Interval Elicitation of Forecasts in a Prediction Market Reveals Lack of Anchoring "Bias"

TL;DR

This study investigates how forecasters update their predictions in a prediction market when they are initially unaware of the current market state, revealing minimal anchoring bias and similar accuracy regardless of visibility conditions.

Contribution

The paper introduces an interval elicitation method in prediction markets and finds that it does not increase anchoring bias or reduce forecast accuracy.

Findings

Forecasters' initial estimates are similar regardless of market visibility.

Forecast accuracy is comparable between conditions with and without market state visibility.

Forecast updates tend to move estimates closer to the market forecast when the state becomes known.

Abstract

In an online prediction market, forecasters who could not see the current state of the market until they made their own separate estimates moved their estimates closer to the market forecast when the current state of the market became known. Their first edits to the market forecast were very similar to the first edits of forecasters who could always see the current state of the market, and forecasters in both conditions had similar accuracy. These results suggest that our more elaborate forecast elicitation method might not improve forecasts and that any anchoring on the state of the market does not constitute an error in judgment.