Affective polarization—the tendency of Democrats and Republicans to dislike and distrust one another—has become an important phenomenon in American politics. Yet, despite scholarly attention to this topic, two measurement lacunae remain. First, how do the different measures of this concept relate to one another—are they interchangeable? Second, these items all ask respondents about the parties. When individuals answer them, do they think of voters, elites, or both? We demonstrate differences across items, and scholars should carefully think about which items best match their particular research question. Second, we show that when answering questions about the other party, individuals think about elites more than voters. More generally, individuals dislike voters from the other party, but they harbor even more animus toward the other party’s elites. The research note concludes by discussing the consequences for both measuring this concept and understanding its ramifications. For nearly two decades, scholars have analyzed voters’ issue positions to determine whether the mass public is, in fact, polarized (Fiorina 2017). In recent years, however, there is a growing awareness that this does not fully capture partisan conflict in the contemporary United States. Regardless of where they stand on the issues, Americans increasingly dislike, distrust, and do not want James N. Druckman is the Payson S. Wilder Professor of Political Science and faculty fellow in the Institute for Policy Research at Northwestern University, Evanston, IL, USA. Matthew S. Levendusky is a professor in the Department of Political Science, and by courtesy, in the Annenberg School of Communication, as well as distinguished fellow in the Institutions of Democracy at Annenberg Public Policy Center, University of Pennsylvania, Philadelphia, PA, USA. The authors thank the Annenberg Public Policy Center for funding this project (M.S.L., Principal Investigator), Sam Gubitz and Natalie Sands for research assistance, and Joe Biaggio, Shanto Iyengar, Yanna Krupnikov, Yphtach Lelkes, the anonymous referees, and the editors for helpful comments. The study was preregistered with AsPredicted.org as study #7041. *Address correspondence to James N. Druckman, Northwestern University, Scott Hall, 601 University Place, Evanston, IL 60208, USA; email: druckman@northwestern.edu. Public Opinion Quarterly doi:10.1093/poq/nfz003 D ow naded rom http/academ ic.p.com /poq/advance-articleoi/10.1093/poq/nfz003/5486527 by U niersity of Penylvania Liraries user on 08 M ay 2019 to interact with those from the other party, a tendency known as affective polarization (Iyengar, Sood, and Lelkes 2012). This divisiveness vitiates political trust (Hetherington and Rudolph 2015), hampers interpersonal relations (Huber and Malhotra 2017), and hinders economic exchanges (McConnell et al. 2018). Yet, two significant measurement lacunae remain. First, scholars use a wide-ranging assortment of items to measure affective polarization, but there is little sense of how these items relate to one another: Are they interchangeable? Second, these measures ask respondents to evaluate “the Democratic Party” or “the Republican Party.” But whom do voters imagine when they answer such questions: ordinary voters or elected officials? Addressing these questions with an original survey experiment, we document how the different measures relate to one another, finding that nearly all of them are strongly interrelated. The exception is the social-distance measures, which we argue tap a distinctive aspect of affective polarization. Further, the results show that when people think about the other party, they think primarily about political elites rather than voters. While they dislike both elites and ordinary voters from the other party, they especially dislike the other party’s elites. These findings have important implications for how scholars measure affective polarization, and for our understanding of its underlying dynamic. What Is Affective Polarization, and How Do We Measure It? Affective polarization stems from an individual’s identification with a political party. Identifying with a party divides the world into a liked ingroup (one’s own party), and a disliked outgroup (the opposing party; Tajfel and Turner 1979). This identification gives rise to ingroup favoritism and bias, which is at the heart of affective polarization: the tendency of people identifying as Republicans or Democrats to view opposing partisans negatively and copartisans positively (Iyengar, Sood, and Lelkes 2012, 406; Iyengar and Westwood 2015, 691). Scholars typically measure affective polarization via survey instruments (Iyengar et al. 2019). The most common is a feeling thermometer rating that asks respondents to rate how cold or warm they feel toward the Democratic Party and the Republican Party (Lelkes and Westwood 2017, 489). A second instrument asks respondents to rate how well various traits describe the parties. Positive traits include patriotism, intelligence, honesty, open-mindedness, and generosity; negative traits include hypocrisy, selfishness, and meanness (Iyengar, Sood, and Lelkes 2012; Garrett et al. 2014). A third approach is to ask citizens to rate the extent to which they trust the parties to do what is right (Levendusky 2013). A final set of questions gauge how comfortable people are having close friends from the other party, having neighbors from the other Druckman and Levendusky Page 2 of 9 D ow naded rom http/academ ic.p.com /poq/advance-articleoi/10.1093/poq/nfz003/5486527 by U niersity of Penylvania Liraries user on 08 M ay 2019 party, and having their children marry someone from the other party (Iyengar, Sood, and Lelkes 2012; Levendusky and Malhotra 2016). These items are known as social-distance measures, as they gauge the level of intimacy (distance) individuals are comfortable having with those from the other party. How do these various measures of affective polarization relate to one another? Prior studies provide little insight into this question; most studies include only one or two measures and do not explicitly compare them. Two general types of these measures exist: While thermometers, trait ratings, and trust measures are general attitudes about broad objects (i.e., parties), socialdistance items capture attitudes about particular behavioral outcomes (e.g., your child marrying someone from the other party). These two should be only marginally related, given how “[e]mpirical research has shown repeatedly that the relation between general attitudes and specific behaviors [and related measures] tends to be very low” (Fishbein and Ajzen 2010, 278). A distinct question concerns the targets of all of these measures: When someone rates “the Democratic Party” on a feeling thermometer, or rates whether “Democrats” are selfish, whom are they considering? Is it Democratic voters or elected officials like Nancy Pelosi and Chuck Schumer? As Iyengar and his colleagues (2012, 411) acknowledge, the existing measures are ambiguous on this point: “We will not be able to clarify whether respondents were thinking of partisan voters or party leaders when providing their thermometer scores.” The same is true for any of the other items; if someone says Republicans are untrustworthy, is that their Republican neighbor, or is that an assessment of President Trump? This distinction is not only crucial to understanding what people affectively envision when asked about the “party,” but it also underlines that people might feel differently toward other voters than they do toward elites.

中文翻译：

---

当我们测量情感极化时，我们测量的是什么？

情感两极分化——民主党人和共和党人彼此厌恶和不信任的倾向——已成为美国政治中的一个重要现象。然而，尽管学术界关注这个话题，但仍然存在两个测量缺陷。第一，这个概念的不同度量是如何相互关联的——它们是否可以互换？其次，这些项目都向受访者询问了当事人。当个人回答他们时，他们会想到选民、精英，还是两者兼而有之？我们展示了项目之间的差异，学者应该仔细考虑哪些项目最适合他们的特定研究问题。其次，我们表明，在回答有关另一方的问题时，个人更多地考虑精英而不是选民。更普遍地，个人不喜欢另一方的选民，但他们对另一方的精英怀有更多的敌意。研究报告最后讨论了衡量这一概念和理解其后果的后果。近二十年来，学者们一直在分析选民的问题立场，以确定大众是否实际上是两极分化的（Fiorina 2017）。然而，近年来，人们越来越意识到这并不能完全反映当代美国的党派冲突。无论他们在这些问题上的立场如何，美国人越来越不喜欢、不信任和不想要 James N. Druckman 是佩森 S. Wilder 政治学教授和西北大学政策研究所的教员，埃文斯顿，伊利诺伊州，美国。Matthew S. Levendusky 是政治学系的教授，顺便说一下，在安纳伯格传播学院，以及美国宾夕法尼亚州费城宾夕法尼亚大学安嫩伯格公共政策中心民主制度的杰出研究员。作者感谢安纳伯格公共政策中心资助该项目（MSL，首席研究员）、Sam Gubitz 和 Natalie Sands 提供研究援助，感谢 Joe Biaggio、Shanto Iyengar、Yanna Krupnikov、Yphtach Lelkes、匿名审稿人和编辑提供的帮助注释。该研究已在 AsPredicted.org 上预先注册为研究 #7041。*地址与 James N. Druckman, Northwestern University, Scott Hall, 601 University Place, Evanston, IL 60208, USA 的通信地址；电子邮件：druckman@northwestern.edu。公众意见季刊 doi:10.1093/poq/nfz003 D ow naded rom http/academ ic.p.com /poq/advance-articleoi/10。1093/poq/nfz003/5486527，宾夕法尼亚大学图书馆用户于 2019 年 5 月 8 日与对方进行互动，这种趋势被称为情感极化（Iyengar、Sood 和 Lelkes，2012 年）。这种分裂损害了政治信任（Hetherington 和 Rudolph 2015），阻碍了人际关系（Huber 和 Malhotra 2017），并阻碍了经济交流（McConnell 等，2018）。然而，仍然存在两个重要的测量缺陷。首先，学者们使用各种各样的项目来衡量情感极化，但很少有人了解这些项目之间的关系：它们可以互换吗？其次，这些措施要求受访者评价“民主党”或“共和党”。但选民在回答这些问题时想象一下谁：普通选民还是民选官员？通过原始调查实验解决这些问题，我们记录了不同的度量如何相互关联，发现几乎所有度量都密切相关。例外是社交距离测量，我们认为它挖掘了情感极化的一个独特方面。此外，结果表明，当人们想到另一方时，他们主要想到的是政治精英而不是选民。他们既讨厌对方的精英，也讨厌对方的普通选民，尤其讨厌对方的精英。这些发现对学者如何衡量情感极化以及我们对其潜在动态的理解具有重要意义。什么是情感极化，以及我们如何衡量它？情感两极分化源于个人对政党的认同。认同某个政党会将世界分为喜欢的内群体（自己的政党）和不喜欢的外群体（对方；Tajfel 和 Turner 1979）。这种认同导致了内群体偏袒和偏见，这是情感两极分化的核心：人们倾向于认为共和党人或民主党人消极地看待反对党派而积极地看待同党派（Iyengar, Sood, and Lelkes 2012, 406; Iyengar and Westwood 2015, 691)。学者通常通过调查工具来衡量情感极化（Iyengar 等人，2019 年）。最常见的是感觉温度计评级，要求受访者评价他们对民主党和共和党的感觉有多冷或多暖（Lelkes and Westwood 2017，489）。第二个工具要求受访者对各种特征对政党的描述程度进行评分。积极的特质包括爱国主义、智慧、诚实、思想开放和慷慨；负面特征包括虚伪、自私和卑鄙（Iyengar、Sood 和 Lelkes，2012 年；Garrett 等人，2014 年）。第三种方法是要求公民对他们相信各方做正确事情的程度进行评分（Levendusky 2013）。最后一组问题衡量了人们对另一方的亲密朋友、另一方 Druckman 和 Levendusky 的邻居的舒适度 第 2 页，共 9 页 Dow naded rom http/academ ic.p.com /poq/advance-articleoi/10.1093 /poq/nfz003/5486527 宾夕法尼亚大学图书馆用户在 2019 年 5 月 8 日的聚会上，让他们的孩子嫁给另一方的某人（Iyengar、Sood 和 Lelkes，2012 年；Levendusky 和 ​​Malhotra 2016）。这些项目被称为社交距离措施，因为它们衡量个人与对方的亲密程度（距离）。这些情感极化的各种衡量标准如何相互关联？先前的研究对这个问题几乎没有提供任何见解。大多数研究只包括一两个措施，并没有明确比较它们。这些度量存在两种一般类型：温度计、特质评级和信任度量是对广泛对象（即派对）的一般态度，而社交距离项目则捕捉对特定行为结果的态度（例如，您的孩子与另一方的某人结婚）。这两者应该只是略微相关，鉴于“[e] 实证研究一再表明，一般态度与具体行为 [和相关措施] 之间的关系往往非常低”（Fishbein 和 Ajzen 2010, 278）。一个明显的问题涉及所有这些措施的目标：当有人用温度计评价“民主党”，或者评价“民主党”是否自私时，他们在考虑谁？是民主党选民还是像南希佩洛西和查克舒默这样的民选官员？正如 Iyengar 和他的同事 (2012, 411) 承认的那样，现有的措施在这一点上是模棱两可的：“我们无法澄清受访者在提供温度计分数时是否考虑了党派选民或政党领导人。” 任何其他项目也是如此；如果有人说共和党人不可信，那是他们的共和党邻居，或者这是对特朗普总统的评估？这种区别不仅对于理解人们在被问及“政党”时的情感想象至关重要，而且还强调人们对其他选民的感受可能与对精英的感受不同。