Ch. 11

1
Ch. 11

• Attitude Measurement

2
Fishbeins Behavioral Intentions Model

• Attitude is degree of like-dislike
• A(act) versus A(object)
• Behavior f(Behavioral Intention) f(A(act)
  Subjective Norm)
• A(act) f(beliefsevaluations of beliefs)
• Subjective Norm f(normative beliefs motivation
  to comply

3
A(object) vs. A(act)

• Interested in measuring attitude towards a
  Mercedes --gt A(object) appropriate
• Interested in measuring attitude towards BUYING a
  Mercedes --gt A(act of buying) appropriate

4
Measurement Scales

• Nominal
• Are you satisfied with X? ___YES ___ NO
• Ordinal
• Rank order X, Y, and Z according to how satisfied
  you are with them.
• Interval
• How satisfied? __Not very to __Very satisfied
• Ratio
• How satisfied? Very dissatisfiedtoVery
  satisfied

5
SPSS and Scales

• SSPS recognizes three types of scales
• Nominal, Ordinal, and Scale
• Scale includes both interval and ratio scales
• Analytic procedures options are the same for
  interval and ration

6
Constructs and Research

• Specify constructs
• Constructs roughly equivalent to concepts
• Satisfaction, loyalty, trust, small businesses
• Conceptual definitions and operational
  definitions
• Conceptual (constitutive) definitions Like a
  dictionary
• Operational definitions How to measure?

7

• Identify/postulate relationships. EX
• Commitment is a positive function of satisfaction
• Decide on a model. EX
• Commitment f(satisfaction, risks, switching
  costs and benefits)
• Satisfaction f(perceived outcomes normative
  expectations)

8
Classification of Attitude Scales
Attitude Scales
Multi-Item Scales
Single-Item Scales
Continuous Scales
Itemized Category Scales
Comparative Scales
Associative Scales
Semantic Differential Scale
Paired Comparison Scales
Q-sort Scales
Stapel Scales
Thurstone Scales
Likert Scales
Pictorial Scales
Constant Sum Scales
Rank Order Scales
9
Single Item Scales (Contd.)

• Itemized-category Scales
• Scales in which the respondent selects from a
  limited number of categories
• Comparative Scale
• A judgment comparing one object, concept, or
  person against one another

10
Single Item Scales (Contd.)

• Rank-order Scales
• Scale in which the respondent compares one item
  with another or a group of items against each
  other and ranks them
• Q-sort Scaling
• Respondents sort comparative characteristics into
  normally distributed groups
• Ten or more groups increases accuracy of results

11
Single Item Scales (Contd.)

• Constant-sum scale
• Respondents allocate a fixed number of rating
  points among serial objects to reflect relative
  preference
• Pictorial scales
• Various categories of the scale are depicted
  pictorially
• Thermometer Scale
• Funny faces scale
• Format must be comprehensible to respond and
  allow accurate response

12
Single Item Scales (Contd.)

• Paired-Comparison Scales
• The brands to be rated are presented two at a
  time, so each brand in the category was compared
  once to every other brand
• Brands are rated on a given 10 pts. that are then
  divided between the two brands
• Advantages
• Performs well
• Limitations
• Cumbersome to administer
• Frame of reference is always the other brand
  being tested these brands may change over time

13
Designing Single Item Scales

• Number of Scale Categories
• 5-7 for satisfaction, Likert, etc. questions work
• Neutral and/or Dont Know alternatives Use when
  approproate.
• Types of Poles Used in the Scale
• One pole only anchored (Stapel)?

14
Designing Single Item Scales

• Strength of the Anchors
• Want a good spread of data
• Labeling of the Categories
• Label all? End points only? Something between?
• Balance of a Scale
• Generally balance scales

15
Multiple-item Scales

• Developed to measure a sample of beliefs toward
  the attitude objects and combine the set of
  answers into an average score
• RELIABILITY a major concern driving the
  development of multi-item scales
• Look for published scales
• Secondary literature
• Books which are collections of scales

16
Multiple-item Scales (Contd.)

• Likert Scale
• Requires respondent to indicate degree of
  agreement or disagreement with a variety of
  statements related to the attitude object
• Can be used as a single item scale, OR
• Summated Scale
• Scores on individual items are summed to give
  total score for respondents
• Likert Scale Is Uni-dimensional

17
Multiple-item Scales (Contd.)

• Thurstone Scales
• Also known as the method of equal-appearing
  intervals objective is to obtain a
  unidimensional scale with interval properties
• Step 1
• Generate a large number of statements or
  adjectives reflecting all degrees of
  favorableness toward the attitude of objects
• Step 2
• A group of judges is given this set of items and
  asked to classify them according to their degree
  of favorableness or unfavorableness

18
Multiple-item Scales (Contd.)

• Thurstone Scales (Cont.)
• Advantages
• Easy to administer
• Requires minimum instructions
• Limitations
• Time consuming
• Expensive to construct
• Not as much diagnostic value as a Likert scale
• Values depend on the attitudes of the original
  judges

19
Multiple-item Scales (Contd.)

• Semantic-differential Scale
• Respondents rate each attribute object on a
  number of five or seven-point rating scales
  bounded by polar adjectives or phrases
• With bipolar scale, the midpoint is a neutral
  point

20
Characteristics of Semantic Differential

• Scales in Semantic Marketing Applications
• Pairs of objects or phrases must be meaningful in
  market being studied and often correspond to
  product/service attributes
• Avoid "halo" effect by placing negative pole on
  either side
• This applies to Likert items, as well.
• Category increments are treated as interval
  scales so group mean values can be computed for
  each object on each scale
• May also be analyzed as a summated rating scale

21
Characteristics of Semantic Differential (Contd.)

• Profile Analysis
• Application of semantic differential scale
• Plot mean ratings of each object on each scale
  for visual comparison
• Overall comparison of brands hard to grasp with
  many brands and attributes
• Not all attributes are independent

22
Multiple-item Scales (Contd.)

• Stapel Scales
• Uses one pole rather than two opposite poles
• Respondents select a numerical response category
• High positive score reflects good fit between
  adjective and object
• Easy to administer and construct
• No need to assure bipolarity

23
Multiple-item Scales (Contd.)

• Associative Scaling
• Most effective for markets where respondent is
  knowledgeable only about a small subset of a
  large number of choices
• Appropriate to choice situations that involve a
  sequential decision process
• Best suited to market tracking where the emphasis
  is on understanding shifts in relative
  competitive positions

24
Multiple-item Scales (Contd.)

• Continuous Rating Scales
• Respondents rate objects by placing a mark at
  appropriate position on a line running from one
  extreme of the criterion variable to the other
• Values can be interpreted as interval or ratio
  scaled data
• It is easy to construct
• Scoring is cumbersome and unreliable

25
General Guidelines For Developing A Multiple-Item
Scale
Determine clearly what you are going to measure
Generate as many items as possible
Ask experts in the field to evaluate the initial
pool of items
Determine the type of attitudinal scale to be used
26
Include some items that will help in the
validation of the scale
Administer the items to an initial sample
Evaluate and refine the items
Finally, optimize the scale length
27
Some Crucial Considerations
28
Validity

• Validity
• An attitude measure is valid if it measures what
  it is supposed to measure
• Face Validity
• The extent to which the content of a measurement
  scale appears to tap all relevant facets of the
  construct Does it appear to measure what we want
  to measure? (experts)
• Criterion Validity
• Based on empirical evidence that the attitude
  measure correlates with other criterion
  variables

29
Validity (Contd)

• Concurrent validity
• Two variables are measured at the same time
• Predictive validity
• The attitude measure can predict some future
  event
• Convergent validity
• Demonstrated when measure correlates highly with
  other measures of construct (e.g., a Likert scale
  measure versus a multiple choice measure)

30
Validity (Contd)

• Discriminant validity
• Discriminant validity is demonstrated when the
  measure does not correlate highly (or correlates
  negatively) with measures regarding other
  constructs (e.g., loyalty versus satisfaction)
• Construct Validity
• Construct validity is difficult to assess, and is
  essentially a function of the quality of the
  theory. That is, do theory based hypotheses turn
  out as predicted?

31
Reliability

• Reliability
• The consistency with which the measure produces
  the same results with the same or comparable
  population
• Assessing reliability
• Test-retest
• Split half (multiple items scale)
• Coefficient alpha a.k.a. Cronbachs alpha
  (multiple items scale)

32
Generalizability

• Question To what extent are the results
  generalizable to other groups in other
  situations?
• Example You find out that permanent Cedar City
  residents view economic development in terms of
  job stability. Can you extend this to SUU
  students, or to residents of other cities in
  Utah? To all U.S. citizens?
• Big issue in experimental research