Early Signals in Decentralized Prediction Markets

Can Prediction Markets Anticipate Election Results More Effectively?

For decades, opinion polls have been the main tool used to understand the political preferences of a population. In recent years, however, prediction markets have become increasingly popular. On these platforms, participants buy and sell positions related to the probability of a particular outcome.

This paper examines whether prediction markets can identify electoral changes earlier than traditional polls. It does so by studying platforms such as Polymarket and Kalshi during Colombia’s 2026 presidential election.

The case is particularly interesting because the final published polls showed a clear advantage for Iván Cepeda. Prediction markets, by contrast, began to reflect the rise of Abelardo de la Espriella, who eventually won an exceptionally close election.

The results suggest that the market detected the change in the race roughly three weeks before the final outcome. It also produced estimates that were closer to the election result than the available polls, which considerably underestimated the support received by the winning candidate.

An important part of the case was the interruption of new polling during the final stage of the campaign. While the traditional polling signal remained frozen, prediction markets continued to update as participants reacted to new information, political events, and changing expectations.

The paper does not conclude that markets should completely replace polls. Prediction platforms may have limited liquidity, can be influenced by large traders, and may react too strongly to particular news events. Nevertheless, they can serve as a valuable complementary source of information, especially when polling is limited or unavailable. A more complete understanding of an election may therefore come from combining both instruments while recognizing the strengths and weaknesses of each.

Read the full paper on SSRN.