The March Report
Gambit Forecaster’s March Report: 2026 Senate and Gubernatorial Races
View Senate Data – View Gubernatorial Data
With two months of the cycle now in the books, the 2026 midterms are moving from baseline mapping into early calibration. The start of primaries in only two days signal the true start of the season. The March simulations largely reaffirm what January and February suggested: the Senate remains structurally tilted toward Republicans, most gubernatorial races continue to sit inside stable partisan environments, and the real action is still concentrated in a relatively small set of contests rather than spread evenly across the board. What has changed is not the center of gravity of the forecast, but the model’s willingness to admit uncertainty in places that previously resolved as near-certainties.
At roughly nine months out from Election Day, the model still leans heavily on historical fundamentals and primary polling data. Candidate fields are still settling, high-quality general election polling remains uneven, and campaign spending has not yet produced the kind of durable signals that force big re-ratings. But compared to February, March provides a sharper diagnostic of where volatility is emerging. The topline probabilities have only moved incrementally, yet variance has expanded in several key races (especially in the Senate where a handful of states are now generating coherent, if still unlikely, Democratic paths that were largely absent earlier in the cycle).
In this March Report, we’ll break down which races are consolidating, where dispersion is expanding, and which early assumptions are already being stress-tested by new information. From Senate contests where Democratic upside is appearing through narrow but repeatable pathways, to gubernatorial races where institutional and candidate effects still dominate the fundamentals, the next phase of the cycle will be about whether or not these pockets of uncertainty widen.
Changes To Methodology
Gambit Forecaster has made two major structural changes to our working election models:
First, I made structural changes to the calibration of weighted polling averages to be more rigorous and consistent across separate races. This ensures that comparable polling conditions translate into comparable modeled ranges, and it prevents any one contest from appearing artificially volatile (or artificially certain) due to scaling artifacts. Factors such as time decay and poll source quality are now weighted and examined more systematically, while other inputs have been loosened to avoid overfitting early-cycle noise.
Second, I added new conditions including a national environment and “state blocs,” which are meant to emulate correlated movement between states. Previously, State A and State B in a simulation could move without any relationship to one another (even if they are geographically adjacent and historically respond similarly to national shifts). More broadly, states were allowed to behave too independently from simulation to simulation, producing maps that could be internally inconsistent even when the national environment was clearly favorable to one party. Under the new structure, each simulation draws a national-level shock that shifts the environment across all states, and then layers in additional regional bloc-level shocks so that groups of states with similar political behavior tend to move together within the same simulated universe. State blocs include the Midwest (Minnesota, Wisconsin, Michigan, Ohio, Pennsylvania), New England (Maine, New Hampshire, Vermont), the Southeast (Georgia, North Carolina, South Carolina, Virginia), and the Central & Southwest (Texas, New Mexico, Arizona, Nevada, Colorado). A final state-specific shock remains in place of course to preserve idiosyncratic variation, but outcomes are no longer treated as entirely independent across the map. This produces a more realistic simulation space: when the model generates a strong Democratic or Republican environment, it is more likely to appear as a coherent regional-and-national pattern rather than scattered, uncorrelated flips.
Senate Forecast & Analysis
Thirty-five seats are on the ballot, including all of Class II and two special elections in Florida and Ohio. Republicans are defending 22 of those seats, Democrats 13, with most states having solid ideological leanings. The 2026 Senate map is structurally favorable to Republicans, and Democrats would need a near-perfect night and at least four pickups to reclaim the majority. Again, as mentioned in previous reports (see January and February here), hitting 50 seats would not be enough for Democrats, since the vice presidency currently rests with the GOP, leaving them in the minority unless they reach 51 outright.
As of the March simulations, the picture is looking better for Democrats, all things considered. Republicans are still favored at 95.4 percent to retain the Senate, while Democrats sit at just 4.6 percent. Democrats are showing incremental gains from previous months, from the 99.8-0.2 split of the January Report, and the 98.11-1.89 split of the February Report.
Rating Changes: Alaska (From Likely R -> Lean D), North Carolina (Solid D -> Likely D)
The most common outcome has shifted 1 into the Democrats favor: Republicans 53, Democrats 47. That result appears with enough frequency to serve as the modal outcome of the model, and offers new pathways for Democrats to potentially assemble a governing majority. With more tangible information, several of these same contests now generate meaningful dispersion when previously outputting certainties. North Carolina is a perfect example, as in previous months they’ve been near-certain in Democrats hands. Only now in March has there been movement in Republicans favor in the respective race.
The most consequential change again finds itself in Alaska, which has shifted from a Republican stronghold into a Dem-leaning, contentious race. Former U.S. Rep. Mary Peltola’s Senate campaign’s popularity in the polls, alongside the states unique ranked-choice voting system, is what makes this possible. Alaska, who have adopted ranked-choice voting in recent years, now functions as one of the primary volatility drivers within the model and accounts for a number of of Democratic win scenarios. Georgia, by contrast, has nested in firmly, consolidating as a firmer Democratic hold and reducing downside risk rather than creating upside. Democrats cannot afford to lose Georgia while also chasing multiple Republican pickups, especially given the national popularity of Jon Ossoff.
Again, states like Texas and Ohio with Republican advantages are finding Democratic opponent gains, which will make latter-year rather exciting. In Texas specifically, the Democratic primary for U.S. Senate is crucial to further performance. Talarico undeniably performs better in general election polling and majority-GOP counties; Crocket has incredible credibility with black voters, especially within the Dallas-Fort Worth metropolitan area. Whoever wins will greatly affect the future outlook of the race. These states are not toss-ups in a conventional sense, but they now appear frequently enough in competitive simulations to matter. Their role is additive rather than decisive.
As stated in previous reports, states like Michigan are seemingly more than in play for Republicans, and the same can be said for Democrats in Maine. Therefore, Democrats still require a clean sweep of competitive states and at least one or two genuine upsets beyond Maine to reach their needed threshold. As of March, it seems more than likely the Senate will remain in Republican hands for the rest of the second Trump administration.
Check out Gambit’s newsletter!
Gubernatorial Forecast & Analysis
The 2026 gubernatorial map remains broadly stable in comparison to the Senate results. With methodological updates, March’s simulations provide great insight on an emerging flow. Where previous month’s apparent competitiveness was understated or overhyped. Unlike the Senate, the gubernatorial landscape is highlighted by state-specific dynamics and a sharper divide between genuine swing races and structural locks.
At a topline level, the March gubernatorial simulations are similar to previous months, with minimal changes showing marginal gains for either side. That said, more states are showing signs of swing-ability.
New ratings: Kansas (Likely R -> Solid R), Michigan (Solid D -> Likely D), Ohio (Solid R -> Likely R), Vermont (Lean R -> Likely R)
Kansas fully folds under the weight of the GOP. The Laura Kelly effect seems to have worn off in how the model is responding. This is great for Republicans, as it settles the rumblings for a race that has behaved somewhat volatile in the past decade. Kansas is the largest shift by far, and earns a new rating of Solid R.
Several races that appeared marginally interesting in January now look more settled. Michigan, for example, continues to consolidate as a Democratic hold, with Democrats winning over 99 percent of simulations. Minnesota and Wisconsin also remain safely in the Democratic column, but more competitive than previously echoed in previous reports.
Vermont and New Hampshire remain their unusual elastic selves. Despite both state’s consistent Democratic lean at the federal level, the gubernatorial races continues to favor Republicans by a wide margin. March’s results mirror prior months, and that again confirms the point that a candidate-driven and institutionally specific race matters more compared to a mere partisan one.
Pennsylvania also remains a firm Democratic hold in the gubernatorial model. Their governor, Josh Shapiro, seems slated to make a run for the 2028 Presidential election and will be running for re-election to his seat this year. He is performing incredibly well in polling, where he is reaching as much as +18 over potential Republican opponents. On the Republican side, states like Florida and Ohio remain locked down. February does not introduce new uncertainty in any of these contests, reinforcing that their January classifications were not provisional.
March Review
The March simulations have reinforced the information taken from the 2026 cycle so far. Similar to previous simulations, Republicans continue to hold a clear structural advantage in the Senate. And, most gubernatorial races remain anchored in stable partisan environments. The forecasts will move incrementally, and large-scale changes (especially as the year goes on) are not likely to happen.
In the Senate, February marked the first clear evidence of variance expansion that we are only now beginning to explore. While Republican control remains overwhelmingly likely, the model is no longer resolving the chamber as a near-universal certainty. Rare but coherent Democratic paths now appear in the simulations, driven by increased volatility in a handful of races rather than broad-based movement. This does not signal a competitive Senate map, but it does indicate that the cycle is beginning to loosen at the margins, particularly in states where structural assumptions are being stress-tested by emerging uncertainty.
Gubernatorial landscape have found its footing. If February confirmed that January’s polarized outcomes were not a product of early modeling noise, March confirmed the extent of the pattern. Most races remain decisively one-sided, but we are beginning to see more movement and expansion within the model.
For now, the March Report suggests a cycle defined by constraint rather than chaos. The paths to power are narrow, the number of decisive contests is limited, and most outcomes remain highly predictable. Where movement does exist, it is largely confined to the margins of already-competitive environments rather than spreading across the map.
Check out the Senate forecast data here >
Check out the Governor forecast data here >
