Abstract
One of branding’s most important constructs is brand personality (BP). Despite more than 50 years of BP measurement, current scales still produce equivocal results. The latter are caused by several limitations of existing BP scales identified which motivated the present study. Existing BP scales are to different degrees construct invalid, contain ambiguous items, and are biased or incomplete. Scales are also non-generalizable, and are inconsistent with human personality. Factor analytically derived BP scales are furthermore not proper measures. They do not comply with the latter’s unidimensionality, invariance and concatenation requirements. Not only is the power and usefulness of existing BP scales diminished. Most subsequent research based on these scales, particularly on Aaker’s (1997) lauded BPS, is likely to be questionable.
To address these concerns a different BP scale development approach was investigated: Rasch Modelling. In Rasch Modelling mathematical algorithms transform raw item and respondent data into logarithmic probabilities. The application of Rasch Modelling results in continuous scales with standardized units through which items and respondents can then be measured. To investigate the viability of Rasch Modelling for BP scale development an empirical effort was conducted. 332 respondents evaluated two brands using Goldberg’s (1992) 100 Personality Markers. Apple and Microsoft were used as stimuli. These are well known brands, different in terms of nature (symbolic vs. functional), and elicit extreme affinity responses (like-dislike). Instead of deriving measures through factor analysis, as conventionally done for BP, a new three phase, nine-step iterative Rasch Modelling sequence was used.
Ten new Rasch BP scales were developed. Positive and negative valenced ones for each human personality Big Five dimension: Surgency, Openness, Conscientiousness, Agreeableness and Neuroticism. The Rasch BP scales exhibit limitations with respect to the characteristics of ideal measures: Items lack coverage, are unevenly and broadly spaced, are out of alignment with respect to respondents, and are not progressively ordered in terms of intensity. However, the Rasch BP scales do overcome the major limitations of conventional factor analytical BP scales produced to date. The Rasch BP measures are valid from a construct perspective, complete and unbiased, generalizable, and consistent with human personality. Most importantly, the Rasch BP scales are proper measures. They comply with the latter’s unidimensionality, invariance and concatenation requirements. The above suggest that Rasch Modelling is a viable option for BP scale development. However, the present effort and its resultant scales should not be seen as a final result. They are rather a first step towards a new line of enquiry.
This dissertation makes a series of contributions. The most important ones are the critical review of the evolution and current state of BP scale development, the suggestion of a more precise and conceptually correct BP definition; the evaluation of a new BP scale development paradigm, Rasch Modelling; the presentation of a new three phase, nine-step Rasch scale development process; the identification of issues associated to the implementation of the Rasch-based scale development technique; and the provision of preliminary suggestions as to how these issues might be addressed in future research.