2026年美国中期选举与通往2028之路：证据、情景及其对经济走势的含义

独立性声明。 本报告为独立第三方分析，不持任何政治、党派、意识形态或国家立场。报告陈述证据并给出概率估计，不建议任何人如何投票。完整免责声明见报告末尾，并另行以独立文件发布。

内容摘要

美国将于2026年11月3日举行中期选举，众议院435个席位和参议院35个席位同时改选。1 距投票还有三周，现有证据显示出清晰但并不均匀的全国环境：总统支持率处于第二任期以来的最低水平或接近最低水平；主要民调均值显示，在“泛国会投票”（generic ballot）中民主党领先约8个百分点；与伊朗冲突相关的能源冲击推高了消费者物价。与此同时，全国得票差能否转化为席位，取决于选区划分、候选人质量和投票率；而在参议院，民主党需要赢下数个在2024年被特朗普总统以两位数优势拿下的州。

本报告形成五项核心判断。概率为作者的分析性估计，综合了民调均值、预测市场价格、已发布的预测模型和历史误差规律，非属任何官方机构的预测。

序号	核心判断	估计	置信度
1	民主党赢得众议院多数（218席）	约88%（市场区间85%–91%）	高
2	民主党赢得参议院控制权（净增四席）	约56%（市场区间61%–64%；模型区间54%–57%）	中低
3	最可能的结果是民主党两院全胜（约54%），其次是民主党控制众院、共和党控制参院（约34%）	见第七节	中
4	由能源驱动的通胀和生活成本是当前环境背后的主要经济渠道；投票前汽油价格若回落，会缩小但大概率不会消除差距	定性判断	中
5	2028年大选高度开放：两党提名民调中的领跑者领先幅度显著但并不稳固，早期总统大选对决民调的预测力偏低	定性判断	高

另有三点构成分析框架。第一，民调领先与市场价格并非相互独立的证据，许多市场交易者参考的是同样的民调。第二，2024年选举表明，全国民调在摇摆州可能低估共和党得票约3个百分点；这种幅度的共和党超预期表现仍有可能出现，对参议院的影响最大。2 第三，中期选举与总统大选结果之间只是松散关联：2010年共和党大胜之后，民主党总统在2012年连任；2018年民主党大胜之后，共和党现任总统在2020年落败。这说明中期选举结果更多地塑造政治议程，而不是决定下一届总统选举的结果。

一、分析框架

1.1 泛国会投票衡量什么

泛国会投票询问受访者在不点名候选人的情况下会支持哪个政党的国会候选人。3 它是衡量全国环境的标准指标，但在解读时有四个局限。

1. 得票不等于席位。 全国众议院得票是435个独立选区结果的加总。席位结果取决于各党选民在各选区的分布效率。
2. 选区图在变。 2018年使用的选区线在2010年人口普查后划定，而2026年选举使用2020年普查后的选区图，以及若干次任期中途重划。4
3. 样本口径不同。 均值混合了成年人、登记选民和可能投票者。Silver Bulletin 使用全部民调的读数为民主党+8.8，经可能投票者调整后为+9.5。5
4. 离散度大。 目前RealClearPolling表中的单项民调从民主党+2（Harvard-Harris）到+12（Quinnipiac及一项THP/MM民调）不等，均值为民主党+8.3。6

1.2 本报告采用的方法

本报告采用三角验证，而非单一模型：（a）比较三个以上方法多样的民调聚合平台；（b）比较两个预测市场；（c）参考已发布的预测模型；（d）参考全国得票差与席位变化关系的历史基准。当来源之间存在分歧时，报告陈述分歧，并在区间内取值，并行对已知的误差来源作折扣。2028年各节同步遵循交叉核对的原则。

1.3 已知的误差规律

• 最终民调误差。 2024年，最终民调均值在每个摇摆州都低估了唐纳德·特朗普的得票约3个百分点；该误差被认为处于正常范围，且小于2016年和2020年。7
• 方向并不稳定。 民调误差在2016、2020和2024年有利于共和党，但在2022年的部分州级选举中有利于民主党。因此不宜朝任一方向作统一修正。合理的工作假设是：全国得票差的对称不确定性约为3–4个百分点，单个州约为4–5个百分点。
• 投票率效应。 中期选举的投票率低于总统选举年，动员程度更高的一方往往相对登记选民民调有所获益。这是目前可能投票者调整有利于民主党的原因之一。

二、全国政治环境

2.1 总统支持率

总统支持率是预测总统所在党在中期选举中表现的最持久指标。目前各机构读数均偏低，但因方法多样而水平不同。

指标	读数	来源
皮尤研究中心（9月28日–10月4日）	30%支持，68%反对	8
路透/益普索（9月30日–10月5日）	32%支持，65%反对	9
AP-NORC（9月24–28日）	31%支持，69%反对	10
Morning Consult（10月9–11日）	42%支持，56%反对	11
Rasmussen（可能投票者，10月12日）	41%支持，58%反对	12
聚合平台（10月8日）	RealClearPolitics 38.3%；《纽约时报》37%；Silver Bulletin 37.1%	13

各机构之间的差距（30%至42%）表明机构效应和样本口径很重要，但没有任何主要系列把支持率放在40%中段以上。《华尔街日报》37%的读数，是该民调系列自1990年以来总统在中期选举前的最低值。14 作为参照，盖洛普在特朗普第一任期的同一时点测得43%，此后的2018年中期选举中其所在党失去41个众议院席位。15,16

对经济管理的认可度更低：《金融时报》/Focaldata民调中为28%。17 特朗普2024年选民中的支持率，已从任期初的95%降至皮尤调查中的67%。18

也有相反方向的证据。《经济学人》/YouGov系列显示，9月18日至10月5日连续三次民调中，黑人和西班牙裔受访者对总统的净支持率改善约19–20个百分点，但基数很低。19 聚合读数也逐日变动；Silver Bulletin的净支持率在周末触及−23.9后，10月14日为−22.4。20

2.2 泛国会投票与历史对照

主要均值中民主党的领先幅度一致。

聚合平台	民主党	共和党	差距	日期
RealClearPolling	50.2	41.9	民主党+8.3	9月21日–10月11日 21
Silver Bulletin（默认）	无	无	民主党+8.8	10月13日 22
Silver Bulletin（可能投票者）	无	无	民主党+9.5	10月12–13日 23
Decision Desk HQ	49.0	41.3	民主党+7.7	10月13日 24

三个默认聚合读数的中位数约为民主党+8。过去一个月，差距由约+6.5升至+8.9（Silver Bulletin），由+5.7升至+8.5（RealClearPolitics），由+6.7升至+7.7（DDHQ）。25 因此该变动是近期发生的，而近期的变动可能部分回吐。

历史比较说明得票差为何重要，却不能决定席位数。2018年，民主党以53.4%对44.8%（民主党+8.6）赢得全国众议院选票，净增41席。26 2010年，共和党以51.7%对44.9%（共和党+6.7）获胜，净增约63席。27 2022年，共和党以50.0%对47.3%获胜，净增9席。28 因此在这三个周期中，得票差到席位的转化率大约为每个百分点3至9席。即使取该区间的低端并应用于民主党+8的环境，民主党的净增席位也将远高于所需的3席，但2026年的选区图不同，因此报告在第四节单独处理席位转化问题。

2.3 提前投票

截至10月14日，全国已有逾370万张邮寄和提前现场投票，已申请或寄出的邮寄选票约5169万张。29 在佐治亚州，提前投票首日接收了188,667张选票，高于2022年首日总数。30 俄亥俄州报告在距选举还有21天时已有194,481张提前现场投票，而2022年同一时点为50,368张。31 据Ground News转述，CNN的早期分析显示民主党在提前投票中有投票率优势，但提前投票的构成是较弱的预测指标，因为两党选民偏好的投票方式不同。32 提前投票证据显示参与度高；它本身并不表明哪一方将受益。

三、经济与生活成本

3.1 当前状况

经济图景是多维的而非一概负面，这在能源供给冲击中很典型。

指标	最新读数	来源
9月CPI通胀率（同比）	3.7%（8月：3.4%）	33
核心CPI（不含食品和能源）	2.6%（8月：2.4%）	34
9月汽油CPI环比	+9.1%，占当月涨幅一半以上	35
能源CPI，12个月	+20.4%；汽油+34.2%	36
AAA全国汽油均价	10月14日4.33美元，去年同期3.07美元；今年已有创纪录的158天高于4美元	37
2026年二季度实际GDP（第三次估计）	年化2.2%；一季度修正为2.5%	38,39
9月就业与失业率	新增2.9万个岗位；失业率4.2%	40
联邦基金利率目标区间	3.75%–4.00%，9月16日加息25个基点，为2023年以来首次加息	41
10月上中旬布伦特原油	约98–106美元/桶，冲突前约72美元	42,43

经济正增长、劳动力市场处于低位但趋于走软、总体通胀上升、央行处于紧缩之中，这一组合构成投票的主要经济背景。汽油是最显眼的价格。AAA的4.33美元读数，较一周前的4.36美元略有回落，但仍处于秋季的纪录高位。44

3.2 机制：从能源价格到投票意向

能源冲击通过三条渠道传导到选举环境。

1. 显著性。 汽油价格公开张贴、购买频繁，因此对生活成本认知的影响大于其在CPI篮子中的权重。
2. 归因。 在AP-NORC的一项调查中，65%的受访者把高生活成本的责任归于总统。45 无论这种归因在分析上是否完整，它都会塑造行为。
3. 政策反应。 美联储加息提高家庭和企业的借贷成本，因而在旨在降低通胀的同时，也可能加剧对生活成本的担忧。

3.3 经济解释的局限

经济状况只能解释环境的一部分。二季度实际消费支出和投资强劲（私人国内购买者实际最终销售增长4.6%），初请失业金人数处于历史低位。46,47 9月仅新增2.9万个岗位，且7月和8月合计下修6万个岗位，显示降温。48 选民的评价还受其他因素影响，包括对伊朗冲突、移民和贸易政策的看法。因此，报告把生活成本视为可识别的最大因素，而非唯一因素。

3.4 石油与战争

2026年2月28日开始的冲突仍未解决。路透社报道，美国能源信息署目前预测第四季度布伦特原油均价约105美元，2026年全年约98美元。49 Kpler船舶追踪数据显示，截至10月7日当周霍尔木兹海峡原油出口均值约850万桶/日，较战前低约40%，国际能源署把这次中断称为石油市场历史上最大的供应中断。50 这些情况意味着能源价格对选举的风险是双向的：外交突破会迅速压低价格，而对航运的进一步攻击会推高价格。关于会谈的消息已使布伦特在单个交易日内波动数美元。51

四、众议院

4.1 起点与评级

民主党目前持有214席，需要218席，即净增约3至4席。52 Cook政治报告10月9日的评级为：稳固民主党185席、可能民主党16席、倾向民主党12席；稳固共和党174席、可能共和党18席、倾向共和党9席；另有21席为胜负难料（Toss Up）。53 以至少倾向民主党计，共213席，在任何一个Toss Up决出之前，民主党距多数仅差5席。Cook在10月9日将一批席位向民主党调整，包括艾奥瓦、宾夕法尼亚和华盛顿州的五个Toss Up调为倾向民主党。54

按党别比较“Toss Up或更差”的席位可见结构：2026年10月为共和党持有23席、民主党持有11席；2018年10月为45席和3席；2022年10月为12席和30席。55 2026年的失衡程度小于2018年，这与选区重划的影响一致，也与处于高度竞争区域的共和党席位少于2018年一致。

4.2 选区重划

2025–26年的任期中途选区重划是与2018年最主要的结构性差异。共和党曾希望通过新选区图获得13席，但众议院多数党党鞭汤姆·埃默（Tom Emmer）表示他们不会拿下全部13席，Cook目前预测仅选区重划一项使共和党净增约4席。56 埃默还表示，加利福尼亚和犹他两州的重划可能使共和党合计失去五席。57 实际含义是：选区图使民主党的席位增幅相对于无重划基线减少约四席，但并未抹去民主党+8的优势。

4.3 席位估计

一个追踪网站引述的Cook最可能结果是民主党净增9至19席，总数224至234席。58 Polymarket的预测席位为234对201。59 综合这些区间以及全国得票差数个百分点的对称不确定性，报告的中心区间为民主党222–238席，对应民主党赢得多数的概率约为88%。不宜大幅调高的三个理由：均值可能部分回吐；即使共和党民调偏差达3个百分点，民主党仍是热门，但优势会收窄；仍有约21个Toss Up。不宜调低的三个理由：自8月以来领先幅度一直稳定在民主党+6以上；民主党方面的早期参与度指标为正面；到多数的席位差距很小。

五、参议院

5.1 选情地图

共和党在100个参议院席位中持有53席。35个席位将改选，含佛罗里达和俄亥俄两个特别选举，其中23个由共和党持有。60 民主党需要净增四席达到51席，因为副总统J.D.万斯在50比50时投决定票。61 NBC新闻的分析指出，民主党必须守住密歇根、佐治亚和新罕布什尔，并在若干2024年特朗普以两位数优势获胜的州中取胜。62

Cook目前将阿拉斯加、艾奥瓦、缅因、俄亥俄和得克萨斯评为Toss Up，北卡罗来纳为倾向民主党，并称参议院控制权为真正的Toss Up。63 在《纽约时报》/锡耶纳（Siena）民调显示共和党参议员罗杰·马歇尔与民主党人亚当·汉密尔顿45比45持平之后，Cook现在也把堪萨斯描述为Toss Up。64

5.2 民调与结构性差距

《纽约时报》/锡耶纳对可能投票者的民调显示：阿拉斯加，玛丽·佩尔托拉（Mary Peltola）50比43领先丹·沙利文（Dan Sullivan）；得克萨斯，詹姆斯·塔拉里科（James Talarico）51比45领先肯·帕克斯顿（Ken Paxton）；俄亥俄，谢罗德·布朗（Sherrod Brown）49比46领先乔恩·赫斯特德（Jon Husted）；堪萨斯45比45持平；艾奥瓦，阿什利·欣森（Ashley Hinson）48比47领先乔什·图雷克（Josh Turek）。65 在Election Central汇总的RealClearPolitics均值中，民主党候选人在所追踪的12场竞选中的10场领先，但关键州的差距很小：艾奥瓦图雷克+0.6，缅因杰克逊（Troy Jackson）+1.2，阿拉斯加佩尔托拉+1.4，得克萨斯塔拉里科+2.6，俄亥俄布朗+3.1。66

由此得出两点分析。第一，多数战场州的差距小于3–5个百分点的典型民调误差，因此每场竞选都接近抛硬币，略偏向民主党。第二，这些竞选相互关联。如果全国民调高估民主党3个百分点，几个最胶着的竞选会同向变动，因此在七个Toss Up中恰好赢下四个或更多的概率，低于把它们视为独立事件相乘所得的概率。这些州2024年的总统选举差距为：阿拉斯加和艾奥瓦共和党+13，俄亥俄共和党+11，得克萨斯共和党+14，缅因民主党+7。67

部分参议院竞选还显示候选人层面的韧性。Cook指出，缅因州参议员苏珊·柯林斯（Susan Collins）在唐纳德·特朗普三次落败的州内仍具竞争力，Cook以此说明民主党的党派品牌相对于总统声望较弱。68 这是有益的制衡：全国环境对民主党有利，但该党自身的好感度并不同样强势。

5.3 市场与模型

预测市场对民主党控制参院的定价约为62%–64%：据10月13–14日发布的汇总，Polymarket为63.5%，Kalshi约61%。69,70 已发布的模型更低：Silver Bulletin的FLIPR模型在一次市场评论发布时为57.3%，Decision Desk HQ 10月1日的预测为54%。71,72 报告约56%的估计处于模型区间的上沿、低于市场区间。相对市场打折的依据是上述相关性误差论证，以及近几周市场向民主党移动，这既可能反映新信息，也可能反映动量。

六、资金、动员体系与外部行为体

竞选资金是相关的结构性变量。埃隆·马斯克在2025–26周期向联邦委员会捐款约3.66亿美元，其中7月以来2.76亿美元；1.58亿美元给了参议院领导基金（Senate Leadership Fund），约7600万美元给了America PAC，后者在七个参议院战场和约三十几个众议院选区开展动员行动。73 他此前曾威胁成立第三党“美国党”（America Party），但并未付诸实施。74 另据报道，亲特朗普的超级政治行动委员会MAGA Inc.在8月持有4.16亿美元。75

这种规模的支出能在边际上影响结果，尤其是在得克萨斯这样成本高昂的州，参议院领导基金在那里已花费远超1亿美元。76 关于广告和动员支出在信息充分的竞选中能在多大程度上改变结果，证据并不一致，其影响可能小于总额所显示的规模。报告把资金优势视为：尽管全国环境如此，仍不把民主党两院全胜的概率定在约54%以上的理由之一。

七、结果情景与概率

情景	预测市场	本报告	主要驱动因素
A. 民主党两院全胜	60%–62%	约54%	泛国会投票领先持续在民主党+7以上；民主党在阿拉斯加、艾奥瓦、缅因、俄亥俄、得克萨斯和堪萨斯战场中拿下四个，且守住密歇根、佐治亚和新罕布什尔
B. 民主党控众院、共和党控参院	28%–30%	约34%	全国民主党领先约5–8个百分点，而参院地图有利于在特朗普以两位数优势获胜的州中的共和党现任者
C. 共和党两院全胜	10%–11%	约10%	民调出现4个百分点以上的倾向共和党的偏差，或汽油价格在后期下降并通过外交解决伊朗冲突
D. 共和党控众院、民主党控参院	约1%	约2%	需要不特殊的分裂，例如民主党赢下参院各州，而选区图使共和党守住众院

市场数字来自77和78。报告的数字合计100%，属于判断，每项周围的置信区间很宽：情景A的合理区间为42%–64%。

7.1 各情景对施政意味着什么

以下为分析性含义，而是对实际政策的预测。

• 民主党控制众院（情景A和B）。 众院将拥有传唤和监督权，并将塑造拨款。分立政府在历史上减少了重大新立法的通过，并提高了预算截止日期的重要性。包括依据现行法定权限实施的贸易措施在内的行政行动无需众院批准，仍将是贸易政策的主要渠道。
• 民主党控制参院（情景A和D）。 参院掌握任命确认和条约批准，并可影响战争权力表决。控制权还将改变司法任命的流程。
• 共和党两院全胜（情景C）。 政策延续，执政联盟的议程延续到任期最后两年，但政治信息是：全国环境并未转化为损失。

八、中期选举之后：值得关注的动态

1. 对结果的解读。 两党都会把决定性结果解读为授权或裁决；微弱的结果则会被争议。民主党的增幅相对于2018年41席基准的大小，将为2028年的叙事定调。
2. 选举的执行。 2026年的计票将在时间上受到关注：佐治亚、俄亥俄、内华达和北卡罗来纳的提前投票系统报告时间不同，因此胶着的参议院竞选可能无法在选举夜确定。
3. 跛脚会期。 新一届国会将于2027年1月就任。11月至1月之间的拨款和债务上限决定将确定财政基线。
4. 2028年的开始。 《洛杉矶时报》指出，2028年的角逐在11月3日投票结束时非正式开始。79 宣布参选可能在中期选举后很快出现，尤其是来自不在选票上的人选。

九、2028年总统选举

9.1 已知情况

尚无主要候选人正式宣布参选。辩论日程尚未确定：辩论由竞选团队和媒体机构组织，而非由法律规定，预计初选辩论将从2027年开始。80 因此目前的证据仅限于早期提名民调、大选对决测试和少数公开表态。

9.2 民主党提名

均值显示卡玛拉·哈里斯（Kamala Harris）排名第一，领先约12个百分点。RealClearPolling均值（7月19日–10月6日）为：哈里斯26.4%，加文·纽森（Gavin Newsom）14.4%，皮特·布蒂吉格（Pete Buttigieg）12.6%，亚历山德里亚·奥卡西奥-科尔特斯（Alexandria Ocasio-Cortez）10.6%，乔恩·奥索夫（Jon Ossoff）8.9%。81《金融时报》/Focaldata 10月2–5日的调查为：哈里斯30%，纽森15%，布蒂吉格9%，奥卡西奥-科尔特斯9%，奥索夫8%，14%未决定。82 Harvard CAPS/Harris在9月下旬为哈里斯38%、纽森17%；The Center Square的Noble Predictive民调（10月5–7日）中哈里斯为29%。83,84

纽森在9月14日的CNN采访中说：“如果她参选，我不会参选。”85 他把两人之间的竞争称为“相互确保毁灭”，理由是双方支持者高度重叠。86 分析人士提出两种解读：真诚的礼让，或是促使哈里斯作出决定的策略性举动；二者均未得到证实。87,88 哈里斯尚未宣布是否参选。89

预测市场与民调明显分歧。据9月底的报道，Polymarket赔率中奥卡西奥-科尔特斯以逾19%领先，奥索夫16%，纽森12.7%，哈里斯8%。90 有两种可能的解释：市场可能预期哈里斯决定不参选；或者市场参与者对大选实力和初选动态的权重不同于只说出熟悉名字的调查受访者。知名度是早期初选民调的已知驱动因素，而选举前两年的民调屡次未能识别最终的提名人：希拉里·克林顿在为2008年所做的2006年民调中领先，杰布·布什在为2016年所做的2014–15年的许多民调中居于前列。

9.3 共和党提名

RealClearPolling均值为：J.D.万斯（JD Vance）39.1%，马可·卢比奥（Marco Rubio）23.8%，小唐纳德·特朗普11.1%，罗恩·德桑蒂斯（Ron DeSantis）7.3%，小罗伯特·肯尼迪4.6%。91《金融时报》/Focaldata民调为：万斯38%，卢比奥24%，德桑蒂斯9%，小特朗普7%，肯尼迪7%，12%未决定。92 《国会山报》援引消息人士报道，卢比奥告诉捐款人和盟友他不会与万斯竞争，万斯则表示他专注于中期选举。93 这些是报道，而非正式宣布。

9.4 大选对决测试

Emerson学院民调的早期对决测试显示出狭窄且不均衡的图景：哈里斯49%对万斯45%；卢比奥48%对哈里斯43%；纽森49%对万斯44%；纽森48%对卢比奥46%；布蒂吉格以49比44同时领先万斯和卢比奥。94 奥索夫和纽森领先万斯5个百分点，但分别落后卢比奥3个和2个百分点，奥卡西奥-科尔特斯则落后于两位共和党人。95 Morning Consult在较早的调查中显示哈里斯对万斯领先1个百分点，对卢比奥领先2个百分点。96

据此得出三点结论。第一，没有任何候选人在对决中有稳定的领先；多数在5个百分点以内，这处于常规误差范围和两年内正常变化范围之内。第二，跨调查的模式同步：共和党人卢比奥在对阵若干民主党人时表现稍优于万斯，这表明在该样本中，候选人个人的好感度比党派标签更重要。第三，知名度和当前对执政党不利的中期选举环境，很可能以慢消退的方式影响这些数字。

9.5 2024年基准

2024年的结果是2028年建模的必要参照。唐纳德·特朗普获得49.80%的普选票，卡玛拉·哈里斯为48.32%，差距2,284,967票，即1.48个百分点；选举人票312对226。特朗普赢下全部七个摇摆州，差距从威斯康星的0.86个百分点到亚利桑那的5.53个百分点不等。97 结果在选举人团层面是决定性的，在普选票层面则很接近。这对2028年的分析很重要：候选人的全国好感度与选票在七个州的分布是不同的变量，二者都会进入结果。

9.6 辩论

辩论可能影响大选，但关于影响有多大的证据有限。2024年，乔·拜登与唐纳德·特朗普6月27日的首场大选辩论被普遍认为对总统不利，拜登在党内压力不断增大后于7月21日退出。98,99 随后哈里斯与特朗普于9月10日举行了一场辩论，因此该周期共举行两场大选辩论，而非三场。100 自1960年以来，没有任何一年的总统辩论超过三场。101 总统辩论委员会原计划2024年举行三场辩论，但两个竞选团队对形式和时间提出异议，委员会因此被绕开。102 因此，2028年辩论的数量、时间和主办方尚未确定。

分析上的教训是：辩论是关键的信息事件，尤其是在揭示选民此前未见观察到的情况时，如2024年的那场辩论。当两位候选人的表现都在常规范围内时，较少证据表明辩论能可靠地改变大量选民的选择。从而，报告把辩论视为结果方差的来源，而非决定性因素。

9.7 独立选民与选民结构

一种普遍的简化把选民分为各约40%的两个党派基本盘和约20%的可争取的中间选民。这一简化有用但有局限。自认独立的选民是异质群体，其中很大一部分稳定地倾向某一政党。关于投票行为的研究总体发现，总统选举中实质可说服的选民占比小于自认独立选民的占比，而边缘选民的投票率与说服亦然重要。实际含义是：2028年将由基本盘动员、少数可说服选民的倾向和约七个州的选举人团地理格局合谋决定。

候选人属性，包括沟通能力、好感度、行政经验、联盟的广度和联合政党的能力，与所有这些渠道相关。属性在两年前很难衡量，且支持者与反对者的评价往往各异。故而，本报告不按个人素质对候选人排名，转而指出分析人士将跟踪的可观察变量：独立选民中的好感度；在西班牙裔、新生和郊区选民中的表现；筹款；以及初选日程的结果。

十、与全球经济走势的联系

美国选举通过四条渠道影响全球经济。以下陈述为带有清晰置信度的假设，并与InsightBridge关于“被迫型区域化”的配套论文中的框架相衔接。

1. 贸易政策的持久性（中等置信度）。 2026年2月20日，最高法院裁定《国际紧急经济权力法》不授权征收关税，贸易政策随后转向其他法定权限，包括针对加拿大的1930年《关税法》第338条。103,104 民主党控制的众院不会直接掌握关税权限，但可通过监督、拨款和立法影响之。贸易伙伴更可能把中期选举结果解读为关于政策持久性的信号，而不是当前关税的变化。
2. 能源与伊朗冲突（中等置信度）。 国际货币基金组织2026年7月更新预测2026年全球增长3.0%、2027年3.4%，并称战争冲击对能源进口国形成压力。105 中期选举投票不会改变霍尔木兹的流量，但高燃油价格的政治成本会影响达成谈判结果的激励。油价波动仍将是冲突传导到美国和海外家庭物价的主要渠道。106
3. 货币条件与美元（低至中等置信度）。 美联储已开始加息，美元占已分配外汇储备的份额在2026年二季度为56.70%，一季度为57.18%。107 中期选举后财政或贸易方向的变化，只会通过对赤字和政策可预测性的预期影响美元需求，且这类影响将是渐进的。
4. 盟友对冲（中等置信度）。 欧盟、加拿大、日本和韩国等盟友经济体，一直在维持核心安全关系的同时，通过建设区域和多边替代安排进行对冲。108 在此背景下2028年大选之所以重要，是因为盟友的规划期限长于一届政府。11月若出现表明政策波动性更大的结果，将提高对冲的价值；表明政策连续性的结果则会降低之。无论如何，第9.2至9.4节所示的2028年不确定性，很可能使对冲行为至少持续到2028年。

十一、分析的风险与反证

• 民调误差。 像2024年那样，共和党统一偏差3个百分点，会降低民主党控制众院的概率，并使民主党控制参院变得不太可能。朝民主党方向的偏差则会产生更大的众院多数，并使两院全胜更有可能。
• 后期经济变化。 重新开放霍尔木兹的外交解决会降低燃油价格，尽管向消费者价格的传导会滞后。10月CPI将于11月10日公布，在投票之后。109
• 民主党品牌偏弱。 Cook警告说，民主党的党派好感度弱于总统声望所显示的水平，缅因州的竞选即为例证。110 选民可以拒绝执政党而不拥抱反对党，这可能降低边缘选民的投票率。
• 选区重划与法律变化。 选区图仍可能因诉讼而改变，新选区图的实际效果也不确定，因为两党都不指望赢下所有目标席位。111
• 市场动量。 预测市场已迅速向民主党移动，两院全胜概率从2月的39%升至约62%。112 动量可能逆转。
• 确定性论断。 一些评论把约7–8个百分点的泛国会投票领先视为民主党在两院获胜的保证。证据支持众院的高概率和参院的中等概率，但不支持任何一院的确定性。

十二、观察指标

观察指标	提高民主党胜算的情形	降低民主党胜算的情形
泛国会投票均值	持续在民主党+8以上	低于民主党+6
总统支持率（均值）	低于38%	高于42%
汽油（AAA）	保持在4.30美元以上	跌破4.00美元
参院战场	得克萨斯、俄亥俄、阿拉斯加的领先保持	缅因或艾奥瓦的领先消失
提前投票构成	年轻和边缘选民投票率高	战场州共和党提前投票追赶
9月和10月CPI	总体通胀高于3.7%	总体通胀和汽油价格同时下降
伊朗会谈	破裂或对航运的进一步攻击	签署重新开放霍尔木兹的协议
2028年宣布参选	11月3日后的早期宣布	重要宣布推迟

十三、结论

证据支持对11月3日的明确中心预期：民主党赢得众议院多数的可能性很高；民主党赢得参议院的可能性略高于一半，但取决于在2024年大比分投给共和党的州中的少数相关竞选；整体上最可能的结果是民主党两院全胜。这一预期建立在泛国会投票约8个百分点的全国差距、总统支持率处于低位和能源驱动的生活成本上升之上。它受到以下因素的限制：得票与席位之间的差距、选区重划、历来常有利于共和党的民调误差，并包括民调中好感度低于全国差距所显示水平的民主党品牌。

对于2028年，证据只支持两项确切陈述。两党的人选都是开放的，领跑者的支持率约在26%–39%之间，且未决定比例显著。早期对决测试没有显示任何一党的持续优势。对全球读者在方面，中期选举的现实意义，不如说在于眼前的席位数，不如说在于结果对政策连续性、贸易权限、能源价格并包括盟友规划期限所发出的信号。

作者简介

殷彤博士（Dr Tong Yin），美国洞见桥全球公司（InsightBridge Global LLC）及其姊妹研究机构洞见桥全球实验室（InsightBridge Global Lab LLC）创始人兼首席研究官，常驻美国阿拉巴马州奥本市。殷彤博士获奥本大学博士学位，并曾在该校任教；另获美国东伊利诺伊大学工商管理硕士（MBA）学位。创立洞见桥之前，他在美国、中国、澳大利亚和欧洲的国际企业担任高级管理职务长达二十年，行业涵盖酒店、专业服务和全球制造，负责运营、商业战略和跨境扩张。

他的研究领域包括地缘政治风险、宏观经济再平衡、人工智能战略和组织韧性，提出了核心密码理论（Core Code Theory）、家园模型（Home Model）、治理负债（Governance Debt）和动态驱动替代理论（DDRT）等原创分析框架。他主笔双语战略情报周刊《洞见桥全球情报网》（InsightBridge Global Intelligence），读者包括主权财富基金、中央银行、政府部委和机构投资人；同时是 Hospitality Net、Hotel News Resource 和 PhocusWire 的长期撰稿人。他以英语和普通话工作。

联系方式：tongyin@insightbridge.global ｜ https://intelligence.insightbridge.global

免责声明

本报告是由 InsightBridge Global 及 InsightBridge Global Lab 发布的独立第三方分析。本报告与美国或其他任何国家的任何政党、候选人、竞选团队、政府、政治行动委员会或倡导组织均无隶属、资助或指导关系，不带有任何政治、党派、意识形态或国家倾向。报告中对政党、候选人和官员的提及仅为描述性的，目的仅在于报告公开事实和已发布的数据，不构成支持、反对或批评。

概率和情景是作者基于2026年10月14日可得信息作出的分析判断，不是对确定结果的预测，选举结果仍具不确定性。民调结果、预测市场价格和预测模型输出均按其来源报告引用，可能变动或被修订；本报告未对其进行独立审计。本报告不构成投资、法律、税务或投票建议。文中观点仅代表作者殷彤博士个人，不必然代表 InsightBridge Global 或 InsightBridge Global Lab 的立场。

Independence statement. This report is an independent, third-party analysis. It takes no political, partisan, ideological or national position. It describes evidence and assigns probabilities; it does not recommend how anyone should vote. The full disclaimer appears at the end of the report and is also published as a standalone document.

---

Executive Summary

The United States will hold midterm elections on 3 November 2026, with all 435 House seats and 35 Senate seats on the ballot.1 Two weeks before the vote, the available evidence points to a clear but not uniform national environment. The presidential approval rating is at or near the lowest level of the second term, the generic congressional ballot favours the Democratic Party by roughly 8 to 10 points across the main averages and recent polls, and consumer prices have been pushed up by an energy shock linked to the Iran conflict. At the same time, translating a national vote margin into seats depends on district maps, candidate quality and turnout, and the Senate map requires Democrats to win several states that President Trump carried by double digits in 2024.

The report reaches five key judgments. Probabilities are the author’s analytical estimates, built from polling averages, prediction-market prices, published forecast models and historical error patterns. They are not forecasts by any official body.

#	Key judgment	Estimate	Confidence
1	Democrats win a House majority (218 seats)	About 90% (markets about 90%; models 77–94%; this study’s district-level model about 86%; historically calibrated scenario about 91%)	High
2	Democrats win Senate control (net gain of four seats)	About 65%, range 55–72% (markets about 70%; outside models 62–78%; this study’s state model 51–69%)	Moderate
3	Most likely outcome is a Democratic sweep (about 61%); the second most likely is a Democratic House with a Republican Senate (about 29%)	See Section 7	Moderate
4	Energy-driven inflation and the cost of living are the main economic channel behind the current environment; a late fall in gasoline prices would narrow, but probably not close, the gap before 3 November	Qualitative	Moderate
5	The 2028 contest is wide open: leading names in both nomination polls hold large but soft leads, and early general-election polls are weak predictors	Qualitative	High

This edition adds Section 4.5, which reruns the district-level model with the polls of 31 October and reports the five scenarios registered before the election. It also adds Section 5.6, which cross-checks the Senate judgment against a new state-level Senate model built for this study, and corrects the seat composition in Section 5.1 (22 Republican-held and 13 Democratic-held seats up). The previous edition added Section 4.4, which cross-checks the House judgment against this study’s district-level statistical model and sets out the model’s limitations and the data-collection plan for 2028. It also adds Section 5.5, which examines candidate quality and local economics in three key states: North Carolina, Kansas and Maine. The previous edition added an author’s vote-arithmetic cross-check (Section 5.4). It derives Republican voters’ disapproval of the president from national and partisan disapproval figures, and it sets out how many Republican-held Senate seats Democrats must win. The check supports the direction of the judgments above; the probabilities themselves combine it with independent models and markets. It also corrects two Senate ratings in the earlier edition: Michigan is a Toss Up and North Carolina is rated Likely Democratic.2

Three further points frame the analysis. First, poll-based leads and market prices are not independent evidence; many market traders read the same polls. Second, the 2024 election showed that national polls can understate the Republican share by about 3 points in swing states, so a Republican over-performance of that size remains plausible and would matter most in the Senate.3 Third, the midterm result and the 2028 result are only loosely linked: the 2010 Republican wave was followed by the 2012 re-election of the Democratic president, and the 2018 Democratic wave was followed by the 2020 defeat of the Republican incumbent, which shows that midterm outcomes shape the political agenda more than they determine the next presidential result.

1. Analytical Framework

1.1 What the generic ballot measures

The generic congressional ballot asks respondents which party they would support for Congress without naming candidates.4 It is the standard summary of the national environment, but it has four limits that matter for interpretation.

1. Votes are not seats. The national House vote is the sum of 435 separate contests. Seat outcomes depend on how efficiently each party’s voters are distributed across districts.
2. Maps change. District lines used in 2018 were drawn after the 2010 census, whereas 2026 elections use post-2020 maps and several mid-decade redrawings.5
3. Population definitions differ. Averages mix adults, registered voters and likely voters. Silver Bulletin reports D+8.8 using all polls and D+9.5 after a likely-voter adjustment.6
4. Dispersion is large. Individual polls in the current RealClearPolling table range from D+2 (Harvard-Harris) to D+12 (Quinnipiac and a THP/MM poll), around an average of D+8.3.7

1.2 Method used in this report

The report uses a triangulation approach rather than a single model. It compares (a) three or more polling aggregators with different methods, (b) two prediction markets, (c) published forecast models, (d) historical benchmarks for the relationship between national margin and seat change, and (e) an author’s vote-arithmetic cross-check that tests the result against partisan disapproval and the Senate seat count (Section 5.4). Where sources disagree, the disagreement is stated and the report adopts a value inside the range, with a discount for known sources of error. The same principle of cross-checking opposing or independent sources applies to the 2028 sections.

1.3 Known error patterns

• Final-poll error. In 2024 the final polling averages understated Donald Trump’s share by roughly 3 points in each swing state; the error is described as within the normal range and smaller than in 2016 and 2020.8
• Direction is not stable. Polling error has favoured Republicans in 2016, 2020 and 2024 but favoured Democrats in some state races in 2022. A uniform correction in either direction is therefore not warranted. A symmetric uncertainty of about 3–4 points on the national margin and 4–5 points in individual states is a reasonable working assumption.
• Primary and turnout effects. Midterm turnout is lower than in presidential years, and the party with the more motivated base often gains relative to registered-voter polls. This is one reason the likely-voter adjustments at present favour Democrats.

2. The National Environment

2.1 Presidential approval

Presidential approval is the most consistent predictor of the president’s party’s midterm performance. Current readings are low across pollsters, though levels differ by method.

Measure	Reading	Source
Pew Research Center (28 Sep–4 Oct)	30% approve, 68% disapprove	9
Reuters/Ipsos (30 Sep–5 Oct)	32% approve, 65% disapprove	10
AP-NORC (24–28 Sep)	31% approve, 69% disapprove	11
Fox News (9–12 Oct, registered voters)	36% approve, 64% disapprove	12
Morning Consult (9–11 Oct)	42% approve, 56% disapprove	13
Rasmussen (likely voters, 12 Oct)	41% approve, 58% disapprove	14
Aggregators (8 Oct)	RealClearPolitics 38.3%; New York Times 37%; Silver Bulletin 37.1%	15
Silver Bulletin (19 Oct)	37.6% approve, 59.8% disapprove	16

The spread between pollsters (30% to 42%) shows that house effects and population definitions matter, but no major series places approval above the mid-40s. The Wall Street Journal’s 37% reading is the lowest for a president heading into a midterm in that poll’s series since 1990.17 For context, Gallup measured 43% for President Trump at the same point of his first term, which preceded a 2018 midterm in which his party lost 41 House seats.18,19

Approval of economic management is lower still: 28% in a Financial Times/Focaldata poll.20 Approval among Trump’s own 2024 voters has fallen from 95% early in the term to 67% in the Pew survey.21 In the Fox News poll, 23% of Republican voters disapproved of the president’s performance, against 78% of independents and 98% of Democrats.22

There is also evidence moving in the other direction. The Economist/YouGov series shows net approval among Black and Hispanic respondents improving by about 19–20 points across three consecutive polls from 18 September to 5 October, albeit from very low levels.23 Aggregates also move day to day; Silver Bulletin’s net figure was −22.4 on 14 October after touching −23.9 on 9 October, and stood at −22.3 (37.6% approve, 59.8% disapprove) on 19 October.24,25

2.2 The generic ballot against history

Democratic leads are consistent across the main averages.

Aggregator	Democrats	Republicans	Margin	Date
RealClearPolling	50.2	41.9	D+8.3	21 Sep–11 Oct 26
Silver Bulletin (default)	n.a.	n.a.	D+8.8	13 Oct 27
Silver Bulletin (likely voters)	n.a.	n.a.	D+9.5	12–13 Oct 28
Silver Bulletin (latest)	51.3	41.7	D+9.6	19 Oct 29
Decision Desk HQ	49.0	41.3	D+7.7	13 Oct 30

The default aggregator readings range from D+7.7 to D+9.6, and the Fox News poll of 9–12 October shows 54% to 45% (D+9) among registered voters and 52% to 46% among those certain to vote.31 The central reading is therefore about D+8 to D+9. Over the last month the margin moved from roughly D+6.5 to D+8.9 (Silver Bulletin), from D+5.7 to D+8.5 (RealClearPolitics) and from D+6.7 to D+7.7 (DDHQ).32 The movement is therefore recent, and recent movement can partly reverse.

Historical comparison shows why the margin matters but does not determine the seat count. In 2018 Democrats won the national House vote 53.4% to 44.8% (D+8.6) and gained 41 seats.33 In 2010 Republicans won 51.7% to 44.9% (R+6.7) and gained about 63 seats.34 In 2022 Republicans won 50.0% to 47.3% and gained 9 seats.35 The conversion of votes to seats therefore varied from roughly 3 to 9 seats per point of margin across these three cycles. Applying even the lower end of that range to a D+8 environment would imply a Democratic gain well above the three seats needed, but the 2026 maps are different, and the report therefore treats seat conversion separately in Section 4.

2.3 Early voting

More than 3.7 million mail-in and early in-person ballots had been cast nationally as of 14 October, and about 51.7 million mail ballots had been requested or sent.36 In Georgia, 188,667 ballots were accepted on the first day of early voting, above the first-day total in 2022.37 Ohio reported 194,481 early in-person votes with 21 days left, against 50,368 at the same point in 2022.38 CNN’s early analysis, as summarised by Ground News, shows an early Democratic turnout advantage, but early-vote composition is a weak predictor because the two parties’ voters now differ in their preferred voting method.39 The early-vote evidence indicates high engagement; it does not by itself indicate which party will benefit.

3. The Economy and the Cost of Living

3.1 Current conditions

The economic picture is mixed rather than uniformly negative, which is typical of an energy supply shock.

Indicator	Latest reading	Source
CPI inflation, September (year on year)	3.7% (August: 3.4%)	40
Core CPI (excluding food and energy)	2.6% (August: 2.4%)	41
Gasoline CPI, monthly change in September	+9.1%, over half the monthly increase	42
Energy CPI, 12 months	+20.4%; gasoline +34.2%	43
AAA national average gasoline price	$4.33 (14 Oct), versus $3.07 a year ago; above $4 on a record 158 days this year	44
Real GDP growth, Q2 2026 (third estimate)	2.2% annualised; Q1 revised to 2.5%	45,46
Payrolls and unemployment, September	+29,000 jobs; unemployment 4.2%	47
Federal funds rate target	3.75%–4.00% after a 25 basis point rise on 16 September, the first hike since 2023	48
Brent crude, early–mid October	Roughly $98–106 per barrel, against about $72 before the conflict	49,50

The combination of positive growth, a low but softening labour market, rising headline inflation and a central bank that is tightening is the main economic context for the vote. Gasoline is the most visible price. The AAA reading of $4.33 compares with $4.36 a week earlier, so there is a mild easing at the pump, but prices remain at record levels for autumn.51 By 19 October the national average had eased further to about $4.30, while Brent crude closed the previous week at $109.86, its highest level since May; because pump prices follow crude with a lag, a renewed rise before the vote cannot be ruled out.52

3.2 Mechanism: from energy prices to voting intentions

Three channels link the energy shock to the electoral environment.

1. Salience. Gasoline prices are posted publicly and purchased frequently, so they shape perceptions of living costs more than their weight in the CPI basket would suggest.
2. Attribution. In an AP-NORC survey, 65% of respondents assigned responsibility for high living costs to the president.53 Whether or not that attribution is analytically complete, it shapes behaviour.
3. Policy response. The Federal Reserve’s rate increase raises borrowing costs for households and firms, and so can intensify cost-of-living concerns even as it aims to reduce inflation.

3.3 Limits of the economic explanation

Economic conditions explain part of the environment but not all of it. Real consumer spending and investment were strong in Q2 (real final sales to private domestic purchasers rose 4.6%), and initial jobless claims have been at historically low levels.54,55 The September payroll figure of 29,000 jobs, with revisions that removed 60,000 jobs from July and August, suggests a cooling.56 Voters’ ratings are also influenced by other factors, including views on the Iran conflict, immigration and trade policy. The report therefore treats the cost of living as the largest identifiable factor, not the only one.

3.4 Oil and the war

The conflict that began on 28 February 2026 remains unresolved. Reuters reports that the U.S. Energy Information Administration now forecasts Brent at about $105 per barrel in the fourth quarter and about $98 for 2026.57 Kpler ship-tracking data show Hormuz crude exports averaging about 8.5 million barrels per day in the week to 7 October, about 40% below pre-war levels, and the International Energy Agency has described the disruption as the largest supply disruption in the history of the oil market.58 These conditions imply that the energy-price risk to the election is two-sided: a diplomatic opening would lower prices quickly, whereas further attacks on shipping would raise them. Announcements of talks have already moved Brent by several dollars in a single session.59

4. The House of Representatives

4.1 Starting position and ratings

Democrats hold 214 seats and need 218, a net gain of about three to four.60 Cook Political Report’s ratings of 9 October place 185 seats in Solid Democratic, 16 in Likely Democratic and 12 in Lean Democratic, against 174 Solid Republican, 18 Likely Republican and 9 Lean Republican, with 21 Toss Ups.61 Counting the 213 seats at least leaning Democratic, Democrats are within five seats of a majority before any Toss Up is decided. Cook shifted a large block of seats toward Democrats on 9 October, including five Toss Ups moved to Lean Democratic in Iowa, Pennsylvania and Washington.62

A comparison of seats rated Toss Up or worse by party shows the structure: 23 Republican-held and 11 Democratic-held seats in October 2026, against 45 and 3 in October 2018, and 12 and 30 in October 2022.63 The 2026 imbalance is smaller than the 2018 one. That is consistent with the effect of redistricting and with fewer Republican seats in highly competitive territory than in 2018.

4.2 Redistricting

The 2025–26 mid-decade redistricting round is the main structural difference from 2018. Republicans had hoped to gain 13 seats from new maps, but House Majority Whip Tom Emmer said they would not win all 13, and Cook now projects a net gain of about 4 seats from redistricting alone.64 Emmer also said Republicans could lose five seats combined because of redistricting in California and Utah.65 The practical implication is that maps reduce the Democratic seat gain by roughly four seats relative to a no-redistricting baseline, but do not remove a D+8 advantage.

4.3 Seat estimate

Cook’s most likely outcome, as cited by one tracker, is a Democratic gain of 9 to 19 seats, for 224 to 234 in total.66 Polymarket’s projected seat count is 234–201.67 Taking those ranges and a symmetric uncertainty of several points of national margin, the report’s central range is 222–238 Democratic seats. That implies a probability of a Democratic majority of about 90%. Prediction markets price the House at about 90% (Kalshi and Polymarket), while published models range from 77% (Decision Desk HQ) to 94% (Silver Bulletin).68,69,70 Three reasons argue against going much higher: the average could partly reverse, a 3-point Republican polling miss would still leave Democrats as favourites but narrow the margin, and about 21 Toss Ups remain. Three reasons argue against going lower: the lead has been stable above D+6 since August, early Democratic engagement indicators are positive, and the seat gap to a majority is small.

4.4 Cross-check with a district-level statistical model (added 24 October)

The House judgments in the previous three sections come from the author’s own analytical logic, reasoning and calculation. They take the generic ballot as the measure of the national environment, combine it with Cook ratings, redistricting estimates, prediction markets and published forecast models, and then set a range and a probability using historical error patterns. To test those judgments, this study built a separate district-level statistical model from large public datasets and froze its run on 24 October. The two approaches use different data and different steps from votes to seats, so they can be compared.

How the model is built.

• Voter survey: public data from the Cooperative Election Study (CES) for 2018–2024.71 Turnout is estimated from the validated 2022 midterm vote of about 60,000 respondents. Vote choice comes from pre-election House vote intention in 2024, with 37,887 two-party respondents.
• Estimation: a weighted ridge logistic regression with partial pooling across education × race × state groups. Each district’s two-party 2024 presidential vote enters as its partisan lean.72
• District population: computed on the actual 120th Congress map. The 173 districts in the nine states with new maps were rebuilt from Census block equivalency files, 2020 Census block data and 2023 American Community Survey education data.73
• Incumbency: measured in the 392 districts with candidates from both parties in 2024. An incumbent running gains about 1.72 points over an open seat held by the same party (standard error 0.19).74 In 2026, 184 Democratic and 177 Republican incumbents are running, and 74 seats are open.
• National environment: the mean of the Silver Bulletin, RealClearPolitics and Decision Desk HQ generic-ballot averages, a Democratic lead of 8.3 points, or a two-party share of about 54.2%.75,76,77
• Uncertainty: errors at the national, state and district levels, with a fat-tailed national term; 40,000 simulations, seats counted district by district, and 218 seats as the majority threshold.

Main results.

Measure	Author’s judgment (Section 4.3)	District-level model
National environment	Democrats ahead by 8–9 points	Democrats ahead by 8.3 points
Democratic seats	Central range 222–238	Median 240; 10th–90th percentile 209–274
Probability of a majority (≥218)	About 90%	About 83%
Competitive districts	About 21 rated Toss Up by Cook	72 districts with a Democratic win probability of 20–80%
Redistricting	Cook estimates a net Republican gain of about 4 seats	Computed on the actual 2026 map; subtracting a further 4 seats as a conservative case gives a median of about 236 and a probability of about 79%
Empirical seat-vote curve (fitted to 2006–2024)	—	238 seats

The district win probabilities sum to about 241 seats, and Democrats are favoured in 235 districts. Because district lean has already been recalculated on the new maps, the “further 4 seats” case partly double-counts redistricting and serves only as a conservative lower reference.

Sensitivity. As the national environment moves from a tie to a Democratic lead of 10.5 points, the model’s median rises steadily from 205 to 250 seats and the probability of a majority from about 28% to about 90%. At a Democratic lead of about 3 points the model puts the majority at roughly even odds. On the current maps, Democrats therefore need a national lead of about 3 points for the two parties to have roughly equal chances of a majority.

Backtest. On the two-party measure, the model’s mean absolute error for the national House vote in 2018–2024 is 1.18 points, with a root mean square error of 1.24 points and a mean bias of −0.19 points. Leave-one-out estimates from the empirical curve understate Democratic seats by 6.8 on average. With only four elections, and with 2018 based on post-election recall rather than pre-election intention, these results are a preliminary calibration only.

Where the two approaches agree.

1. Direction: both find a Democratic House majority clearly more likely than not.
2. Central value: the top of the author’s range is 238 seats, the model median is 240, the empirical curve gives 238 and Polymarket’s projected count is 234.78
3. Redistricting: the new maps reduce the Democratic seat advantage but do not offset a national lead of about 8 points.
4. The national environment is the main variable: both show that the outcome depends first on whether the Democratic national lead holds at a moderate level or above, and only second on individual district contests.

Where they differ.

1. Width of uncertainty: the model’s 10th–90th percentile range spans about 65 seats, much wider than the author’s central range. The model knows only the national environment and district structure. It has no information on candidate quality, fundraising, local polls or rating agencies’ district calls, so it treats more districts as uncertain: 72 competitive districts against about 21 Cook Toss Ups. The author’s judgment absorbs that district-level information and its range is narrower as a result.
2. Lower probability: the model’s centre is slightly higher, but its wider distribution puts more weight below 218 seats, so its majority probability is about 83%, below the author’s 90%. Published models range from 77% to 94%, and both figures fall inside that range. The author’s 90% matches prediction markets; the model suggests that this figure sits at the upper side of the plausible range.
3. Slightly higher centre: the model converts the 8.3-point lead directly into seats and reaches a median of 240. The author lowered the centre to allow for partial reversion of the lead and for past polling errors that have often favoured Republicans. In the backtest the empirical curve tends to understate Democratic seats, so there is no sign that the model overstates Democrats, but four elections are too few to settle the point.
4. Limited independence: the model takes the generic-ballot average as its national input, so agreement on the national environment holds by construction. Its independent contribution is the translation from votes to seats. A broad national polling error would affect both approaches in the same way.

Limitations of the model.

• The sample was not purpose-collected: the model uses public CES data. CES obtains representativeness by matching an online panel to population targets; it is not a probability sample. Vote choice rests on about 38,000 two-party respondents, fewer than 90 per district on average across 435 districts, so district estimates rely mainly on district lean and group-level extrapolation. This is not the sample the author has in mind: 100,000 to 200,000 real respondents, designed and collected to reflect each state’s electorate. A collection on that scale could not be completed in the short time before this election, and this is the model’s main limitation.
• Timing of the data: vote choice is 2024 intention. Change since then enters only as a uniform national shift, which cannot capture different movements across states and districts over two years. Population figures are from the 2020 Census and education figures from 2023.
• Turnout level is high: the model’s implied midterm turnout is about 58.4%, while the Census Bureau measured 2022 turnout among eligible citizens at 52.2%.79 The survey sample leans towards politically engaged people. District shares are calibrated to the national target, but the composition of the electorate may still be off.
• Few backtest elections: only four elections are available, and the intervals around the error measures are themselves unstable.
• Factors left out: candidate characteristics, campaign finance and local issues are not in the model, and it covers the House only, not the Senate.

Plan for 2028. For the 2028 presidential election, the author plans to design and collect a real sample of 100,000 to 200,000 respondents built on each state’s electorate, with stratified sampling within states, oversampling of competitive states and districts, and panel re-interviews, to describe the structure of the roughly 240 million eligible voters in the United States.80 This design addresses the model’s three main limitations directly: the source of the sample, state-level resolution and the timeliness of the data. With a design effect of 1.5, a sample of 150,000 would reduce sampling error to about ±0.3 points nationally and about ±2.2 points in the average state. In the average district it would still be about ±6.5 points, so partial pooling and district covariates would still be needed.

What this means for the report’s judgment. The district-level model supports the direction and the central value of the report’s House judgment, but suggests the probability may be somewhat lower than the author’s estimate. The report keeps its estimate of about 90% for a Democratic House majority and uses the model’s 83% as a more conservative reference; a plausible range is about 80–90%.

4.5 Final pre-election update (added 31 October)

Election day is three days away, on 3 November. This section reruns the district-level model of Section 4.4 with the latest polls of 31 October and reports the five scenarios registered before the election. Where the figures differ, this section supersedes Section 4.4.

National environment. The model uses three generic-ballot averages:

Source	Democrats	Republicans	Margin
Silver Bulletin (computed with its poll-database weights, 31 October)81	51.9%	43.3%	D+8.6
RealClearPolitics (8–26 October)82	51.4%	43.2%	D+8.2
Decision Desk HQ (29 October)83	51.2%	41.8%	D+9.4
Mean of the three			D+8.73

The mean was D+8.3 in the 24 October edition. Decision Desk HQ updated to 51.5% against 42.6% (D+8.9) on 30 October;84 substituting it gives a mean of about D+8.6, which on the model’s sensitivity analysis moves the median by less than one seat.

Two changes to the model. Compared with Section 4.4, the model adds two components:

• Expert-rating layer: it adds the information in the Cook, Inside Elections and Sabato ratings that the model does not already contain (the rating surprise). In the 2006–2024 backtest it improves the Brier score for district outcomes by about 39% on the validation set and 35% on the sealed set. The form of this layer was settled after the first validation results had been seen, so 2026 is its first genuinely out-of-sample test. With the layer, competitive districts (a Democratic win probability of 20–80%) fall from 72 to 53, which addresses the gap noted in Section 4.4: the lack of district-level information.
• Issue-salience layer: state-level search interest in gas prices, inflation, abortion and immigration failed the pre-registered leave-one-election-out test for all four issues, so the model makes no adjustment.

Pre-registration. To rule out adjustment after the fact, the model was frozen and four scenarios were registered on 30 October; a fifth was added before the election on 31 October. The code, data and results of each scenario carry digital fingerprints, and all five will be scored after the election by four rules set in advance: seat error, the log score of the majority probability, district Brier score and interval coverage.

Scenario	National environment	Median seats	10th–90th pct.	Majority	With redistr.
R1 main	Mean raw margin of the three; no systematic poll bias	239	212–270	85.7%	235 / 81.1%
R2 poll discount	Discount fitted to 2006–2016; national environment about D+3.5	229	199–255	71.2%	225 / 64.5%
R3 Cook ratings	Derived from 23 October ratings and historical hit rates	228	220–236	95.9%	—
R4 two-party	Two-party share computed within each pollster, then averaged	240	214–271	86.6%	236 / 82.1%
R5 historical calibration	Historical mean poll bias of 0.96 points removed; error sd 2.0 points	236	218–253	90.7%	232 / 85.2%

R5 uses the inputs frozen on 30 October (D+8.63). Its basis is that in the nine elections from 2006 to 2024 with data, the eve-of-election generic-ballot average overstated Democrats in seven, by 1.17 points of two-party share on average, with a standard deviation of 1.38 points.85

External comparisons. On 31 October the prediction market Kalshi put the probability of Democratic control of the House at 93.9%;86 the ElectIt model of Ebanks, Katz and King puts it at 88%;87 and in Cook’s ratings of 23 October Democrats are favoured outright in 218 districts.88

How to read these figures.

1. Direction: all five scenarios, the empirical seat-vote curve (239 seats), the prediction market and the external model show a Democratic House majority as more likely than not. They differ only in the size of the probability, from about 71% to 96%.
2. Central values cluster: the scenario medians lie between 228 and 240 seats, largely overlapping the author’s central range of 222–238.
3. Source of the difference in probability: it turns mainly on one assumption, whether national polls overstate Democrats as they have in most past elections. R2 represents a clear overstatement and gives the lowest probability. R5 removes the historical average bias but uses a narrower error band, so its probability is slightly higher than R1’s.
4. A change in sensitivity: with the rating layer, the model gives Democrats about a 55% chance of a majority with the national vote tied, about 70% with a 3-point lead and about 77% with a 5-point lead. The statement in Section 4.4 that a lead of about 3 points gives the two parties roughly equal chances no longer holds in the updated model. The change comes from the rating layer and will need to be tested after the election.

What this means for the report’s judgment. The report keeps its estimate of about 90% for a Democratic House majority, with a plausible range of about 80–90%. The model’s main scenario (85.7%) and its historically calibrated scenario (90.7%) both fall within that range. The most cautious scenario, in which polls clearly overstate Democrats, gives about 71% and is the report’s main downside risk. One election provides only one test; firm conclusions on method will rest on the backtests together with the 2026 result.

5. The Senate

5.1 The map

Republicans hold 53 of the 100 Senate seats. Thirty-five seats are up for election, including special elections in Florida and Ohio; 22 of the 35 are Republican-held and 13 Democratic-held.89 Democrats need a net gain of four seats for 51, because Vice President JD Vance breaks 50–50 ties.90 The NBC News analysis notes that Democrats must hold Michigan, Georgia and New Hampshire, and win in several states Trump carried by double digits in 2024.91

Cook’s ratings of 6 October place seven races in the Toss Up category: six Republican-held seats (Alaska, Iowa, Kansas, Maine, Ohio and Texas) and one Democratic-held seat (Michigan). Georgia and the open Republican-held seat in North Carolina are rated Likely Democratic, and Cook projects a Democratic gain of two to six seats.92 The earlier edition of this report omitted Michigan from the Toss Up list and described North Carolina as Lean Democratic; both points are corrected here. Cook describes control of the chamber as a true toss-up.93 Kansas moved into the Toss Up group after a New York Times/Siena poll showed a 45–45 tie between Republican Senator Roger Marshall and Democrat Adam Hamilton.94

5.2 Polling and the structural gap

New York Times/Siena polls of likely voters show Mary Peltola ahead of Dan Sullivan 50–43 in Alaska, James Talarico ahead of Ken Paxton 51–45 in Texas, Sherrod Brown ahead of Jon Husted 49–46 in Ohio, a 45–45 tie in Kansas, and Ashley Hinson ahead of Josh Turek 48–47 in Iowa.95 In the RealClearPolitics averages compiled by Election Central, Democratic candidates lead in 10 of 12 tracked races, but the margins in the decisive states are small: Turek +0.6 in Iowa, Jackson +1.2 in Maine, Peltola +1.4 in Alaska, Talarico +2.6 in Texas and Brown +3.1 in Ohio.96

Two analytical points follow. First, the margins in most battleground states are smaller than the typical polling error of 3–5 points, so each race is close to a coin flip with a Democratic tilt. Second, the races are correlated. If polls overstate Democrats by 3 points nationally, several of the closest races would move together, so the chance of winning exactly four or more of seven toss-ups is lower than a product of independent probabilities suggests. The 2024 presidential margins in these states were R+13 in Alaska and Iowa, R+11 in Ohio, R+14 in Texas and D+7 in Maine.97 New York Magazine’s review of current Senate polling makes the same point: if every race ended where the polls now stand, Democrats would gain seven seats, but a Republican over-performance of about 3 points could save the Republican seats in Iowa, Kansas, Maine and Texas and leave Republicans in control with 50 seats and the vice-presidential tie-break.98

Some Senate races also show candidate-level resilience. Cook notes that Senator Susan Collins of Maine remains competitive in a state Donald Trump lost three times, which Cook cites as an illustration of a weaker Democratic party brand relative to the president’s standing.99 This is a useful counterweight: the national environment is favourable to Democrats, but the party’s own favourability is not equally strong.

5.3 Markets and models

Prediction markets price a Democratic Senate at about 61–62%: Polymarket 61% and Kalshi 62% in prices checked on 19–20 October.100 Published models differ widely. Silver Bulletin’s FLIPR model gives 75%, Race to the White House 72.4% and Decision Desk HQ 56%, with DDHQ projecting a 51-seat Democratic caucus.101,102 Silver Bulletin’s own scenarios show why the range is wide: if polls are broadly accurate nationally, the Democratic probability is 88%; if they overstate Democrats by 3–4 points, it falls to 38%; if they understate Democrats by the same amount, it rises to 99%.103

The report’s estimate is about 65% (62% in the 19 October edition), with a range of 55–72%. It sits slightly below the market price of 31 October (about 70%), below Silver Bulletin and above Decision Desk HQ. The discount to the most favourable model reflects the correlated-error argument above, the 2024 record of polls understating Republicans by about 3 points, and the Democratic seats that must also be held. The estimate is above the most cautious model because the author’s cross-check in Section 5.4 and the consistency of Democratic leads across the seven Republican-held battlegrounds point in the same direction.

5.4 Vote-arithmetic cross-check (author’s method)

This section sets out the author’s own cross-check. It does not replace published polls; it asks whether the national figures are internally consistent with a substantial loss of support among the president’s own voters, and what that would imply for the Senate.

Step 1: deriving Republican disapproval from national figures. If the electorate is divided into Democrats, independents and Republicans in shares of about 40%, 20% and 40%, national disapproval of the president is the weighted sum of the three groups’ disapproval. Solving for the Republican group gives:

Republican disapproval = (national disapproval − 0.4 × Democratic disapproval − 0.2 × independent disapproval) ÷ 0.4

The Fox News poll of 9–12 October (1,202 registered voters, margin of error ±2.5 points for the full sample) reports national disapproval of 64%, with 98% among Democrats and 78% among independents.104,105 Disapproval of the president’s handling of specific issues is higher: 77% on inflation, 70% on the economy, 69% on tariffs and 64% on foreign policy.106

National disapproval used	Democratic / independent disapproval assumed	Derived Republican disapproval
64% (overall job rating)	98% / 78%	23%
64%	95% / 75%	about 28%
67% (midpoint)	98% / 78%	about 31%
67%	95% / 75%	35%
70% (economy rating)	98% / 78%	38%
70%	95% / 75%	about 43%

Three properties of the calculation matter for interpretation. First, the formula is arithmetically sound, and with the overall job rating as input it reproduces the 23% that Fox News measured directly; the two approaches agree when given the same input. Second, the result is highly sensitive to inputs: because the Republican figure is the residual, each 1-point change in national disapproval moves it by 2.5 points. Third, the 70% figure is an issue-specific rating rather than an overall job rating, so it tends to overstate general disapproval. The Fox Republican sub-sample is only a few hundred respondents, so its 23% carries a wider margin of error than the national figure, which is one reason for testing it with a derived estimate.

On this basis the report adopts a central estimate of about 30% for Republican voters’ disapproval of the president, within a range of about 25–38%. Disapproval is not the same as a vote for the other party. Some disapproving Republicans will still vote for their party’s candidates, and some will not vote. The figure should therefore be read as an upper bound on the potential erosion of the Republican base, not as a measure of votes lost.

Step 2: what the Senate requires. Republicans hold 53 seats. Democrats need a net gain of four, because a 50–50 chamber would be organised by the Republican vice-president. Losing three seats would therefore leave Republicans in control; they must lose four. Most of the Republican seats on the ballot are not competitive. The relevant group is the seven Republican-held seats rated Toss Up or better for Democrats: Alaska, Iowa, Kansas, Maine, North Carolina, Ohio and Texas.107 Democrats must win four of these seven while holding Michigan, the one Democratic-held Toss Up, and Georgia and New Hampshire.

The table below shows the probability of winning at least four of the seven seats if each seat has the same Democratic win probability and the outcomes are independent.

Democratic win probability per seat	Probability of winning at least four of seven
50%	50%
55%	61%
60%	71%
65%	80%
70%	87%

Decision Desk HQ’s seat-level probabilities in mid-October were about 82% for North Carolina, 62% for Alaska, 57% for Texas, 54% for Maine, 51% for Ohio, 48% for Iowa and 35% for Kansas, an average of about 56% per seat; it gave Democrats about 70% in Michigan.108,109 Two adjustments are needed before the table can be read as a probability of control. First, the races are correlated: a common national polling error would move most of them together, which pulls the higher values in the table towards the middle. Second, a Democratic loss in Michigan, Georgia or New Hampshire would raise the requirement to five Republican-held seats. Correlation also works in the other direction here, because Democratic losses in their own seats are most likely in the same environments in which they also lose the Republican-held battlegrounds. Taken together, the adjustments lower the independent-seat values by roughly 5–10 points.

Result. The arithmetic supports the direction of the report’s judgment: under conservative inputs, a meaningful share of Republican voters disapprove of the president, and Democrats lead in enough Republican-held seats to make a net gain of four more likely than not. It does not support certainty. The Senate outcome depends on four or more correlated races in states that the president carried by double digits in 2024, and on the size and direction of any national polling error. The report’s Senate estimate of about 65% (range 55–72%) combines this cross-check with the models and markets in Section 5.3.

5.5 Three key states: candidates and local economics (added 23 October)

The national environment sets the direction, but Democrats need a net gain of four seats, so control will be decided in a handful of individual races. This section examines three states where Democrats have their best chances of a breakthrough, each in different circumstances: North Carolina, Kansas and Maine. In all three, candidates’ personal qualities and local economic conditions play a role independent of national partisanship.

North Carolina: a clear gap between the candidates. The seat is open because Republican Senator Thom Tillis did not seek re-election. The candidates are:

• Democrat: Roy Cooper, a two-term former governor.
• Republican: Michael Whatley, a former Republican National Committee chairman who entered the race at Donald Trump’s public urging. This is his first run for elected office.110

Of the three, this is the seat where the Democratic advantage is clearest:

• Polls: Cooper has led every poll since both men entered the race in July 2025. An Elon University poll in October put him ahead by 13 points, 52% to 39%.111
• Money: In early October Cooper had US$12.4 million cash on hand against Whatley’s US$916,000. Senate Republicans’ main super PAC cancelled tens of millions of dollars of advertising at the start of the month.112
• Presidential support: At a rally for Whatley in Greenville on 20 October, Trump said, “I’m the one that got him into this,” adding, “I hope it works out.”113
• Rating: NBC News rates the seat “Likely Democrat”.114

Kansas: a centrist candidate in a deep-red state. Trump carried Kansas by about 16 points in 2024, and the state has not elected a Democratic US senator in nearly a century.115 The candidates are:

• Democrat: Adam Hamilton, founder and senior pastor of the United Methodist Church of the Resurrection, running as a centrist.116
• Republican: Senator Roger Marshall, seeking a second term.117

Election analysts see candidate quality as the main reason for Marshall’s slide and Hamilton’s rise.118 The polling picture is as follows:

• Averages: As of mid-October, the polling averages put Hamilton ahead by about 0.8 points, with individual averages ranging from Marshall +0.3 to Hamilton +1.5.119
• Latest poll: InsiderAdvantage had Marshall at 48% and Hamilton at 47%.120
• Enthusiasm: Democrats report markedly higher enthusiasm.121

This is the most uncertain of the three seats. The polls show a tie, but the state’s partisan base and voting history favour the incumbent. Decision Desk HQ earlier put the Democratic chance at about 35%.122

Maine: a moderate incumbent’s dilemma. Susan Collins has represented Maine since 1997 and is the only Republican senator seeking re-election in a state Trump lost in 2024.123 Her voting record cuts both ways:

• Breaks with the president: In 2021 she was one of seven Republican senators who voted to convict Trump. In his second term she voted against confirming Defense Secretary Pete Hegseth and FBI Director Kash Patel, and against final passage of the One Big Beautiful Bill Act.124
• Support for the president: CQ found that she supported the president’s position on 94.6% of votes in 2025.125

The Democratic side changed candidates mid-campaign:

• Graham Platner, who won the June primary, withdrew on 10 July after a sexual-assault allegation, which he denies.
• On 25 July, at the first replacement nominating convention in the state’s history, held in Bangor, Democrats nominated Troy Jackson, a former Maine Senate president and logger, by 566 votes to 5.126

Collins is under pressure from both sides. Some Republican voters resent her independent votes, while Democrats focus their campaign on the votes in which she sided with the president. The Fox News poll of 16–20 October (1,514 likely voters, margin of error ±2.5 points) found:127

• Vote intention: the candidates are tied at 49% each.
• Concerns: 53% of voters worry that Collins votes with the president too often, and 46% worry that Jackson is too liberal.
• Presidential approval: 62% of Maine voters disapprove of the president’s job performance, and more than half strongly disapprove.
• Enthusiasm: 79% of Democrats say they are extremely interested in the election, against 61% of Republicans.
• Timing of votes: Jackson leads by 40 points among those who have already voted; Collins leads by 12 points among those who plan to vote on Election Day.

Collins also has clear sources of resilience:

• Crossover support: She leads among independents by 13 points, and about one in five likely voters who disapprove of the president back her.128
• Appropriations seniority: As chair of the Senate Appropriations Committee, she stresses the value of her seniority in securing federal funds.129
• A record of beating the polls: In 2020 she trailed in nearly every public poll yet won by 50.98% to 42.39%.130

Maine uses ranked-choice voting in federal general elections. If no candidate wins a majority of first choices, ballots are retabulated by voters’ rankings. In 2020 Collins won a first-round majority, so no further rounds were needed.131 This year there are six declared write-in candidates, and the secretary of state considers a ranked-choice tabulation “a huge likelihood”. The result could take days or weeks to determine.132 Decision Desk HQ currently gives Jackson a 56% chance of winning.133

Local economics: high costs and low incomes combined. Maine’s economic conditions form an important backdrop to the race:

• Income: Maine’s median household income of US$71,773 is the lowest in New England and below the national median of US$78,538.134
• Heating costs: Maine is the most heating-oil-dependent state in the country; about half of households heat primarily with fuel oil. On 19 October the statewide average price reached a record US$6.04 a gallon, 83% higher than a year earlier.135
• Winter bills: The National Energy Assistance Directors Association (NEADA) estimates that average spending by Maine households heating with oil will rise from US$1,664 last winter to US$2,939 this winter, the highest in the Northeast.136

Low incomes combined with high energy costs make the cost of living a sharper issue in Maine than in most states. This fits the judgement in Section 3 of this report: energy-driven increases in the cost of living are the main economic channel behind the current political environment. With Republicans holding the White House and both chambers of Congress, voters’ economic discontent is more likely to be directed at the governing party.

The polling also shows that this mechanism does not work in one direction only:

• Age: Voters under 45 favour Collins by 6 points, while those 45 and older favour Jackson by 3.137
• Education: The divide is sharper still. College graduates back Jackson by 23 points; voters without a degree back Collins by 16.138

Economic pressure does not translate automatically into support for one party. Local service, candidate image and the educational make-up of the electorate also matter.

The author’s assessment.

• North Carolina: the seat Democrats are most likely to win of the three.
• Kansas: the most uncertain, with a Democratic chance below one-half.
• Maine: the pivotal seat for Senate control. Three factors lead the author to put the Democratic chance slightly above current models, at about 55–65%: low presidential approval, higher Democratic enthusiasm, and the heavy impact of energy costs on low-income households. However, Collins’s lead among independents, her record of beating the polls by a wide margin in 2020, and the uncertainty of a ranked-choice count mean the seat should still be treated as a toss-up.

One path to a net gain of four seats for Democrats is:

• win North Carolina and Maine;
• win two more of Alaska, Texas, Ohio, Iowa and Kansas;
• hold Michigan, Georgia and New Hampshire.

This points in the same direction as the report’s Senate estimate (about 62% on 23 October, raised to about 65% in this edition on the basis of Section 5.6; range 55–72%).

5.6 A state-level Senate model (added 31 October)

The preceding sections base the Senate judgment on polls, markets, outside models and the author’s vote arithmetic. To give the Senate the same kind of testable quantitative check as the House, this study built a state-level Senate model on 31 October and registered it before the election. It is independent of the House model in Sections 4.4 and 4.5 and shares only the national environment (the three-source generic-ballot mean, D+8.73) and the same historical poll-error settings.

How the model works.

• Data: all 35 races; ratings from 11 forecasters published on 28–30 October; head-to-head polls since September in 17 competitive states, using each pollster’s latest poll only.139
• Central estimate for each state: a poll average (more recent polls weigh more; polls sponsored by one party get half weight) combined with a prior. Half of the prior comes from the average of the 11 ratings and half from the national environment, the state’s partisan lean (PVI) and incumbency. The more polls a state has, the more weight the polls get, up to 85%.
• Error: two layers. A national layer is shared by all states and is the same size as in the House R5 scenario (standard deviation 2.0 two-party points); each state adds its own error, larger where polls are scarce. The shared layer captures the risk that all states miss in the same direction, which Senate forecasts most often understate.
• Seat count: among seats not up, the Democratic caucus (including two independents) holds 34 and Republicans 31. The Democratic caucus needs 17 of the 35 races to reach 51; at 50–50 the vice president breaks ties for Republicans. Dan Osborn in Nebraska and Seth Bodnar in Montana are independents who have not committed to either caucus and are counted separately.

Scenario	National error	D caucus median	10th–90th pct.	P(51+)
S1 main	No systematic poll bias	52	48–55	68.8%
S2 historically calibrated	Historical mean poll bias removed, as in House R5	51	46–54	51.1%
S3 no polls	Ratings and fundamentals only	51	48–54	53.3%
S4 wider national error	National standard deviation raised by 50%	52	46–56	63.7%

Key states. The table lists the main competitive races. “Poll average” is the margin of the Democratic side (including independents) over the Republican.

State	Held by	Poll average (pollsters)	Model centre	D-side win S1	S2
North Carolina	R	D+11.4 (9)	D+11.0	96%	92%
Georgia	D	D+9.0 (5)	D+9.6	92%	87%
Minnesota	D	D+7.1 (8)	D+8.0	89%	83%
New Hampshire	D	D+6.8 (4)	D+7.9	87%	81%
Michigan	D	D+4.8 (14)	D+5.0	79%	69%
Maine	R	D+2.1 (10)	D+3.1	69%	57%
Ohio (special)	R	D+3.5 (14)	D+2.9	68%	57%
Texas	R	D+2.6 (24)	D+2.1	63%	51%
Kansas	R	D+3.7 (8)	D+1.9	62%	50%
Alaska	R	D+2.8 (6)	D+1.5	59%	47%
Iowa	R	R+0.3 (12)	R+0.5	47%	35%
South Carolina	R	R+1.5 (8)	R+2.5	35%	25%
Nebraska	R	R+1.9 (7; independent challenger)	R+3.8	28%	19%

The Democratic path is therefore to hold Michigan, New Hampshire, Minnesota and Georgia, win North Carolina, and take three of Maine, Ohio, Texas, Kansas, Alaska and Iowa. The poll margins in all six are within 4 points, smaller than an ordinary polling error.

External comparisons. On 31 October the prediction markets Kalshi and Polymarket priced Democratic Senate control at about 70% and 71%;140,141 Silver Bulletin’s FLIPR model gives 77.7%, Race to the White House 77.4% and Decision Desk HQ 62%.142 The main scenario (68.8%) is close to the markets, below the first two models and above Decision Desk HQ.

How to read these figures.

1. Same direction: the main scenario, the markets and the outside models all make Democratic control more likely than not, but by a much smaller margin than in the House.
2. High sensitivity to national error: if national polls overstate Democrats by 2 points on the margin, the main-scenario probability falls from about 69% to about 50%; by 4 points, to about 31%; if they understate Democrats by 2 points, it rises to about 83%. The historically calibrated scenario (51%) is an instance of the first case.
3. Cautious fundamentals: ratings and fundamentals alone give 53%, which shows that the Democratic leads in several states the Republicans won comfortably in 2024 come mainly from this year’s polls rather than from the states’ long-run lean.
4. Limitations: the model has not been backtested on past Senate cycles and does not correct for pollster house effects. S2 applies the historical bias of the House generic ballot directly to state Senate polls, which is an assumption. Candidate quality enters only indirectly, through ratings and polls.

Implications for the report’s judgment. The report raises its estimate for Democratic Senate control from about 62% to about 65%, keeping the range of 55–72%. The reasons are that since late October the market price has risen from about 62% to about 70%, outside models have generally moved up by 5–8 points, and polls in the main competitive states have continued to move towards the Democratic side. The estimate stays below the main scenario (68.8%) and the market because the historically calibrated scenario (51.1%) shows that, if polls again overstate Democrats across the board, the outcome would be close to even; the report keeps a discount for that risk without assuming it will materialise. Like the House model, this model was registered before the election and will be scored afterwards on state-level Brier scores, the log score for control, seat-interval coverage and median error.

6. Money, Turnout Operations and Outside Actors

Campaign finance is a relevant structural variable. Elon Musk has given about $366 million to federal committees in the 2025–26 cycle, including $276 million since July, of which $158 million went to the Senate Leadership Fund and about $76 million to America PAC, which runs a turnout operation in seven Senate battlegrounds and nearly three dozen House races.143 He had earlier threatened to form a third party, the America Party, but has not done so.144 Separately, MAGA Inc., the pro-Trump super PAC, was reported to hold $416 million in August.145

Spending of this scale can affect outcomes at the margin, particularly in expensive states such as Texas, where the Senate Leadership Fund has spent well over $100 million.146 The evidence on how much advertising and turnout spending moves outcomes in high-information races is mixed, and it is likely to be smaller than the headline totals suggest. The report treats the funding advantage as one reason for not assigning more than about 61% to the Democratic sweep despite the national environment.

7. Outcome Scenarios and Probabilities

Scenario	Prediction markets (13–14 Oct)	DDHQ model (19 Oct)	This report	Main drivers
A. Democratic sweep (House and Senate)	60–62%	54%	About 61%	Generic ballot above D+7 persists; Democrats win four of the seven Republican-held battlegrounds (Alaska, Iowa, Kansas, Maine, North Carolina, Ohio and Texas) while holding Michigan, Georgia and New Hampshire
B. Democratic House, Republican Senate	28–30%	About 23%	About 29%	A national Democratic margin of about 5–8 points with a Senate map that favours Republican incumbents in states carried by Trump by double digits
C. Republican sweep	10–11%	About 21%	About 6%	A polling miss of 4 or more points in Republican favour, or a late fall in gasoline prices and a diplomatic resolution of the Iran conflict
D. Republican House, Democratic Senate	About 1%	About 2%	About 4%	Requires an unusual split, for example Democratic wins in Senate states with a map-driven Republican House hold

Market figures are from 147 and 148. DDHQ figures are derived from its published probabilities of 77% for the House, 56% for the Senate and 54% for both chambers.149 The report’s figures sum to 100% and are consistent with its chamber estimates: about 90% for the House (A + B) and about 65% for the Senate (A + D). The report column gives the 31 October estimates: with the Senate raised from about 62% to about 65%, scenario A moves from about 58% to about 61% and scenario B from about 32% to about 29%. They are judgments, and the confidence interval around each is wide: a reasonable range for scenario A is 51–71%.

7.1 What each scenario would mean for governance

The following are analytical implications, not predictions of specific policies.

• Democratic House (scenarios A and B). The House would hold subpoena and oversight powers and would shape appropriations. Divided government has historically reduced the passage of major new legislation and increased the importance of budget deadlines. Executive actions, including trade measures under statutory authorities, would not require House approval and would remain the main channel of trade policy.
• Democratic Senate (scenarios A and D). The Senate controls confirmations and treaty ratification, and can affect war-powers votes. Control would also alter the judicial confirmation pipeline.
• Republican sweep (scenario C). Policy continuity, with the governing coalition’s agenda continuing into the last two years of the term, but with a political message that the national environment has not translated into losses.

8. After the Midterms: Dynamics to Watch

1. Interpretation of the result. Both parties will read a decisive result as a mandate or a verdict; a narrow result will be contested. The scale of the Democratic gain relative to the 2018 benchmark of 41 seats will frame the 2028 narrative.
2. Administration of the election. The 2026 count will be watched for timing: the Georgia, Ohio, Nevada and North Carolina early-vote systems report on different schedules, so a close Senate may not be called on election night.
3. The lame-duck period. The new Congress is seated in January 2027. Appropriations and debt-limit decisions between November and January will set the fiscal baseline.
4. The start of 2028. The Los Angeles Times notes that the 2028 contest begins unofficially when polls close on 3 November.150 Announcements could come quickly after the midterms, especially from candidates who are not on the ballot.

9. The 2028 Presidential Race

9.1 What is known

No major candidate has formally announced. Debate schedules have not been set: debates are organised by campaigns and media organisations, not set by law, and primary debates are expected from 2027.151 The current evidence is therefore limited to early nomination polls, general-election trial heats and a small number of public statements.

9.2 Democratic nomination

Averages put Kamala Harris first with a lead of about 12 points. The RealClearPolling average (19 July–6 October) shows Harris 26.4%, Gavin Newsom 14.4%, Pete Buttigieg 12.6%, Alexandria Ocasio-Cortez 10.6% and Jon Ossoff 8.9%.152 The Financial Times/Focaldata survey of 2–5 October found Harris 30%, Newsom 15%, Buttigieg 9%, Ocasio-Cortez 9% and Ossoff 8%, with 14% unsure.153 Harvard CAPS/Harris in late September had Harris at 38% and Newsom 17%, and The Center Square’s Noble Predictive poll (5–7 October) had Harris at 29%.154,155

Newsom said in a CNN interview on 14 September, “I wouldn’t run if she ran.”156 He described a contest between them as “mutual assured destruction,” citing overlapping supporters.157 Analysts have offered two readings, genuine deference or a strategic move that presses Harris to decide, and neither is confirmed.158,159 Harris has not announced whether she will run.160

Prediction markets and polls diverge markedly. According to a report at the end of September, Ocasio-Cortez led the Polymarket odds at over 19%, Ossoff at 16%, Newsom at 12.7% and Harris at 8%.161 Two explanations are possible: markets may anticipate a decision by Harris not to run, or market participants may weigh general-election strength and primary dynamics differently from survey respondents who name a familiar figure. Name recognition is a known driver of early primary polling, and polls two years before an election have repeatedly failed to identify the eventual nominee; Hillary Clinton led the 2006 polls for 2008 and Jeb Bush was among the leaders in many 2014–15 polls for 2016.

9.3 Republican nomination

The RealClearPolling average shows JD Vance 39.1%, Marco Rubio 23.8%, Donald Trump Jr. 11.1%, Ron DeSantis 7.3% and Robert F. Kennedy Jr. 4.6%.162 The Financial Times/Focaldata poll had Vance 38%, Rubio 24%, DeSantis 9%, Trump Jr. 7% and Kennedy 7%, with 12% unsure.163 The Hill reported, citing sources, that Rubio told donors and allies he will not run against Vance, and Vance has said he is focused on the midterms.164 These are reports, not announcements.

9.4 General-election trial heats

Early trial heats from Emerson College Polling show a narrow and uneven picture: Harris 49%, Vance 45%; Rubio 48%, Harris 43%; Newsom 49%, Vance 44%; Newsom 48%, Rubio 46%; Buttigieg leading both Vance and Rubio 49–44.165 Ossoff and Newsom lead Vance by 5 points but trail Rubio by 3 and 2 points respectively, and Ocasio-Cortez trails both Republicans.166 Morning Consult, in an earlier survey, showed Harris +1 against Vance and Harris +2 against Rubio.167

Three conclusions follow. First, no candidate has a durable lead in matchups; most are within 5 points, which is inside the range of normal error and of normal change over two years. Second, the pattern is consistent across surveys that Republican Rubio performs somewhat better than Vance against several Democrats, which suggests that candidate-specific favourability matters more than party label in this sample. Third, name recognition and the current midterm environment, which is unfavourable to the party in power, are likely to influence these numbers in ways that will fade.

9.5 The 2024 benchmark

The 2024 result is a necessary reference for 2028 modelling. Donald Trump won 49.80% of the popular vote to Kamala Harris’s 48.32%, a margin of 2,284,967 votes or 1.48 points, and 312 electoral votes to 226. Trump won all seven swing states, with margins ranging from 0.86 points in Wisconsin to 5.53 points in Arizona.168 The result was decisive in the Electoral College and close in the popular vote. This matters for the analysis of 2028: a candidate’s national favourability and the distribution of votes across seven states are separate variables, and both enter the outcome.

9.6 Debates

Debates are likely to influence the general election, but the evidence on how much is limited. In 2024, the first general-election debate on 27 June between Joe Biden and Donald Trump was widely judged to have gone poorly for the president, and Biden withdrew on 21 July after growing pressure from within his party.169,170 Harris and Trump then held one debate on 10 September, so the cycle included two general-election debates in total, not three.171 Since 1960, no year has had more than three presidential debates.172 The Commission on Presidential Debates, which had planned three debates for 2024, was bypassed after both campaigns objected to format and timing.173 The number, timing and host of 2028 debates are therefore open.

The analytical lesson is that debates are an important information event, particularly when they reveal something voters had not previously observed, as with the 2024 debate. There is less evidence that debates reliably shift large numbers of voters when both candidates perform within the normal range. The report therefore treats debates as a source of variance in the outcome rather than as a deterministic determinant.

9.7 Independents and the structure of the electorate

A common simplification divides the electorate into two partisan bases of about 40% each and a persuadable middle of about 20%. This simplification is useful but has limits. Self-identified independents are a heterogeneous group, and a large share lean consistently toward one party. Studies of voting behaviour generally find that the share of truly persuadable voters in presidential elections is smaller than the share of self-identified independents, and that turnout among marginal voters is as important as persuasion. The practical implication is that 2028 will be decided by a combination of base mobilisation, the preferences of a small group of persuadable voters, and the Electoral College geography of seven or so states.

Candidate attributes, including communication skill, favourability, executive experience, coalition breadth and the ability to hold a party together, are relevant to all of these channels. Attributes are hard to measure two years ahead and tend to be assessed differently by supporters and opponents. For that reason this report does not rank candidates by personal qualities; it identifies the observable variables that analysts will track: favourability among independents, performance among Hispanic, young and suburban voters, fundraising, and the outcome of the primary calendar.

10. Links to the Global Economic Outlook

The U.S. election matters for the global economy through four channels. The statements below are hypotheses with explicit confidence levels, and they connect to the framework in the companion InsightBridge paper on forced regionalization.

1. Trade policy durability (moderate confidence). On 20 February 2026 the Supreme Court held that the International Emergency Economic Powers Act does not authorise tariffs, and trade policy then moved to other statutory authorities, including Section 338 of the Tariff Act of 1930 against Canada.174,175 A Democratic House would not control tariff authority directly, but it could use oversight, appropriations and legislation to influence it. Trading partners are likely to read the midterm result as a signal about durability rather than a change in current tariffs.
2. Energy and the Iran conflict (moderate confidence). The IMF’s July 2026 update projects global growth of 3.0% in 2026 and 3.4% in 2027, and says the war shock weighs on energy importers.176 The midterm vote will not change Hormuz flows, but the political cost of high fuel prices influences the incentives for a negotiated outcome. Oil-price volatility will remain the main transmission channel from the conflict to household prices in the United States and abroad.177
3. Monetary conditions and the dollar (low to moderate confidence). The Federal Reserve has begun raising rates, and the dollar’s share of allocated foreign-exchange reserves was 56.70% in the second quarter of 2026, down from 57.18% in the first quarter.178 A change in fiscal or trade direction after the midterms would matter for dollar demand only through expectations about deficits and policy predictability, and such effects would be gradual.
4. Allied hedging (moderate confidence). Allied economies, including the European Union, Canada, Japan and South Korea, have been hedging by building regional and plurilateral alternatives while maintaining core security relationships.179 The 2028 race matters in this context because allies plan on a time horizon longer than one administration. A result in November that signals greater policy volatility would raise the value of hedging; a result that signals continuity would lower it. Either way, the 2028 uncertainty shown in Sections 9.2 to 9.4 is likely to sustain hedging behaviour through at least 2028.

11. Risks to the Analysis and Counter-Evidence

• Polling error. A uniform Republican miss of 3 points, as in 2024, would cut the Democratic House probability and make a Democratic Senate unlikely. A miss in the Democratic direction would produce a larger House majority and make a Senate sweep more likely.
• Late economic change. A diplomatic resolution that reopens Hormuz would lower fuel prices, although a pass-through to consumer prices would lag. The CPI for October will be released on 10 November, after the vote.180
• Democratic brand weakness. Cook warns that Democratic party favourability is weaker than the president’s standing would suggest, as the Maine race illustrates.181 Voters can reject the incumbent party without embracing the opposition, which can reduce turnout among marginal voters.
• Redistricting and legal change. Maps may still be altered by litigation, and the practical effect of the new maps is uncertain because neither party expects to win all of its target seats.182
• Market momentum. Prediction markets have shifted rapidly toward Democrats, with sweeps rising from 39% in February to about 62%.183 Momentum can reverse.
• Claims of certainty. Some commentary treats a generic-ballot lead of about 8–10 points, or a simple count of vulnerable seats, as a guarantee of a Democratic result in both chambers. The evidence supports a high probability for the House and a moderate one for the Senate, not certainty in either.

12. Signposts

Signpost	What would increase Democratic odds	What would decrease them
Generic ballot averages	Sustained above D+8	Below D+6
Presidential approval (averages)	Below 38%	Above 42%
Gasoline (AAA)	Stays above $4.30	Falls below $4.00
Senate battlegrounds	Leads in Texas, Ohio, Alaska hold	Leads in Maine or Iowa disappear
Early-vote composition	High turnout among younger and marginal voters	Republican early-vote catch-up in battlegrounds
September and October CPI	Headline above 3.7%	Falling headline and falling gasoline
Iran talks	Collapse or further shipping attacks	A signed reopening of Hormuz
2028 announcements	Early declarations after 3 November	Delay in major announcements

13. Conclusion

The evidence supports a clear central expectation for 3 November: a Democratic House majority is highly likely, a Democratic Senate is more likely than not (about 65%) but depends on a small number of correlated races in states that voted Republican by large margins in 2024, and the likeliest overall result is a Democratic sweep. The expectation rests on a national margin of about 8 to 9 points in the generic ballot, erosion of approval among the president’s own voters that the author’s cross-check places at about 25–38%, low approval ratings for the president, and an energy-driven rise in the cost of living. It is limited by the gap between votes and seats, redistricting, polling error that has often favoured Republicans, and a Democratic brand that polls less favourably than the national margin suggests. On 31 October this study’s district-level statistical model puts the Democratic House median at 239 seats with a majority probability of about 86% in its main scenario, and at 236 seats and about 91% after calibration for historical poll bias. That agrees with the direction and centre of the author’s judgment. The new state-level Senate model puts the Democratic caucus median at 52 seats with a probability of about 69% of reaching 51, and at 51 seats and about 51% after calibration for historical poll bias; on that basis the report raises its Senate estimate from about 62% to about 65%, between the two and slightly below the market price. These models rely on public survey data rather than a purpose-collected large sample; the 2028 forecast will use a sample of 100,000 to 200,000 respondents collected to reflect each state’s electorate.

For 2028, the evidence supports only two firm statements. The fields in both parties are open, with leaders at about 26–39% and large undecided shares. Early trial heats show no durable advantage for either party. For global audiences, the practical significance of the midterms lies less in the immediate seat count than in what the result signals about policy continuity, trade authority, energy prices and the horizon over which allies must plan.

About the Author

Dr Tong Yin is Founder and Chief Research Officer of InsightBridge Global LLC and its sister research entity, InsightBridge Global Lab LLC, based in Auburn, Alabama. He holds a Ph.D. in Hospitality Management from Auburn University, where he also taught, and an M.B.A. from Eastern Illinois University. Before founding InsightBridge, he spent two decades in senior executive roles with international firms in hospitality, professional services and global manufacturing in the United States, China, Australia and Europe, leading operations, commercial strategy and cross-border expansion.

His research covers geopolitical risk, macroeconomic realignment, AI strategy and organizational resilience. He developed the Core Code Theory, the Home Model, Governance Debt and Dynamic Driver Replacement Theory (DDRT) frameworks. He writes InsightBridge Global Intelligence, a weekly bilingual strategic briefing for sovereign wealth funds, central banks, ministries and institutional investors, and contributes regularly to Hospitality Net, Hotel News Resource and PhocusWire. He works in English and Mandarin Chinese.

Contact: tongyin@insightbridge.global | https://intelligence.insightbridge.global

Disclaimer

This report is an independent, third-party analysis published by InsightBridge Global and InsightBridge Global Lab. It is not affiliated with, funded by or directed by any political party, candidate, campaign, government, political action committee or advocacy organisation, in the United States or elsewhere. It carries no political, partisan, ideological or national leaning. References to parties, candidates and officials are descriptive and are made solely to report public facts and published data; they do not constitute endorsement, opposition or criticism.

Probabilities and scenarios are the author’s analytical judgments based on information available on 31 October 2026. They are not predictions of certain outcomes, and electoral outcomes remain uncertain. Poll results, prediction-market prices and forecast-model outputs are quoted as reported by their sources and may change or be revised; the report has not independently audited them. This report does not constitute investment, legal, tax or voting advice. The views expressed are those of the author, Dr Tong Yin, and do not necessarily reflect the positions of InsightBridge Global or InsightBridge Global Lab.

---

1. NBC News, “Midterms 2026: Congress Polls and More.” https://www.nbcnews.com/politics/2026-elections/midterms↩︎

2. Action Network, “2026 Midterm Predictions: Polls vs. Prediction Markets,” updated 15 October 2026, market prices checked 20 October 2026 (Cook Political Report Senate ratings of 6 October 2026; Kalshi and Polymarket prices). https://www.actionnetwork.com/politics/2026-midterm-predictions-what-prediction-markets-say-vs-the-polls↩︎

3. Wikipedia, “2024 United States presidential election” (national results, swing-state margins and polling error). https://en.wikipedia.org/wiki/2024_United_States_presidential_election↩︎

4. El Cronista, “Generic ballot 2026: Why Democrats’ lead may not win the Senate,” citing RealClearPolitics, 14 October 2026. https://www.cronista.com/en/today/democrats-lead-the-generic-ballot-a-month-before-the-midterms-heres-why-it-may-not-be-enough-to-win-the-senate/↩︎

5. Newsweek, “How midterm forecast maps compare to previous elections,” 14 October 2026 (Cook Political Report ratings of 9 October 2026; comparisons with 2018 and 2022). https://www.newsweek.com/how-midterm-forecast-maps-compare-previous-elections-12565502↩︎

6. Silver Bulletin, “Who’s ahead on the generic congressional ballot?,” updated 13 October 2026. https://www.natesilver.net/p/generic-ballot-average-2026-nate-silver-bulletin-congress-polls↩︎

7. RealClearPolling, “2026 Generic Congressional Vote,” average for 21 September–11 October 2026. https://www.realclearpolling.com/polls/state-of-the-union/generic-congressional-vote↩︎

8. Wikipedia, “2024 United States presidential election” (national results, swing-state margins and polling error). https://en.wikipedia.org/wiki/2024_United_States_presidential_election↩︎

9. Pew Research Center, “Trump’s approval rating drops as fewer Americans are confident in his handling of the economy, other issues,” 8 October 2026. https://www.pewresearch.org/politics/2026/10/08/trumps-approval-rating-drops-as-fewer-americans-are-confident-in-his-handling-of-the-economy-other-issues/↩︎

10. Reuters, “Hispanic voters sour on Republicans, Reuters/Ipsos poll finds,” 5 October 2026. https://www.reuters.com/world/us/trump-approval-record-low-hispanic-voters-sour-on-republicans-reutersipsos-poll-2026-10-05/↩︎

11. USA Today, “Trump approval rating stays at record low in new poll as midterms near” (AP-NORC and Reuters/Ipsos), October 2026. https://www.usatoday.com/story/news/politics/2026/10/05/trump-approval-rating-remains-record-low-latest-poll/92109112007/↩︎

12. Fox News, “Fox News Poll: Cost of living concerns weigh on GOP ahead of midterms,” 14 October 2026 (fieldwork 9–12 October, 1,202 registered voters). https://www.foxnews.com/politics/fox-news-poll-cost-living-concerns-weigh-gop-midterms↩︎

13. NBC News, “President Donald Trump’s approval rating: Latest Polls,” updated 11 October 2026. https://www.nbcnews.com/politics/2026-elections/presidential-approval-polls↩︎

14. Rasmussen Reports, “Daily Presidential Tracking Poll,” 13 October 2026. https://www.rasmussenreports.com/public_content/politics/trump_administration_second_term/prez_track_oct12↩︎

15. Yahoo News, “Trump’s approval rating sinks to new second-term low, Pew poll shows” (aggregator readings of 8 October). https://www.yahoo.com/news/politics/articles/trumps-approval-rating-sinks-second-211242590.html↩︎

16. The Conversation, “Democrats remain clear favourites to gain control of Congress at midterms,” 19 October 2026 (Silver Bulletin forecast, approval and generic-ballot averages; gasoline and oil prices). https://theconversation.com/democrats-remain-clear-favourites-to-gain-control-of-congress-at-midterms-294465↩︎

17. Yahoo News, “Trump approval rating hits new career low in latest AP-NORC poll” (Wall Street Journal 37% reading). https://www.yahoo.com/news/politics/articles/trump-approval-rating-hits-career-165316228.html↩︎

18. Forbes, “Trump’s 39% approval rating below his first term and Biden’s ahead of midterms,” 5 October 2026. https://www.forbes.com/sites/saradorn/2026/10/05/trumps-39-approval-rating-below-his-first-term-and-bidens-midterm-ratings/↩︎

19. Wikipedia, “2018 United States House of Representatives elections.” https://en.wikipedia.org/wiki/2018_United_States_House_of_Representatives_elections↩︎

20. The Hill, “Just 1 in 4 approve of Trump handling of economy: Survey” (Financial Times/Focaldata, 2–4 October 2026). https://thehill.com/homenews/administration/6145073-inflation-rising-midterm-odds/↩︎

21. Pew Research Center, “Trump’s approval rating drops as fewer Americans are confident in his handling of the economy, other issues,” 8 October 2026. https://www.pewresearch.org/politics/2026/10/08/trumps-approval-rating-drops-as-fewer-americans-are-confident-in-his-handling-of-the-economy-other-issues/↩︎

22. The Hill, “Trump’s approval rating hits record low in Fox News poll,” 14 October 2026 (approval and disapproval by party). https://thehill.com/homenews/administration/6149735-trumps-approval-rating-hits-record-low-in-fox-news-poll/↩︎

23. Newsweek, “Trump’s approval rating changes direction with non-white voters,” 11 October 2026. https://www.newsweek.com/donald-trump-approval-rating-changes-non-white-voters-12549047↩︎

24. Silver Bulletin, “Trump Approval Rating: Latest Polls.” https://www.natesilver.net/p/trump-approval-ratings-nate-silver-bulletin↩︎

25. The Conversation, “Democrats remain clear favourites to gain control of Congress at midterms,” 19 October 2026 (Silver Bulletin forecast, approval and generic-ballot averages; gasoline and oil prices). https://theconversation.com/democrats-remain-clear-favourites-to-gain-control-of-congress-at-midterms-294465↩︎

26. RealClearPolling, “2026 Generic Congressional Vote,” average for 21 September–11 October 2026. https://www.realclearpolling.com/polls/state-of-the-union/generic-congressional-vote↩︎

27. Silver Bulletin, “Who’s ahead on the generic congressional ballot?,” updated 13 October 2026. https://www.natesilver.net/p/generic-ballot-average-2026-nate-silver-bulletin-congress-polls↩︎

28. Silver Bulletin, “Who’s ahead on the generic congressional ballot?,” updated 13 October 2026. https://www.natesilver.net/p/generic-ballot-average-2026-nate-silver-bulletin-congress-polls↩︎

29. The Conversation, “Democrats remain clear favourites to gain control of Congress at midterms,” 19 October 2026 (Silver Bulletin forecast, approval and generic-ballot averages; gasoline and oil prices). https://theconversation.com/democrats-remain-clear-favourites-to-gain-control-of-congress-at-midterms-294465↩︎

30. Decision Desk HQ, “Generic Congressional Ballot Polling Average.” https://votes.decisiondeskhq.com/polls/generic-ballot/national↩︎

31. Fox News, “Fox News Poll: Cost of living concerns weigh on GOP ahead of midterms,” 14 October 2026 (fieldwork 9–12 October, 1,202 registered voters). https://www.foxnews.com/politics/fox-news-poll-cost-living-concerns-weigh-gop-midterms↩︎

32. New York Magazine, “6 Midterms Takeaways With a Month to Go,” 14 October 2026. https://nymag.com/intelligencer/article/2026-midterms-takeaways-with-a-month-to-go-blue-wave.html↩︎

33. Wikipedia, “2018 United States House of Representatives elections.” https://en.wikipedia.org/wiki/2018_United_States_House_of_Representatives_elections↩︎

34. Wikipedia, “2010 United States House of Representatives elections.” https://en.wikipedia.org/wiki/2010_United_States_House_of_Representatives_elections↩︎

35. Wikipedia, “2022 United States House of Representatives elections.” https://en.wikipedia.org/wiki/2022_United_States_House_of_Representatives_elections↩︎

36. NBC News, “Tracking the 2026 Early Vote: Midterm Elections.” https://www.nbcnews.com/politics/2026-elections/early-vote↩︎

37. WSB-TV, “First day of early voting beats 2022 numbers, according to unofficial tally,” October 2026. https://www.wsbtv.com/news/local/first-day-early-voting-beats-2022-numbers-according-unofficial-tally/LWXWSOE4NNBD5I7VZ5Y3NGNMKA/↩︎

38. 10TV, “Early in-person voting in Ohio nearly 4 times higher than 2022 midterm,” October 2026. https://www.10tv.com/article/news/politics/elections/early-in-person-voting-ohio-nearly-4-times-higher-than-2022/530-7570a606-58d8-4552-b926-53c0bd54bf5a↩︎

39. Ground News, “Early voting data shows emerging signs of a Democratic advantage in the midterms” (summarising CNN analysis), 13 October 2026. https://ground.news/article/early-voting-data-shows-emerging-signs-of-a-democratic-advantage-in-the-midterms↩︎

40. CNN, “Inflation was hotter than expected in September,” 14 October 2026. https://edition.cnn.com/2026/10/14/economy/us-cpi-inflation-september↩︎

41. CNN, “Inflation was hotter than expected in September,” 14 October 2026. https://edition.cnn.com/2026/10/14/economy/us-cpi-inflation-september↩︎

42. U.S. Bureau of Labor Statistics, “Consumer Price Index Summary,” release of 14 October 2026 (next release 10 November 2026). https://www.bls.gov/news.release/cpi.nr0.htm↩︎

43. Finance Channel, “U.S. Inflation Rises to 3.7% as Gasoline Prices Surge in September,” 14 October 2026. https://finchannel.com/u-s-inflation-rises-to-3-7-as-gasoline-prices-surge/138086/america/2026/10/↩︎

44. Cleveland.com, “Gas prices for 10/14/26,” citing AAA. https://www.cleveland.com/news/2026/10/gas-prices-for-101426-price-remains-above-4-per-gallon-for-record-breaking-158th-day-this-year.html↩︎

45. U.S. Bureau of Economic Analysis, “GDP (Third Estimate), Industries, Corporate Profits, State GDP, and State Personal Income, 2nd Quarter 2026.” https://www.bea.gov/news/2026/gdp-third-estimate-industries-corporate-profits-state-gdp-and-state-personal-income-2nd↩︎

46. Associated Press, “US economy grew a solid 2.2% in the second quarter.” https://apnews.com/article/economic-growth-inflation-trump-trade-b759367e9e6149fe145652cc95887aa0↩︎

47. CNBC, “Jobs report September 2026.” https://www.cnbc.com/2026/10/02/jobs-report-september-2026.html↩︎

48. CNBC, “Fed rate decision September 2026: Rates rise to 3.75%-4%.” https://www.cnbc.com/2026/09/16/fed-rate-decision-september-2026.html↩︎

49. CNBC-TV18, “Crude oil prices today,” October 2026. https://www.cnbctv18.com/market/commodities/crude-oil-prices-today-brent-wti-saudi-houthi-attacks-us-iran-war-strait-of-hormuz-oil-tankers-drones-missiles-supply-shocks-20009526.htm↩︎

50. Transport Topics, “Strait of Hormuz News Updates.” https://www.ttnews.com/tags/strait-hormuz↩︎

51. AAA, “Pump Prices Remain at Record Highs for Autumn.” https://gasprices.aaa.com/pump-prices-remain-at-record-highs-for-autumn/↩︎

52. The Conversation, “Democrats remain clear favourites to gain control of Congress at midterms,” 19 October 2026 (Silver Bulletin forecast, approval and generic-ballot averages; gasoline and oil prices). https://theconversation.com/democrats-remain-clear-favourites-to-gain-control-of-congress-at-midterms-294465↩︎

53. USA Today, “65% of Americans blame Trump for high cost of living, new poll shows” (AP-NORC), 1 October 2026. https://www.usatoday.com/story/news/politics/2026/10/01/donald-trump-approval-rating-economy-cost-of-living/92035389007/↩︎

54. Yahoo Finance, “US economy grew at a faster pace than expected in Q2.” https://finance.yahoo.com/economy/articles/us-economy-grew-faster-pace-123905750.html↩︎

55. Lee & Associates, “2026 Q2 Economic Report.” https://www.lee-associates.com/research-article/2026-q2-economic-report/↩︎

56. CFO Dive, “All Fed officials backed vote to hike benchmark rate: FOMC minutes.” https://www.cfodive.com/news/fed-officials-backed-vote-hike-benchmark-rate-fomc-minutes-inflation-unemployment-monetary-policy/832421/↩︎

57. Reuters, “US EIA hikes oil price forecasts again as Iran war drains global stockpile,” 6 October 2026. https://www.reuters.com/business/energy/us-eia-hikes-oil-price-forecasts-again-iran-war-drains-global-stockpile-2026-10-06/↩︎

58. Strategic Energy Map, “Energy infrastructure news: strikes, closures and disruptions” (Kpler data; IEA characterisation). https://strategicenergymap.org/events/↩︎

59. CNBC-TV18, “Crude oil prices today,” October 2026. https://www.cnbctv18.com/market/commodities/crude-oil-prices-today-brent-wti-saudi-houthi-attacks-us-iran-war-strait-of-hormuz-oil-tankers-drones-missiles-supply-shocks-20009526.htm↩︎

60. Rio Times, “Midterm Odds and the Limits of Trump’s Power,” 14 October 2026. https://www.riotimesonline.com/midterm-odds-trump-power-democratic-congress-2026/↩︎

61. Newsweek, “How midterm forecast maps compare to previous elections,” 14 October 2026 (Cook Political Report ratings of 9 October 2026; comparisons with 2018 and 2022). https://www.newsweek.com/how-midterm-forecast-maps-compare-previous-elections-12565502↩︎

62. 270toWin, “Cook Political Report 2026 House Ratings,” as of 9 October 2026. https://www.270towin.com/2026-house-election/table/cook-political-report-2026-house-ratings/↩︎

63. Newsweek, “How midterm forecast maps compare to previous elections,” 14 October 2026 (Cook Political Report ratings of 9 October 2026; comparisons with 2018 and 2022). https://www.newsweek.com/how-midterm-forecast-maps-compare-previous-elections-12565502↩︎

64. Axios, “House GOP leaders admit redistricting fell short,” 6 October 2026. https://www.axios.com/2026/10/06/house-republicans-redistricting-midterm-elections↩︎

65. Axios, “House GOP leaders admit redistricting fell short,” 6 October 2026. https://www.axios.com/2026/10/06/house-republicans-redistricting-midterm-elections↩︎

66. Election Tracker Live, “2026 Midterm Elections Dashboard” (citing Cook’s expected range). https://electiontracker.live/↩︎

67. Bitcoin.com News, “Polymarket Gives Dems 62% Midterm Sweep Odds,” 14 October 2026. https://news.bitcoin.com/featured/polymarket-gives-dems-62-midterm-sweep-odds-clarity-act-in-jeopardy/↩︎

68. Action Network, “2026 Midterm Predictions: Polls vs. Prediction Markets,” updated 15 October 2026, market prices checked 20 October 2026 (Cook Political Report Senate ratings of 6 October 2026; Kalshi and Polymarket prices). https://www.actionnetwork.com/politics/2026-midterm-predictions-what-prediction-markets-say-vs-the-polls↩︎

69. Decision Desk HQ, “2026 Midterm Election Forecast,” updated 19 October 2026. https://votes.decisiondeskhq.com/forecast/2026↩︎

70. The Conversation, “Democrats remain clear favourites to gain control of Congress at midterms,” 19 October 2026 (Silver Bulletin forecast, approval and generic-ballot averages; gasoline and oil prices). https://theconversation.com/democrats-remain-clear-favourites-to-gain-control-of-congress-at-midterms-294465↩︎

71. Cooperative Election Study (CES), Common Content 2018, 2020, 2022 and 2024, Harvard Dataverse; 2024 release doi:10.7910/DVN/X11EP6. https://cces.gov.harvard.edu/ ; https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/X11EP6↩︎

72. The Downballot, “2024 presidential results by congressional district” (2024 maps and 2026 maps, updated 9 July 2026). https://www.the-downballot.com/p/2024-presidential-results-by-congressional↩︎

73. U.S. Census Bureau, “Congressional Districts” (120th Congress boundaries; nine states with new maps for 2026), with 2020 Census PL 94-171 block data, ACS 2023 five-year Table B15002 and IPUMS USA microdata. https://www.census.gov/programs-surveys/decennial-census/about/rdo/congressional-districts.html ; https://usa.ipums.org/↩︎

74. MIT Election Data and Science Lab, “U.S. House 1976–2024,” Harvard Dataverse, doi:10.7910/DVN/IG0UN2; incumbency effect estimated against district-level 2024 presidential vote. https://doi.org/10.7910/DVN/IG0UN2↩︎

75. Silver Bulletin, “Who’s ahead on the generic congressional ballot?,” updated 13 October 2026. https://www.natesilver.net/p/generic-ballot-average-2026-nate-silver-bulletin-congress-polls↩︎

76. RealClearPolling, “2026 Generic Congressional Vote,” average for 21 September–11 October 2026. https://www.realclearpolling.com/polls/state-of-the-union/generic-congressional-vote↩︎

77. Decision Desk HQ, “Generic Congressional Ballot Polling Average.” https://votes.decisiondeskhq.com/polls/generic-ballot/national↩︎

78. Bitcoin.com News, “Polymarket Gives Dems 62% Midterm Sweep Odds,” 14 October 2026. https://news.bitcoin.com/featured/polymarket-gives-dems-62-midterm-sweep-odds-clarity-act-in-jeopardy/↩︎

79. U.S. Census Bureau, “2022 Voting and Registration Data Now Available,” 2 May 2023. https://www.census.gov/newsroom/press-releases/2023/2022-voting-registration.html↩︎

80. U.S. Census Bureau, Current Population Survey report P20-590 (citizen voting-age population of 236.1 million in November 2024). https://www2.census.gov/library/publications/2026/demo/p20-590.pdf↩︎

81. Silver Bulletin, generic congressional ballot average and poll database, 31 October 2026; two-party components computed from the 27 most recent polls using Silver Bulletin’s influence weights. https://www.natesilver.net/p/generic-ballot-average-2026-nate-silver-bulletin-congress-polls↩︎

82. RealClearPolling, “2026 Generic Congressional Vote,” average for 8–26 October 2026, accessed 31 October 2026. https://www.realclearpolling.com/polls/state-of-the-union/generic-congressional-vote↩︎

83. Decision Desk HQ, “Generic Congressional Ballot Polling Average,” values of 29 October and 30 October 2026. https://votes.decisiondeskhq.com/polls/generic-ballot/national↩︎

84. Decision Desk HQ, “Generic Congressional Ballot Polling Average,” values of 29 October and 30 October 2026. https://votes.decisiondeskhq.com/polls/generic-ballot/national↩︎

85. Eve-of-election generic-ballot averages, FiveThirtyEight, “generic_topline_historical” and “generic_ballot_averages” data files (2006–2024; 2014 not available), compared with official results from the Office of the Clerk, U.S. House of Representatives. https://github.com/fivethirtyeight/data/tree/master/congress-generic-ballot ; https://clerk.house.gov/Members/ElectionInformation↩︎

86. Kalshi, “Who will control the House? 2026,” accessed 31 October 2026. https://kalshi.com/markets/controlh/house-winner↩︎

87. Danny Ebanks, Jonathan N. Katz and Gary King, “ElectIt: 2026 House and Senate forecasts,” accessed 31 October 2026. https://gking.harvard.edu/electit/↩︎

88. The Cook Political Report, “The Cook Political Report Moves 19 More House Seats Towards Democrats,” 23 October 2026. https://thecookpoliticalreport.substack.com/p/the-cook-political-report-moves-19↩︎

89. 270toWin, “Cook Political Report 2026 Senate Race Ratings,” 12 October 2026. https://www.270towin.com/2026-senate-election/cook-political-report-2026-senate↩︎

90. Rio Times, “Midterm Odds and the Limits of Trump’s Power,” 14 October 2026. https://www.riotimesonline.com/midterm-odds-trump-power-democratic-congress-2026/↩︎

91. NBC News, “Midterms 2026: Congress Polls and More.” https://www.nbcnews.com/politics/2026-elections/midterms↩︎

92. Action Network, “2026 Midterm Predictions: Polls vs. Prediction Markets,” updated 15 October 2026, market prices checked 20 October 2026 (Cook Political Report Senate ratings of 6 October 2026; Kalshi and Polymarket prices). https://www.actionnetwork.com/politics/2026-midterm-predictions-what-prediction-markets-say-vs-the-polls↩︎

93. Washington Examiner, “Senate Republicans playing defense heading into Election Day,” 14 October 2026. https://www.washingtonexaminer.com/premium/4754150/senate-republicans-defense-democratic-majority-2026-elections/↩︎

94. Yahoo News, “Charlie Cook: Democrats ‘badly damaged,’ Kansas now a toss-up,” 11 October 2026. https://www.yahoo.com/news/politics/articles/charlie-cook-democrats-badly-damaged-171825696.html↩︎

95. U.S. Political Rank, “Times Siena puts Democrats ahead or even in five Trump states,” 12 October 2026 (New York Times/Siena, 21 September–1 October). https://uspoliticalrank.com/articles/times-siena-five-red-state-senate-polls-october-2026↩︎

96. Election Central, “2026 Senate Polls: Latest Averages and Battleground Races,” 14 October 2026 (RealClearPolitics averages). https://www.uspresidentialelectionnews.com/2026-senate-polls/↩︎

97. The New York Times, “Tracking Competitive Races in the House and Senate,” updated 9 October 2026. https://www.nytimes.com/interactive/2026/us/elections/midterms-house-senate-key-races.html↩︎

98. New York Magazine, “Polls Show a Blue Wave, But Beware All These Close Races,” 19 October 2026. https://nymag.com/intelligencer/article/polls-show-a-midterms-blue-wave-but-beware-all-the-close-races.html↩︎

99. Yahoo News, “Charlie Cook: Democrats ‘badly damaged,’ Kansas now a toss-up,” 11 October 2026. https://www.yahoo.com/news/politics/articles/charlie-cook-democrats-badly-damaged-171825696.html↩︎

100. Action Network, “2026 Midterm Predictions: Polls vs. Prediction Markets,” updated 15 October 2026, market prices checked 20 October 2026 (Cook Political Report Senate ratings of 6 October 2026; Kalshi and Polymarket prices). https://www.actionnetwork.com/politics/2026-midterm-predictions-what-prediction-markets-say-vs-the-polls↩︎

101. New York Magazine, “Polls Show a Blue Wave, But Beware All These Close Races,” 19 October 2026. https://nymag.com/intelligencer/article/polls-show-a-midterms-blue-wave-but-beware-all-the-close-races.html↩︎

102. Decision Desk HQ, “2026 Midterm Election Forecast,” updated 19 October 2026. https://votes.decisiondeskhq.com/forecast/2026↩︎

103. The Conversation, “Democrats remain clear favourites to gain control of Congress at midterms,” 19 October 2026 (Silver Bulletin forecast, approval and generic-ballot averages; gasoline and oil prices). https://theconversation.com/democrats-remain-clear-favourites-to-gain-control-of-congress-at-midterms-294465↩︎

104. Fox News, “Fox News Poll: Cost of living concerns weigh on GOP ahead of midterms,” 14 October 2026 (fieldwork 9–12 October, 1,202 registered voters). https://www.foxnews.com/politics/fox-news-poll-cost-living-concerns-weigh-gop-midterms↩︎

105. The Hill, “Trump’s approval rating hits record low in Fox News poll,” 14 October 2026 (approval and disapproval by party). https://thehill.com/homenews/administration/6149735-trumps-approval-rating-hits-record-low-in-fox-news-poll/↩︎

106. Fox News, “Fox News Poll: Cost of living concerns weigh on GOP ahead of midterms,” 14 October 2026 (fieldwork 9–12 October, 1,202 registered voters). https://www.foxnews.com/politics/fox-news-poll-cost-living-concerns-weigh-gop-midterms↩︎

107. Action Network, “2026 Midterm Predictions: Polls vs. Prediction Markets,” updated 15 October 2026, market prices checked 20 October 2026 (Cook Political Report Senate ratings of 6 October 2026; Kalshi and Polymarket prices). https://www.actionnetwork.com/politics/2026-midterm-predictions-what-prediction-markets-say-vs-the-polls↩︎

108. Decision Desk HQ on X, “DDHQ Senate forecast update (10/12),” 12 October 2026. https://x.com/DecisionDeskHQ/status/2109695637717504316↩︎

109. NetNewsLedger, “NewsHawks: U.S. midterm polls, the battle for Senate control …,” 12 October 2026 (Decision Desk HQ race probabilities, updated 11 October). https://www.netnewsledger.com/2026/10/12/newshawks-u-s-midterm-polls-the-battle-for-senate-control-canadian-job-losses-thunder-bay-rail-plans-and-regional-news-for-october-11-2026/↩︎

110. Roll Call, “Trump-backed Whatley faces balancing act in North Carolina as polls tilt blue,” 21 October 2026. https://rollcall.com/2026/10/21/north-carolina-senate-whatley-cooper-election/↩︎

111. The News & Observer, “Trump stumps in Greenville for Republican candidates,” 20 October 2026. https://www.newsobserver.com/news/politics-government/election/article317703733.html↩︎

112. The Washington Post, “Trump rallies in North Carolina, even after GOP pulls its money,” 20 October 2026. https://www.washingtonpost.com/politics/2026/10/20/gop-has-pulled-north-carolina-money-trump-will-rally-whatley/↩︎

113. Roll Call, “Trump-backed Whatley faces balancing act in North Carolina as polls tilt blue,” 21 October 2026. https://rollcall.com/2026/10/21/north-carolina-senate-whatley-cooper-election/↩︎

114. NBC News, “Trump holds rally in North Carolina for midterm GOP candidates” (live blog), 20 October 2026. https://www.nbcnews.com/politics/trump-administration/live-blog/trump-midterm-rally-north-carolina-2026-election-congress-live-updates-rcna604465↩︎

115. Newsweek via Yahoo News, “Trump’s Kansas approval rating ahead of rally as GOP Senate seat at risk,” 19 October 2026. https://www.yahoo.com/news/politics/articles/trumps-kansas-approval-rating-rally-175548493.html↩︎

116. Wikipedia, “2026 United States Senate election in Kansas” (polling averages and polls, accessed 23 October 2026). https://en.wikipedia.org/wiki/2026_United_States_Senate_election_in_Kansas↩︎

117. Wikipedia, “2026 United States Senate election in Kansas” (polling averages and polls, accessed 23 October 2026). https://en.wikipedia.org/wiki/2026_United_States_Senate_election_in_Kansas↩︎

118. High Plains Public Radio, “Election analyst: Candidate quality key to Roger Marshall’s slide, Adam Hamilton’s rise in Kansas,” 16 October 2026. https://www.hppr.org/hppr-news/2026-10-16/election-analyst-candidate-quality-key-to-roger-marshalls-slide-adam-hamiltons-rise-in-kansas?_amp=true↩︎

119. Wikipedia, “2026 United States Senate election in Kansas” (polling averages and polls, accessed 23 October 2026). https://en.wikipedia.org/wiki/2026_United_States_Senate_election_in_Kansas↩︎

120. Wikipedia, “2026 United States Senate election in Kansas” (polling averages and polls, accessed 23 October 2026). https://en.wikipedia.org/wiki/2026_United_States_Senate_election_in_Kansas↩︎

121. KMBC, “Kansas Senate poll: Less than 1 point separates Marshall, Hamilton,” 11 October 2026. https://www.kmbc.com/article/kansas-senate-poll-roger-marshall-adam-hamilton-october-2026/74104185↩︎

122. NetNewsLedger, “NewsHawks: U.S. midterm polls, the battle for Senate control …,” 12 October 2026 (Decision Desk HQ race probabilities, updated 11 October). https://www.netnewsledger.com/2026/10/12/newshawks-u-s-midterm-polls-the-battle-for-senate-control-canadian-job-losses-thunder-bay-rail-plans-and-regional-news-for-october-11-2026/↩︎

123. Cleveland.com, “New poll shows Susan Collins in a neck-and-neck battle to win a sixth term,” 23 October 2026. https://www.cleveland.com/politics/2026/10/new-poll-shows-susan-collins-in-a-neck-and-neck-battle-to-win-a-sixth-term.html↩︎

124. NEWS CENTER Maine, “Sen. Susan Collins cites work ethic and results in run for sixth term,” October 2026; Press Herald, “Susan Collins votes no as Kash Patel narrowly wins confirmation as FBI chief,” 20 February 2025. https://www.newscentermaine.com/article/news/politics/elections/sen-susan-collins-cites-work-ethic-and-results-in-run-for-sixth-term/97-ca765d0c-c0e5-4f9d-9ea0-d6267d0f2a9d ; https://www.pressherald.com/2025/02/20/susan-collins-to-oppose-kash-patel-as-fbi-chief/↩︎

125. NEWS CENTER Maine, “Sen. Susan Collins cites work ethic and results in run for sixth term,” October 2026; Press Herald, “Susan Collins votes no as Kash Patel narrowly wins confirmation as FBI chief,” 20 February 2025. https://www.newscentermaine.com/article/news/politics/elections/sen-susan-collins-cites-work-ethic-and-results-in-run-for-sixth-term/97-ca765d0c-c0e5-4f9d-9ea0-d6267d0f2a9d ; https://www.pressherald.com/2025/02/20/susan-collins-to-oppose-kash-patel-as-fbi-chief/↩︎

126. Reuters, “Maine Democrats nominate Troy Jackson to replace Platner in US Senate race,” 25 July 2026; ABC News, “Troy Jackson secures Democratic nomination for Maine Senate race,” 25 July 2026. https://www.reuters.com/legal/government/maine-democrats-nominate-troy-jackson-replace-platner-us-senate-race-2026-07-25/ ; https://abcnews.com/Politics/troy-jackson-secures-democratic-nomination-maine-senate-race/story?id=135085406↩︎

127. Fox News, “Fox News Poll: Nail-biter in Maine Senate race,” 22 October 2026. https://www.foxnews.com/politics/fox-news-poll-nail-biter-maine-senate-race↩︎

128. Bangor Daily News, “Fox News poll shows Susan Collins and Troy Jackson tied less than 2 weeks out,” 23 October 2026. https://www.bangordailynews.com/2026/10/22/politics/fox-news-poll-susan-collins-troy-jackson-tied/↩︎

129. Fox News, “Collins insists socialist rival won’t beat her as Maine’s money machine,” 15 October 2026. https://www.foxnews.com/politics/collins-insists-socialist-rival-wont-beat-her-maines-money-machine-youll-lose-big-time.amp↩︎

130. Wikipedia, “2020 United States Senate election in Maine”; Newsweek, “Susan Collins poll numbers worse than in 2020,” 3 October 2026. https://en.wikipedia.org/wiki/2020_United_States_Senate_election_in_Maine ; https://www.newsweek.com/susan-collins-polls-numbers-worse-than-in-2020-can-she-defy-odds-again-12003873↩︎

131. Wikipedia, “2020 United States Senate election in Maine”; Newsweek, “Susan Collins poll numbers worse than in 2020,” 3 October 2026. https://en.wikipedia.org/wiki/2020_United_States_Senate_election_in_Maine ; https://www.newsweek.com/susan-collins-polls-numbers-worse-than-in-2020-can-she-defy-odds-again-12003873↩︎

132. Seacoastonline, “The outcome of Maine’s US Senate race may not be known for weeks. Here’s why,” 23 October 2026. https://www.seacoastonline.com/story/news/politics/elections/2026/10/23/maines-us-senate-race-outcome-may-not-be-known-for-weeks-collins-jackson/92377041007/↩︎

133. Bangor Daily News, “Fox News poll shows Susan Collins and Troy Jackson tied less than 2 weeks out,” 23 October 2026. https://www.bangordailynews.com/2026/10/22/politics/fox-news-poll-susan-collins-troy-jackson-tied/↩︎

134. Sun Journal, “Does Maine have the lowest median income in New England?” 14 January 2026 (Census ACS 2019–2023). https://www.sunjournal.com/2026/01/14/does-maine-have-the-lowest-median-income-in-new-england-fact-brief/↩︎

135. PenBay Pilot, “Heating oil in Maine reaches a statewide average of $6.04 per gallon,” 22 October 2026 (Maine Department of Energy Resources). https://www.penbaypilot.com/article/heating-oil-maine-reaches-statewide-average-604-gallon/275713↩︎

136. National Energy Assistance Directors Association, “Heating Oil Press Release,” 25 September 2026. https://neada.org/wp-content/uploads/2026/09/NEADA-Heating-Oil-Press-Release-Sept-25-2026A-1.pdf↩︎

137. Fox News, “Fox News Poll: Nail-biter in Maine Senate race,” 22 October 2026. https://www.foxnews.com/politics/fox-news-poll-nail-biter-maine-senate-race↩︎

138. Fox News, “Fox News Poll: Nail-biter in Maine Senate race,” 22 October 2026. https://www.foxnews.com/politics/fox-news-poll-nail-biter-maine-senate-race↩︎

139. Wikipedia, “2026 United States Senate elections” (ratings of 28–30 October and seat composition) and the individual state election pages (polling tables), accessed 31 October 2026. https://en.wikipedia.org/wiki/2026_United_States_Senate_elections↩︎

140. Kalshi, “Who will control the Senate? 2026,” accessed 31 October 2026. https://kalshi.com/markets/controls/senate-winner↩︎

141. Polymarket, “Which party will win the Senate in 2026?,” accessed 31 October 2026. https://polymarket.com/midterms/midterms-control↩︎

142. New York Magazine, “Final midterms polls show Democrats poised to flip Congress,” 30 October 2026. http://nymag.com/intelligencer/article/final-midterms-polls-show-democrats-poised-to-flip-congress.html↩︎

143. CNBC, “Elon Musk gives $276 million to Republican groups, becomes biggest known 2026 midterm donor,” 12 October 2026. https://www.cnbc.com/2026/10/12/elon-musk-republican-donations-2026-midterm-election.html↩︎

144. NBC News, “Elon Musk’s super PAC dives into the midterms to bolster Republicans in battleground races,” 7 October 2026. https://www.nbcnews.com/politics/2026-election/elon-musk-super-pac-midterms-republicans-rcna595909↩︎

145. Al Jazeera, “Elon Musk spends $366m to bolster Republican Party in US midterm elections,” 12 October 2026. https://www.aljazeera.com/news/2026/10/12/elon-musk-spends-366m-to-bolster-republican-party-in-us-midterm-elections↩︎

146. CNN, “Elon Musk emerges as the top booster for Republicans with $276 million in donations,” 12 October 2026. https://www.cnn.com/2026/10/12/politics/elon-musk-midterm-spending-276-million↩︎

147. Rio Times, “Midterm Odds and the Limits of Trump’s Power,” 14 October 2026. https://www.riotimesonline.com/midterm-odds-trump-power-democratic-congress-2026/↩︎

148. Bitcoin.com News, “Polymarket Gives Dems 62% Midterm Sweep Odds,” 14 October 2026. https://news.bitcoin.com/featured/polymarket-gives-dems-62-midterm-sweep-odds-clarity-act-in-jeopardy/↩︎

149. Decision Desk HQ, “2026 Midterm Election Forecast,” updated 19 October 2026. https://votes.decisiondeskhq.com/forecast/2026↩︎

150. Los Angeles Times, “Newsom and Harris continue circling, may finally collide,” 30 September 2026. https://www.latimes.com/politics/newsletter/2026-09-30/gavin-newsom-kamala-harris-2028-democratic-presidential-campaign↩︎

151. 2028 Tracker, “2028 Presidential Debates: Schedule and What to Expect.” https://2028tracker.com/debates-2028↩︎

152. RealClearPolling, “2028 U.S. Presidential Election,” averages for 19 July–6 October 2026. https://www.realclearpolling.com/elections/president/2028/national↩︎

153. The Hill, “Harris, Vance leading in 2028 presidential polling” (Financial Times/Focaldata, 2–5 October 2026). https://thehill.com/homenews/campaign/6144802-harris-vance-lead-2028-poll/↩︎

154. Newsweek, “Kamala Harris Leads Newsom in 2028 Democrat Front-Runners Poll” (including Harvard CAPS/Harris and Echelon Insights). https://www.newsweek.com/kamala-harris-leads-newsom-2028-democrat-front-runners-poll-12559762↩︎

155. The Center Square via KDH News, “Poll: Two years out, Harris remains frontrunner in 2028 presidential primary” (Noble Predictive Insights, 5–7 October 2026). https://kdhnews.com/news/nation/poll-two-years-out-harris-remains-frontrunner-in-2028-presidential-primary/article_488f37d3-66a7-5901-b558-b6cf9a57b489.html↩︎

156. Politico, “Newsom says he won’t run for president if Harris runs again,” 14 September 2026. https://www.politico.com/news/2026/09/14/newsom-says-he-wont-run-for-president-if-harris-runs-again-01075320↩︎

157. The Washington Post, “Gavin Newsom says he won’t run for president if Kamala Harris does,” 14 September 2026. https://www.washingtonpost.com/politics/2026/09/14/gavin-newsom-says-he-wont-run-president-if-kamala-harris-does/↩︎

158. Los Angeles Times, “Newsom and Harris continue circling, may finally collide,” 30 September 2026. https://www.latimes.com/politics/newsletter/2026-09-30/gavin-newsom-kamala-harris-2028-democratic-presidential-campaign↩︎

159. ABC News 4, “Why experts see Newsom’s pledge to avoid running against Harris as strategic move.” https://abcnews4.com/news/nation-world/why-experts-see-newsoms-pledge-to-avoid-running-against-harris-as-strategic-move-gavin-newsom-kamala-harris-democrats-2028-presidential-race↩︎

160. Newsweek, “Kamala Harris Is Already Shaping the 2028 Democratic Primary.” https://www.newsweek.com/kamala-harris-2028-democratic-primary-newsom-12548632↩︎

161. Yahoo News, “Gavin Newsom says he won’t run in 2028 if Kamala Harris does” (including Polymarket odds), 30 September 2026. https://www.yahoo.com/news/politics/articles/gavin-newsom-says-wont-run-224246364.html↩︎

162. RealClearPolling, “2028 U.S. Presidential Election,” averages for 19 July–6 October 2026. https://www.realclearpolling.com/elections/president/2028/national↩︎

163. The Hill, “Harris, Vance leading in 2028 presidential polling” (Financial Times/Focaldata, 2–5 October 2026). https://thehill.com/homenews/campaign/6144802-harris-vance-lead-2028-poll/↩︎

164. The Hill, “Harris, Vance leading in 2028 presidential polling” (Financial Times/Focaldata, 2–5 October 2026). https://thehill.com/homenews/campaign/6144802-harris-vance-lead-2028-poll/↩︎

165. Newsweek, “Jon Ossoff vs. Pete Buttigieg’s chances of defeating JD Vance in 2028” (Emerson College Polling). https://www.newsweek.com/jon-ossoff-vs-pete-buttigiegs-chances-of-defeating-jd-vance-in-2028-12349300↩︎

166. Ground News, “GOP Trails Democrats by 8 Points on Congressional Ballot in New Poll” (Emerson College Polling summary). https://ground.news/article/gop-trails-democrats-by-8-points-on-generic-congressional-ballot-in-new-poll↩︎

167. RealClearPolling, “Presidential General Election Polls.” https://www.realclearpolling.com/latest-polls/president/general-election↩︎

168. Wikipedia, “2024 United States presidential election” (national results, swing-state margins and polling error). https://en.wikipedia.org/wiki/2024_United_States_presidential_election↩︎

169. Wikipedia, “2024 United States presidential debates.” https://en.wikipedia.org/wiki/2024_United_States_presidential_debates↩︎

170. Associated Press, “Biden drops out of 2024 race after disastrous debate inflamed age concerns.” https://www.ap.org/news-highlights/elections/2024/biden-drops-out-of-2024-race-after-disastrous-debate-inflamed-age-concerns-vp-harris-gets-his-nod/↩︎

171. Wikipedia, “2024 United States presidential debates.” https://en.wikipedia.org/wiki/2024_United_States_presidential_debates↩︎

172. Wikipedia, “United States presidential debates.” https://en.wikipedia.org/wiki/United_States_presidential_debates↩︎

173. Wikipedia, “Commission on Presidential Debates.” https://en.wikipedia.org/wiki/Commission_on_Presidential_Debates↩︎

174. Congressional Research Service, “Supreme Court Rules Against Tariffs Imposed Under the International Emergency Economic Powers Act,” LSB11398. https://www.congress.gov/crs-product/LSB11398↩︎

175. Congressional Research Service, “U.S. Tariffs on Canadian Imports: Section 338 of the Tariff Act of 1930,” R49349, 14 September 2026. https://www.congress.gov/crs-product/R49349↩︎

176. International Monetary Fund, World Economic Outlook Update, July 2026, 8 July 2026. https://www.imf.org/en/publications/weo/issues/2026/07/08/world-economic-outlook-update-july-2026↩︎

177. Congressional Research Service, “U.S. Conflict with Iran,” R48887, updated 18 September 2026. https://www.congress.gov/crs-product/R48887↩︎

178. International Monetary Fund, “Currency Composition of Official Foreign Exchange Reserves (COFER),” IMF Data Brief, 30 September 2026. https://data.imf.org/en/news/imf%20data%20brief%20september%2030↩︎

179. Reuters, “EU chief wants Canada to become an ‘associate member’. What does that mean?,” 16 September 2026. https://www.reuters.com/world/eu-chief-wants-canada-become-an-associate-member-what-does-that-mean-2026-09-16/↩︎

180. U.S. Bureau of Labor Statistics, “Consumer Price Index Summary,” release of 14 October 2026 (next release 10 November 2026). https://www.bls.gov/news.release/cpi.nr0.htm↩︎

181. Yahoo News, “Charlie Cook: Democrats ‘badly damaged,’ Kansas now a toss-up,” 11 October 2026. https://www.yahoo.com/news/politics/articles/charlie-cook-democrats-badly-damaged-171825696.html↩︎

182. Axios, “House GOP leaders admit redistricting fell short,” 6 October 2026. https://www.axios.com/2026/10/06/house-republicans-redistricting-midterm-elections↩︎

183. Covers, “2026 Midterm Election Odds: Live Prediction Market Tracker.” https://www.covers.com/politics/2026-midterm-prediction-market-odds-tracker↩︎