Belief Elicitation without Reduction

Beliefs are important in a wide range of decision-making scenarios and accurately eliciting them is critical to studying those decisions. In recent years, belief elicitation work has focused on developing procedures that are theoretically incentive-compatible (expected utility is maximized by truth-telling) for increasingly varied types of preferences. The Binarized Scoring Rule (BSR), generalized by Hossain & Okui (2013), is widely recognized as state-of-the-art in this literature; however, other papers raise concerns about applications of the BSR yielding relatively low rates of truth-telling Danz et al. (2020). We propose to test one explanation for the BSR's failure to produce higher rates of truth-telling through comparing its performance to an alternative procedure.

While not explicitly stated in Hossain & Okui (2013), the reduction of compound lotteries is necessary to incentivize the truthful revelation of beliefs under the BSR. The Rank-Ordered Elicitation (ROE) procedure we introduce does not require the reduction of compound lotteries for incentive compatibility, but rather a weaker statewise monotonicity requirement Azrieli et al. (2018). We aim to test 1) whether the ROE outperforms the BSR in truth-telling, and 2) whether non-reducers of compound lotteries account for the difference in performance.