Grade B — ranks ahead of 85% of House members. Highest: Donor Capture (90th percentile) — their money comes more from individuals, and from their own state, than most. Lowest: Legislative Activity (28th percentile) — they sponsor and advance less real legislation than most.
Representative Craig Goldman, a Texas Republican, sponsored 13 bills this Congress, mostly on energy, taxes, and health care. His furthest bill, a resolution that cancels a Department of Energy rule setting stricter energy standards for commercial refrigerators and freezers, became law. Goldman cast 138 recorded votes on policy, voting Yea 72% of the time and Nay 28%, including a Yea on the GENIUS Act, which sets rules for digital coins tied to a fixed value. He raised $1,756,535 this cycle, with 67% from individual donors rather than PACs, and 85% of those donations with an address on file came from Texas.
First we rank every district by need — the same 7-factor score (poverty, low income, no health insurance, unemployment, food assistance, high rent costs, and high mortgage costs) shown in the District Need card. Then we rank every district by how much federal contract and grant money it actually receives per person. Delivery looks at the gap between those two rankings: is this district's funding rank higher or lower than its need rank? A high-need district with a low funding rank has a big gap — money isn't reaching where it's needed most. Because rural and urban areas receive federal money very differently (one big highway grant looks huge per person in a small rural district, tiny in a large city), we never compare that gap nationally — only against other districts with similar population density, rural vs. rural, city vs. city. The score is where a district's gap ranks among those similar-density peers: 100 means it's getting the most funding relative to its need among districts like it, 0 means the least. A known limitation: because the gap starts from need, the highest-need districts have to be among the best-funded just to score in the middle, so Delivery scores, and overall grades, run lower in high-need districts. In our own check, 21% of House members from the highest-need fifth of districts get an F, compared with 1% from the lowest-need fifth. Because high-need districts are disproportionately Black, 6 of the 8 members representing majority-Black districts get a D or F. Federal contract and grant money also tends to follow existing institutions (contractors, universities, large hospitals) that a member can't create on their own. We publish the score as built and flag this here; how Delivery accounts for need may change in a future revision.
Two questions, averaged together. First: what share of a member's campaign money comes from individual people rather than PACs? Second: of the individual donations we can trace to an address, what share comes from the member's own state — the people they actually represent? That second number isn't compared raw: a member from a big, wealthy state like California naturally raises more at home than one from Wyoming, and members in close races attract money from across the country because their seats are targeted. So we compare each member's home-state share with what's typical for a member from a similar-sized donor state in a similarly competitive seat, and score how far above or below that they land. We used to score small-dollar donations instead, but that mostly rewarded national fame — the highest scorers were members with big national followings, most of whom raise the large majority of their money outside their own state. Party leaders are scored on the first question only, since raising money nationally is part of their job.
We look at every bill a member has sponsored this Congress, straight from Congress.gov, and keep only real legislation — ceremonial resolutions, post office namings, gold medals, and "National ___ Day" designations are left out. The score has two halves. The first is how many real bills a member introduced, compared with everyone in their chamber — anyone can write and introduce a bill. The second is how far those bills got — real committee action (a hearing or markup, not just a referral), passing their chamber, becoming law — but compared only with members of their own party, because whichever party holds the majority controls which bills advance. Comparing across parties there would mostly measure who's in the majority, not who's working. The Speaker and each party's floor leader and whip aren't scored on this: they run the chamber, and by long-standing practice bills carry other members' names.
We rank every member nationally on each of the three scores above, then average those three rankings together for one composite score (two, for the party leaders noted above). Members are then graded on a curve: the top 10% get an A, the next 25% get a B, the middle 30% get a C, the next 25% get a D, and the bottom 10% get an F.
This letter grade is not a judgment of a member's character or overall job performance — it measures three specific, factual things: whether federal funding reaches their district relative to its need, whether their campaign money comes from individuals in their own state, and how much real legislation they're moving. A member can score low here while doing plenty well outside what's measured, and score high here while their supporters and critics still disagree sharply about them.
District Need is a single 0 to 100 score: we combine 7 signs of hardship — poverty, low income, no health insurance, unemployment, food-assistance (SNAP) use, high rent costs, and high mortgage costs — ranking each one against every other district nationally and averaging those rankings. A higher score means more combined economic hardship than most other districts — it's relative to peers, not a raw severity scale. It's the yardstick Delivery is measured against above (does funding match need?), but it never scores the member directly or pushes their grade up or down on its own. Voting Alignment (shown above, next to the composite score) measures the gap between a member's ideology rank, based on their real voting record, and their district or state's own partisan-lean rank. It used to be part of the composite, but a small gap doesn't necessarily mean good representation — it can just mean a landslide seat — and a large gap doesn't necessarily mean bad representation — it can mean a swing-district member voting independently of local political pressure. Because the score itself can't tell those cases apart, it's shown as a fact about the member, not folded into a value judgment. Legislative Effectiveness Score data (bills introduced vs. enacted, from the Center for Effective Lawmaking at thelawmakers.org, covering both the House and Senate) is shown for context but not scored: the most recent published scores cover the 118th Congress, so they describe the previous Congress, not this one — the grade's own Legislative Activity component uses current-Congress bills instead, and never borrows that peer-reviewed name. The Bill Activity card's raw counts (every bill and resolution a member sponsored this Congress, pulled directly from Congress.gov) are shown as-is; the filtered, weighted version described above is what actually counts toward the grade. Vote Attendance (missed roll-call votes this Congress, House and Senate) is shown too, but not part of the grade for a different reason than Voting Alignment above: the direction here isn't ambiguous — showing up is part of the job — but the record can't distinguish a member who simply didn't show up from one on medical leave or running a campaign for another office, and we'd rather show the real number than guess at which explanation applies to whom.
Every race starts from how that district or state has actually voted in recent elections — its baseline partisan lean. We shift that baseline three ways: a small, measurable bonus for whoever already holds the seat (incumbents really do outperform the baseline, and we adjust the bonus for open seats where nobody is running as the incumbent); a shift toward whichever party is doing better nationally right now, based on generic-ballot polling; and, for Senate races, a smaller adjustment for differences in candidate fundraising. Add those up and you get one party's expected share of the two-party vote, which converts directly into a win probability and a rating label (Safe, Likely, Lean, or Toss-up).
The seat-count ranges shown for the full House and Senate aren't just those probabilities added up. National mood doesn't move district by district independently — when it shifts, it tends to shift in the same direction almost everywhere at once. We simulate that shared shift 10,000 times to get a realistic range, instead of a falsely narrow one.
None of this uses our own Accountability Index grade, and it isn't a guess — the model is backtested against what actually happened in 2018, 2020, and 2024 before being trusted on 2026, and it correctly called each of those elections' outcomes over 90% of the time.
Each member page includes a short paragraph explaining their grade in plain English. It's generated once per member (not created live when you load the page) by an AI model grounded in real data — one real, recent vote and the bill's own official summary from the Congressional Research Service, the district's real Census numbers, and the member's real campaign-finance totals — rather than asked to interpret a bare score on its own. It states what happened (how the member voted, what the bill actually does, where the campaign money came from) and is deliberately instructed never to characterize whether a vote was good or bad for the district; that judgment is left to the reader.