We study market scoring rule (MSR) prediction markets in the presence of risk-averse or risk-seeking agents that have unknown yet bounded risk preferences. It is well known that if agents can be prescreened, then MSRs can be corrected to elicit agents’ beliefs. However, agents cannot always be screened, and instead, an online MSR mechanism is needed. We show that agents’ submitted reports always deviate from their beliefs, unless their beliefs are identical to the current market estimate. This means, in most cases it is impossible for a MSR prediction market to elicit an individual agent’s exact belief. To analyze this issue, we introduce a measure to calculate the deviation between an agent’s reported belief and personal belief. We further derive the necessary and sufficient conditions for a MSR to yield a lower deviation relative to another MSR. We find that the deviation of a MSR prediction market is related to the liquidity provided in the MSR’s corresponding cost-function prediction market. We use th...

中文翻译：

---

走上“艺术”科学之路：市场评分规则中的风险偏见

我们在存在规避风险偏好但风险偏好有限的规避风险或寻求风险的代理人的情况下研究市场评分规则（MSR）预测市场。众所周知，如果可以对代理进行预筛选，则可以对MSR进行校正以引起代理的信念。但是，不能总是筛选代理，而是需要一种在线MSR机制。我们表明，代理商提交的报告始终偏离他们的信念，除非他们的信念与当前市场估计相同。这意味着，在大多数情况下，MSR预测市场不可能激发出单个代理商的确切信念。为了分析此问题，我们引入一种措施来计算代理人报告的信念和个人信念之间的偏差。我们进一步得出一个MSR产生相对于另一个MSR较低偏差的必要和充分条件。我们发现，MSR预测市场的偏差与MSR相应的成本函数预测市场中提供的流动性有关。我们用...