What criteria do databases like the Heritage Election Fraud Database use to count a 'case' or 'prosecution'?
Executive summary
The Heritage Foundation’s Election Fraud Database presents what Heritage calls a “sampling of proven instances” of election misconduct and treats each entry as an individual “case” that—by Heritage’s own language—ended in a finding of wrongdoing or left results “sufficiently in question to be overturned” [1]. Independent reviewers and news organizations, however, say the database mixes convictions, prosecutions, charged individuals, and sometimes items with limited public documentation, and therefore users must read entries and primary sources closely [2] [3].
1. How Heritage labels a “case”
Heritage describes each database entry as a “proven instance of election fraud,” framing a “case” as an instance that culminated in an official finding of wrongdoing or a result deemed sufficiently tainted to be overturned; that is the explicit definitional anchor the Foundation uses to justify including an item [1].
2. What constitutes a “prosecution” or conviction in the database
Heritage’s public presentation emphasizes prosecutions and convictions—its entries often note charges, indictments, and court outcomes—but independent analysts warn that the database’s public-facing summaries do not always distinguish between mere allegations, pending prosecutions, plea deals, acquittals, or convictions, so the simple presence of an entry does not uniformly mean a final conviction without consulting the linked primary source [2] [3].
3. Sources Heritage relies on and how entries are documented
Heritage compiles items from news reports, court filings and other public records and provides case pages with citations to primary sources in many instances; the Brennan Center and third-party compilers have used those links to audit Heritage’s claims, underscoring that the Foundation relies on publicly available documents rather than unseen datasets [2] [4] [3].
4. Scope, sampling and temporal framing
The database is explicitly a sampling rather than a census: it was launched in 2017 and reaches back decades to collect illustrative examples across states and election types, which means Heritage can and does aggregate cases from many years to reach its totals rather than limiting itself to a single recent election cycle [1] [5] [3].
5. Independent critiques and methodological limits
Analysts at the Brennan Center, USA Today and academic commentators have repeatedly flagged methodological issues: the database mixes offense types (from individual absentee-ballot errors to multi-person conspiracies), sometimes lacks enough public documentation on particular entries for outsiders to verify Heritage’s characterization, and can overstate the policy-relevance of isolated incidents; the Brennan Center concluded the dataset is a “grab‑bag” and cautioned against using it as proof of widespread, outcome‑changing fraud without case‑by‑case verification [6] [2] [3]. Brookings and Newsweek note that Heritage’s long time horizon and broad inclusion criteria produce totals that look large only because they pool decades of small-scale cases across hundreds of millions of ballots [5] [7].
6. How to interpret counts responsibly
A responsible reading recognizes three things the sources converge on: Heritage counts discrete entries as “cases” and emphasizes prosecutions/convictions in its framing [1]; the compilation is drawn from public reporting and court materials but is not a statistically representative survey of all elections [4] [2]; and independent reviewers have found items with limited documentation or mixed outcomes, so every headline number should be checked back to the case page and primary sources before using it to support broader claims about systemic fraud [3] [2].