1: \begin{abstract}

2: Prediction markets are long known for prediction accuracy. However,

3: there is still a lack of systematic understanding of how prediction

4: markets aggregate information and why they work so well. This work

5: proposes a multivariate utility (MU)-based mechanism that unifies

6: several existing prediction market-making schemes. Based on this

7: mechanism, we derive convergence results for markets with myopic,

8: risk-averse traders who repeatedly interact with the market maker.

9: We show that the resulting limiting wealth distribution lies on the

10: Pareto efficient frontier defined by all market participants'

11: utilities. With the help of this result, we establish both

12: analytical and numerical results for the limiting price for

13: different market models. We show that the limiting price converges

14: to the geometric mean of agents' beliefs for exponential

15: utility-based markets. For risk measure-based markets, we construct a

16: risk measure family that meets the convergence requirements and show

17: that the limiting price can converge to a weighted power mean of

18: agent beliefs. For markets based on hyperbolic absolute risk

19: aversion (HARA) utilities, we show that the limiting price is also a

20: risk-adjusted weighted power mean of agent beliefs, even though the

21: trading order will affect the aggregation weights. We further

22: propose an approximation scheme for the limiting price under the

23: HARA utility family. We show through numerical experiments that our

24: approximation scheme works well in predicting the convergent prices.

25: \begin{keyword}

26:     Dynamic mechanisms, Information elicitation, Prediction markets, Econometrics, Machine learning

27: \end{keyword}%    aversion.

28: \end{abstract}

29: