Comparing Prediction Market Mechanisms Using An Experiment-Based Multi-Agent Simulation

Prediction markets are an interesting instrument to draw on the “wisdom of the crowds”, e.g., to forecast sales or project risks. So far, mainly two market mechanisms have been implemented in prediction markets, the continuous double auction and logarithmic market scoring rule. However, the effects of the choice between these two market mechanisms on relevant variables such as prediction market accuracy are not fully understood. These effects are relevant as faulty prediction market outcomes might cause wrong decisions. This work contributes via an experiment-based simulation model to understand the mechanism-related effects and to direct further laboratory experiments. Our results show, that the mechanism decision does matter. Due to the higher amount of trades and the lower standard deviation of the price, the logarithmic market scoring rule seems to have a clear advantage on a first view. Taking the accuracy error as an independent variable, the effects are not as straightforward and depend on the environment and actors.