Research Data
Point-in-time index and constituent data for academic research on prediction-market indices. Free for non-commercial research use.
To request access, email research@beliefsystems.xyz – details below.
The Program
Belief Systems publishes rules-based indices built on prediction markets. This page describes the research data program: what data exists, how it is versioned, and how to request it.
The program exists because the questions this data raises – how event contracts should be aggregated, what belief indices measure, how they behave as an asset class – are open, and they will be answered by independent researchers, not by us. Researchers who use the data stay fully independent: no compensation, no advisory titles, no strings.
The Data
Complete index-level history for every index from each index's inception, at intraday resolution (observations approximately every 30 minutes). Timestamps in UTC. Download instantly from the Index Data page – no request needed.
Basket membership and target weights per composition version, plus a dated, immutable log of every composition change with the chain-linking values that keep index levels continuous. Also downloadable instantly.
Per-window component breakdowns – the weight, price, and resolution status of each constituent at every observation, not just the aggregate level. Available to academic researchers on request, under a data agreement.
Custom cuts for a specific research design – particular windows, particular events, particular fields – delivered as flat files with field definitions. Anything delivered is yours to keep for replication.
The program currently covers 13 published indices; the earliest observations begin January 2026. Index levels, compositions, and reconstitution histories are self-serve downloads on the Index Data page; the on-request tier below is for constituent-level detail and custom extracts. Index construction rules are documented in the published methodology.
Reproducibility and Durability
Every observation carries the composition in force when it was computed, so you can reconstruct what an index held – and at what weights – at any timestamp in the sample, not just today.
Resolved markets stay in the history at their settlement values, and removed markets persist in prior compositions. Historical records are never edited, restated, or deleted.
Computation-rule changes receive a new version string, recorded with every subsequent observation and documented in the published methodology.
Data granted for research use stays available. We understand a working paper can take years to publish; access and the underlying series will not disappear underneath it. And because data is delivered as flat files, anything already delivered stays in your hands regardless.
Open Questions
Questions we think the data can speak to. We have views; we would rather have evidence. Critique is as welcome as use – if you think the construction is wrong, we want to hear precisely how.
Our indices weight constituents by a function of market depth and traded volume. Whether that is the right way to aggregate event-contract prices into a single belief measure is an open question – we would like to see it answered.
How do index-level measures of macro and policy expectations lead, lag, or complement survey-based measures of the same quantities?
What are the return, volatility, and correlation properties of event-risk indices relative to traditional asset classes?
How do aggregate, index-level measures perform against polls, models, and expert forecasts on realized political and macro outcomes?
How to Cite
In your acknowledgments:
Data provided by Belief Systems (beliefsystems.xyz).
In a data or references section:
Belief Systems (2026). Belief Index Data, snapshot [YYYY-MM-DD] [data set]. Retrieved from https://beliefsystems.xyz/data/archive/[YYYY-MM-DD]/
Dated snapshots are permanent, so the archive URL you cite will resolve to exactly the files you used. Machine-readable citation metadata (CITATION.cff) ships in every bundle.
If you use the data in a paper, tell us – we read everything, and we will list it here.
Request Access
For index levels, compositions, and reconstitution histories, skip the email – download them now. For constituent-level detail or custom extracts, email us with your name, institution, and a rough description of the research question. No form – just email.
research@beliefsystems.xyzAccess is free for non-commercial academic research. For commercial data or methodology licensing, see institutional licensing.