Scaling
Techniques in Social Research

Paired Comparison Method of L.L.
Thurstone, 1928

 Select a set of attitude statements
toward an issue in question
 Each judge in a panel of 25 or more
reads every possible pair of items
 Each judge decides which item is
more favorable to the issue
 Obtain for each pair the
proportion of times one item is more favorable to the
issue
 Using these data and applying some
mathematics, one computes precise scores for each
item
 A precise but unwieldly method of
attitude scaling:
 One must assemble a panel of
judges, who make many pairedcomparisons:
 For k items there are
.5(k)(k1) pairsi.e., 300 paired
judgments for 25 items.
 Download a PowerPoint
Lecture on Thurstone
scaling from this site (and other forms of
scaling)

Method of Summated Ratings of Rensis
Likert, 1932

 Likert devised terms for five
categories of attitude ratings:
 Strongly approve, Approve,
Undecided, Disapprove, Strongly Disapprove
 Like Thurstone, Likert devised a
statistical model to generate scores for these
categories
 He soon found that his complex
scores correlated .99 with the assigned scores:
54321.
 He typically added the scores on set
of items, producing "The Method of Summated
Ratings"
 For an application in politics, see
British
political attitudes

Scalogram Analysis of Louis Guttman,
1944

 Guttman assumed that some attitude
items were more extreme than others
 Respondents who agree to the most
extreme statement agree to the less extreme
 Scalogram analysis determines the
consistency of the pattern of agreement: See
Guttman
scale analysis
 The coefficient of
reproducibility assesses the consistency of the
pattern
 Reproducibility can be quite high
still have many errors
 Guttman scaling has become less
popular in recent years

Semantic Differential of Charles
Osgood, 1957

 Respondents are asked to choose
between opposite sets of symbols when rating a
concept.
 Example from a Eurobarometer survey
of citizens' feelings toward "violence":
 Respondents faced eight pairs of
words relating to "violencee.g.,
 beautiful v. ugly
 strong v. weak
 Reponse patterns are analyzed in
scoring.
 See this site for semantic
differential


