Inferred and Stated Attribute Non‐attendance in Food Choice Experiments

Abstract

This paper uses a hypothetical choice experiment to investigate Italian consumer preferences and willingness-to-pay (WTP) for organic, conventional, and Gentically Modified (GM)-fed beef, utilizing intrinsic, search cues (price, colour and visible fat) and extrinsic, credence cues. Data is gathered from three different locations in Northern, Central and Southern Italy using a sequential Bayesian approach. Results showed that consumers attach higher value to organic meat. WTP for GM-fed beef, which is not yet sold in Italy, is well below current conventional beef prices. Organic beef is attractive to consumers because it is associated with higher animal welfare standards and environmental-related issues (food miles, biodiversity preservation). No differences are found in marginal WTP estimates by gender, age, education, being a parent or having a higher level of knowledge about organic production. Ethical/environmental issues (credence cues) appear to be more relevant in explaining variation in WTP for organic beef than ordinary product characteristics (search cues).

Are Fast Responses More Random? Testing the Effect of Response Time on Scale in an Online Choice Experiment

“…It is hypothesised that respondents who state that they have considered several or all attributes when stating their choices have taken more time for this task. Further, an equality-constrained latent class model is used to infer attribute attendance from choice data and test the influence of response time on class membership (Scarpa et al 2009(Scarpa et al , 2012. These analyses are intended to shed greater light on the question of whether response time is an appropriate proxy for cognitive effort in stated preference surveys.…”

“…The two econometric models employed in this study are the equality-constrained latent class (ECLC) (Scarpa et al 2009(Scarpa et al , 2012 and the generalised multinomial logit (GMNL) models (Fiebig et al 2010;Gu et al 2013). In all models, utility of respondent n from choosing option i in choice occasion t is assumed to consist of representative utility explained by the vector x nit consisting of alternative-and respondent-specific attributes and its coefficient vector β n and an unobserved error term ε nit according to…”

“…The vector z n contains individual-specific variables (such as response time) influencing class membership. What distinguishes the ECLC from a normal latent class model are the restrictions of the parameters in different classes (Scarpa et al 2009(Scarpa et al , 2012. These are threefold: (i) if an attribute is not attended to its coefficient is zero as this attribute does not affect choice probabilities; (ii) if not zero, estimated parameters are equal across classes except for (iii) the alternative-specific constant indicating the change options which is not constrained to be equal across classes (Glenk et al 2014).…”

New Paradigms for the Old Question: Challenging the Expectation Rule Held by Risky Decision-Making Theories

“…Process data are richer than inputoutput data and provide important evidence for explanatory mechanisms (Schulte-Mecklenbeck et al, 2010). Various technologies used to study information processing of people during decision making include behavioural experiments, functional magnetic resonance imaging (fMRI), event-related potential (ERP), mouse tracking, and eyetracking (Dshemuchadse, Scherbaum, & Goschke, 2013;Glöckner & Herbold, 2011;Rao, Li, Jiang, & Zhou, 2012;Rao et al, 2011;Scarpa, Zanoli, Bruschi, & Naspetti, 2013;Su et al, 2013). However, a few researchers have decomposed the decision-making process and directly assessed the computational steps assumed by the expectation rule by weighting, summing, and maximising the overall values.…”

Behavioral implications of providing real incentives in stated choice experiments

“…Up until now, such differences have mainly been concentrated on hypothetical bias in terms of WTP differences (an exception is found in Scarpa et al 2013, who examines attribute non-attendance), where findings have been ambiguous. For example, Carlsson and Martinsson (2001) find no differences in WTP 1 , whereas Grebitus et al, (2013), Johansson-Stenman and Svedsäter (2008), Taylor et al, (2010); Broadbent et al, (2010) and Ready et al, (2010) all find differences between hypothetical and incentivized WTP estimates.…”