I am going to make an environmental index to be used as an explanatory variable in a regression model. For making this index, I asked respondents a set of questions about their environmental attitudes. Each question has 5 response options from 1=“completely agree” to 5=“completely disagree”. I'm going to summarize each these categories as one index.

What general advice can be offered for creating such an index?

  • $\begingroup$ Maybe look at factor analysis or principal components analysis? $\endgroup$
    – fmark
    May 6, 2011 at 8:28

2 Answers 2


People often use the phrase "Likert scale" erroneously, not realizing that it originally described the coherent method Rensis Likert developed to do just what you're describing. Key steps are

  1. Seeing how well each item correlates with "the whole"--the average of all other items
  2. Checking variability, since an item with low variability is unlikely to contribute much information to such a scale
  3. Using steps 1 and 2, as well as tests of Cronbach's alpha (and sometimes factor analysis, as @fmark has said), to narrow down the list of items to be kept
  4. Averaging the items ultimately selected to compute the scale score for each person.

An excellent and very accessible guide to this process can be found in Paul Spector's Summated Rating Scale Construction, a little green Sage book that's available for about $18 new.


You need to check how others have built indexes with similar questions. My guess is that Inglehart and Norris, in Rising Tide, have built their Gender Equality/Empowerment index in a way that you can emulate (the construction of the index escapes me but I remember it's in the Technical Appendix).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.