本周早些时候，我和内森·里德谈过了我的一个“朋友”(实际上，是我另一个自我的一部分)，他试图让我相信共和党也可以甚至比538预测的还要好.今天，你会见到纳撒尼尔·布鲁，他是我想象中的另一个虚构人物，我在布鲁克林鹅卵石山的一个酒吧和他聊过，他和他的丈夫以及两只法国斗牛犬住在一起。布鲁是纽约一家大型出版公司的广告销售主管，他表示相信民主党会赢，尽管你会看到他的信心很容易动摇。下面是我们谈话的简要记录。

蓝色:选举季快乐内特。我从未见过每个人如此恐惧。

银色:嗯，如果我能戴上我的关心市民的帽子一会儿—

蓝色:那看起来更像底特律老虎队的帽子。

银色:非常有趣。 如果我能暂时戴上关心市民的帽子，我有理由担心。所有这些选票上否认选举的人.这攻击保罗·佩洛西.我不知道这通向哪里，但它看起来不像是一个好地方。

蓝色:我说的不是这个。我从未见过每个人如此害怕承认民主党会赢。

银色:嗯——

蓝色:即使是你内特。你知道我非常尊重你和538。为什么，如果可以的话，我会有一整个动物园的五只狐狸毛绒玩具娃娃。但是你的预测是民主党会赢，而你害怕承认这一点。

银色:的确如此不让民主党获胜。你在读那些吗来自DCCC的假538警报?

蓝色:这高级的预报共和党领先了吗.但是我不相信它引入的所有主观性—

银色:郑重声明，我反对对豪华预测的这种描述—

蓝色:精简版预测认为民主党会赢。那是我信任的人。

银色:Lite拥有它们几乎不在参议院领先。共和党人的支持率下降得如此之快，到周二的时候，这种领先优势可能会消失。更不用说众议院了，在那里共和党人是最受欢迎的。此外，豪华是更好，更准确的产品。

蓝色:前几天共和党人在豪华城的时候，我几乎可以看到你脸上的笑容。只是太多了更安全的对你来说是这样的。

银色:你认为我不喜欢冒险?你还是拒绝在室内吃晚餐。

蓝色:不是真的。只有在去年冬天奥米克隆涨潮的时候。那是我们最后一次见面吗?此外，这是一个美好的夜晚。

银色:我几乎感觉不到我的脚趾。

蓝色:平心而论，不只是你。是所有人。你和我以及所有媒体人。每个人都害怕预测这个任何事对民主党人来说，好事情发生了，重演了2016年，看起来又很愚蠢。

银色:你不在“媒体”工作你在一家媒体公司做广告销售。仅仅因为你在康泰纳仕餐厅看到大卫·雷姆尼克并不意味着你就是玛吉·哈伯曼。

蓝色:有人需要再来一杯酒。事实是，没有人知道这次选举结果如何。这只是人们四处摸索，试图听起来聪明。这是拍摄的供给满足拍摄的需求。现在，每个人都希望“共和党会赢。”

银色:我真的一点都不相信。打开MSNBC，你就会看到对民主党人来说，提前投票和投票率以及其他任何事情都充满了希望.人们喜欢奉承他们感情的镜头。但我对你的室友的心理不感兴趣。我感兴趣的是为什么你认为我们的预测可能是错的，而你对此却只字未提。另外，我以为你会加入忧郁队。我以为你会气炸了。

蓝色:英语字母表中第九个字母是忧郁团队。为什么，我可能已经是四分卫了。然后我去了堪萨斯。

银色:你到底在堪萨斯做什么?我知道你和科里在讨论买一栋乡间别墅，但那太久了—

蓝色:科里在托皮卡长大。所以我们在那里说“2号不要”——你知道，就是堕胎公投.我们甚至做了点调查。我知道没有人会赢，我尤其知道年轻女性对此有多么强烈的感受。但连我都没想到它会以18个百分点的优势胜出。

银色:毫无疑问，令人印象深刻。如果周二是某种全民公投鱼卵我相信民主党会做得很好。但是—

蓝色:让我打断你一下。因为它不是只有堪萨斯。也有一些特殊的选举。你熟悉FiveThirtyEight dot com这个网站吗?在这里，我要拿出我的手机，因为我将页面加入书签只是为了确保我没弄错。根据纳撒尼尔·拉基奇先生(顺便说一下，这是个可爱的名字)的说法，在罗伊案被推翻后，事情发生了变化。“自那次决定以来……民主党在(四次众议院特别选举)中的领先优势平均超出了他们的预期9个百分点。”这还不算阿拉斯加!你是在告诉我，当民主党赢得堪萨斯州和阿拉斯加州的选举时，他们将度过糟糕的一年?

银色:我告诉你我们已经不在堪萨斯了。

我告诉你，从那以后，民主党的地位恶化了。你在八月举行这次选举，是的，我认为民主党会保住参议院，甚至众议院。但是民意调查已经非常清楚地显示了共和党的反弹。

蓝色:所以现在我们兜了一圈。民意测验已经改变了吗，还是说振动偏移?这种转变是真实的，还是只是媒体报道的假象?因为我看不出现在和八月相比有什么不同。

内特:天啊，这听起来开始有点阴谋了。还有一个许多自8月份以来，情况发生了变化。通货膨胀的坏消息.大约无数次关于犯罪的福克斯新闻频道片段和“解除警察的资金”实际上，民主党并没有像共和党那样节节败退添加未决定的选民—可能会一直投票给他们的人。一切都过去了某种意义上非常正常.

蓝色:我当然没有指控任何阴谋，或纵容，或阴谋集团。我是说民意测验专家有动机，而这些动机倾向于尽可能给共和党人最乐观的数字，因为他们会如果他们再一次错过民主党，我会非常难过但是没有人会在意他们是否高估了共和党一两个百分点。

银色:这里有几件事，因为这是你的主张。第一，你可以在2020年提出完全相同的论点，然而民意调查显示比2016年更大的民主偏见!第二，我不确定动机是否如此明显最清楚的激励是你只想准确。第三，我认为你低估了民意测验专家。当然，他们中的一些人愤世嫉俗，有党派偏见，但大多数人认为他们所做的是一项公共服务—

蓝色:为什么这么可爱!让我们送一个水果篮到皮尤研究中心，感谢他们的服务!你认为特拉法尔加和拉斯姆森关心公共服务吗?他们是淹没该区域党派民意调查!

银色:事实上，他们是2020年一些最准确的民意测验专家.

蓝色:我不在乎。一个坏了的钟一天对两次。你有很多共和党的民意调查，却没有民主党的。它会扭曲平均值。

银色:我不确定那是真的，至少在五点三十八分。我们的模型有一个房屋效应调整因此，如果一位民意调查者持续显示出对共和党过于乐观的结果——或对民主党过于乐观——它会考虑这一点。此外，对于你关于激励的观点，这是一个自由市场。如果有一家公司的投票率模型显示民主党取得了巨大的成果，如果他们对这些数据有信心，他们可以公布这些数据。

蓝色:我可以告诉你整个讨论让你感到不安。你整晚看起来都不舒服。

银色:我只是有点冷。下次在室内?

听着，现在情况不太好。随着美国党派色彩越来越浓，对机构的信任逐渐消失，民意调查机构将面临许多下游的负面后果。首先，大多数人基本上不再接陌生号码的电话了.所以回答民意调查的人在某种意义上是奇怪的，他们可能不代表选民.您过去可以默认使用更多的投票中的黄金标准—这很贵，但你可以做到。现在，有更多的选择要做。最重要的是，信任媒体是有史以来最低迷的.这是否让民意测验专家更有可能跟着人群走而不是公布可能给他们带来巨大痛苦的数字?也许吧，但在周二之前，我不会对此做太多假设。

蓝色:所以如果我们不能相信民意调查，也许我们应该看看早期的投票数据

银色:哦，不不不不。让我打断你一下。这是个陷阱.有很少使用可靠的基准，以及分析不可避免地反映了人们的党派倾向.在提前投票中，我在某种程度上唯一信任的人是内华达州的乔恩·罗尔斯顿他认为那里的数字看起来很漂亮严重的对民主党来说.

蓝色:你同意我说的任何一句话吗，内特?我曾经可以指望你挑战传统智慧。现在你听起来和其他人一样。你做了什么事实上觉得?

银色:我最不喜欢的问题!我没有自己的一套信仰!我相信我们的预测，它是基于我四年前写的一个计算机程序，而不是我坐在这里喝着一杯皮诺的心情!我们的预测说，参议院对民主党人来说是一个胜负难料的事情，共和党人一直保持着势头。但我不确定我们到底在争论什么。我同意特别选举对民主党有利。我非常同意民主党人能够击败他们的投票。这是一个完全现实的场景.但这不是最有可能的情况。此外，总统的政党在中期选举中表现不佳将是最可以想象的正常事情尤其是在通货膨胀率高达8%的情况下。

蓝色:你一直在用“正常”这个词，但我们已经不是正常时代了。这个国家要完蛋了。

银色:所以你最终还是加入了忧郁队!

蓝色:告诉我还有机会，民主党保住参议院的机会。

银色:有一个45%的几率.

蓝色:我要了。

The Case For A Democratic Surprise On Election Night

Earlier this week, I spoke with Nathan Redd, a “friend” of mine (actually, a fragment of my alter ego) who attempted to convince me that Republicans will do even better than the FiveThirtyEight forecast suggests. Today, you’ll meet Nathaniel Bleu, another figment of my imagination who I spoke with at a wine bar in Cobble Hill, Brooklyn, where he lives with his husband and two French bulldogs. Bleu, an ad sales executive with a major New York publishing company, expressed confidence that Democrats will win, although as you’ll see his confidence was easily shaken. An abbreviated transcript of our conversation follows.

Bleu: Happy election season, Nate. I’ve never seen everyone so terrified.

Silver: Well, if I can put on my concerned citizen hat for a moment —

Bleu: That looks more like a Detroit Tigers hat.

Silver: Very funny. If I can put on my concerned citizen hat for a moment, there’s good reason to be worried. All those election deniers on the ballot. The attack on Paul Pelosi. I’m not sure where this leads, but it doesn’t seem like it’s a good place.

Bleu: That’s not what I’m talking about. I’ve never seen everyone so terrified to admit that Democrats are going to win.

Silver: Ummmmm —

Bleu: Even you, Nate. You know I have the utmost respect for you and FiveThirtyEight. Why, I’d have a whole menagerie of Fivey Fox plush dolls if I could. But your forecast has Democrats winning and you’re terrified to admit it.

Silver: It very much does not have Democrats winning. Are you reading those fake FiveThirtyEight alerts from the DCCC?

Bleu: The Deluxe forecast has Republicans ahead. But I don’t trust all the subjectivity it introduces —

Silver: For the record, I’m going to object to that characterization of the Deluxe forecast —

Bleu: The Lite forecast has Democrats winning. That’s the one I trust.

Silver: Lite has them barely ahead in the Senate. And Republicans are closing so quickly, that lead might disappear by the time we get to Tuesday. And that’s to say nothing of the House, where Republicans are heavy favorites. Besides, Deluxe is the better, more accurate product.

Bleu: I could almost see the grin on your face when Republicans went ahead in Deluxe the other day. It’s just so much safer for you that way.

Silver: You think of me as risk-averse? You’re the one who still refuses to eat dinner indoors.

Bleu: Not true. Only during that Omicron surge last winter. Was that the last time we saw each other? Besides, it’s a lovely evening.

Silver: I can barely feel my toes.

Bleu: To be fair, it’s not just you. It’s everybody. You and me and everyone in the media. Everybody is terrified of predicting that anything good happens for Democrats, having a replay of 2016 and looking foolish again.

Silver: You don’t work in “the media.” You work in ad sales for a media company. Just because you see David Remnick in the Condé Nast cafeteria doesn’t make you Maggie Haberman.

Bleu: Somebody needs another glass of wine. The truth is, nobody knows how this election is going to turn out. It’s just people groping around trying to sound smart. It’s the supply of takes meeting the demand for takes. And right now, the take everyone wants is “Republicans are going to win.”

Silver: I don’t really buy that at all. Turn on MSNBC, and you’ll see plenty of hopium for Democrats about early voting and turnout and whatever else. People like takes that flatter their sensibilities. But I’m not interested in the psychology of your brunchmates. I’m interested in why you think our forecast might be wrong, and you haven’t said a word about that yet. Besides, I expected you to be on Team Gloom. I thought you’d be apoplectic by this point.

Bleu: I was on Team Gloom. Why, I might as well have been the quarterback. And then I went to Kansas.

Silver: What the hell were you doing in Kansas? I know you and Corey were talking about buying a country home, but that’s a long —

Bleu: Well, Corey grew up in Topeka. So we were out there for “No on 2” — you know, the abortion referendum. We were even doing a little canvassing. I knew that No was going to win, I knew in particular how strongly younger women felt about it. But even I didn’t expect it to win by, what was it, 18 percentage points?

Silver: It was impressive, no doubt. And if Tuesday was some sort of national referendum on Roe, I’m sure Democrats would do pretty well. But —

Bleu: Let me stop you right there. Because it wasn’t just Kansas. There were a bunch of special elections, too. Are you familiar with the website FiveThirtyEight dot com? Here, I’m going to pull my phone out because I bookmarked the page just to make sure I got this right. According to Mr. Nathaniel Rakich — lovely first name, by the way — things changed after Roe was overturned. “Since that decision … Democrats have outperformed their expected margins in [four special House elections] by an average of 9 points.” And that doesn’t even count Alaska! You’re telling me Democrats are going to have a bad year when they’re winning elections in Kansas and Alaska?

Silver: I’m telling you that we’re not in Kansas anymore.

I’m telling you that Democrats’ position has deteriorated since then. You hold this election in August, and yeah, I think Democrats keep the Senate and maybe even the House. But the polls have been pretty clear in showing a Republican rebound.

Bleu: So now we come full circle. Have the polls shifted, or has there been more of a vibe shift? And is the vibe shift real, or is it just an artifact of media coverage? Because I don’t see what’s so different now compared with August.

Nate: Oh boy, this is starting to sound conspiratorial. And there’s a lot that’s changed since August. Bad news on inflation. About a zillion Fox News segments on crime and “defund the police.” And it’s not really even that Democrats are losing ground so much as that Republicans are adding undecided voters— people who were probably going to vote for them all along. It’s all been very normal in some sense.

Bleu: I’m certainly not alleging any conspiracy, or connivance, or cabal. I’m saying that pollsters have incentives, and those incentives run toward giving Republicans the most optimistic numbers they can, because they’ll get a ton of grief if they miss high on Democrats again, but nobody will care if they overestimate Republicans by a point or two.

Silver: Well, a few things here, because this is quite the claim you’re making. Number one, you could have made exactly the same argument in 2020, and yet the polls had an even bigger Democratic bias than in 2016! Number two, I’m not sure the incentives are so obvious, and the clearest incentive is that you just want to be accurate. Number three, I think you’re underestimating the pollsters. Sure, some of them are cynical and partisan, but most of them see what they do as a public service —

Bleu: Why how lovely! Let’s send a fruit basket to the Pew Research Center to thank them for their service! Do you think Trafalgar and Rasmussen care about public service? They’re flooding the zone with partisan polls!

Silver: Actually, they were some of the most accurate pollsters in 2020.

Bleu: I don’t care. A broken clock is right twice a day. You have tons of Republican polls, and no Democratic ones. It’s going to skew the averages.

Silver: I’m not so sure that’s true, at least for FiveThirtyEight. Our model has a house-effects adjustment, so if a pollster consistently shows overly rosy results for Republicans — or for Democrats, for that matter — it takes that into account. And besides, to your point about incentives, it’s a free market. If there’s a firm with a turnout model that shows great results for Democrats, they can publish those numbers if they have confidence in them.

Bleu: I can tell this whole discussion is making you uneasy. You’ve looked uncomfortable all night.

Silver: I’m just a little chilly. Indoors next time?

Look, it’s not a great situation. As America gets more partisan, and trust in institutions erodes, there are a lot of downstream, negative consequences for pollsters. To start with, most people basically don’t answer phone calls from unknown numbers anymore. So people who do answer polls are weird in some sense, and they may not be representative of the electorate. You used to be able to default to more of a gold standard in polling — it was expensive, but you could do it. Now, there are a lot more choices to make. On top of that, trust in the media is about as low as it’s ever been. Does that make pollsters more likely to stick with the herd instead of publishing numbers that could cause them a lot of grief? Maybe, but I’m not going to make too many assumptions about that until Tuesday.

Bleu: So if we can’t trust the polls, maybe we should look at early voting data instead —

Silver: Oh, no no no no no. Let me stop you right there. It’s a trap. There are rarely reliable benchmarks to use, and the analysis inevitably reflects people’s partisan priors. About the only person I trust to any degree at all on early voting is Jon Ralston in Nevada, and he thinks the numbers there look pretty bad for Democrats.

Bleu: Have you agreed with a single thing I’ve had to say, Nate? I could once count on you to defy conventional wisdom. Now you sound just like everybody else. What do you really think?

Silver: My least favorite question! I don’t have some private set of beliefs that I keep to myself! I trust our forecast, which is based on a computer program I wrote four years ago and not my mood as I’m sitting here with a glass of pinot! Our forecast says that the Senate is a toss-up at best for Democrats, and the momentum has been with Republicans. But I’m not sure what we’re really arguing about. I agree that the special elections were good for Democrats. And I very much agree that Democrats could beat their polls. It’s an entirely realistic scenario. But it’s not the likeliest scenario. Besides, the president’s party doing poorly in the midterms would be about the most normal thing imaginable, especially with inflation at 8 percent.

Bleu: You keep using that word “normal,” but we’re not in normal times anymore. This country is going to hell.

Silver: So you’re on Team Gloom after all!

Bleu: Just tell me there’s a chance, a chance that Democrats keep the Senate.

Silver: There’s a 45 percent chance.

Bleu: I’ll take it.