My research area is in communications for development. I am looking at attitudes to women leadership and voting behaviours in elections, for a small Pacific Islands ethnic group. I am interested in the structure of social attitudes/representations because it may point to where communications interventions can be most effective.
For the class of general attitudes to women leadership, one might expect some consensual features, but for attitudes to voting it is less clear. The evidence shows that a major influence on voting is the expectation that the winning candidate will provide material assistance to his/her supporters.
I have understood that Guttman scaling has an affinity with social representations (eg. Roiser, 1996) and would like to explore this approach, but I am unsure that I have properly grasped the concept. Let's say that I have some survey data with responses to the question "in which places do women make the decisions (i.e. are effective leaders)", and from another survey, responses to why people voted for particular candidates in the last election, and how gender affected their decision. I understand that Guttman scaling using statements from these survey responses will indicate how commumulative the underlying attitudes are, and hence the consenual features. My question is really whether I have grasped the concept properly. I would appreciate any replies from social statisticians. Thank you!
I am not very familiar with the Guttman scaling procedure you mention. However unless the different constructs were assessed among the same set of subjects and using similar Likert scale structured questions with the same number of response options, I dont see how one could combine measures from different surveys to assess the consensual nature of the various attitudinal and behavioral measures you are examining.
Hi thanks David.
I am looking at the two seperately. Firstly at general attitudes to women leadership. This may not predict specific behaviours such as voting, but may be associated with a domain of behaviours that generally support women leaders or otherwise. I think that there may be some cumulative relation to the responses, for instance, most people are likely to say that women are effective leaders in the home. There will be less who agree that women can be effective parliamentarians, but those that do will probably also agree with women being good leaders in the home. So one might draw a diagram with concentric circles. From what little I know is that there may be a "valid" Guttman scale, showing the cumulative nature of those attitudes.
In the case of attitudes to voting, it is not clear how the gender of candidates mediates in regard to other variables such as the voter's expectations. I.e. does gender mediate once expectations are satisfied (women can be counted on to provide...),or is it more effective to represent women candidates in a new, transformative light. Could a Guttman scaling exercise help to show how/if gender operates in the construction of voting intentions (as in reasoned action). This will have implications for a communications stratgy designed to make women more electable.
I am still a bit muddled about this as you can tell, but I was hoping to look at these two cases seperately (not combining them statistically). My project is very modest, but there is an opportunity to conduct both surveys.
As I do not know exactly what you have it is a little hard to tell what you would need, but instead of look to Guttman (which is deterministic), you could try Mokken-scaling (which is a non-parametric probabilistic extension of Guttman) or Rasch to find the structure within the attitudes. Mokken has a search procedure to look at items that are uni-dimensional and have an cumulative structure. Mokken and Rasch models are implemented in R. (see https://www.jstatsoft.org/article/view/v020i11).
- Badges
- Science topic
- Similar topics
- Psychology
- Social Psychology