Blue Chip Quality Polling
As serious midterm election watchers await the results in the Michigan Democratic Senate primary, I wanted to get several documents off my plate and into your hands. Many readers of my Substack, Charlie Cook Politics, will find them helpful, or at least interesting. I confess to being a polling snob and see way too many people, including political reporters and news organizations that should know better, cite and publicize polls that are mediocre at best, and others downright shoddy, when good ones are now available. Polls are not commodities; they are not of equal quality or should be given equal credibility, no more than if someone said all houses or automobiles are of equal quality. Some polls are far more methodologically rigorous than others, their sponsors spending serious money to get it right; sometimes the cost of a single survey can be into six digits. In my judgment, at least 70 percent of the publicly available polling in key Senate and gubernatorial races can be considered junk.
Below please find:
A compilation of 14 blue-chip quality polls in seven critical Senate races, with links to questionnaire toplines and crosstabs, or in two cases, slide decks with important information. Unfortunately, the New York Times/Siena and Fox News polls were conducted during the original Maine Senate race configuration, between Susan Collins and then-Democratic nominee Graham Platner, who has since been replaced on the ballot by former state Senate President and unsuccessful 2026 gubernatorial candidate Troy Jackson. These numbers can be thought of as placeholder numbers, or for comparison, when high-quality polling comes out of a Collins-Jackson trial heat. Once the Democratic nomination is settled in Michigan, expect quality polling to be released in that contest within the next three or four weeks.
A copy of what I call my Swiss Army knife chart, a ranking of the 50 states from most to least Republican using the major, two-party vote (third party and independent votes omitted) for the two major party nominees in 2016, 2020, and 2024. My view is that averaging and weighting equally weighting the last three elections, when state voting patterns have been fairly stable, diminishes the impact of idiosyncratic factors that can crop up in any single election. The Senate seats up this year as well as in 2028 and 2030 are included, giving hints of which races down the road are likely to be most competitive. Also included is the same chart in alphabetical order for easy reference.
Partisanship has never been stronger in our elections, and the outcome of the most recent presidential election is the single best predictor of the outcome of any senatorial, congressional, or gubernatorial general election.
Finally, some numbers that demonstrate just how narrow the U.S. House playing field is this year and how few competitive races there are. With so few races in either the Toss Up or Lean categories, and those that are less representative than ever, the utility of the generic Congressional ballot test as an indicator of the national House election outcome is less than it used to be; the footprint of races is just too small. The generic ballot test has long been helpful in showing which direction the political winds are blowing and, very roughly, whether they are light, moderate, or heavy, but it is of little use in determining the velocity of those winds and a likely outcome. If someone can guarantee me that a race is being held in a generic district, I would be more likely to give a generic ballot test more credence.
National polling
In terms of national polling, in my subjective opinion, the very top of the pecking order constitutes the New York Times/Siena, NBC News (and CNBC), and Pew Research Center. Not that far behind are the AARP, Associated Press/NORC, CNN/SSRC, Fox News, and the Wall Street Journal. Both the Marist University, Marquette University Law School, and Quinnipiac University polls are quite competent and are the cream of the crop among academic pollsters. AARP, Fox News, NBC News, CNBC, and the Wall Street Journal use bipartisan teams of pollsters.
State-level polling
There was a time when most states had one or more major news organizations, a newspaper or television station that commissioned high-quality statewide polls, but in most states, that day is long gone, victims of newsrooms with minimal funding for doing polling the right way. Some news organizations have teamed up with polling firms or universities, with their quality generally ranging from rubbish to mediocre. Yet many, including political reporters and news organizations that ought to know better, continue to cite and quote these polls, often with little regard for the quality of the polling and methodological rigor. Life is too short to waste time looking at bad polls.
Thankfully, several major organizations have begun releasing high-quality, blue-chip statewide polls, in some cases among the only, or the only competent, polling in the public domain. The New York Times/Siena and Fox News have each published six, and the AARP has released its first two battleground state polls, in some cases. Below is a summary table and compilation of links to The New York Times/Siena, Fox News, and AARP in key Senate race states; this list will be updated as new surveys are released.
The narrow House battlefield
The United States is a remarkably evenly divided country; the U.S. House of Representatives, even with the degree of partisan sorting and gerrymandering we have today, is still, more or less, a representative institution, we should expect the House to be relatively evenly divided, as it is now and has been after the last three elections, in 2020, 2022, and 2024.
Control of the U.S. House will be mostly determined by the outcome of fewer than two dozen hotly contested races around the country; the broader battlefield of competitive races is fewer than four dozen. This is a substantially narrower House battlefield than used to be the case, when there were frequently over 100 and sometimes over 150 competitive seats in play.
There are 17 states with one or more competitive House races; seven have a combined 21 competitive contests, with California, Florida, Iowa, Michigan, New York, Pennsylvania, and Texas each with three races rated Toss Up, Lean Democrat, or Lean Republican. Three more states have two competitive races: Arizona, Ohio, and Virginia, with Colorado, Nebraska, Nevada, New Jersey, New Mexico, New York, North Carolina, and Washington having one each. There are currently no competitive races in 37 states.
With the competitive race playing field so narrow and concentrated, an over-reliance on the generic Congressional ballot test question seems awfully problematic. The generic is still helpful in ascertaining a general direction of the political winds and a rough approximation of the extent of those winds, but its utilization in modeling a national House outcome seems highly problematic, more so than in the past.