Response Times and Subjective Complexity of Food Choices: A Web-Based Experiment Across 3 Countries

Abstract

Accurate collection of response times is one of the main advantages of web-administered stated choice experiments and it can be thought of as a behavioral indicator of cognitive effort. We use data from a food choice experiment administered across three countries and estimate a panel Mixed Multinomial Logit Model to obtain individual-specific utility weights. These are used to construct two utility-based measures of contextual choice complexity, which are combined with subjective measures of cognitive resources as well as indicators of opt-out selection. We first develop and then test hypothesized effects of complexity at the level of single choice task and choice sequence on response times. By using a log-linear random effects model with choice task response-time as dependent variable, we isolate these effects from other background variables. Results suggest that as our measures of complexity increase, so do response times and such effects are robust across the three countries. We argue that these results broadly support the validity of web-based choice surveys to measure food preference. We suggest that computers can help improve survey design by implementing algorithms to improve the overall efficiency of choice tasks design, for example, by using adaptive design algorithms that control cognitive challenges in accordance with the respondent’s predicted ability to tackle cognitive effort.

My History or Our History? Historical Revisionism and Entitlement to Lead

“…14 Results indicate that randomization was effectively implemented, as participants are balanced across treatments on observable characteristics (see Table A1 in the Supplementary Material). 15 When designing and pre-registering our experiment, we decided to flag and remove time outliers, an approach that has been adopted by many researchers as standard protocol (Atzori et al 2021;Fodeman, Snook, and Horgan 2020;Jedinger 2018;Matjašič, Vehovar, and Manfreda 2018). There has been some scholarly debate regarding the consequences of excluding time outliers, in particular, since inattentive respondents have been found to be different (younger and less educated) than the broader population (Alvarez et al 2019;Berinsky, Margolis, and Sances 2014).…”