Prediction Markets for Academic Researchers

Prediction markets have moved from curiosity to legitimate research tool across multiple academic disciplines. Political scientists use them to study election dynamics. Economists use them to study information aggregation. Epidemiologists used them during COVID to track pandemic expectations. This guide explains how academic researchers can leverage prediction markets in their work, whether as data sources, experimental tools, or subjects of study.

Research Application 1: Forecasting Accuracy Studies

The most established academic use of prediction markets is studying their forecasting accuracy relative to other methods. Key research questions include:

• Do prediction markets outperform polls for election forecasting?
• How well-calibrated are prediction market probabilities?
• Does increased liquidity improve accuracy?
• What factors cause prediction markets to fail?

Available Data

Platforms like Polymarket provide historical price data, trading volumes, and resolution outcomes. This data can be used to construct calibration curves, Brier scores, and other accuracy metrics. The data is publicly accessible and machine-readable.

Research Application 2: Information Aggregation

Prediction markets are living laboratories for studying how markets aggregate dispersed information into prices. Research in this area explores:

• How quickly do prediction markets incorporate new information?
• Do prediction markets suffer from information cascades?
• How does market design (continuous vs. periodic trading, fees, anonymity) affect information aggregation?
• Can prediction markets elicit private information more effectively than surveys?

Research Application 3: Replication and Metascience

One of the most innovative uses of prediction markets is in metascience. Researchers have used prediction markets to forecast the outcome of replication studies. The key finding: prediction markets where researchers bet on whether studies would replicate were significantly more accurate than expert surveys or base rates.

This has implications for research funding, peer review, and the credibility of scientific findings. If prediction markets can identify which studies are likely to replicate, they could improve the efficiency of scientific resource allocation.

Research Application 4: Policy Evaluation

Prediction markets can serve as real-time measures of expected policy impact. Researchers studying policy can use prediction market prices before and after policy announcements to estimate the market's assessment of policy effects. This is similar to event studies in finance but applied to public policy.

Research Application 5: Behavioral Economics

Prediction markets provide a natural setting for studying behavioral biases in real-money environments. Research topics include:

• Favorite-longshot bias: Do traders overprice longshot outcomes?
• Anchoring: Do initial prices influence subsequent trading?
• Overconfidence: Do traders overestimate their accuracy?
• Herding: Do traders follow the crowd at the expense of their own information?

Research Application 6: Decision Markets for Organizations

Internal prediction markets (sometimes called "decision markets") can be used within research organizations to allocate resources, prioritize research questions, and improve forecasting. Studies of corporate prediction markets at companies like Google and Intel have shown that internal markets can outperform management committees for forecasting business outcomes.

Methodological Considerations

Key Academic Literature

• Arrow et al. (2008): "The Promise of Prediction Markets" - The seminal case for prediction market research, published by a group of leading economists.
• Wolfers & Zitzewitz (2004): "Prediction Markets" - Foundational survey of prediction market theory and evidence.
• Dreber et al. (2015): Using prediction markets to forecast replication outcomes in psychology.
• Page (2007): "The Difference" - Theoretical foundation for why diverse crowds outperform experts.
• Manski (2006): "Interpreting the Predictions of Prediction Markets" - Critical analysis of what prediction market prices actually mean.

Getting Started with Prediction Market Research

• Step 1: Familiarize yourself with major prediction market platforms and their data availability.
• Step 2: Define your research question and identify which prediction market data is relevant.
• Step 3: Collect and validate historical data. Most platforms provide API access or downloadable datasets.
• Step 4: Apply appropriate statistical methods for your analysis (calibration tests, Brier scores, regression analysis, event studies).
• Step 5: Consider the limitations of prediction market data (selection bias in which markets exist, liquidity variation, potential manipulation).

Frequently Asked Questions

Can I use prediction market data in peer-reviewed research?

Yes. Prediction market data has been published in top economics, political science, management, and computer science journals. The data is considered a legitimate source for empirical research.

Is prediction market data publicly available?

Most major platforms provide public access to current and historical market data. Some offer APIs for programmatic access. Data availability varies by platform, but Polymarket is among the most accessible.

How do I address the limitations of prediction market data?

Common limitations to acknowledge: selection bias in which markets exist, variation in liquidity, potential manipulation of thin markets, and regulatory changes that affect market participation. Standard robustness checks (excluding thin markets, comparing across platforms) help address these concerns.

Can I run my own prediction market experiment?

Yes, though real-money prediction markets may require regulatory compliance and IRB approval. Play-money prediction markets (like those used in replication studies) face fewer regulatory hurdles and can still produce useful data, though with weaker incentive properties.