters from cluster analysis
Answer: C
Orthogonal designs are used because all possible trade-off combinations would be too
cumbersome for respondents, therefore, researchers can drastically reduce the number.
12) When using _______________, a respondent’s prior choices or responses are used to
determine future comparisons.
a. choice-based profiles
b. rankings
c. adaptive conjoint
d. the k-means method
Answer: C
With the adaptive conjoint approach, the program adapts the comparisons to the respondent’s
prior choices so that time is not spent on comparisons that the respondent does not believe is
important.
13) In adaptive conjoint, if the first few conjoint responses indicate that a respondent is
insensitive to some aspect of a product,
a. he or she should be questioned on that attribute in greater detail to determine why
b. he or she should be questioned on that attribute in greater detail to achieve more finelygrained data
c. he or she should be questioned on trade-offs more relevant to that person
d. he or she should be questioned exactly as planned in the research design
Answer: C
In adaptive conjoint, if the first few conjoint responses strongly indicate that a person is
insensitive to some aspect of a product, he or she should be questioned about trade-offs more
relevant to that person.
14) In a nonadaptive conjoint program, even when all the data are collected,
a. researchers don’t know if it will match the information needs of the project
b. researchers don’t know if these respondents are the best respondents to query
c. researchers have no guarantee of sufficiently accurate results
d. all of the above
Answer: C
An adaptive conjoint program will calculate part worths after each new piece of data is
supplied by the respondent and use special algorithms to better measure those part worths that
are not yet determined with sufficient accuracy. The researcher can specify exactly how much
accuracy is required, and respondents can continue to reply until that level of accuracy is
achieved. In a nonadaptive conjoint program, even when all the data are collected,
researchers have no guarantee of sufficiently accurate results.
15) Advantages of adaptive conjoint include all of the following except
a. it allows one to quickly zero in on accurate utility measurements for a particular respondent
b. the researcher can specify exactly how much accuracy is required
c. it is especially suitable if there are many attribute levels
d. adaptive conjoint routines are included in most general statistics programs
Answer: D
Adaptive conjoint routines are highly specialized and are not included in commonly used
general statistics programs.
16) Disadvantages of adaptive conjoint include all of the following except
a. greater costs
b. it sometimes requires dedicated lab facilities
c. the number of attribute levels that it can handle is limited
d. greater complexity
Answer: C
Adaptive conjoint’s suitability to handling many attribute levels is a primary advantage of the
method.
17) In _______________, respondents are given a set of items and asked which they most
prefer, without any rankings, ratings, or comparisons.
a. choice-based conjoint
b. profiling
c. adaptive conjoint
d. ranking
Answer: A
The choice-base conjoint analysis tries to mimic real life situations by asking respondents
which they prefer and not ask them to make comparisons, rankings, or ratings.
18) Until _______________ models were developed, no statistical methods rigorously
measured part worths based on choice data. This made choice-based conjunct possible.
a. McFadden
b. discrete choice
c. Heckman
d. continuous choice
Answer: B
Until discrete choice models were developed, no statistical methods rigorously measured part
worths based on this sort of choice data.
19) Advantages of choice-based conjunct include all of the following except
a. it is dramatically more “real world” than other conjoint tasks
b. consumers are accustomed to judging and calibrating their own preferences
c. it is transparent and easy for respondents
d. it can be highly efficient and handling multiple attributes and product profiles
simultaneously
Answer: B
Choice-based conjunct is more “real world,” because consumers are accustomed to making
choices rather than judging and calibrating their own preferences.
20) In choice-based conjunct, descriptions of highly technical products should include
a. an explanation of the metrics used to quantify the attribute
b. the precise technical specifications
c. an interpretative descriptor of the technical specifications
d. all of the above
Answer: C
When studying highly technical products, it is often a good idea to describe an attribute level,
rather than include its precise technical specifications.
21) In choice-based conjunct, if an important attribute is omitted from the study,
a. respondents may make inferences about the missing attribute
b. respondents may become increasingly uncooperative with the researchers
c. respondents may show a preference for higher price levels
d. both a and c
Answer: D
If an important attribute is excluded from the study, respondents may assume that a higher
price means a higher quality on the missing attributes and show what seems to be a
preference for higher price levels.
22) Conjoint helps predict all of the following except
a. consumer preference
b. optimal product attributes
c. market share
d. optimal price
Answer: C
Conjoint predicts preference, not sales or market share.
23) Typical applications of conjoint include all of the following managerial questions except:
a. How does the choice probability of this product change if we change its appearance and ad
message?
b. What demographics are shared by the target segment for this product?
c. Should we bring out a single product or an entire line?
d. How should we price a new version of our product with upgraded features?
Answer: B
Segmentation studies would be better served by cluster analysis, discriminant analysis, or
multidimensional scaling.
24) What _______________ is designed to do is estimate what drives purchases by
calibrating the sources and strengths of consumers’ preferences.
a. multidimensional scaling
b. conjoint analysis
c. discriminant analysis
d. factor analysis
Answer: B
Conjoint analysis is excellent for understanding and quantifying consumer preferences over
product attributes.
25) Sales and share are affected by all of the following except
a. promotion
b. customer expertise and knowledge
c. distribution
d. conjoint results
Answer: D
Sales and market share are affected by many factors, including distribution, advertising,
promotion, awareness of product, customer expertise and knowledge, and purchase readiness.
26) To determine the optimal product from conjoint results, researchers
a. choose the levels of each attribute with the highest part worths, including the price attribute
b. incorporate information about production and distribution costs with the conjoint results to
determine product attributes and price that result in profit
c. fit continuous curves through the part worths, including price, and use the levels at the
highest points of these curves
d. none of the above
Answer: B
It is possible that the optimal configurations chosen by respondents would involve negative
profit. (The consumer will want “the best of everything at the lowest price,” and the
manufacturer will want “the lowest production cost and the highest price.”) Detailed
information about production and distribution costs may thus result in a higher price than
suggested by initial conjoint results.
27) _______________ summarizes data about associations between a fixed set of objects to
reveal relationships between them, such as brands in a particular class.
a. Multidimensional scaling
b. Discriminant analysis
c. Cluster analysis
d. Segmentation analysis
Answer: A
Multidimensional scaling is an excellent method for creating pictures of markets based on
associations between a fixed set of objects.
28) Multidimensional scaling
a. relies on a prior decision about which dimensions are important
b. depicts categories and brands in terms of attributes displayed on a perceptual map
c. requires collection of data on each attribute of similarity or difference
d. all of the above
e. none of the above
Answer: E
MDS is a way to construct “pictures” of markets without a prior decision as to which
dimensions are important or collection of any attribute-based data. This short circuits the idea
of depicting categories and brands only in terms of their attributes, instead displaying item
similarity.
29) When depicting a higher-dimensional object in lower-dimensional space,
a. this is immediately apparent when one looks at the resultant map
b. severe distortion is possible
c. external information cannot be overlain on the resultant map
d. a perfectly accurate rendering results
Answer: B
Whenever one tries to depict a higher-dimensional object in a lower-dimensional space, one
is allowing for the possibility of distortion, sometimes severely so.
30) MDS will always provide as faithful a representation of _______________ as possible.
a. relative similarity or distance
b. relative attribute levels
c. absolute map positions
d. absolute attribute levels
Answer: A
MDS will always provide as faithful a representation of relative similarity or distance as
possible.
31) In an MDS routine, for n objects, _______________ possible pair-wise comparisons are
used to estimate 2n coordinates in two-dimensional space.
a. n(n – 1)/2
b. n2 – 1
c. n!
d. cannot be determined
Answer: A
For n objects, n(n – 1)/2 possible pair-wise comparisons are used to estimate 2n coordinates
in 2-D space.
32) A stress score of 0.1 in a multidimensional scaling is considered
a. poor
b. fair
c. good
d. excellent
Answer: B
For stress scores, 0.1 is fair; 0.05 is good; and 0.01 or less is excellent.
33) An MDS routine
a. operates through a set of formulas, as in regression, that yield the best fit
b. always attempts to maximize the stress level of the distance data
c. is iterative, determining a good initial guess and continuing to derive better estimates
d. none of the above
Answer: C
Unlike regression, an MDS routine is iterative, determining a good initial guess and
continuing to derive better estimates to lower stress and better represent the data.
34) A poor stress level in an MDS output indicates
a. considerable disparity between the derived map and the original distance data
b. that the researchers have high cortisol levels
c. considerable similarity between the derived map and the original distance data
d. none of the above
Answer: A
A stress level over 0.2 is considered quite poor, indicating a good deal of disparity between
the derived map and the original distance data
35) To interpret the dimensions in MDS, each of the dimensions
a. is left to the judgment of the researchers
b. has its numbers correlated with external data that has been previously shown to relate to
particular dimension
c. has its numbers correlated with any external data
d. none of the above
Answer: C
Each of the dimensions is a column of numbers that is correlated with any other external data.
If the correlation is near zero, it means that the new information has nothing to do with that
particular dimension; if the correlation is high, the new information is helpful in getting a
sense of what that dimension means.
36) The axes of an MDS solution
a. are determined uniquely from the data by the MDS routine
b. have a special, privileged meaning
c. can be rotated and/or flipped and reinterpreted
d. are pre-determined by researchers before the data is entered
Answer: C
The entire map can be rotated in any direction, or flipped on any axis, and the inter-item
distances would remain identical. The two new axes would shift into place and be interpreted
by researchers.
37) _______________ multidimensional scaling takes rank-ordered input measures and
generates the rank order of each object on each dimension.
a. Fully metric
b. Fully nonmetric
c. Metric
d. Nonmetric
Answer: B
This is the definition of fully nonmetric MDS.
38) _______________ multidimensional scaling requires interval- or ratio-scaled input
measures and generates a set of relationships among objects that is also interval or ratio.
a. Fully metric
b. Fully nonmetric
c. Metric
d. Nonmetric
Answer: A
This is the definition of fully metric MDS.
39) _______________ multidimensional scaling methods take rank-ordered input measures
and generates a set of relationships among objects that is interval.
a. Fully metric
b. Fully nonmetric
c. Metric
d. Nonmetric
Answer: D
This is the definition of nonmetric MDS.
40) All of the following are among the most frequent applications of MDS except
a. identifying products that are viewed as substitutes for one another
b. identifying the underlying structure of variables
c. identifying salient product attributes perceived by buyers in a market
d. exposing the gaps in a market that can support a new product venture
Answer: B
Factor analysis is used to identify the underlying structure of variables. The other three can
benefit from MDS.
SHORT ANSWER
NOTE: Refer to this output regarding part worths for brightness, bulb life, and price in light
bulbs to answer the following questions.
1) Given the part worths above, what is the utility for a 75-watt bulb that burns for 1500
hours and costs $2.50?
Answer: 65
The sum of the part worths = 60 + 0 + 5
2) Which bulb would be preferred by more customers, a 100-watt bulb that burns for 1500
hours and cost $3.00 or a 75-watt bulb that burns for 2000 hours and costs $2.50?
Answer: The first bulb.
The utility for the first bulb is 100 + 0 + 0 = 100; for the second is 60 + 25 + 5 = 90. So the
first bulb is somewhat preferred.
3) What kind of design is generated by specialized routines in statistical programs to
dramatically reduce the number of profiles required for a conjoint study?
Answer: An orthogonal design.
4) To best emulate the way people evaluate product options in the actual market, what
method should be used to estimate the part worths of those options?
Answer: Choice-based conjunct.
5) What method should be used to generate a rank order of each object along each dimension,
given rank-ordered input measures.
Answer: Fully nonmetric multidimensional scaling.
ESSAY
1) Discuss the basic steps in a conjoint analysis.
Answer: 1) collecting trade-off data from consumers
2) estimating value systems (e.g., utilities, importances, part worths)
3) making choice predictions
2) What does an adaptive conjoint program do?
Answer: In an adaptive conjoint, the response to earlier questions are used to help generate
attribute levels for all subsequent questions. An adaptive conjoint program will calculate part
worths after each new piece of data is supplied by the respondent, using special algorithms to
better measure those part worths that are not yet determined with sufficient accuracy.
3) Discuss the advantage of using the choice-based conjoint over the other types of conjoint
analysis.
Answer: The choice-based has the following advantages:
1) it is dramatically more “real-world” because consumers are simply asked to make a choice
2) it is easy and transparent for respondents
3) because several profiles can be offered and compared at once, it is very efficient
4) Discuss some of the managerial questions where conjoint analysis is applied in marketing
research.
Answer: Student answers will vary; some possible examples below.
5) Discuss the uses of multidimensional scaling.
Answer: Applications of MDS include:
• identifying salient product attributes perceived by buyers in a market
• determining the combination of product attributes consumers most prefer
• identifying products that are viewed as substitutes for one another
• noting viable segments that exist in a market
• exposing the “holes” in a market that can support a new product venture
• application to problems of
- product life-cycle analysis
- market segmentation
- vendor evaluation
- advertising evaluation
- test marketing
- salesperson and store image
- brand switching research
- attitude scaling
Test Bank for Modern Marketing Research: Concepts, Methods, and Cases
Fred M. Feinberg, Thomas Kinnear, James R. Taylor
9781133188964, 9781133191025, 9780759391710
