Mario Mazzocchi (Dept. of Statistics, University of Bologna)
Marco Vassallo (National Institute for Research on Food and Nutrition, INRAN)

This presentation is part of the workshop "Structural equation and Path Modelling: a useful tool in sensory and consumer science" (Chair persons: Dr Mario Mazzocch and Dr Hal MacFie)

The presentation is structured in two parts. First, a “practitioners-targeted” discussion on the costs and benefits of employing structural equation modelling to solve consumer research and marketing problems, with a special focus on applications to food consumption behaviour and food marketing. Through an empirical example on the impact of socio-economic status on dietary behaviour, obesity and health, based on UK data, a set of issues relevant to SEM are introduced. These include endogeneous relationships in food choice, measurement issues and errors-in-variables situations. The difficulties in performing effective SEM and the proper approach to model assessment will be discussed.

Based on the case study, the second part of the presentation will get into a selection of methodological issues that are especially relevant. First, the problem of the poor specification of the factors in terms of convergent and discriminant validity in food choice. For example, the issues faced when a factor relies on only two or even one manifest indicators. Second, the positive and negative implications of multi-collinearity in SEM. Third, the presence of correlated error terms in measurement models – when is it appropriate to assume correlation? Based on a technical insights, practical cut-off criteria and other rule of thumbs will be illustrated.