Utility maximization versus regret minimization in health choice behavior: Evidence from four datasets

Abstract

Behavioral evidence has emerged as an important form of evidence which can inform health policy. Data from discrete choice experiments (DCEs) and non-experimental data, analyzed using choice models, have proven popular for understanding many health behaviors, from purchasing consumer goods, to clinical decision making and lifestyle choices (

Can decision field theory enhance our understanding of health‐based choices? Evidence from risky health behaviors

“…Therefore, it is important to distinguish public health significance from the statistical significance of differences (see e.g., Hess et al., 2020). These results suggest that the choice of decision rule has little bearing on predicted choice shares, which accords with empirical evidence elsewhere (Buckell et al., 2021).…”

“…Then, the use of DFT outputs may be preferred, as it appeared to better depict behavior in this dataset. Moreover, using multiple decision rules is frequently recommended (Boeri et al., 2013; Buckell et al., 2021; de Bekker‐Grob & Chorus, 2013). Model outputs differed from single‐rule models when multiple decision rules were simultaneously used.…”

“…The typical assumption in choice modeling studies using stated preference data (also referred to as DCEs) is random utility maximization (RUM) (Soekhai et al., 2019), due to its basis in economic consumer theory and ease of estimation (Boeri et al., 2013; Soekhai et al., 2019). Random utility maximization has also been used with revealed preference (RP) data (e.g., Buckell et al., 2021; Buckell & Hess, 2019; De Corte et al., 2021). In health sciences, nearly 99% of studies make the RUM assumption 1 .…”

“…Previous studies introduced random regret minimization (RRM) to health economics (Boeri et al., 2013; Buckell et al., 2021; de Bekker‐Grob & Chorus, 2013; de Bekker‐Grob, Donkers, Bliemer, Veldwijk, & Swait, 2020; Dennis et al., 2020; Paul et al., 2018; Ryan et al., 2018). These studies indicate that whilst RUM is a good general assumption, allowing for RRM behaviors can offer improved model performance in some cases.…”

“…These studies indicate that whilst RUM is a good general assumption, allowing for RRM behaviors can offer improved model performance in some cases. In addition, RRM may generate different, policy‐relevant explanations of behavior, and may uncover that different individuals are best described by different decision rules (Boeri et al., 2013; Buckell et al., 2021; de Bekker‐Grob & Chorus, 2013). Random regret minimization model outputs differed, such as predicted choice shares (Dennis et al., 2020).…”