Control & Ownership of the US Federal Government
Federal money and federal votes, 2024 cycle. An attempt to measure the distance between the people who authorise a government and the people who finance it. (State and local layers are not yet built.)
What one adult costs
Every dollar in the 2024 federal cycle — contributions and independent expenditure together — divided by the voting-age population of the state it was spent in. The dark portion of each bar is money spent for or given to candidates; the red portion is money spent to defeat one.
Who supplies it
The same money, re-sorted by where it came from rather than what it was spent on. One square is $5 million raised by a sitting member of Congress in the 2024 cycle. Teal is money from outside the state or district that member represents.
What the money actually looks like
In the 2024 federal cycle, the dominant channel of political money was not contribution to candidates. It was expenditure about them, by committees the candidate does not control and is not accountable for.
Put plainly: for every dollar given to a federal candidate in this dataset, $8.25 was spent about them by someone else. The single largest category is money spent to defeat a candidate rather than to elect one. This is the structural legacy of Citizens United v. FEC (2010) and SpeechNow.org v. FEC (2010), which together permitted unlimited independent expenditure by committees that accept unlimited contributions.
Concentration
Largest independent-expenditure committees, 2024
| Committee | FEC type | Spent | Txns |
|---|---|---|---|
| FF PAC | V | $509.6M | 490 |
| Make America Great Again Inc. | O | $377.0M | 465 |
| WinSenate | O | $311.3M | 578 |
| Congressional Leadership Fund | W | $216.8M | 2,096 |
| Senate Leadership Fund | O | $211.1M | 317 |
| HMP (House Majority PAC) | W | $199.8M | 1,420 |
| America PAC | O | $173.7M | 1,186 |
| Americans for Prosperity Action | V | $162.0M | 3,081 |
| Preserve America PAC | O | $112.3M | 30 |
| SFA Fund, Inc. | V | $81.8M | 360 |
Ten committees account for roughly $2.36 billion of independent expenditure — more than four times the total contributed directly to every federal candidate combined. Note the transaction counts: Preserve America PAC moved $112.3 million in 30 transactions, an average of $3.7 million each.
The most important result on this page contradicts the thesis the report set out to explore. Across the 2024 cycle, money correlates only weakly with roll-call behaviour (r = 0.20 Republican, 0.27 Democrat — roughly 4–7% of variance explained). Party affiliation accounts for almost everything roll-call voting has to explain. It is stated here, in full, ahead of the test itself, because a report engineered to confirm its own premise would be worthless as evidence. The full record of what was predicted and what survived sits in the Falsifiability Register.
The test that could have refuted this report
A naive reading of the IOCE paradigm predicts that concentrated money buys concentrated control — that heavily-funded legislators should vote differently. We pre-registered that prediction and tested it against 1,497,926 individual roll-call votes across the 117th–119th Congresses.
| Party | Members | Mean party unity | r (PAC share × unity) | r² |
|---|---|---|---|---|
| Republican | 262 | 92.3% | 0.195 | 0.038 |
| Democrat | 253 | 96.9% | 0.267 | 0.071 |
The prediction failed. PAC funding share explains between 4% and 7% of the variation in how often a member votes with their party. Party affiliation alone explains the overwhelming majority. If this report existed to confirm that money buys votes on the floor of Congress, it would have to report that it did not find it.
It does not follow that money has no effect. It follows that money does not operate through defection on recorded votes — which is the one channel this dataset can see. Roll calls occur after agenda-setting, candidate selection, committee assignment and bill drafting have already filtered what can be voted on. A measurement that finds nothing at the last stage of a pipeline says little about the earlier stages. We treat this as unresolved, not as exoneration, and we say so in the Uncertainty Taxonomy.
The register at a glance
Five hypotheses, each with its refutation condition fixed before the data was queried. Two failed their own test and are published as failed. Every verdict links to the working:
The three that survived are all structural — they describe where the money comes from and who controls it. Both that failed were behavioural — they predicted that the money changes what legislators do. That is the shape of the result, and it is not the shape the report set out to find.
Where divergence actually shows up
The ownership/control gap in this data is not primarily behavioural. It is structural — visible in who supplies the resources rather than in who defects on votes:
- Authorisation is equal; finance is not. Every eligible voter holds an identical share of the mandate. The financing distribution is concentrated enough that the largest 1% of expenditures carry half the dollars.
- Most money is unaccountable by construction. 89% of the money measured here flows through committees a candidate may not legally coordinate with.
- The dominant use of money is negative. More was spent to defeat candidates than to support them.
This is what OCDI-Gov is designed to quantify: the distance between the distribution that authorises an office and the distribution that finances it.
Financed by people who cannot vote for you
Aggregating all 58,208,756 individual contribution records to committee × state, and matching 519 sitting members to their committees, gives the clearest divergence signal in the report:
This one survives the honesty test that H2 failed. It was pre-registered with a refutation condition — median below 25% — and the observed value clears it in both chambers. The median senator draws roughly seven of every ten itemised dollars from people who hold no share of the mandate they are financing.
One caveat, stated plainly: this is a share of itemised contributions. Sub-$200 donations are not universally itemised, so this population skews toward larger donors. The ratio is unaffected — both numerator and denominator come from the same itemised pool — but it should not be read as a claim about all donors. Full detail in the Falsifiability Register.
Scope and honesty about coverage
Version 1.1 covers federal money and federal roll calls, and the title now says so. An earlier draft was titled "Federal, State, Local and Individual", which described the intended programme rather than what had been built; it was corrected on 10 August 2026. Specifically not yet included: state campaign-finance filings, lobbying disclosures (Senate/House LDA), 501(c)(4) issue spending, media-ownership concentration, and ballot-measure funding. Each is a planned layer with an identified public bulk source; see the roadmap.
FEC individual-contribution records are public and include donor names, street addresses, employers and occupations. This report deliberately stores and publishes aggregates only — by committee, state, ZIP and amount band. Re-publishing a searchable donor dossier would add nothing analytically and would expose private individuals to targeting. The database schema has nowhere to put a donor's name.
Read this to get an overview of the report. The data was assembled for a separate project, to test how these numbers hold up against a thesis - and the test is reported above whichever way it came out.