International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
Investigating alternative non-attendance and other decision processing strategies in food choice
Ole Bonnichsen, Danny Campbell, Morten Raun Mørkbak, Søren Bøye Olsen

Last modified: 18 May 2015

Abstract


While many different decision processing strategies have been identified in the Choice Experiment literature (see for example Hensher (2007) and Hensher (2010) for an overview), one that has received increasing attention in the recent literature is attribute non-attendance, i.e. where respondents ignore one or more attributes of the alternatives when making their choices (see for example Hensher et al. (2005), Campbell et al. (2008) and Hensher (2010)). While respondents might invoke such a heuristic for different reasons (Alemu et al. 2012), another related heuristic is when respondents ignore one or more of the choice task alternatives rather than just ignoring attributes. Inspired by the often used approach to incorporate stated attribute non-attendance, we investigate the issue empirically in a food choice context by asking respondents follow-up questions targeting alternative non-attendance. Specifically, after each choice task respondents are asked to state whether they have ignored any of the four alternatives that they were presented with in the choice task. Then, in the econometric modelling, observations of an alternative that has been stated to be ignored are assigned a zero LL-contribution in the likelihood function.

 

Based on responses from approximately 1000 respondents in an empirical case concerning consumer preferences for cheese characteristics, we find that in some cases more than half of the respondents stated to have ignored an alternative in a choice task. We generally find significant improvements in model performance when conditioning the choice models on attendance to alternatives as well as on attributes. However, using this approach leads to striking reductions in the numerical value of the price parameter estimates, which in turn causes WTP estimates to become unrealistically large. Further investigations reveal that this most likely is the result of respondents stating to ignore high-priced alternatives. Specifically we find significant correlation between stated alternative non-attendance and price levels, whereas there is no significant correlation with non-price attribute levels. So, when respondents state to have ignored an alternative, they have in fact in many cases done the opposite. Ignoring high-priced alternatives could be part of an elimination-by-aspects heuristic but it might as well be reflecting fully rational behaviour when the price exceeds the individual’s choke price.

 

Additionally, in an attempt to infer alternative non-attendance from the econometric models rather than merely relying on respondents’ statements, we find inspiration in the literature concerning choice set formation literature. We relax the basic assumption that all alternatives in a choice task are considered by adapting a modified version of Swait’s (1984) independent availability logit (IAL) model. We find significant improvements compared to the standard MNL model. We furthermore find noteworthy differences between the inferred and the stated levels of alternative non-attendance, in terms of inferred levels being smaller. This supports the suspicion raised above regarding respondents perceiving the typical non-attendance question differently from how the researcher expects it to be perceived.


Conference registration is required in order to view papers.