过去十年美国政治的主题是两极分化:具体来说，我们已经看到了很多。例如，一些权威人士哀叹摇摆选民的消失，以及“消极党派”对理性的统治。正如《纽约时报》所说问今年2月:“还有没有可以说服的选民?”
本周，当乔·拜登(Joe Biden)总统和前总统唐纳德·特朗普(Donald Trump)在周四晚上举行的总统大选第一场辩论中交锋时，准备好再次听到许多这样的话语。我预测这场辩论的影响会让人绝望——或者更准确地说，它可能无关紧要。但是任何认为辩论在总统竞选中无关紧要的人都不了解他们的历史。
竞选活动仍然有小的影响
两极分化意味着辩论无关紧要的理论很简单:当95%的选民投票给与他们有相同党派身份的候选人时，失言就没那么重要了，因为失去支持的人会更少;强劲的表现也没那么重要，因为要讨好的选民更少了。
然而，虽然现在公众观点的转变确实比过去要小，但有说服力的选民确实存在观点仍然会发生变化——在势均力敌的选举中，小的变化会带来大的不同。最近引起这种变化的事件可能包括时任美国联邦调查局导演詹姆斯·科米在2016年大选前11天宣布对前国务卿希拉里进行调查可能让她在民意调查中损失不多)和2020年乔治·弗洛伊德的谋杀案(这导致了持久的变化在公众舆论或回答民意调查的人的变化—或者两者都有)。
2024年，你可以在这个名单上加上特朗普对伪造商业记录的34项重罪的定罪。根据538的平均投票率目前，拜登在全国民调中仅以0.1个百分点的微弱优势领先特朗普。这种差距是如此之小，它是坚实的在不确定区间内但这确实代表了过去几周民意调查的真正变化。6月10日，特朗普平均领先拜登1个百分点。5月30日，特朗普被定罪的当天，他领先1.7个百分点——与今天的数字相差1.8个百分点。拜登的支持率也下降了打勾几个百分点。
然而，拜登在民调中的上升并没有转化为他在我们的选举预测它超越了目前民调结果的问题，试图回答四个月后谁可能赢得大选的问题。在过去的两周里，当全国民意调查转向总统时，他赢得选举的机会实际上从53/100下降到50/100。
那是因为离选举日还有很长的路要走;我们的模型正在定价大约9分的变化根据民意调查的历史波动，从现在到选举日的全国民意调查差额。该模型还将民意调查与对每位候选人最终投票份额的预测相结合基于经济指标。这一预测在过去一周对拜登来说略有恶化，抵消了他在民意调查中取得的进展。
总统辩论通常会改变投票结果
周四的辩论为候选人提供了一个打破这种平衡的机会，并通过系统发出冲击。同样，你可能认为两极分化意味着辩论无关紧要，但我对538的历史总统民意调查数据集的研究表明并非如此。
从1976年开始，我们收集了所有的选举民意调查，我用了538个当前民意测验平均方法计算所有主要政党候选人在两个时间点的历史大选平均值:举行辩论的当天和两周后。然后，我计算了辩论导致的民调差距变化(承认我们无法将辩论效应与民调中的其他故事、新闻事件或动态完全隔离开来)。
平均而言，在每个周期的第一场总统辩论后的两周内，民调移动了2.4个百分点。这个平均值隐藏了很多差异:有时，就像1988年，候选人已经众所周知，民意调查没有太大变化。其他时候，如1976年和2008年，一名候选人表现特别出色，或者利用竞选中的现有运动攀登新的高度。
虽然从历史上看，这并不是一个大的转变，但如今，在我们这个竞争激烈的时代，竞选中2.4个百分点的变化将对每个候选人的胜算产生有意义的影响。特朗普获得2.4个百分点，将使他在亚利桑那州和内华达州等阳光地带的战场上明显超过拜登，给现任总统带来更大压力，以横扫三个北部摇摆州——密歇根州、宾夕法尼亚州和威斯康星州。另一方面，拜登在所有地方都上升了2.4个百分点，这将使他在密歇根州、宾夕法尼亚州和威斯康星州取得稳固的领先优势，并使一些目前势均力敌的州，如明尼苏达州、缅因州和新罕布什尔州，进一步远离共和党。
然而，占主导地位的模式是，首次辩论通常对现任不利。自1976年以来的七次选举中，有一名当选的在任者参加投票，在任者在其中六次的投票中平均落后2个百分点。这对拜登来说将是一个重大挫折，他今年上半年的大部分时间都在巩固民主党人的支持，以在全国赛马中领先于特朗普。但也许拜登没什么可担心的;过去的挑战者在辩论后获得的大部分原因是，他们在第一次大规模介绍给大众时给美国人民留下了积极的印象。今年，拜登的挑战者是一个已知的数量，所以公众对他的了解较少。
对拜登或特朗普来说，一个小小的推动可能会改变整个局面
然而，两位候选人都面临一些严重的下行风险。如果你认为潜在的辩论效果是由统计分布产生的(谁不这样想呢!?)，你可能会说每个候选人的净效应是歪斜的消极的，潜在的肥胖的尾巴—说白了，翻牌的几率高于预期。例如，一场旷日持久的辩论有可能向美国展示一幅年迈、疲惫的总统画像——而他们是他们最关心的问题之一拜登的第二个任期。另一方面，在被定罪后，特朗普看到了支持度下降最多从最不积极的选民那里——那种你真的只能通过总统辩论产生的广泛报道才能接触到的人。
此外，它甚至不需要一个大的辩论效应来改变竞选的轮廓或11月的赢家。538总统预测目前认为，拜登需要在11月赢得全国普选1.5个百分点，才能赢得大多数选举人团的选票。今天他在全国民调中领先0.1个百分点，大约1个百分点的提升将使拜登非常接近他需要在11月5日赢得激烈胜利的位置。
对于你们当中的愤世嫉俗者来说，这听起来可能有些不可思议。熟悉的候选人之间的辩论事实上改变主意?在…里这极度两极分化的政治格局?民意调查显示今年到目前为止非常稳定，你可能会反驳。但对此我想说:更疯狂的事情已经发生了，包括最近。比赛通常是稳定的直到他们不再是.
最近美国政治两极分化的加剧没有理由期待周四的辩论不会产生任何影响——甚至任何一位候选人的小幅上升都可能对竞选产生有意义的影响。
Why the Biden-Trump debate matters more than you think
The theme of American politics over the last decade is polarization: Specifically, we've seen a lot of it. Some pundits have bemoaned the disappearance of swing voters, for example, and the rule of "negative partisanship" over rationality. As The New York Timesaskedthis February: "Are There Any Persuadable Voters Left?"
Prepare to hear a lot of this discourse again this week, when President Joe Biden and former President Donald Trump face off in the first debate of the presidential general election on Thursday night. I predict much hand-wringing about the effects of this debate — or, to be more precise, how little it could matter. But anyone who thinks debates don't matter in presidential campaigns isn't up on their history.
Campaign events still have small effects
The theory that polarization means debates don't matter is straightforward: When 95 percent of voters vote for the candidate with the same partisan identity as them, gaffes matter less because fewer supporters will be turned off; strong performances, too, matter less because there are fewer voters to court.
Yet while it is true that shifts in public opinion are smaller now than they used to be, persuadable votersdo existand opinions do still change — and in closely divided elections, small changes make big differences. Recent events causing such changes may include then-FBI Director James Comey announcing an investigation into former Secretary of State Hillary Clinton 11 days before the 2016 election (whichprobably cost her modestly in the polls) and the murder of George Floyd in 2020 (which caused either adurable changein public opinion or achange in who was answering polls— or both).
For 2024, you can add to that list Trump's conviction on 34 felony counts of falsifying business records. According to538's polling averages, Biden currently has a razor-thin, 0.1-percentage-point lead over Trump in national polls. That margin is so small that it is solidlywithin the uncertainty intervalof our average, but it does represent a real change in the polls over the last few weeks. On June 10, Trump was leading Biden by 1 point in that average. And on May 30, the day Trump was convicted, he had a 1.7-point lead — a 1.8-point difference from today's numbers. Biden's approval rating has alsoticked upa few percentage points.
However, Biden's rise in the polls has not translated into a significant gain inour election forecast, which pushes beyond the question of what the polls say today and attempts to answer the question of who is likely to win the election four months from now. Over the last two weeks, when national polls shifted toward the president, his chances of winning the election actually decreased from 53-in-100 to 50-in-100.
That's because there is still a long way to go until Election Day; our model is pricing inabout 9 points' worth of changein the national polling margin from now to Election Day, based on historical volatility in the polls. The model also combines the polls with a forecast of the final vote share for each candidatebased on economic indicators. That forecast has gotten slightly worse for Biden over the last week, canceling out the gains he has made in the polls.
Presidential debates usually move the polls
The debate on Thursday provides an opportunity for the candidates to interrupt this equilibrium and to send a shock through the system. Again, you may think that polarization means debates don't matter, but my study of 538's historical dataset of presidential polls says otherwise.
Taking all the election polls we have collected since 1976, I used 538'scurrent polling average methodologyto calculate historical general election averages for all major party candidates at two points in time: the day a debate was held and two weeks later. Then I calculated how much the polling margin changed as a result of the debate (acknowledging that we cannot completely isolate a debate effect from other stories, news events or dynamics in the polls).
On average, polls have moved by 2.4 points in the two weeks after the first presidential debate of each cycle. That average hides a lot of variance: Sometimes, as in 1988, the candidates are already well known and polls don't change much. Other times, as in 1976 and 2008, one candidate puts on an especially strong performance or capitalizes on existing movement in the race to climb to new heights.
While not a large shift historically, nowadays, in our era of closely contested elections, a 2.4-point change in the race would meaningfully impact each candidate's odds. A 2.4-point gain for Trump would have him clearly pulling away from Biden in Sun Belt battlegrounds such as Arizona and Nevada, putting more pressure on the incumbent to sweep the three northern swing states — Michigan, Pennsylvania and Wisconsin. On the other hand, Biden moving up everywhere by 2.4 points would give him a solid lead in Michigan, Pennsylvania and Wisconsin and put some currently close states like Minnesota, Maine and New Hampshire further out of Republicans' reach.
The predominant pattern, however, is that first debates are usually bad for incumbents. In the seven elections since 1976 with an elected incumbent on the ballot, that incumbent lost ground in the polls in six of them, by an average of 2 points. This would be a significant setback for Biden, who has spent most of the first half of this year consolidating support among Democrats to pull ahead of Trump in the national horse race. But perhaps Biden has little to worry about; much of the reason past challengers have gained after debates is that they made a positive impression on the American people in their first big introduction to the masses. This year, Biden's challenger is a known quantity, so there's less for the public to learn about him.
A small boost for Biden — or Trump — could be the whole ballgame
Both of the candidates face some serious downside risk, however. If you think of potential debate effects as being generated by a statistical distribution (and who doesn't think like that!?), you might say the net effect for each candidate isskewednegatively, potentially with afat tail— in plain English, there's a higher-than-expected chance of flopping. A long debate risks showing America a portrait of an old, tired president, for example — which theyrate among their highest concernsfor a second Biden term. On the other side, in the wake of his criminal conviction, Trump has seenthe most degradation in supportfrom the least engaged voters — the type you can really only reach with extensive coverage generated by a presidential debate.
Furthermore, it would not even take a big debate effect to change the contours of the race or the winner in November. The 538 presidential forecast currently thinks Biden needs to win the national popular vote by 1.5 points in November to win a majority of Electoral College votes. With his 0.1-point margin in national polls today, a roughly 1-point boost would put Biden very close to where he needs to be to pull off a white-knuckle victory come Nov. 5.
This may all sound incredibly far-fetched to the cynics among you. Can a debate between familiar candidatesreallychange minds? Inthishyper-polarized political landscape? The polls have beenvery stable so far this year, you might retort. But to that I would say: Crazier things have happened,including recently. Races are usually stableuntil they're not.
The recent rise of political polarization in America is no reason to expect absolutely no effect from the debate Thursday — and even a small bump for either candidate could have a meaningful effect on the race.
