CASE: Modelling the Non-linear Relationship between Customer Satisfaction and Loyalty


Recently MITTAL et al. (1998) and MATZLER et al. (2003) have shown that the relationship between attribute level performance and overall satisfaction is non-linear. Based on prospect theory MITTAL et al. argued for a stronger impact of negative events and negative attribute performance on overall satisfaction. Based on the Kano-model, MATZLER et al. (2003) argued for the existence of different types of asymmetric relationships (excitement factors, basic factors) as well as linear relationships (performance factors). Some of the results from MITTAL et al. indicate that a re-examination of the link between satisfaction and its consequent behaviours, such as loyalty, word-of-mouth and retention is necessary. Previous conceptualizations have assumed these relationships to be linear and symmetric (YI 1990). However satisfaction s impact on these behaviours could be asymmetric and nonlinear. High levels of satisfaction might not increase loyalty, while high levels of dissatisfaction might have a strong and negative impact on loyalty. This could account for the empirical fact that many firms experience high rates of customer defection despite high rates of customer satisfaction (REICHHELD 1996). Both cited papers have used dummy regressions to estimate the non-linear effect of attribute satisfaction on overall satisfaction. This approach is problematic since both satisfaction judgements and also loyalty are clearly latent variables. Not correcting for measurement error leads to inconsistent and attenuated parameter estimates and dummy coding leads to a loss of information. Recently Klein has developed a statistically efficient and practicable estimation method for structural equation models with multiple latent non-linear effects that has been shown to outperform other available approaches with respect to efficiency (KLEIN & MUTHEN 2004). Furthermore Klein s Quasi-ML method allows for the estimation of multiple interaction and quadratic effects. We apply the Quasi-ML method to the substantive research question of non-linear effects of customer satisfaction on loyalty. We demonstrate the applicability of the Quasi-ML method in two studies: one in the automotive industry and one in the telecommunication industry and compare the results with previously utilized methods.


Principal Investigators
Paulssen, Marcel Prof. Dr. (Details) (Quantitative Climate, Weather and Energy Analysis)

Duration of Project
Start date: 01/2005
End date: 02/2005

Publications
Modelling the non-linear relationship between satisfaction and loyalty with structural equation models (submitted to GfKl conference)

Last updated on 2020-09-03 at 23:05