Are Final Market Prices Sufficient for Information Aggregation? Evidence from Last-Minute Dynamics in Parimutuel Betting

Hiroaki Hanyu, Shunsuke Ishii, Suguru Otani, Kazuhiro Teramoto · 2025

Read Are Final Market Prices Sufficient for Information Aggregation? Evidence from Last-Minute Dynamics in Parimutuel Betting by Hiroaki Hanyu, Shunsuke Ishii, Suguru Otani, Kazuhiro Teramoto (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-14645.txt.

Introduction

Hiroaki Hanyu and three co-authors show in 2025 that final prices in parimutuel betting markets do not tell the whole information story. In such markets all stakes pool and winners share the pool, so odds move until betting closes. Analysts often read beliefs from final odds alone.

Using interim odds from horse racing, the authors find returns depend on the odds path, not only the endpoint. Horses whose odds fell in the last minutes earned higher realised returns than horses with similar final odds. A two period information model explains the pattern: informed bettors wager late on private signals. Final price sufficiency fails when timing itself carries information, a caution for market based belief measurement.

Catalog record

CREATORHiroaki Hanyu, Shunsuke Ishii, Suguru Otani, Kazuhiro Teramoto
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE55 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-14645

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Are Final Market Prices Sufficient for Information Aggregation? Evidence from Last-Minute Dynamics in Parimutuel Betting Thanks: The authors are grateful to Pierre-André Chiappori, Nobuyuki Hanaki, Yuichiro Kamada, Hayato Kanayama, Michihiro Kandori, Shunya Noda, Yasuhiro Shirata, and Ryoji Yanashima for their insightful comments and suggestions. We also thank the participants of the SWET2025 Market Design session for helpful feedback and discussions. This research uses data from the JRA-VAN database (a licensed service providing structured JRA data), with official permission granted by JRA-VAN for academic use. Financial support from JST ERATO Grant Number JPMJER2301 is gratefully acknowledged. Hiroaki Hanyu, Shunsuke Ishii, Suguru Otani, Kazuhiro Teramoto Abstract Most betting market models employ static frameworks that condition decisions on final odds. Using a unique dataset of interim odds from Japanese horse racing, this study examines the validity of such static analyses by asking whether there is a systematic relationship between expected returns and the trajectory of odds. We find that returns are negatively related to last-minute changes in odds, and that these late movements attenuate the favorite-longshot bias by weakening the correlation between final odds and returns. These patterns suggest that informed bettors place wagers at the final stage based on private signals, leaving surprises in final odds. Keywords : Information aggregation, Parimutuel mechanism, Betting markets, Favorite-longshot bias JEL code : G14, D47, D83, L83 1 Introduction Parimutuel betting markets offer a well-defined empirical setting for analyzing equilibrium behavior under uncertainty. One of the most enduring puzzles in these markets is the favorite-longshot bias (FLB), first documented by Griffith 1949 . The FLB refers to the systematic tendency for wagers on longshots—horses with high odds and low perceived winning probabilities—to generate lower average returns than wagers on favorites—horses with low odds and high perceived chances of success. This pattern stands in sharp contrast to conventional notions of market efficiency ( Thaler and Ziemba 1988 ) . A large body of research has attempted to account for this anomaly through equilibrium models that link final odds to win probabilities ( Weitzman 1965 , Ali 1977 , Hurley and McDonough 1995 , Jullien and Salanié 2000 , Bradley 2003 , Snowberg and Wolfers 2010 : e.g.,) . 1 1 1 A comprehensive review of empirical work estimating risk preferences using field data is provided by Barseghyan et al. 2018 , who highlight betting markets as a key setting for identifying such preferences from aggregate behavior. See also the surveys by Thaler and Ziemba 1988 , Hausch and Ziemba 1995 , Ottaviani and Sørensen 2008 , and Jullien and Salanié 2008 . For experimental and theoretical perspectives, see Plott et al. 2003 , Koessler et al. 2012 , Kajii and Watanabe 2017 , and Gillen et al. 2017 . More recent studies, including Gandhi and Serrano-Padial 2015 and Chiappori et al. 2019 , develop models with heterogeneous bettors and use structural estimation to recover risk preferences, probability perceptions, and subjective beliefs from the relationship between final odds and realized outcomes. This study is motivated by the central question whether the static relationship between final odds and realized outcomes is sufficient for identifying bettors’ risk preferences and beliefs. While existing studies have substantially advanced our understanding of pricing in betting markets, they largely rely on static frameworks in which bettors are assumed to condition their choices on final odds. However, this assumption is questionable as parimutuel betting markets, unlike financial markets where limit orders are feasible, do not allow bettors to act contingent on final odds. In such markets, strategic delays by informed bettors in placing their wagers can prevent efficient information aggregation and thereby generate the FLB, independently of bettors’ risk preferences or probability perceptions ( Ottaviani and Sørensen 2009 ) . To address our research question, we empirically test the validity of inferences that rest on the assumption that final odds serve as sufficient statistics for bettors’ decisions. More specifically, we analyze the entire evolution of odds throughout the betting period to assess whether outcomes depend not only on final odds but also on the trajectory by which those odds are reached. If final odds are truly sufficient statistics, then horses with identical final odds should, on average, yield identical expected returns regardless of the path of odds movements. Any systematic deviation from this benchmark would indicate that odds dynamics contain predictive information beyond what final odds alone can capture. For our analysis, we employ a comprehensive dataset from the Japan Racing Association (JRA) that covers all centrally administered races from 2004 to 2023. A distinctive feature of this dataset is the availability of high-frequency interim odds, recorded at five-minute intervals from the opening of the betting window until one minute before post time. These interim observations allow us to trace the real-time evolution of market expectations and to examine whether, and in what ways, the trajectory of odds—particularly in the final stages of betting—affects subsequent realized returns. In parimutuel betting markets, final odds and interim odds are calculated in the same way—based on the cumulative distribution of wagers—but they serve markedly different functions. Final odds are determined at the close of betting and directly govern payout calculations, thereby playing a dual role as both a summary of aggregate market expectations and as payoff-determining prices. Interim odds, by contrast, are snapshots of cumulative bets at intermediate points. Although they may capture evolving expectations, they do not enter the payout rule and thus lack a direct pricing function. Because of this, conventional models of betting markets have typically abstracted from interim odds and focused exclusively on final odds. The contribution of our study is to move beyond this convention by testing whether expected returns depend not only on final odds but also on the trajectory of odds movements during the betting period. To this end, we extend the standard regression framework used to detect the FLB—where realized returns are regressed on final odds—by incorporating measures of interim odds dynamics as explanatory variables. This approach enables us to ask whether horses with identical final odds but distinct interim trajectories yield systematically different average returns, thereby providing a direct empirical test of path dependence. Exploiting the five-minute update structure of the JRA data, we construct multiple indicators of odds changes and identify the phases of the betting process in which path dependence is most pronounced. Our main findings reveal a systematic relationship between expected returns and odds movements in the final five minutes before post time. In particular, realized returns are negatively related to last-minute changes in odds: horses experiencing a late surge in popularity—manifested as declining odds—tend to yield higher realized returns than those with identical final odds but no such movement. This pattern suggests that late-stage odds dynamics reflect the behavior of informed bettors who strategically place their wagers at the very end of the betting period in response to private signals not yet incorporated into market prices ( Ottaviani and Sørensen 2006 , Ottaviani and Sørensen 2009 ) . 2 2 2 The systematic relationship we document between expected returns and last-minute odds movements should be interpreted as an in-sample correlation rather than evidence of ex-ante predictability or arbitrage opportunities. Bettors cannot revise their actions based on final odds, nor can they submit limit orders contingent

Read the complete text