Belief Elicitation: Limiting Truth Telling with Information on Incentives

Abstract

Belief elicitation is central to inference on economic decision making. The recently introduced Binarized Scoring Rule (BSR) is heralded for its robustness to individuals holding risk averse preferences and for its superior performance when eliciting beliefs. Consequently, the BSR has become the state-of-the-art mechanism. We study truth telling under the BSR and examine whether information on the offered incentives improves reports about a known objective prior. We find that transparent information on incentives gives rise to error rates in excess of 40 percent, and that only 15 percent of participants consistently report the truth. False reports are conservative and appear to result from a biased perception of the BSR incentives. While attempts to debias are somewhat successful, the highest degree of truth telling occurs when information on quantitative incentives is withheld. Consistent with incentives driving false reports, we find that slow release of information decreases truth telling. Perversely, our results suggest that information on the BSR incentives substantially distorts reported beliefs.