Prediction markets have become increasingly visible in financial markets over the past two years, fuelled by rapid growth in platforms such as Kalshi and Polymarket and their prominence around elections, monetary policy and geopolitical events. For institutional investors, however, their most immediate value may lie less in trading the contracts themselves than in the stream of continuously updated probability data they generate.
That distinction emerged strongly from a panel on “Prediction Markets and Forecasting Platforms” at the recent A-Team Group/Eagle Alpha Alternative Data Conference in New York. While relatively few delegates indicated they were actively using or evaluating prediction-market data, the discussion pointed to growing interest in incorporating it into macroeconomic forecasting, political analysis, investment research and risk management.From trading instrument to data source
Direct institutional trading remains limited. The panel discussion highlighted insufficient order-book depth as one constraint, with much of the professional activity currently concentrated around liquidity provision and market making rather than institutions using event contracts directly to express investment views.
The data generated by those markets presents a different proposition. Unlike a poll or periodic survey, prediction-market prices adjust continuously as participants respond to new information. Participants also have money attached to their forecasts, potentially providing a different measure of conviction from stated opinions.
That has made the data increasingly visible in discussions around elections and macroeconomic events. One observation from the panel was that references to prediction markets have become commonplace when investment committees and research teams discuss political developments or macro expectations. Their appeal lies partly in providing a continuously changing probability that can sit alongside conventional market indicators, polling and economic forecasts.
Election forecasting provides an example of how the data can complement existing sources. Traditional models can incorporate previous election results, demographics, party registration, campaign finance information and polling. Prediction-market data adds a faster-moving input. The panel heard how introducing it into an election forecasting model had helped the model respond more quickly to late-breaking events, where waiting for new polling could otherwise take days or weeks.
Testing the signal for alpha
For investment firms, however, useful forecasting information doesn’t necessarily translate into an exploitable trading signal.
Research discussed during the panel examined whether Polymarket prices around individual companies contained information that could subsequently be used to trade equities. The results were disappointing. Rather than prediction markets consistently identifying information before equity markets, the research suggested that information generally travelled in the opposite direction, with developments already known to stock-market investors subsequently reflected in prediction-market prices.
That finding puts prediction-market data into territory familiar to alternative-data users. A dataset doesn’t necessarily have to generate standalone alpha to be valuable. Its contribution may come from improving an existing model, providing an independent check on another signal or allowing investment and risk teams to quantify changes in expectations more quickly.
Building probabilities into existing models
Prediction-market probabilities could also serve as a continuous prior within existing investment models. Contracts around CPI releases or Federal Reserve decisions, for example, could become inputs into macro signals. Event contracts can potentially extend the concept further. A market estimating the probability of disruption in the Strait of Hormuz doesn’t have to be traded directly for its changing probability to inform positions in oil or commodity derivatives.
This creates a potentially different route into institutional investment processes. Rather than building strategies around prediction-market contracts, firms can incorporate the information contained in their prices into models that ultimately express positions through deeper and more liquid traditional markets.
Deciding when to trust the price
Using prediction markets as data sources also raises questions about what the price actually represents. Higher volume doesn’t automatically produce a better forecast, and market structure, position limits and participant composition can all influence outcomes. The panel discussed evidence suggesting that smaller position limits can sometimes produce accurate forecasts, while a relatively small group of well-informed participants may play a disproportionate role in moving incorrectly priced contracts towards fair value.
For institutional data teams, that creates a due-diligence challenge extending beyond simply acquiring a price feed. Liquidity, trading volume, contract construction, position limits and concentration of activity may all provide important context when assessing whether a probability is robust enough to use within an investment process.
Regulation adds another layer of uncertainty. Recent US court cases have produced conflicting decisions over federal and state jurisdiction for sports-related event contracts, with further litigation potentially determining where the boundaries lie. The panel drew a distinction between sports contracts and contracts with a clearer economic purpose, while also identifying elections and entertainment as areas where jurisdictional questions could persist.
For institutions consuming the data rather than trading the underlying contracts, questions around governance and provenance are likely to become increasingly important as prediction-market information moves closer to production investment processes.
The panel’s expectations for the next three to five years centred overwhelmingly on macroeconomic and political forecasting, with risk management another potential application. Prediction markets may therefore find their institutional role first as another component of the alternative-data stack: a continuously updating measure of probabilities that can be combined with market prices, economic statistics, polling, news and other signals.
The contracts themselves may take longer to establish a significant place in institutional portfolios. The information generated by the markets is already beginning to travel further.
Subscribe to our newsletter
