Poliscope
Beta
Login

Coverage & limitations

What's accurate, what's documented, and what's being investigated

Poliscope is a transparency platform — which means we owe you honesty about what we're confident in and what's still rough. This page enumerates every documented limitation, with the context that makes the trade-offs visible.

Launch coverage window

Currently surfaced cycles: 2016–2026.

Cycles before 2016 are excluded from the default browse experience while we resolve the structural FEC earmark-disclosure gap that predates the standard tp=15E tag adopted in the mid-2010s. Pre-2016 candidates with significant Joint Fundraiser Committee operations undercount by 30–50% in our reconstruction (Obama 2012, Romney 2012, etc.) — the donor→ ultimate-recipient dollars don't exist in our data sources for those cycles.

The historical data is still in our database and accessible via direct API calls, but it's not surfaced in cycle dropdowns or list views until we can deliver an honest pre-2016 reconstruction. Modern cycles (2016+) must match FEC's published totals within ±0.5% on our golden acceptance set; the launch coverage window is where we stand by that claim.

Headline accuracy commitment

Politician headline totals use FEC's own candidate totals where available. Committee headline totals use FEC committee totals where available, otherwise they are labeled as Poliscope FEC-derived rollups.

  • Candidate pages — Total Raised, Individual Contributions, Itemized, Unitemized: Sourced directly from FEC's /v1/candidate/{cand_id}/totals/ API rollup and stored in fec_candidate_cycle_totals. This is the authoritative roll-up FEC publishes (principal committee + authorized committees + JFC allocation, all combined). The production launch gate requires the golden candidate set to match within ±0.5%; it passed on May 22, 2026.
  • Standalone committee pages: Preferred source is FEC's /v1/committee/{cmte_id}/totals/ receipts field for that cycle. Where that row is unavailable, the page falls back to fact_contribution_summary and should not be read as a FEC-exact headline.
  • Donor breakdown panel (Top Donors, Industries, Employers): Aggregated from the FEC bulk and electronic contribution filings we have ingested, including earmark disclosures via tp=15E where available. The donor list is refreshed through the same pipeline that rebuilds FCS, edges, metrics, and serving tables; launch validation checks the public serving tables against their fact-table derivation.

Documented limitations

Pre-2016 conduit/JFC pass-through donations

Documented

Affects: Politicians from 2008–2014 cycles

What we show

Direct individual contributions and PAC contributions for the candidate's principal committee.

What's limited

Joint Fundraiser Committee (JFC) pass-through donations from before ~2016 are undercounted. Pre-2016 cycles undercount candidates with significant JFC operations by 30–50%.

Context

FEC's earmark disclosure code (transaction_tp=15E) only became standard practice around 2014–2016. Before then, donations routed through joint fundraisers were filed only under the JFC's books, with no separate disclosure linking the original donor to the ultimate recipient. We exclude JFC pass-throughs to avoid double-counting on modern cycles where 15E disclosures exist; older cycles get under-counted because the disclosure simply doesn't exist in our data sources.

501(c)(4) and 501(c)(6) donor identities

Documented

Affects: Dark money, nonprofits

What we show

Nonprofit-to-PAC grants (Schedule I), nonprofit officers (Part VII), nonprofit political spending (Schedule C), Super PAC inflows from named c4s.

What's limited

Original individual or corporate donors who fund 501(c)(4) and (c)(6) organizations are not disclosed by IRS rule. We can show what the c4 spent and where it sent grants, but not who paid into it.

Context

This is a federal disclosure-rule limitation, not a data-quality issue. No public-facing transparency platform — including OpenSecrets, FollowTheMoney, or ProPublica — has this data unless it's surfaced via journalistic investigations or court filings. We surface those when available.

Nonprofit officer cross-linking

Investigating

Affects: /nonprofits/[ein] and /dark-money/source/[ein] (currently hidden from main nav)

What we show

All 2.9M Form 990 officer records are loaded with names, titles, and compensation.

What's limited

Officer names are not yet linked to their corresponding person entities, which means clicking through from a nonprofit officer to their other roles (committee treasurer, lobbyist, legislator) is currently broken. Pages are accessible by direct URL but are temporarily hidden from public navigation until the entity-resolution work lands.

Context

Naive name matching produces 53% false matches because of common-name collisions across centuries of public service. We're building a multi-signal matcher (employer overlap, address, role timing) before approving cross-domain merges. This is a deliberate slowdown, not laziness — bad merges silently corrupt relationship data.

Gift sponsor identification

Documented

Affects: Gifts and travel pages

What we show

Member name, gift description, value, sponsor name as filed.

What's limited

Gift and travel sponsors appear as filed-text names rather than entity_id-linked organizations. Cross-linking from a gift sponsor to the sponsor's other activities (lobbying, donations, contracts) requires manual lookup.

Context

Sponsor names on gift filings are free-text and inconsistent across years. Resolving them to canonical entities requires the same multi-signal matching used elsewhere; queued behind 990-officer ER.

FEC bulk-data publishing delay (donor breakdown freshness)

Documented

Affects: Donor breakdown panel on politician and committee detail pages, recent/current cycles (2024 and 2026)

What we show

Headline Total Raised, Individual Contributions, Itemized, and Unitemized numbers come from FEC's API rollup endpoint and match FEC.gov exactly. The Top Donors list shows individual donors aggregated from FEC's bulk-data publications.

What's limited

On recent/current cycles, the running sum of the Top Donors panel can still run behind the headline FEC itemized total if FEC's API has fresher late-filed totals than the latest public bulk files. The headline FEC numbers are current where FEC API totals are loaded; the donor list follows the bulk files we have ingested.

Context

The launch remediation refreshed 2024/2026 bulk data and the production launch gate passed the strict ±0.5% FEC golden checks. The daily pipeline timer is installed and active; the first unattended scheduled run still needs to be checked on May 23, 2026 before we describe it as proven daily operation.

Standalone committee FEC-total coverage

Documented

Affects: Committee detail pages

What we show

When an authoritative FEC committee totals row exists for a committee/cycle, the headline uses that FEC API receipts value. Otherwise the page falls back to Poliscope's FEC-derived contribution summary for that committee/cycle.

What's limited

Some committee/cycle pairs have no FEC totals row returned by the FEC API, or are conduit/JFC cases where a single receipts field is not the right semantic benchmark. Those pages are labeled as FEC-derived rather than FEC-exact.

Context

The committee-total loader now covers active and surfaced committees beyond the older e-file-only worklist; the May 22 launch gate saw 13,278 current-cycle committee-total rows for 2026. Candidate pages use the candidate totals API and are gated separately.

Non-campaign source freshness

Documented

Affects: Lobbying, government contracts, SEC filings, trades, and company pages

What we show

Federal lobbying registrations, USAspending contract awards, SEC filings and insider transactions, congressional trades, gifts, disclosures, nonprofits, pensions, and dark-money traces where available.

What's limited

These sources have different publication schedules and source-level quirks. USAspending, LDA, and SEC are not FEC-style cycle rollups, so accuracy depends more on current source refresh, row counts, entity linkage, and date sanity than on a single public total.

Context

Launch remediation published LDA 2026 facts, USAspending FY2026 awards, SEC filings and insider transactions, and blocked impossible future dates from public metrics. The production launch gate passed these freshness checks on May 22, 2026.

ActBlue / WinRed standalone-committee benchmarks

Documented

Affects: Conduit committee detail pages

What we show

Pass-through earmarks (15E disclosures) are correctly attributed to ultimate recipients, not to the conduit. Donor pages correctly credit the ultimate recipient candidate.

What's limited

ActBlue and WinRed's own headline 'total raised' is intentionally not benchmarked against a single FEC field, because the right comparison metric is 'receipts minus earmark transfer disbursements' which has fuzzy edges. We've left the conduits' standalone-committee pages without a strict numeric gate for now.

Context

This affects the conduit pages themselves, not any candidate or donor page. The semantics for measuring conduit accuracy will get tightened in a follow-up sprint with a transaction-level FEC API comparison.

2026 cycle data sparsity

Active

Affects: Politicians and committees with 2026 as their current cycle

What we show

2026 filings loaded from the public source refreshes that have completed.

What's limited

The 2026 cycle is in progress. Mid-cycle FEC data is sparse by design — committees haven't filed Q2/Q3 reports yet. Page views for 2026 candidates may show smaller numbers than expected because the cycle isn't done.

Context

Numbers will fill in as the cycle progresses. Switch the cycle filter to 2024 for a complete picture of the most recent finished cycle.

Bill semantic search coverage

Active

Affects: Bill semantic search and bill detail pages for older Congresses

What we show

Full text and metadata for all bills 108th–119th Congress. Roll-call votes for the 118th House.

What's limited

Vector embeddings for semantic search currently cover ~70% of bills (118th and 119th Congress are complete; 117th and earlier are still being processed in the background). Older bills can be found via keyword search but won't surface in 'similar bills' or topic-based discovery until the embedding catch-up completes.

Context

Embedding generation is a one-time cost per bill; the catch-up is in flight on a continuous worker.

Why we ship anyway

Every existing political-money tool has the same trade-off: pick a subset of disclosures to track well, document the rest as gaps. We made different choices about what to track well, and we're explicit about the gaps that result. Where the existing tools differ from us:

OpenSecrets

Strong editorial context and long-running campaign-finance coverage. Poliscope's candidate headline totals are tied directly to FEC candidate-total rollups, so our stated number is the FEC-reported receipt total rather than an editorially adjusted campaign status value.

FollowTheMoney

Strong on state-level data, thin on federal. No IRS 990 integration. We integrate both layers in the same graph.

ProPublica Nonprofit Explorer

Best-in-class 990 search but zero campaign finance data. We connect nonprofit grants → Super PAC inflows → independent expenditures → election outcomes.

FEC.gov

Authoritative on the disclosed transactions. No tax data, no lobbying, no contracts, no entity resolution. The same donor with three address variations appears as three different donors. We resolve them.

Going deeper