Choice blindness manipulation detection as indicator of risk preference stability? An online experiment to inform urban drainage and flood risk policy.

Online consultation is a convenient format for public risk preference elicitation. Substantial instability and inconsistencies cast doubts about the reliability of the obtained preferences for policymaking. Given the parallels between choice blindness manipulation (CBM) detection and risk preference (in)stability, I explore in this study explores whether CBM detection can indicate risk preference. An online randomized controlled experiment with 330 UK residents served to test the coincidence of CBM and risk preference stability along with common predictors, incl. task-, domain-, and individual-specific variables. Using 2 different stimuli to communicate risks and outcomes in elicitation gambles, covering five urban drainage and flood prevention outcomes, the results do not support CBM detection as an indicator of preference stability. Only about 1 in 4 participants detected at least one CBM and 1 in 10 manipulations were detected. Risk preferences were moderately stable across trials but inconsistent across treatment conditions; icon array stimuli led to risk-seeking and numeric stimuli led to risk-averse preferences. The significant predictors of CBM detection included task stimuli, choice latency, risk numeracy, age, and household flooding or waterbody pollution domains. Significant predictors of risk preference stability were gamble distinctness and risk numeracy, as also the only joint predictors of CBM detection and preference stability. By contrasting with theoretical explanations and earlier findings in literature, gaps to address in future research were identified. The empirical findings highlight a strong influence of choice architecture on preference construction also in this study, questioning neutrality and reliability of online risk preferences elicitation for policymaking.