Methods: To assess violence outcome expectancies (a cognition pertinent to the study of aggression), a sample 136 adolescents were asked what they thought would happen if they acted aggressively towards their best friend, boy/girlfriend and caregiver. Respondents could provide up to 24 short answers. To create a quantitative measure, the following process was implemented: (1) preliminary coding by researchers, (2) grouping of categories and creation of coding strategy, (3) preliminary coding by research assistants, (4) testing of inter-rater reliability, (5) discussion of issues with research assistants, e.g., inter-rater disagreements, ambiguous responses, unclassifiable responses, (6) refinement of the coding strategy, (7) final coding of responses, and (8) final testing of inter-rater reliability.
Results: The respondents generated a number of unexpected responses; this necessitated the creation of several new categories. After refinement, the final coding strategy contained a total of 39 categories spread across 8 broad themes. Yet, despite this complexity, satisfactory agreement was obtained in the final testing of inter-rater reliability (Kappa= .834).
Implications: This study demonstrated that answers to complex open-ended questions could be reliably coded for use in quantitative research. Moreover, the iterative process required to create the coding scheme allows for a new understanding of the concept being measured, and of differences in how people perceive the social world. Given the time and energy required to transform the data, researchers must nevertheless judge whether the potential benefits merit the investment.