|
Title
|
|
|
|
Efficient prediction markets
| |
|
Author
|
|
|
|
| |
|
Abstract
|
|
|
|
| A core question in financial economics is whether markets are informationally efficient i.e., whether asset prices accurately reflect all available information. Although the empirical literature on market efficiency is vast, the subject is still heavily debated among academics and practitioners. A fundamental issue with informational efficiency is that it is virtually untestable in traditional financial markets. For example, in an ideal world, researchers would compare stock prices on stock markets with their true values to check whether they are aligned or not. However, as true values of stocks are never available, this is not possible. This untestability of market efficiency is referred to as the joint hypothesis problem. In this dissertation, we try to make an original, unconventional contribution to the market efficiency debate by studying prediction markets. Prediction markets are platforms where people can bet on the outcome of future events, like a presidential election or a football game. Prediction markets have many characteristics that make them interesting research labs. Their main advantage is that the outcomes of the events are exogenously revealed, the market prices collide with reality. This allows researchers to systematically compare market prices with terminal values to detect mispricing, which is not possible on stock markets and circumvents the joint hypothesis problem. This dissertation contains three empirical chapters. In the first, we review 40 years of literature on mechanical trading strategies in sports prediction markets. Many individual studies claim to have found profitable trading strategies which implies inefficient market pricing. However, when we consider the entire literature, the evidence is consistent with an efficient market where profit opportunities are chance results. Furthermore, we argue for more meta-scientific reflexes and put forward a hurdle rate of |z|>3 to benchmark the statistical significance of empirical results. The second empirical chapter studies the evolution of the UK fixed odds betting market between 2000 and 2018. This period is of particular interest as it coincides with the rise of online gambling. We find that over this period, transaction costs decreased very significantly, both statistically and economically. Furthermore, we document a decrease in the favorite-longshot bias, a persistent anomaly in prediction market research. The third empirical chapter tests whether time series momentum, a well-known irregularity in traditional financial markets, is also present in prediction market data. We find that a time series momentum effect is indeed present and by leveraging the prediction market characteristics, we show it is consistent with behavioral underreaction and not a rational premium for variance or skewness. |
| |
|
Language
|
|
|
|
English
| |
|
Publication
|
|
|
|
Antwerp
:
University of Antwerp, Faculty of Business and Economics
,
2022
| |
|
ISBN
|
|
|
|
978-90-5728-750-3
| |
|
Volume/pages
|
|
|
|
191 p.
| |
|
Note
|
|
|
|
:
Annaert, Jan [Supervisor]
:
De Ceuster, Marc [Supervisor]
| |
|
Full text (open access)
|
|
|
|
| |
|