选举日终于来临，民调、提前投票和总体选举动态现在几乎都达成一致：你应该预期中期选举将出现一场巨大的“蓝色浪潮”，其规模足以让民主党赢得众议院，甚至可能大到足以让他们翻转参议院。但我们该对这一共识抱有多少信心？
Election Day is finally here, and the polls, early voting and general electoral dynamics are practically now all in agreement: You should expect a big “blue wave” in the midterm elections that is large enough to give Democrats the House, and might be large enough for them to flip the Senate. But how much faith should we put into that consensus?
以下是我们在 FiftyPlusOne 的选举预测模型的最终展望，MS NOW 独家合作 FiftyPlusOne 获取民调数据和选举分析，以帮助读者为今年选举的潜在结果做好准备。
Here’s a final look at our election forecasting model at FiftyPlusOne, MS NOW’s exclusive partner for polling data and election analysis, to prepare readers for the potential outcomes of the elections this year.
我们的最终预测显示，民主党赢得众议院的几率为 99%，中位数预测为 240 个席位。
Our final forecast gives Democrats a 99% chance of winning the House, with a median of 240 seats.
该预测意味着民主党在全国众议院得票率上的优势约为 9 个百分点，与我们的普选民调平均值 D+8.7 一致。这将超过 2008 年以来任何一届众议院选举中任一党派的优势幅度。虽然可能性不大，但如果民调低估了民主党的支持程度，他们的得票优势可能接近“水门事件”发生后的水平，当时他们的得票优势在十几个百分点。
The forecast implies a national House vote margin for the Democrats of about nine percentage points, in line with our generic ballot average of D+8.7. This would be larger than the margin for either party in any House election since 2008. Though unlikely, if the polls understate Democrats modestly, they could approach the levels reached in the immediate aftermath of Watergate, when their vote margin lead was in the mid-teens.
共和党需要民调的偏差大于 2016 年、2020 年或 2024 年的偏差，才能改变这一局面。
Republicans would need the polls to be off by more than they were in 2016, 2020 or 2024 to change that.
在这几年里，众议院普选民调高估了民主党约 3.9 个百分点。将这一数值代入我们的预测模型，即便出现如此大的民调偏差，民主党预计仍将获得 226 个席位，并有 82% 的几率赢得多数席位。若偏差达到 2020 年的水平——约 5 个百分点——民主党获胜几率仍为 71%。除了竞选民调，我们还可以衡量选民情绪。
Across those years, the polls overestimated Democrats on the House generic ballot by about 3.9 percentage points. Plugging that into our forecasting model, a poll miss that large still leaves Democrats with a projected 226 seats and an 82% chance of a majority. A miss the size of 2020’s — about 5 points — left them at 71%. Aside from horse-race polls, we can gauge voter sentiment as well.
Verasight 是一家混合模式民调机构，其受访者来源包括互联网和短信等多种渠道。该机构于 10 月 13 日至 19 日对 2028 名美国成年人进行了民调，为我的 Substack 通讯“Strength in Numbers”提供数据。结果显示，58% 的成年人认为，选出一个能够制衡唐纳德·特朗普总统的国会更为重要，而只有 25% 的人希望国会帮助他推行议程。这一比例较 9 月的 56% 对 26% 有所上升。
Verasight, a mixed-mode pollster that draws respondents from a mix of sources, including the internet and text messages, conducted a poll of 2,028 U.S. adults from Oct. 13-19 for my Substack newsletter, “Strength in Numbers.”It found that 58% of adults said it’s more important to elect a Congress that acts as a check on President Donald Trump, versus just 25% who want one that helps him pass his agenda. That’s up from 56% to 26% in September.
特朗普的支持率降至 35.6%，接近历史最低点。总统支持率与众议院选举结果的历史关联表明，民主党将在中期选举中净增 30 多个席位。考虑到重新划分选区的因素，这一预测可能过于乐观，但民主党仅需翻转三个席位即可赢得多数。
And Trump’s approval rating is at a near all-time low of 35.6%. The historical relationship between presidential approval and House election outcomes suggests a Democratic seat gain in the mid-30s. This is likely too optimistic because of redistricting, but Democrats only need to flip three seats to win the majority.
因此，从这位预测师和民调专家的角度来看，关键因素不仅在于民调显示大量选民倾向于投票给民主党，更在于民调指向了一种预计民主党将大获全胜的选举环境。
So from the perspective of this forecaster and pollster, the relevant factor is not just that the polls show voters they’ll vote Democratic in large numbers, but that they point to the type of environment where Democrats should be expected to win big.
然而，我们对众议院的把握并不延伸至国会上议院——参议院。参议院的民调要胶着得多。
Our certainty about the House does not extend to the upper chamber of Congress, however. The polls in the Senate are much closer.
民主党目前掌握 47 个席位。由于 50 比 50 的僵局由共和党副总统 J.D. 万斯（JD Vance）决定，民主党名义上需要 51 个席位才能控制参议院。实际上，他们可能需要 52 个席位，以防万一其成员、宾夕法尼亚州参议员约翰·费特曼（John Fetterman，民主党）倒向共和党（或者至少出现费特曼变成乔·曼钦（Joe Manchin）式民主党人的可能性，即凭借阻挠民主党议程的能力施加不成比例的影响力）。
Democrats currently hold 47 seats. Because a 50-50 split is decided by the Republican Vice President JD Vance, Democrats need 51 to nominally take control. They may in fact need 52 to safeguard against one of their members, Sen. John Fetterman, D-Pa., potentially becoming a Republican (or at least the possibility that Fetterman becomes a Joe Manchin-type Democrat, one who wields disproportionate influence thanks to his ability to stymie Democrats’ agenda).
我们的最终预测显示，民主党有 77% 的机会达到这一目标，中位数为 52 个席位。他们有 87% 的机会至少获得 50 个席位，但 50 个席位是不够的。
Our final forecast gives Democrats a 77% chance of getting there, with a median of 52 seats. They have an 87% chance of reaching at least 50, but 50 isn’t enough.
参议院控制权极有可能取决于七个胶着席位：六个由共和党把持的席位，以及密歇根州这一唯一由民主党把持、共和党有实质翻盘机会的席位。根据我们从 FiftyPlusOne 获取的各选区民调数据库及平均值，民主党在所有这七个席位的民调平均值中均处于领先：缅因州：前州参议员特洛伊·杰克逊平均领先现任参议员苏珊·柯林斯 5.1 个百分点。《纽约时报》/锡耶纳学院于 11 月 1 日完成的民调显示，杰克逊以 51% 对 45% 领先。
The Senate will most likely be decided in seven close seats: six Republican-held seats and Michigan, the one Democratic-held seat Republicans have a solid chance of flipping. Drawing on our database of polls and averages in each race from FiftyPlusOne, Democrats lead the polling average in all seven of these seats: Maine: Former state Sen. Troy Jackson leads incumbent Sen. Susan Collins by 5.1 points on average. A New York Times/Siena College poll completed Nov. 1 had Jackson up 51% to 45%.
密歇根州：阿卜杜勒·埃尔-赛义德平均领先前众议员迈克·罗杰斯 5.3 个百分点。锡耶纳学院 11 月 1 日的民调显示他以 53% 对 42% 领先，但 Research Co. 同一天的民调显示为 48% 对 46%。
Michigan: Abdul El-Sayed leads former Rep. Mike Rogers by 5.3 points. Siena had him up 53% to 42% on Nov. 1, but a Research Co. poll the same day had it 48% to 46%.
得克萨斯州：州众议员詹姆斯·塔拉里科平均领先州检察长肯·帕克斯顿 3.8 个百分点。锡耶纳学院 11 月 1 日的民调显示塔拉里科以 50% 对 43% 领先。特拉法尔加集团同一天的民调显示帕克斯顿以 48.1% 对 47.5% 领先，这是近期唯一一份显示他领先的民调。
Texas: State Rep. James Talarico leads state Attorney General Ken Paxton by 3.8 points. Siena found Talarico ahead 50% to 43% on Nov. 1. A Trafalgar Group poll the same day had Paxton up 48.1% to 47.5%, the only recent poll with him in front.
堪萨斯州：亚当·汉密尔顿平均领先现任参议员罗杰·马歇尔 3.4 个百分点。锡耶纳学院 11 月 1 日的民调显示为 49% 对 44%。
Kansas: Adam Hamilton leads incumbent Sen. Roger Marshall by 3.4 points. Siena had it 49% to 44% on Nov. 1.
俄亥俄州：前参议员谢罗德·布朗平均领先现任参议员乔恩·休斯特德 2.8 个百分点。锡耶纳学院 11 月 1 日的民调显示布朗以 49% 对 44% 领先，但 ActiVote 10 月 31 日的民调显示为 50.2% 对 49.8%。
Ohio: Former Sen. Sherrod Brown leads incumbent Sen. Jon Husted by 2.8 points. Siena had Brown up 49% to 44% on Nov. 1, but ActiVote’s Oct. 31 poll had it 50.2% to 49.8%.
阿拉斯加州：前众议员玛丽·佩尔托拉平均领先现任参议员丹·沙利文 1.8 个百分点。锡耶纳学院 10 月 30 日的民调显示佩尔托拉以 48% 对 46% 领先。
Alaska: Former Rep. Mary Peltola leads incumbent Sen. Dan Sullivan by 1.8 points. Siena had Peltola up 48% to 46% on Oct. 30.
爱荷华州：在一场尤为胶着的竞争中，民主党籍州众议员乔什·图雷克平均领先共和党籍州众议员阿什利·辛森 0.3 个百分点。锡耶纳学院 11 月 1 日的民调显示图雷克以 48% 对 47% 领先，而 co/efficient 10 月 28 日的民调显示辛森以 48% 对 46% 领先。
Iowa: In a particularly close contest, Democratic state Rep. Josh Turek leads Republican state Rep. Ashley Hinson by 0.3 points. Siena had Turek ahead 48% to 47% on Nov. 1, and co/efficient had Hinson up 48% to 46% on Oct. 28.
北卡罗来纳州，一个由共和党把持的席位，实际上已不在竞争行列。前州长罗伊·库珀平均领先前共和党全国委员会主席迈克尔·沃特利 10.3 个百分点，YouGov 10 月 29 日的民调显示他以 55% 对 44% 领先。
North Carolina, a Republican-held seat, is effectively off the board. Former Gov. Roy Cooper leads former Republican National Committee Chair Michael Whatley by 10.3 points on average, and YouGov had him ahead 55% to 44% on Oct. 29.
FiftyPlusOne 的预测显示，民主党在所有七个州的胜算均超过五成：缅因州 77%、密歇根州 76%、得克萨斯州 64%、阿拉斯加州 63%、俄亥俄州 62%、堪萨斯州 57% 和爱荷华州 53%。民主党无需赢下全部七州。如果北卡罗来纳州转为蓝色，他们只需守住密歇根州，并在其余六州中赢下三个，这也是我们的预测看好民主党的原因。
FiftyPlusOne’s forecast gives Democrats better than even odds in all seven: Maine at 77%, Michigan 76%, Texas 64%, Alaska 63%, Ohio 62%, Kansas 57% and Iowa 53%. Democrats don’t need to win all seven. If North Carolina goes blue, they need to hold Michigan and win three of the other six, which is why they are favored in our forecast.
许多民主党人对民调感到焦虑，进而对我们的预测也感到焦虑，这是可以理解的。参议院选举民调在各州平均高估了民主党的支持率：2016 年高估 3.2 个百分点，2020 年高估 5.6 个百分点，2024 年高估 3.3 个百分点。
It is understandable that many Democrats are anxious about the polls, and by proxy, our forecasts. Senate polls overstated Democrats on average across states by 3.2 points in 2016, 5.6 points in 2020 and 3.3 points in 2024.
然而，近几次选举中民调也曾低估民主党。2022 年，民主党在关键参议院竞选中的表现优于民调预测，其中在宾夕法尼亚州，费特曼击败共和党提名人梅赫梅特·奥兹的优势比民调预测的大出约 4 个百分点。2018 年，参议院竞选的民调误差实际上为零。正如任何合格的民调专家都会告诉你的，试图提前预测民调误差的方向是徒劳的。
Yet polls have also understated Democrats in recent elections. In 2022, Democrats outperformed their polls in key Senate races, including in Pennsylvania, where Fetterman beat Republican nominee Mehmet Oz by about 4 points more than the polls forecast. In 2018, the polling error in Senate races was practically zero. And as any pollster worth their salt will tell you, it’s a fool’s errand to try to predict the direction of polling error ahead of time.
这就是为什么我们准备了一个交互式数据可视化工具，可以逐州展示：如果民调像过去两组年份（总统选举年和中期选举年）的平均误差那样出错，选举结果会怎样。您可以在我们的网站上自行探索这些情景。
This is why we have prepared an interactive data visualization that can show you, state by state, what would happen in the election if the polls are as wrong as they were on average in these two sets of years: presidential cycles and midterms. You can explore the scenarios yourself on our site.
这里的结论实际上比您预期的更有利于民主党。即使出现像 2016 年、2020 年和 2024 年平均误差那样大的偏差，民主党仍将赢得北卡罗来纳州、得克萨斯州和堪萨斯州：库珀将在北卡罗来纳州以 6.7 个百分点获胜，塔拉里科将在得克萨斯州以 1.1 个百分点获胜，汉密尔顿将在堪萨斯州以 1.2 个百分点获胜。加上密歇根州，埃尔-赛义德将以 0.8 个百分点守住该州，民主党最终将获得 50 个席位，比目前净增三席。
The takeaway here is actually better for Democrats than you might expect. Even with a miss as big as the average in 2016, 2020 and 2024, Democrats would still pick up North Carolina, Texas and Kansas: Cooper would win North Carolina by 6.7 points, Talarico would win Texas by 1.1 points and Hamilton would win Kansas by 1.2 points. Add Michigan, where El-Sayed would hold on by 0.8 points, and Democrats would end up with 50 seats, a net gain of three from where they are now.
这还不足以确保控制权，但其他州的竞选结果如果再加上另一方向的州级民调误差，可能会把天平倾向民主党。在高误差情景下，共和党将以1.7个百分点赢得缅因州，以0.9个百分点赢得俄亥俄州，以5.2个百分点赢得阿拉斯加州，以5.7个百分点赢得爱荷华州。缅因州和俄亥俄州的差距都在一到两个百分点之内，远低于州级民调平均误差的常规范围。如果其中任何一州的结果相反，民主党将拥有51个席位并控制参议院。
That isn’t enough for control, but other races would be close enough for an additional state-level polling error in the other direction to tilt the scales toward Democrats. In the high-error scenario, Republicans would win Maine by 1.7 points, Ohio by 0.9, Alaska by 5.2 and Iowa by 5.7. Both Maine and Ohio would be within one or two points, well inside the usual error of a state polling average. If either one goes the other way, Democrats would have 51 seats and control of the Senate.
因此，共和党的道路很窄。他们需要民调像2016年、2020年和2024年那样严重失误，然后再保持缅因州和俄亥俄州的优势。
So the Republican path is narrow. They need the polls to miss as badly as they did in 2016, 2020 and 2024, and then hold both Maine and Ohio on top of that.
在这种情景下，民主党仍有可能获胜。假设2024年式的平均民调失误，民主党获得参议院多数席位的概率将从72%降至40%。
Under this scenario, it would still be possible for Democrats to win. Democrats’ chances of a Senate majority would fall from 72% to 40%, assuming a 2024-style average polling miss.
但如果相反，民调与过去两次中期选举的准确度相当，民主党将有望赢得52个席位，翻转缅因州、德克萨斯州、堪萨斯州和阿拉斯加州的红蓝色。除此之外，爱荷华州、俄亥俄州和南卡罗来纳州的竞选结果都将在一个百分点之内，可能轻易左右。若民调实际上偏向民主党——这是该党最乐观的情景——他们可能赢得这些席位以及内布拉斯加州，在那里我们将独立人士丹·奥斯本视为民主党人，以确定参议院控制权。
But if, on the other hand, the polls are as accurate as they were in the past two midterm elections, Democrats would be favored to win 52 seats, flipping Maine, Texas, Kansas and Alaska from red to blue. Additionally, the races in Iowa, Ohio and South Carolina would all be within a point and could easily go either way. If the polls actually Democrats — the bull case for the party — they could win those seats plus Nebraska, where we count independent Dan Osborn as a Democrat for purposes of determining Senate control.
尚不清楚奥斯本是否会与民主党同席，但他击败皮特·里克特斯后，肯定会剥夺共和党一名议员。这将使民主党总共拥有56个席位。
It’s unclear whether Osborn would caucus with Democrats, but by defeating Pete Ricketts, he would definitely deprive Republicans of a member. That would give Democrats a total of 56 seats.
如果民调仅像2018年和2022年那样失误，参议院将与我们的预测大致相同。在这种情景下，民主党控制的概率几乎没有变化——72% 与已公布的77%——他们将拥有52个席位，俄亥俄州和爱荷华州基本持平。所以当你在周二关注结果时，请注意以下几点。
If the polls miss only as much as they did in 2018 and 2022, the Senate looks a lot like the one in our forecast. Democrats’ control odds barely moved in that scenario — 72% versus the published 77% — and they would have 52 seats, with Ohio and Iowa essentially tied. So as you’re watching results come in Tuesday, here’s what to pay attention to.
北卡罗来纳、俄亥俄、佛罗里达和弗吉尼亚今年都举行参议院选举，这些州的投票站在美国东部时间晚上7点30分或之前关闭，计票速度相对较快。关注这些州的结果，看它们是否与我们的民调平均值大致相符。由于各州的民调误差往往相近，这些计票较快的选举中若出现误差，可能预示着其他地方也会出错。如果误差极小，预计民主党将在全国范围内表现良好。
North Carolina, Ohio, Florida and Virginia — all holding Senate races this year — close their polling places at or before 7:30 p.m. ET and count votes relatively quickly. Watch these states for the results to roughly match our polling averages. Since polls tend to be off by a similar amount across states, an error in these quick-counting contests could forebode errors elsewhere. If there is minimal error, expect Democrats to do well across the country.
如果民主党在俄亥俄州和缅因州都保持领先——缅因州将在美国东部时间晚上8点左右开始公布结果——那么共和党要想保住多数席位，就需要在其他州出现比近年来任何一次都更大的民调偏差。
If Democrats hold their leads in both Ohio and Maine, which will start releasing results around 8 p.m. ET, Republicans would need a polling miss bigger than any in recent years in the other states to hold onto their majority.
如果共和党在这两州都获胜，问题就变成民主党是否仍能通过得克萨斯、堪萨斯和密歇根三州达到51席，民调平均值显示民主党在这三州领先3到5个百分点。
If Republicans win both, the question becomes whether Democrats can still reach 51 through Texas, Kansas and Michigan, three states where the polling averages put them ahead by 3 to 5 percentage points.
这些优势足够微弱，其中任何一个州的一个糟糕夜晚都至关重要。但共和党必须几乎赢下所有势均力敌的选战，而且民调必须朝着有利于他们的方向出错才能做到。
Those margins are thin enough that one bad night in any of them matters. But Republicans have to win nearly everything that’s close, and the polls have to be wrong in their favor to do it.
当你唯一的胜选条件取决于所有选前数据都是错的时，这可不是一个好处境。
It is not a good position to be in when your sole win condition for the election hinges on all the pre-election data being wrong.