Statistical properties of volume and calendar effects in prediction markets

Valerio Restocchi, Frank McGroarty, Enrico Gerding

Research output: Contribution to journalArticlepeer-review

Abstract

Prediction markets have proven to be an exceptional tool for harnessing the “wisdom of the crowd”, consequently making accurate forecasts about future events. Motivated by the lack of quantitative means of validations for models of prediction markets, in this paper we analyze the statistical properties of volume as well as the seasonal regularities (i.e., calendar effects) shown by volume and price. To accomplish this, we use a set of 3385 prediction market time series provided by PredictIt. We find that volume, with the exception of its seasonal regularities, possesses different properties than what is observed in financial markets. Moreover, price does not seem to exhibit any calendar effect. These findings suggest a significant difference between prediction and financial markets, and offer evidence for the need of studying prediction markets in more detail.
Original languageEnglish
Number of pages20
JournalPhysica a-Statistical mechanics and its applications
Early online date29 Mar 2019
DOIs
Publication statusE-pub ahead of print - 29 Mar 2019

Keywords

  • Prediction markets
  • Political markets
  • Stylized facts
  • Long memory
  • Power-law behavior

Fingerprint

Dive into the research topics of 'Statistical properties of volume and calendar effects in prediction markets'. Together they form a unique fingerprint.

Cite this