To Scobit or not to Scobit: the Potentials of Scobit Models in Marketing Research
Christelle Garrouste
Econometrics and Applied Statistics Unit, European Commission Joint Research Center, Ispra, Italy

The assumption imposed in the traditional logit and probit models that both the logistic and normal density functions are symmetric around 0 means that what is tested by these models is how much a variation in the dependent variables can challenge the hypothesis of a 0.5 probability that individuals choose between two alternatives. Although this normality hypothesis is methodological convenient and robust, it does not fit the common skewness observed in most logistic distributions when social and behavioural data are concerned. The scobit (or skewed-logit) model developed by Nagler overcomes this weakness by assuming that individuals with any initial probability of choosing either of two alternatives are most sensitive to changes in independent variables. Through two empirical examples based respectively on the EU-SILC data and on the Consumer Empowerment Index, this paper demonstrates that the scobit model constitutes an optimal alternative to the traditional logit and probit models in the fields of labour market micro-analyzes and marketing research.

Keywords: Statistic; Assessment; Skewed-logit; Consumer empowerment

Biography: Dr Christelle Garrouste is a research fellow at the Unit of Econometrics and Applied Statistics of the European Commission's Joint Research Center. She obtained a degree from the Norwegian Business School of Bergen and a PhD in International Education from Stockholm University. Her main research activities focus on economics of education, human capital accumulation, labour economics and multilevel analysis.