Behavioral implications of providing real incentives in stated choice experiments

Abstract

This paper explores the potential impacts of introducing real economic incentives in Choice Experiments (CE). While many others have investigated such impacts before, the majority of the literature has focused solely on mitigation of hypothetical bias. We contribute to this literature by widening the scope of research to other behavioural aspects where consumers in CE are often found to deviate from homo economicus. We develop a theoretical model where not only Willingness to pay (WTP) measures but also decision processing can be affected by the introduction of an economic incentive. Specifically, our model allows for differential impacts on attribute processing, depending on the character of the attribute as well as self-image effects. In an empirical CE survey, we find some, though not unequivocal, support of our model. Even though we find no impact on WTP from introducing an economic incentive, we find marked benefits in relation to a number of behavioural aspects that together would favour the use of an economic incentive regardless of hypothetical bias being present or not.

Intrinsic Valuation of Information in Decision Making under Uncertainty

“…Among cognitive studies explicitly assessing the value of non-instrumental information, Pierson & Goodman (2014) found that participants self-reported a willingness to pay for non-instrumental information [ 39 ]. However, this behaviour was only assessed using a hypothetical survey task, which may have confounded results given the well-documented disparity in behaviour between hypothetical and incentive-compatible choice tasks [ 40 ]. Separately, a behavioural economic study using an incentive-compatible task concluded that observing non-instrumental information was related not to intrinsic valuation of information per se , but to a desire to increase one’s post-hoc confidence regarding an earlier decision [ 41 ].…”

Using eye tracking to account for attribute non-attendance in choice experiments

“…For example, responses may be influenced by how the question is asked or interpreted. Respondents may not recall how they chose, or may not answer the attendance statement truthfully (Kragt, 2013;Scarpa et al, 2013), or may bias their answer in a socially desirable manner (Mørkbak et al, 2014). Additionally, serial stated ANA questions may be difficult to answer because they may have applied different attribute processing strategies for each choice task (Hess and Hensher, 2010;Puckett and Hensher, 2009).…”

“…Rather than relying on self-reported information regarding attribute attendance, a second method infers ANA behaviour using analytical models such as equality constrained latent class (ECLC) models, which impose specific restrictions on the utility functions for each class by constraining some coefficients to zero for selected attributes in a certain class (Campbell et al, 2011;Caputo et al, 2013;Hensher and Greene, 2010;Kragt, 2013;Lagarde, 2013;Scarpa et al, 2009; 2013) Yet another method of inferring ANA is based on the coefficient of variation of individual-specific posterior means (Hess and Hensher, 2010;Scarpa et al, 2013;Mørkbak et al 2014). More research has been called for on other methods to account for ANA (Caputo et al, 2017;Scarpa et al, 2013).…”

Modelling Heterogeneous Preferences for Income Redistribution–An Application of Continuous and Discrete Distributions

“…Research has though shown that the internal ranking of characteristics is not affected by the hypothetical nature of the instrument, whereas no clear‐cut conclusions exist with regards to the absolute values—Carlsson and Martinsson () and Mørkbak et al . () find no differences in WTP, whereas Johansson‐Stenman and Svedsäter (); Ready et al . (); Taylor et al .…”