【PRI】What the GOP’s Internal Memos Reveal About Trump’s Miscalculation
English text appears at the bottom of this post.
【PRI】共和党の内部メモが示す「トランプの誤算」――2026年中間選挙、世論調査では見えない異変
在米ジャーナリスト 髙濱 賛 (Tato Takahama)
Pacific Research Institute
2026年9月18日
世論調査は、選挙を占う材料の一つにすぎない。
調査会社によって数字は違う。質問の仕方でも変わる。まして、投票日までまだ1カ月半ある。だから選挙前の数字には、どうしても「当たるも八卦、当たらぬも八卦」というところがある。
今回、私が注目したいのは、そうした一般向けの世論調査そのものではない。
共和党の内部で、選挙戦を実際に動かしている人たちが何を見ているのかである。
2月、米ニュースサイトAxiosは、共和党関係者が内部調査を見た結果、11月の上院選について従来の激戦州だけでなく、アラスカ、アイオワ、オハイオといった「共和党が勝つのが当然」と考えられてきた州まで競争になっていると認識している、と報じた。
ある共和党関係者は、前年なら「上院選はほぼ確実に勝つ」と言えたが、内部調査を見た現在はそう言えない、と語っている。さらに共和党側では、これまでトランプ大統領の勝利を支えた移民政策と経済問題が、今度は共和党候補にとって負担になり得るとの認識も出ていた。
これは、民主党側が「共和党は危ない」と言っている話とは違う。
共和党自身が、自分たちの内部データを見て警戒し始めたということである。
しかも、その後の動きを見ると、この警戒感は一過性ではない。
7月には、共和党系の政治団体が、これまで民主党が攻め込むのは難しいと考えられてきたオハイオとアイオワで、大規模なテレビ広告を確保した。Axiosは、共和党側が両州を「赤い防壁」として守るために支出を増やしていると報じた。
8月になると、Axiosは、アラスカ、アイオワ、テキサスなど、通常なら共和党が大きな心配をしない州まで民主党が競争に持ち込んでいると報じた。共和党関係者の間では、これらの州はいずれ「トランプ支持の共和党」に戻るとの見方もある一方、現時点では賭けられない、という警戒が広がっていた。
そして9月に入ると、その「内部で見えていた異変」が、公開された選挙データにも現れてきた。
9月17日、RealClearPoliticsがまとめた上院の激戦8州では、最新の平均値で7州が民主党候補のリードとなった。テキサスでは共和党のケン・パクストン氏と民主党のジェームズ・タラリコ氏がほぼ互角。ミシガンでは民主党のエル・サイード氏が共和党のマイク・ロジャーズ氏をリードしている。アラスカでも民主党のメアリー・ペルトラ氏が共和党のダン・サリバン氏を平均でリードしている。
ノースカロライナも象徴的だ。
9月16日、トランプ大統領は共和党候補マイケル・ホワットリー氏を支援するため同州入りした。ところが、その直前に公表されたCarolina Journal/Harperの調査では、民主党のロイ・クーパー氏が49％、ホワットリー氏が34％だった。RCPの最新平均でもクーパー氏がリードしている。
ここで重要なのは、「トランプが応援したから数字が悪くなった」と単純化することではない。
問題は、トランプを前面に出すことが、すべての共和党候補に同じ効果をもたらすとは限らなくなっていることだ。
テキサスはさらに興味深い。
同州では2024年大統領選でトランプ氏に投票したヒスパニック層の一部が、今回の上院選では民主党のタラリコ氏に流れていることを示す調査が出ている。共和党にとって、これまでの「トランプ＋ヒスパニック」という組み合わせをそのまま中間選挙に持ち込めるのかは、重要な問題になる。
アイオワも同じだ。
9月16日の調査では、共和党のアシュリー・ヒンソン氏が民主党のジョシュ・トゥレック氏を48％対43％でリードしている。ただし、共和党にとって安心できるほどの差ではない。アイオワは2024年大統領選でトランプ氏が大差で勝った州であり、そこで上院選が競争になっていること自体が、共和党にとって無視できない変化である。
オハイオでは、共和党のジョン・ハステッド氏と民主党のシェロッド・ブラウン氏が競り合っている。メインでは共和党のスーザン・コリンズ氏と民主党のトロイ・ジャクソン氏、ミネソタでは共和党のミシェル・タフォヤ氏と民主党のペギー・フラナガン氏が接戦となっている。
つまり、共和党内部で2月から警戒されていた「地図の拡大」が、半年後には公開調査にも姿を現しているのである。
もちろん、これだけで11月3日の結果を予測することはできない。
共和党には依然として強固な支持基盤がある。トランプ氏自身も共和党支持層を動員する力を持っている。9月17日には下院共和党のスティーブ・スカリース院内総務が、共和党が下院で最大225議席を獲得する可能性に言及し、民主党の「ブルーウェーブ」という見方を否定した。共和党指導部がなお勝負になると考えていることも事実である。
一方、9月14日に公表されたReuters/Ipsos調査では、トランプ大統領の支持率は35％。民主党44％、共和党37％という「議会一般投票」の数字では、民主党が7ポイント上回った。これは議席数を直接予測する数字ではないが、共和党にとって無視できない環境であることは確かだ。
そして、ここで見逃せないのが「トランプ個人」ではなく「共和党候補にとってのトランプ」という問題である。
大統領としてのトランプ氏には、熱烈な支持者がいる。共和党の選挙戦略にとって、その支持者を投票所に動員することは大きな意味を持つ。
しかし、中間選挙ではトランプ氏自身の名前は投票用紙にない。
共和党候補者一人ひとりが、自分の州、自分の選挙区で有権者から評価される。
そこで、「トランプを前面に出せば共和党票が増える」のか。それとも、トランプ氏への強い反発を持つ無党派層や穏健な有権者を遠ざけるのか。
この問いに対する答えは、州によっても、候補者によっても違うだろう。
実際、9月にはトランプ氏の長年の世論調査担当者トニー・ファブリツィオが、共和党候補に向けてAIデータセンター問題についての内部メモを作成した。そこでは、単純にデータセンター建設を支持するのではなく、地域社会への負担や環境問題などを先に認める必要があると助言している。トランプ氏自身がデータセンター建設を強く支持していることとの間に、微妙な緊張関係が生じている。
ここにも、2026年中間選挙の特徴が表れている。
共和党の候補者たちは、トランプ大統領の政策を否定しているわけではない。しかし同時に、トランプ氏の全国的なメッセージを、そのまま自分の選挙区に持ち込めばいいとも限らない。
これは「反トランプ」の選挙という単純な話ではない。
むしろ、「トランプを使う選挙」と「トランプから少し距離を取る選挙」の間で、共和党がどこまで細かな調整を迫られているか、という話なのである。
共和党の内部データが早くから示していたのは、まさにこの問題だった。
トランプ氏は共和党にとって、最大の動員装置である。同時に、接戦州では候補者が自分自身の選挙を戦う際の制約要因にもなり得る。
その二面性が、11月3日の中間選挙で初めて本格的に試される。
共和党にとってトランプは、政治資産なのか。それとも、すでに「負の遺産」になり始めているのか。
11月3日、その答えが示される。
【参考文献】
- RealClearPolitics, “Senate Polls Narrow Across Eight Battleground States,” September 17, 2026.
- Axios, “Exclusive: GOP issues stark warning to AI companies,” August 19, 2026.
- Axios, “Scoop: Trump pollster gives GOP a data center survival guide,” September 15, 2026.
- Reuters/Ipsos, “Trump approval up from record low, but outlook sours for Republicans,” September 14, 2026.
- The Washington Post, “We polled Michigan’s Senate race. There’s good and bad news for Democrats,” September 17, 2026.
- Carolina Journal, “CJ poll: Cooper leads Whatley by 15 as Republican voters drift to undecided,” September 17, 2026.
- Axios, “Exclusive: Scalise puts GOP high-water mark at 225 House seats,” September 17, 2026.
【Cultural Newsへのご支援のお願い】
本稿は、髙濱賛による連載「A Japanese View from America」の一篇です。
Cultural Newsは、日系社会と米国社会を結ぶ架け橋として、報道、論評、文化活動を続けています。こうした活動を継続するためには、読者の皆様のご支援が欠かせません。
本記事に共感いただけましたら、Cultural Newsへのご寄付をご検討いただければ幸いです。皆様からのご支援は、取材活動、記事制作、地域社会への情報発信を支える重要な基盤となります。
今後ともご愛読とご支援を賜りますよう、心よりお願い申し上げます。
Cultural Newsへのドネーションは、下記リンクからお願いします。
https://www.culturalnews.com/donation
【PRI】What the GOP’s Internal Memos Reveal About Trump’s Miscalculation
By Tato Takahama
Pacific Research Institute
September 18, 2026
Polls are one way of taking the temperature of an election. But they are not the election itself.
Different pollsters get different results. Question wording matters. Samples differ. And with more than six weeks still to go before Election Day, any individual poll can become obsolete quickly.
That is why I am less interested here in what any single public poll says than in what Republicans themselves are seeing behind the scenes.
The most revealing evidence may not be the latest headline poll. It may be the internal polling, message testing and private memoranda being circulated among Republican candidates and campaign strategists.
Those documents offer a different window into the 2026 midterm elections: not simply what voters tell pollsters, but what Republican campaign operatives believe their candidates need to do to compete on November 3.
And what that window reveals is increasingly uncomfortable.
In August, the National Republican Senatorial Committee privately warned major AI companies that the backlash against data centers had become a serious political problem for Republicans, particularly in the Ohio Senate race.
According to an internal memo obtained by Axios, the NRSC said Democrats had made data centers a “centerpiece” of their campaign to defeat Republican Sen. Jon Husted — and that the strategy was working. The memo called data centers a “sleeper issue for the entire election cycle.”
Then, on September 15, Tony Fabrizio, President Donald Trump’s longtime pollster, circulated a two-page memo to Republican candidates offering advice on how to talk about AI data centers.
His warning was unusually blunt: simply saying “I support building more data centers” was not a winning political position.
Fabrizio’s polling suggested that voters were more receptive to data-center development when it came with enforceable conditions addressing electricity costs, water use and other local concerns. His advice was essentially that Republican candidates should acknowledge those concerns before emphasizing jobs, economic growth or national security.
The message was clear.
Republican candidates could not simply repeat a national Republican talking point. They had to adjust their message to the concerns of voters in their own states and districts.
That is more significant than any single public poll.
It tells us what Republican strategists themselves are worried about.
And the same pattern is beginning to appear in the broader Senate landscape.
On September 17, RealClearPolitics reported that seven of eight closely watched Senate races were leaning Democratic in its latest averages. The eight states were Texas, Michigan, Ohio, Maine, Alaska, Iowa, Minnesota and North Carolina.
This is not a prediction of what will happen on November 3. Poll averages can move, undecided voters can break late, and several of these races remain extremely close.
But the geographic spread of the competition is important.
Texas is perhaps the most striking example.
A new Emerson College/Nexstar poll put Democratic state Rep. James Talarico one point ahead of Republican state Attorney General Ken Paxton. Another poll earlier in the week gave Talarico a three-point advantage.
The race itself is essentially a dead heat.
But the demographic movement is more intriguing.
According to Emerson College Polling Executive Director Spencer Kimball, Texas Hispanic voters who supported Trump in 2024 are now breaking for Talarico by 15 points.
That does not mean that Hispanic voters as a whole have abandoned the Republican Party. Nor does it mean that the current movement will necessarily persist until Election Day.
It does, however, illustrate a problem Republicans cannot easily dismiss: the coalition that helped deliver Trump a second term in 2024 does not automatically translate into a Republican advantage in every 2026 Senate race.
Michigan presents another warning sign.
Two polls released on September 17 showed Democratic candidate Abdul El-Sayed narrowly ahead of former Republican Rep. Mike Rogers. An Emerson College/Nexstar poll put El-Sayed ahead 48% to 46%, while a Washington Post-Schar School poll had him ahead 48% to 45%.
Again, these are small margins.
But it is precisely these small margins that campaign strategists spend months trying to understand.
Alaska is another state where the map is becoming more complicated. The latest RCP average has Democratic Rep. Mary Peltola three points ahead of Republican Sen. Dan Sullivan. Because Alaska uses ranked-choice voting, however, a simple two-candidate comparison does not tell the whole story. A recent Fabrizio/Anzalone poll showed Peltola with 46% in the first round and Sullivan with 41%, with other candidates accounting for the remainder; in a simulated final round, Peltola led 53% to 47%.
Maine is similarly unsettled. Republican Sen. Susan Collins remains competitive, but the latest RCP average has Democratic candidate Troy Jackson slightly ahead. A recent Quantus poll put Collins one point ahead, but nearly 70% of respondents remained undecided.
Iowa may be even more revealing because of the state’s recent voting history.
Republican Rep. Ashley Hinson is the only Republican candidate leading in the RCP average among the eight closely watched Senate races, and even there the lead is within the margin of error. Iowa has reliably voted for Trump by wide margins in recent presidential elections, yet 52% of respondents in the latest poll disapproved of his job performance.
The question for Republicans, therefore, is not simply whether Trump remains popular among Republican voters.
He remains a powerful force within the party.
The more complicated question is whether that political strength can be transferred to Republican candidates running their own races in states where independents, suburban voters and other swing constituencies may have very different priorities.
North Carolina offers perhaps the clearest illustration.
Trump went to the state on September 16 to campaign for Republican Senate candidate Michael Whatley. The timing was striking. A new Carolina Journal/Harper poll had Democratic former Gov. Roy Cooper ahead of Whatley by 15 points. The latest RCP average had Cooper ahead by 8.5 points.
The same Carolina Journal/Harper poll found that 57% of respondents disapproved of Trump’s job performance, compared with 41% who approved. Most strikingly, exactly 50% said they would be less likely to vote for a candidate endorsed by Trump.
That does not establish that Trump’s endorsement is hurting Republican candidates everywhere.
It does suggest that the political value of a Trump endorsement is not uniform across the country.
Texas, for example, presents a different picture. In a poll conducted after the Republican midterm convention in Dallas, 59% of Republicans who had heard about the convention said it made them more excited to vote. Among independents, however, about 68% said the convention had no impact on their motivation.
That distinction matters.
Mobilization and persuasion are not the same thing.
A candidate can benefit from Trump’s ability to bring Republican voters to the polls while simultaneously worrying that too much association with Trump could alienate voters who are not already committed Republicans.
That is the strategic tension running through the 2026 midterms.
And it helps explain why Republican strategists are conducting message testing on issues that might seem, at first glance, unrelated to Trump himself.
Health care. Data centers. Electricity prices. Water use. Immigration. The economy.
The internal memos are not necessarily saying that Republicans should abandon Trump.
They are saying something more subtle: Republican candidates cannot assume that Trump’s national message will automatically work in their own political environments.
That is a very different proposition.
The distinction is particularly important because Trump himself will not appear on the ballot in November.
Every Republican Senate and House candidate must ultimately run a separate race, under different local conditions, against a different Democratic opponent, with a different electorate.
The Republican Party therefore faces a strategic problem that cannot be reduced to Trump’s approval rating.
Trump can energize the Republican base. His endorsement can provide money, attention and organizational support. His rallies can dominate the news cycle.
But the same association can also become a liability in a competitive state if voters are more concerned about prices, health care, electricity bills, local development or other issues than about the national political battles Trump prefers to fight.
The internal memos are revealing precisely because they expose this tension before Election Day.
The Republican campaign operation is not simply asking, “Do voters like Trump?”
It is asking a more practical question:
“How should a Republican candidate use Trump — and where should the candidate draw the line?”
That may ultimately be one of the central strategic questions of the 2026 midterms.
National polling provides one picture. Internal campaign research provides another. Neither is infallible, and neither should be treated as a forecast.
But when the party’s own strategists begin changing the language they want candidates to use, when a Senate campaign committee privately warns an industry that its political image is becoming a serious electoral problem, and when Republican candidates in traditionally favorable states suddenly find themselves in competitive races, those are signals worth watching.
There is another reason to pay attention.
A September 14 Reuters/Ipsos poll put Trump’s approval rating at 35%, up from 33% in the previous survey. At the same time, Democrats led Republicans 44% to 37% on the generic congressional ballot. That generic ballot does not translate directly into a prediction of House or Senate seats, but it provides another indication of the broader political environment facing Republicans.
Republican leaders, for their part, are not conceding the battlefield.
House Majority Leader Steve Scalise has argued that Republicans could still reach as many as 225 House seats, rejecting the idea that a Democratic “blue wave” is inevitable. That is the assessment of a Republican leader, not an independent forecast, and it illustrates that the party continues to see a viable path to retaining congressional power.
The point, then, is not that the Republican Party is doomed.
It is that the party’s own internal data appear to be forcing a more complicated conversation about Donald Trump.
For years, the central Republican question was whether Trump could bring new voters into the party and keep the conservative base unified.
In 2026, the question is beginning to look different.
Can Republican candidates remain closely aligned with Trump while also persuading voters who are less enthusiastic about him?
That is the dilemma behind the numbers.
And it is why the internal memoranda may ultimately tell us more about the political choices facing Republican candidates than any single public poll.
For the Republican Party, is Trump still a political asset — or is he beginning to look like a political liability?
November 3 will provide the answer.
Note: English translation was prepared by the author. Translation assistance tools may have been used.
References
- RealClearPolitics, “Senate Polls Narrow Across Eight Battleground States,” September 17, 2026.
- Axios, “Exclusive: GOP issues stark warning to AI companies,” August 19, 2026.
- Axios, “Scoop: Trump pollster gives GOP a data center survival guide,” September 15, 2026.
- Reuters/Ipsos, “Trump approval up from record low, but outlook sours for Republicans,” September 14, 2026.
- The Washington Post, “We polled Michigan’s Senate race. There’s good and bad news for Democrats,” September 17, 2026.
- Carolina Journal, “CJ poll: Cooper leads Whatley by 15 as Republican voters drift to undecided,” September 17, 2026.
- Axios, “Exclusive: Scalise puts GOP high-water mark at 225 House seats,” September 17, 2026.
