RS is a Republican research firm with three arms. Published research, public tools and models, and direct strategy work with campaigns. The methodology below runs through all three. Read where the cycle sits today. Name where it is heading. Reverse-test against history. File the call on the record. RS publishes under the firm name, not an analyst byline. This page is the credential.
Reading where the electorate sits this week is the floor. Any firm with a polling subscription gets there. The read that decides cycles, and the read most firms skip, is where the electorate is heading. Which coalitions are quietly drifting. Which message frames are gaining structural advantage. Which districts come into play when the cycle hits them. RS is built around that question.
The work shows up in three places on this site. Published research files under the firm byline, dated and reverse-tested before publication. Public models tier every U.S. House and Senate seat, rank where the next campaign dollar generates the most marginal margin across the 2026 map, read sixty to ninety days forward against polling drift, and track the frame-elasticity deltas underneath the top-line. Direct campaign work applies the same methodology to one specific race, brief- or full-report tier, 72 hours from kickoff.
Win now is the immediate call. Position for the cycle after this one is the structural call. Strong campaigns do both at once. They move the consensus instead of chasing it. They allocate dollars where the next dollar generates measurable additional margin this cycle, and they invest in the message frames that own the cycle after. That is the firm's organizing principle. The arms below are how it gets applied.
The research arm files monthly. Each note is a dated piece on a coalition shift, a modeling technique, a prediction, or a positioning frame for the cycle ahead. Every piece carries a date stamp and the firm byline. Every piece closes forward, on what a 2026 or 2028 campaign should do about the finding.
The publishing standard is what "Reality before consensus" actually means. RS files before consensus forms. The record either holds or doesn't. No piece is taken down after the fact. Corrections are made in place with a dated correction note appended to the original. No client has veto power over published work, and no model is retroactively edited to align with a later development.
Reverse-testing runs before publication, not after. If a frame cannot explain what already happened, it does not ship. The cycle-level back-tests run against 2018, 2020, 2022, and 2024. Special-election overperformance is the live test between major cycles.
Katana is the firm's flagship national model. It reads the partisan reality of every U.S. House district and every state from precinct-level presidential returns aggregated on current district lines, weighted across the two most recent cycles to capture the current alignment without overfitting to a single year. The output is a signed magnitude per district, scaled against the national center, that anchors every downstream call. Built on our own pull, our own weighting choices, and our own reverse-test calendar. Refreshed cycle to cycle.
Katana is the spine. Everything else modifies around it. Race Ratings sits on top of Katana plus the structural factors (incumbency tenure and 2024 margin, fundraising parity, district economic conditions) and the forward-looking layer (positioning, candidate quality, cycle environment, narrative pressure). The Superseat Map applies Katana plus the persuadability cross-analysis to score where the next campaign dollar generates the most marginal margin. The Generic Ballot uses Katana's national aggregate as the structural anchor against polling drift.
Underneath: precinct-level returns going back four cycles, ACS five-year demographics for all 435 districts and 50 states, FEC quarterly filings refreshed on the cycle calendar, voter-file integration where the file is available, special-election overperformance tracked live between major cycles, and the cycle-trend signal that distinguishes a district drifting from one holding steady. The categories are visible. The weighting choices, segment definitions, and the math that turns the categories into a single number stay private. The methodology shown publicly is enough to verify what RS is doing. It is not enough to copy it.
Every model RS ships runs through the same architecture. Katana anchors the partisan baseline at precinct-level resolution. Structural factors (incumbency tenure and recent margin, fundraising parity, district economic conditions) modify around it. A forward-looking layer (positioning, candidate quality, cycle environment, narrative pressure) modifies again. Every model carries a reverse-test against prior cycles before it ships. Every output closes on a forward call. The math gets refreshed against current cycle inputs on a fixed monthly calendar, not when the result feels right.
One cut tiers every U.S. House and Senate seat across seven tiers, from Solid D to Solid R, with movement logged the moment it changes. Another ranks the 2026 map by where the next campaign dollar generates the most marginal additional margin. Another reads the generic ballot sixty to ninety days forward against polling drift, not just last week's average. Another tracks the frame-elasticity deltas underneath the polling top-line, so a campaign team sees which messages are gaining structural advantage before the cross-tabs catch up. Another decomposes a single state's legislative map to precinct level for a campaign with decisions to make this quarter.
Different cuts of the same underlying read. The factor categories and the resulting scores live on each tool's own page. The weighting choices, the segment definitions, and the math that turns inputs into a single number stay private. Enough to verify what RS is doing. Not enough to copy it.
Published: Race Ratings · Superseat Map · Generic Ballot · Sentiment Gauge · NE 2026 Targeting
The consulting arm applies the same methodology to one specific race. Three engagement shapes.
A district report runs the same inputs that drive the public models against one specific race. Brief tier is one page, one recommendation. Full tier is twenty pages: demographics, persuasion vectors, media market cost, candidate quality, positioning, message frames, and a 90-day forward call. 72 hours from kickoff either way.
See the report scopes