DIME Assimilation Report

Stanford's Database on Ideology, Money in Politics and Elections corrects more of this report's known defects than any other free source. It also has a hard access problem. Both are set out here before any ingestion.

What DIME is

DIME (Bonica, Stanford University Libraries) compiles 850 million itemized political contributions from 1979 to 2024 covering local, state and federal elections, with entity resolution, geocoding, and estimated ideology scores (CFscores) for 156,000+ candidates, 37,000+ committees and 36 million individual donors.

Licensing is unusually permissive for an academic dataset — ODC-BY and CC0 both permit redistribution and derivative work. Attribution is required and is given on this page and in the Provenance Chain.

The single most important fact: it is mostly state and local

This report has an open defect: only 3 of 50 states publish machine-readable campaign finance, so 47 states are missing. DIME's recipient file breaks down as follows.

What will be assimilated

Ingestion is planned in three tranches, easiest and most corrective first.

Tranche 1 — recipients (14 MB compressed)

One row per candidate/committee per cycle. Loads into a new usv_dime_recipients table. The fields that matter:

• Crosswalk keys — ICPSR (100% populated) joins directly to this report's usv_members and 1.5M roll-call records; FEC.ID (47,710 rows) joins to usv_committees; bonica.rid / bonica.cid track the same actor across cycles and across levels of government.
• CFscores — recipient.cfscore (100% populated), contributor.cfscore, recipient.cfscore.dyn, irt.cfscore, dwdime.
• Money totals — total.receipts, total.indiv.contrib, total.pac.contribs, unitemized, and a full independent-expenditure breakdown (ind.exp.for/against, non.party.ind.exp.for/against, comm.cost.for/against, party.coord.exp).
• Geography and outcome — state, district, district.partisanship, district.pres.vs, gen.elect.pct, winner.

Tranche 2 — itemized contributions, 2024 cycle

Contribution-level records with geocoding to congressional district. Streamed and aggregated in flight under the same privacy rule already applied to FEC data: no donor names or addresses are persisted.

Tranche 3 — contributors and multi-cycle history

The donor file and prior cycles back to 1979, enabling the trend analysis this report has so far refused to attempt on a single cross-section.

Defects this corrects

The access problem

So the two access routes are complementary and neither is sufficient alone. The v2 mirror was used to verify the schema, field completeness and seat-type distribution reported on this page — every figure above is measured from the real file, not from documentation. The v4.0 files require one manual browser download by an operator, after which ingestion is fully automated.

What DIME does not fix

• Public itemized files exclude CRP/NIMSP-sourced records. Stanford states this explicitly and directs researchers to request them separately for academic use. State itemized coverage in the public files is therefore partial, even though recipient-level totals remain.
• CFscores are estimates, not measurements. They are inferred from contribution patterns. Under this report's convention they must be rendered as modelled, never as measured, however well validated.
• It is money in, not money spent. DIME records contributions to recipients. It does not record where independent expenditure was targeted, so the presidential-allocation assumption on The Price of a Vote is improved but not eliminated.
• It does not resolve causation. A better ideology measure makes H2 testable, not answered. Money still correlates with position without that establishing direction.