Elsevier

International Journal of Forecasting

Volume 35, Issue 1, January–March 2019, Pages 420-428
International Journal of Forecasting

When are prediction market prices most informative?

https://doi.org/10.1016/j.ijforecast.2018.05.005Get rights and content

Abstract

Prediction markets are a popular platform for the elicitation of incentivised crowd predictions. This paper examines the variation in the information contained in prediction market prices by studying Intrade prices on U.S. elections around the release of opinion polls. We find that poll releases stimulate an immediate uptick in trading activity. However, much of this activity involves relatively inexperienced traders, meaning that the price efficiency declines in the immediate aftermath of a poll release, and does not recover until more experienced traders enter the market in the following hours. More generally, this suggests that information releases do not necessarily improve prediction market forecasts, but instead may attract noise traders who temporarily reduce the price efficiency.

Introduction

Prediction markets allow individuals to trade on the outcome of future events. They are a popular forecasting method, as the participants have a strong incentive to acquire useful information and produce accurate forecasts. In addition, the prices produced by these markets aggregate dispersed information, thus harnessing the ‘wisdom of crowds’ Galton, 1907, Surowiecki, 2004.

But when are prediction market prices most informative? We answer this question by studying a rich dataset from the now-defunct prediction market Intrade. We analyse every transaction on U.S. elections (at both the national and state levels) between 2008 and 2012. Crucially, the data include anonymised trader identification. We then marry this data with information on poll releases from Gallup, the oldest polling firm in the U.S., which allows us to examine both trader activity in the hours surrounding poll releases — who traded and how much — and the effect of this activity on the accuracy, or efficiency, of prices.

We find that poll releases stimulate an increase in trading volume. However, this initial response to the poll release is made by the inexperienced traders (with less prior trading activity), and these traders seem to respond more to the incidence of a poll than to its content. As a result, we observe a significant decline in price efficiency in the hour immediately following a poll release. This decline is then interrupted in the subsequent hours by the arrival of more experienced traders, who correct this mispricing (Hanson & Oprea, 2009). This suggests that information releases may actually harm the prediction market price accuracy temporarily, by attracting the attention of less experienced noise traders (De Long, Shleifer, Summers, & Waldmann, 1990). In short, if an individual is basing their forecast on prediction market prices, it may be worth turning away after a significant information event.

This paper contributes to the literature on the accuracy of prediction market prices. Authors have compared prediction markets to opinion polls Chen et al., 2005, Leigh and Wolfers, 2006, Sjöberg, 2009, Vaughan Williams and Reade, 2016a, Wang et al., 2015, combined prediction markets with opinion polls Graefe et al., 2014, Rothschild, 2015, considered the predictive power of social media content D’Amuri and Marcucci, 2017, Huberty, 2015, Peeters, 2018, and also married prediction market forecasts with social media content Brown et al., 2018, Vaughan Williams and Reade, 2016b. There are perhaps three papers that are particularly closely related to ours. Croxson and Reade (2014) studied the accuracy of betting (prediction) market prices immediately following soccer goals, and found that prices respond almost instantaneously, and indeed accurately, to information. Another closely related paper is that by Page and Clemen (2013), who studied prediction market accuracy over time, and found that prediction market accuracy improved as the event in question approached. However, notwithstanding this general increase in prediction market accuracy over time, Page (2012) showed that in-play betting prices towards the end of matches overestimate the likelihoods of low probability outcomes (e.g., the losing team winning the match). We identify more noise around information events than Croxson and Reade (2014), and also show more bumps in the road to efficiency than Page and Clemen (2013).

The rest of the paper is structured as follows. Section 2 introduces our dataset, Section 3 details our methodology, and Section 4 presents our results. Section 5 concludes.

Section snippets

Data

We use data from the now-defunct Intrade prediction market between 2008 and 2012. The dataset begins after the 2008 US presidential election and runs through to the end of the 2012 presidential election. The overwhelming majority of the contracts over the four-year period are associated with US politics.

The dataset includes a timestamp (in Co-ordinated Universal Time, UTC) for each trade, along with unique identifiers for both the buyer and the seller, the number of contracts traded and the

Methodology

Our methodology involves running a number of linear regressions in order to determine the significance or otherwise of salient information around the time of Gallup poll releases. Considering prediction market movements on days when polls are released is relatively uninformative in general, since polls were released on most days in 2012. However, Gallup always released polls at 1 pm Eastern Time.1

Influence of polling outcomes on prediction market activity

We begin by considering whether the Gallup polls had any impact on the total amount of trading on Intrade. We aggregate by hour of the day and consider the total number of contracts traded each hour. The first column of Table 1 displays the result of regressing the total number traded per hour on dummies for the proximity of a Gallup poll release. The results suggest that there is a significant increase in trading activity for the four-hour window around the release of a Gallup poll, with the

Conclusions

Prediction markets often vie with opinion polls as forecasting tools for elections. This paper has studied the accuracy of prediction market prices around the release of opinion polls. We find an immediate but temporary decline in the price accuracy, as inexperienced traders respond noisily to the incidence of a poll rather than to its content, but more experienced traders then enter and correct the market in the following hours. More generally, it would appear that information releases do not

Alasdair Brown is a Senior Lecturer in Economics at the University of East Anglia. He completed his Ph.D. in Finance and Management at SOAS, University of London in 2012. Prior to this he obtained an MSc in Financial Engineering from Birkbeck College, University of London, and a BSc in Economics from the University of Birmingham. His main research interests are in behavioural finance, market microstructure, and forecasting.

References (28)

  • GaltonF.

    Vox populi

    Nature

    (1907)
  • HansonR. et al.

    A manipulator can aid prediction market accuracy

    Economica

    (2009)
  • HendryD. et al.
  • HendryD. et al.

    Automatic selection of indicators in a fully saturated regression

    Computational Statistics

    (2008)
  • Cited by (15)

    • Human-machine hybrid prediction market: A promising sales forecasting solution for E-commerce enterprises

      2022, Electronic Commerce Research and Applications
      Citation Excerpt :

      Unlike polls or expert panels that require traders to express their independent opinions, each trader in the prediction market can see the net effect of all other traders’ beliefs and the dynamic trend of those beliefs over time so that they can change their beliefs accordingly (Berg et al., 2008). Moreover, in a prediction market, payoffs are tied to the outcomes of future events so that individuals are well-incentivized to acquire information and produce accurate forecasts (Brown et al., 2019). Third, prediction markets can generate efficient market prices, although there are biases and mistakes in the market.

    • Forecasting: theory and practice

      2022, International Journal of Forecasting
      Citation Excerpt :

      Expectation polls try to exloit the first issue (Graefe, 2014), while prediction markets, as efficient aggregators of information, exploit both these issues to yield election forecasts (see also Sections 2.6.4 and 2.11.4). Several studies have proven the performance of these approaches (Berg, Nelson, & Rietz, 2008; Erikson & Wlezien, 2012; Williams & Reade, 2016; Wolfers & Zitzewitz, 2004), even studying their links with opinion polls (Brown, Reade, & Vaughan Williams, 2019). Practice has also developed econometric models (Fair, 1978) that exploit structural information available months before the election (e.g., the evolution of the economy or the incumbent popularity).

    • U.S. Politics from a multifractal perspective

      2022, Chaos, Solitons and Fractals
      Citation Excerpt :

      Most of the academic literature studying the predictive ability of prediction markets agree on the superior forecasting accuracy compared to polls or quotes, examples are [9,10,31,99,114]. Interestingly, [15] document that information efficiency temporally decreases with the release of polling results. This effect is argued to be caused by uninformed traders but diminishes rather quickly as informed traders utilize the caused noise.

    • Sonic Thunder vs. Brian the Snail: Are people affected by uninformative racehorse names?

      2021, Journal of Behavioral and Experimental Economics
      Citation Excerpt :

      Thus, as betting markets are designed to allocate resources and aggregate information, market prices forecast future events (Berg, Nelson & Rietz, 2008). As the participants in these markets profit from accurate predictions, they have strong incentives to acquire useful information, and thus, market prices are expected to accurately forecast the underlying outcome (Brown et al., 2019). In the prediction market literature, there is a wide consensus regarding the high accuracy of prediction market prices (see, e.g., Berg et al., 2008; Rothschild, 2015; Spann & Skiera, 2009; Vaughan Williams & Reade, 2016; Wolfers & Zitzewitz, 2004).

    • Expert performance and crowd wisdom: Evidence from English Premier League predictions

      2021, European Journal of Operational Research
      Citation Excerpt :

      We ask whether the consensus view outperforms the forecasting strategies of laypeople and experts? Evaluating crowd wisdom is a promising branch of the literature and we build on recent work evaluating the judgement of amateur online crowds (Brown & Reade, 2019; Brown, Reade & Williams, 2019; O'Leary, 2017; Peeters, 2018). Third, we investigate the efficiency of the forecasting strategies by evaluating the returns relative to bookmaker odds using different approaches.

    View all citing articles on Scopus

    Alasdair Brown is a Senior Lecturer in Economics at the University of East Anglia. He completed his Ph.D. in Finance and Management at SOAS, University of London in 2012. Prior to this he obtained an MSc in Financial Engineering from Birkbeck College, University of London, and a BSc in Economics from the University of Birmingham. His main research interests are in behavioural finance, market microstructure, and forecasting.

    James Reade is an Associate Professor in Economics at the University of Reading. Prior to this he was a Lecturer in Economics at the University of Birmingham, and completed his Ph.D. in Economics at the University of Oxford in 2007. His research interests are in sports economics and forecasting.

    Dr. Leighton Vaughan Williams is Professor of Economics and Finance and Director of the Betting Research Unit and of the Political Forecasting Unit at Nottingham Business School, Nottingham Trent University.

    View full text