Longshots, overconfidence and efficiency on the Iowa Electronic Market

Abstract

We study the forecast accuracy and efficiency of popular "binary" prediction markets. Such markets forecast probabilities for future states of the world (e.g., election winners) by paying off $0 or $1 depending on the realized state (e.g., who actually won). To assess accuracy, forecast probabilities must be compared to realization frequencies, not individual realizations. We use Iowa Electronic Market (IEM) data to test efficiency against two alternative propositions from behavioral finance: the longshot bias and the overconfidence bias (which yield opposing predictions). No longshot bias appears in IEM markets. Nor does overconfidence influence prices at short horizons. However, overconfident traders may bias prices at intermediate horizons. While the markets are efficient at short horizons, non-market data indicate some intermediate-horizon inefficiency. We calculate Sharpe ratios for static trading strategies and document returns for dynamic trading strategies to assess the economic content of the inefficiencies.

Designing Prediction Markets to Forecast Multi-Stage Elections: The 2022 French Presidential Election

“…Absent significant asymmetric hedging demand, these should approximate true probabilities. Berg and Rietz (2019) studied WTA market accuracy. For price ranges and probabilities typical in elections, they found that WTA markets accurately forecast probabilities.…”

Arbitrage in Political Prediction Markets

“…Betfair ( Sethi, 2015aSethi, , 2016Berg and Rietz, 2019). In particular, markets that impose transaction costs and/or profit fees can lead to mispricing (Berg and Rietz, 2019). Note that the generally popularized notion that prediction markets are efficient stems primarily from the efficiency of the IEM.…”

“…Note that the generally popularized notion that prediction markets are efficient stems primarily from the efficiency of the IEM. However, the IEM has key features, such as account limits for avoiding any one trader to become too large, no trading fees, and a simple arbitrage structure, which allow the market to operate efficiently in spite of irrational trader behavior (Berg and Rietz, 2019). That there exist arbitrageurs in other prediction markets as well appears untested in the recent literature.…”

Moderators for Partition Dependence in Prediction Markets

“…There does exist a great body of work to discuss and explain some of the forecasting errors that appear systematically with the occurrence of (cognitive) biases in prediction markets (Sonnemann et al 2011;Berg and Rietz 2018;Cipriano and Gruca 2014;Cowgill et al 2009;Page 2012;Luckner and Weinhardt 2008;Woodland and Woodland 2011). Some examples are the favorite-longshot bias (tendency to overvalue longshots and undervalue favorites) (Snowberg and Wolfers 2010;Page 2012), confirmation bias (tendency to ignore conflicting information) (Pouget et al 2017;Cipriano and Gruca 2014), the overconfidence bias (overestimation of own skills or results) (Berg and Rietz 2018), or-subject of this research-the partition dependence bias (Sonnemann et al 2013). There are different explanations why each bias occurs based on market mechanism, market liquidity, information spread, or motivation of participants.…”