Gender achievement gaps among Chinese middle school students and the role of teachers' gender
中国中学生的性别成绩差距和教师性别的作用

Gender achievement gaps  性别成就差距
Random assignment  随机分配
student-teacher interaction
师生互动
CEPS

Whether student-teacher demographic match could influence student learning processes and educational outcomes has been an increasingly important topic in social science. If they do exist, they could potentially either alleviate or perpetuate educational inequity by affecting the classroom dynamics, while leaving the marginal distribution of education inputs unchanged. One such demographic alignment is student-teacher gender match. There are several channels through which teacher gender may affect students’ learning process and academic performance. The identity-based motivation (IBM) model, in particular, provides context for understanding how identities are dynamically constructed and made salient by situational cues, such as instructor gender (Oyserman, 2007). Based on this theory, learning under the instruction from a same-gender teacher may thus influence how students interpret his or her experience in the classroom (Gilmartin, Denson, Li, Bryant, & Aschbacher, 2007).
师生匹配是否会影响学生的学习过程和教育成果一直是社会科学中一个越来越重要的话题。如果它们确实存在，它们可能会通过影响课堂动态来缓解或延续教育不平等，同时保持教育投入的边际分配不变。其中一种人口统计学一致性是师生性别匹配。教师性别可能通过多种渠道影响学生的学习过程和学习成绩。特别是基于身份的动机 （IBM） 模型，为理解身份如何通过情境线索（如教师性别）动态构建和突出提供背景（Oyserman，2007）。基于这个理论，在同性老师的指导下学习可能会影响学生在课堂上如何解读他或她的经验（Gilmartin， Denson， Li， Bryant， & Aschbacher， 2007）。

In addition to the general benefit of gender congruence in creating a more inviting learning environment, researchers have particularly emphasized the potential of student-teacher gender match in protecting against negative consequences that could arise from “stereotype threat”, defined as the stereotypical linkage between an individual’s membership in a group and lower ability in an area. Such negative gender norms begin to form early in life and may lead to decreased
除了性别一致性在创造更具吸引力的学习环境方面的普遍好处外，研究人员还特别强调了师生性别匹配在防止“刻板印象威胁”可能产生的负面后果方面的潜力，“刻板印象威胁”被定义为个人在群体中的成员身份与某个领域的较低能力之间的刻板印象联系。这种消极的性别规范在生命的早期就开始形成，并可能导致
human capital investment, confidence, and performance (Bian, Leslie, & Cimpian, 2017; Marx & Roman, 2002; McIntyre et al., 2005; Spencer, Steele, & Quinn, 1999). For example, societal stereotypes often associate girls with lower math ability than boys, which may undermine girls’ confidence in math and their interests in math-intensive fields (Bian et al., 2017). Having a female math teacher, however, can mitigate the stereotype threat by presenting a positive female role model. In addition to role model affects, another student-teacher factor that might influence student learning and performance is teacher bias, where teachers may treat girls and boys differently. For example, a girl may perform better in a class taught by a female teacher because the female teacher favors girls during the learning process and in assigning a grade. The possible existence of multiple channels through which the impacts of student-teacher gender match may unfold highlights the importance for discussions about potential impacts of teacher gender to look beyond test score outcomes and include subjective measures of student perceptions of a subject and their perceptions of the learning processes.
人力资本投资、信心和表现（Bian， Leslie， & Cimpian， 2017;马克思与罗曼，2002 年;McIntyre 等人，2005 年;Spencer， Steele， & Quinn， 1999）。例如，社会刻板印象经常将女孩的数学能力低于男孩联系在一起，这可能会削弱女孩对数学的信心和她们对数学密集型领域的兴趣（Bian et al.，2017）。然而，拥有一位女数学老师可以通过树立积极的女性榜样来减轻刻板印象的威胁。除了榜样影响之外，另一个可能影响学生学习和表现的师生因素是教师偏见，教师可能会区别对待女孩和男孩。例如，一个女生在女教师授课的课堂上可能表现得更好，因为女教师在学习过程和分配成绩时偏爱女生。可能存在多种渠道来展开师生性别匹配的影响，这凸显了讨论教师性别潜在影响的重要性，即超越考试成绩结果，包括学生对学科的看法和他们对学习过程的看法的主观测量。

This study presents new quasi-experimental evidence on the effect of student-teacher gender match on both scholastic achievement and subjective perceptions of a subject by exploiting a unique setting where middle school students in China are randomly assigned to teachers in each of the three main mandatory subject areas: Chinese, English, and math. Using a nationally representative data, China Education Panel
本研究通过利用一个独特的环境，将中国的中学生随机分配到三个主要必修学科领域的教师，即语文、英语和数学，为师生性别匹配对学业成绩和对学科的主观认知的影响提供了新的准实验证据。使用具有全国代表性的数据，中国教育面板

Survey (CEPS), this paper addresses two research questions: (1) Are there gender disparities in academic achievement, self-perceived ability, and interactions with teachers among China’s middle school students in each of the main subjects? (2) What is the effect of teachers’ gender on student learning outcomes, beliefs, and perceived interactions with the subject teacher?
调查 （CEPS） 中，本文解决了两个研究问题：（1） 中国中学生在每个主要科目的学业成绩、自我感知能力和与教师的互动方面是否存在性别差异？（2） 教师的性别对学生的学习成果、信念以及与学科教师的感知互动有什么影响？

Our first goal is to document gender disparities among middle school students in China based on three sets of measures: (a) academic achievement in the three main subjects in middle schools: Chinese, English, and math; (b) self-perceived ability and self-perceived importance of a subject; and (3) self-perceived interactions with the subject teacher teaching each subject. Our results suggest that girls outperform boys in all three subjects, although the gaps are much more pronounced in Chinese and English than in math. More importantly, these gender performance gaps remain throughout the middle school years and the gaps in Chinese and English become larger in later years. The gender gaps in subjective measures generally echo the performance gaps in favor of girls, except for math: despite higher average scores in math, girls tend to report substantially lower self-perceived ability of learning math, and lower levels of perceived teacher praise and questions in math classes. These descriptive patterns align with the societal stereotypes documented in the existing literature that girls are often associated with lower math ability than boys, which may not only influence girls’ confidence in their mathematical ability, but may also influence teachers’ expectations and interactions with students as a result of negative gender norms.
我们的第一个目标是根据三组衡量标准记录中国中学生的性别差异：（a） 中学三门主要科目的学业成绩：语文、英语和数学;（b） 受试者的自我感知能力和自我感知的重要性;（3） 与教授每个科目的学科教师的自我感知互动。我们的结果表明，女生在所有三门科目中的表现都优于男生，尽管中文和英文的差距比数学的差距要明显得多。更重要的是，这些性别表现差距在整个中学阶段仍然存在，而中英文的差距在以后的几年里会变得更大。主观测量中的性别差距通常与有利于女孩的表现差距相呼应，但数学除外：尽管数学平均分较高，但女孩往往报告说自我感知的数学学习能力要低得多，数学课上感知到的老师表扬和问题水平也较低。这些描述性模式与现有文献中记录的社会刻板印象一致，即女孩的数学能力通常低于男孩，这不仅可能影响女孩对自己数学能力的信心，还可能影响教师的期望和与学生的互动由于负面的性别规范。

Given the gender disparity in performance and beliefs, our second goal is to estimate the role of teachers’ gender in influencing these gaps between boys and girls. Among the three subjects examined, we only identify robust and persistent benefits of having a female teacher for girls in math. We also find suggestive evidence that girls receive more praises in a math class taught by a female teacher. In contrast, we find little evidence of any consistent effect of student-teacher gender match on girls in literacy subjects; nor do we find any impact of teacher gender on boys. These results suggest that gender-specific benefits through student-teacher gender congruence may work best in reducing stereotype threat, while the benefit is less salient in subjects that do not have prevalent negative stereotype.
鉴于表现和信念方面的性别差异，我们的第二个目标是估计教师的性别在影响男孩和女孩之间的这些差距方面的作用。在研究的三个科目中，我们只确定了为女子数学教师配备女性教师的强大而持久的好处。我们还发现了一些暗示性的证据，表明女生在女教师教授的数学课上受到的表扬更多。相比之下，我们几乎没有发现证据表明师生性别匹配对识字科目中的女生有任何一致的影响;我们也没有发现教师性别对男孩有任何影响。这些结果表明，通过学生-教师性别一致性实现的性别特异性益处可能在减少刻板印象威胁方面效果最好，而在没有普遍负面刻板印象的科目中，这种益处则不那么明显。

It is important to note that if teacher gender indeed impacts a middle school student’s performance and motivation, then these influences occur at a critical juncture in the life-cycle: An extensive literature indicates that academic choices and career aspirations based on individual aptitudes, self-concept, and values are formulated during early adolescence (Eccles, Vida, & Barber, 2004; Wang & Degol, 2013). More importantly, the nine-year compulsory education ends at grade 9 in China and middle school graduates are then required to choose between a high school that may eventually lead to post-secondary education, or a vocational school that is oriented towards obtaining occu-pation-specific skills. Accordingly, student academic performance, educational aspiration, and attitude toward different subjects formulated in middle school will have a substantial influence on their subsequent decisions of academic paths which, in turn, would also influence an individual’s occupational choice and labor market outcomes.
值得注意的是，如果教师性别确实影响了中学生的表现和积极性，那么这些影响就会发生在生命周期的关键时刻：大量的文献表明，基于个人才能、自我概念和价值观的学术选择和职业抱负是在早期青春期形成的（Eccles， Vida， & Barber， 2004;Wang & Degol，2013 年）。更重要的是，在中国，九年义务教育在 9 年级结束，初中毕业生需要在最终可能获得高等教育的高中或以获得职业特定技能的职业学校之间做出选择。因此，学生在中学时期形成的学习成绩、教育抱负和对不同科目的态度将对他们随后的学术道路决定产生重大影响，这反过来也会影响个人的职业选择和劳动力市场结果。

Taken together, the results from this study make three distinct contributions to the existing literature on gender achievement gaps and the impacts of student-teacher gender congruence on these gaps. First, we add to the small but growing literature on the gender achievement gaps in developing countries. A large volume of studies examining the gender achievement gaps among middle school students in the U.S. and many other Western countries generally found a gender disparity in favor of girls obtaining higher grades throughout elementary, middle, and high schools (Card & Lemieux, 2000; Cho, 2007; Cole, 1997; Duckworth & Seligman, 2006; Fortin, Oreopoulos, & Phipps, 2015; Goldin, Katz, & Kuziemko, 2006; Pomerantz, Altermatt, & Saxon, 2002). Yet, it is not until recently that researchers started attending to this issue in developing countries.
综上所述，本研究的结果对关于性别成就差距的现有文献以及师生性别一致性对这些差距的影响做出了三个不同的贡献。首先，我们补充了关于发展中国家性别成就差距的少量但不断增长的文献。大量研究调查了美国和许多其他西方国家的中学生的性别成就差距，普遍发现，在小学、初中和高中，女孩获得更高的成绩存在性别差异（Card & Lemieux，2000;Cho， 2007;Cole， 1997;Duckworth & Seligman， 2006;Fortin， Oreopoulos， & Phipps， 2015;Goldin， Katz， & Kuziemko， 2006;Pomerantz， Altermatt， & Saxon， 2002）。然而，直到最近，研究人员才开始在发展中国家关注这个问题。

Using data from middle school students in Beijing, China’s capital city, Lai (2010) found a more pronounced female dominance than in the U.S. studies. This might be partly due to China’s greater emphasis on discipline in teacher-dominated classrooms. An extensive literature from psychology suggests that girls tend to be more self-disciplined than boys, which give girls the “edge” (Cole, 1986; Davis, 1995; Duckworth & Seligman, 2006; Saarni, 1984). A test-oriented education system like China that requires a high level of self-discipline and cooperative attitude may put boys in an especially disadvantaged situation. However, the data set used in Lai (2010)'s study only includes one cohort of students in one central administrative district in Beijing and therefore has limited external validity. We examine whether the gender disparities documented by Lai are also observed with a nationally representative data.
使用中国首都北京中学生的数据，Lai （2010） 发现女性主导地位比美国研究更明显。这可能部分是由于中国在教师主导的课堂上更加强调纪律。来自心理学的大量文献表明，女孩往往比男孩更自律，这给了女孩“优势”（Cole，1986 年;Davis，1995 年;Duckworth & Seligman， 2006;Saarni， 1984）。像中国这样以考试为导向的教育体系，要求高度的自律和合作的态度，可能会使男孩处于特别不利的境地。然而，Lai （2010） 研究中使用的数据集仅包括北京一个中央行政区的一组学生，因此外部效度有限。我们研究了 Lai 记录的性别差异是否也通过具有全国代表性的数据来观察。

Second, our study adds to the literature that has mainly focused on objective academic indicators by extending to subjective measures of beliefs about one’s ability in different subject areas. A number of studies have explored student-teacher gender interaction on student course performance at both the K-12 level (e.g. Ammermüller & Dolton, 2006; Dee, 2007; Lim & Meer, 2017) and at the postsecondary education level (e.g. Bettinger & Long, 2005; Carrell, Page, & West, 2010; Fairlie, Hoffmann, & Oreopoulos, 2014; Hoffmann & Oreopoulos, 2009; Nixon & Robinson, 1999). Overall, these studies found small positive effects for having a same-gender teacher in regard to academic indicators of success. However, while academic measures are important, a rich body of literature also shows that motivational believes, such as self-perceived ability in different domains of knowledge and perceived importance of these domains of knowledge, are stronger predictors of educational and occupational choice compared to an individual’s academic achievement (e.g. Joyce & Farenga, 2000; Wang, Eccles, & Kenny, 2013). The CEPS not only includes student test scores in each subject, but also directly collected information on student self-perceived ability in each subject and perceptions of how important a subject is for their future. Based on a subset of the CEPS data, Gong, Lu, and Song (2018) found that having a female teacher raises girls’ test scores averaged across three subject areas and improves their mental status. Our paper builds on their study but extends it in two ways. First, the CEPS recently released a second wave of data, which allows us to examine how the impacts of teacher gender on student outcomes evolve over time. Additionally, instead of averaging across the three main subject, we estimate the impacts of teacher gender on test scores and subject measures in each of the three subjects separately and identify noticeable between-subject differences.
其次，我们的研究通过扩展到对一个人在不同学科领域的能力的主观信念的测量，增加了主要关注客观学术指标的文献。许多研究探讨了学生-教师性别互动对 K-12 级别学生课程表现的影响（例如 Ammermüller 和 Dolton，2006 年;Dee，2007 年;Lim & Meer， 2017） 和高等教育阶段 （例如 Bettinger & Long， 2005;Carrell， Page， & West， 2010;Fairlie， Hoffmann， & Oreopoulos， 2014;Hoffmann & Oreopoulos， 2009;Nixon & Robinson， 1999）。总体而言，这些研究发现，在学术成功指标方面，拥有同性教师的积极影响很小。然而，尽管学术衡量标准很重要，但丰富的文献也表明，动机信念，如在不同知识领域的自我感知能力和这些知识领域的重要性，比起个人的学术成就，更强地预测了教育和职业选择（例如 Joyce & Farenga，2000;Wang， Eccles， & Kenny， 2013）。CEPS 不仅包括学生在每个科目的考试成绩，还直接收集有关学生在每个科目中的自我感知能力以及对某个科目对他们未来有多重要的看法的信息。根据 CEPS 数据的子集，Gong、Lu 和 Song （2018） 发现，拥有女教师可以提高女生在三个学科领域的平均考试成绩，并改善她们的精神状态。我们的论文以他们的研究为基础，但以两种方式对其进行扩展。首先，CEPS 最近发布了第二波数据，这使我们能够研究教师性别对学生成绩的影响如何随着时间的推移而变化。 此外，我们不是对三个主要科目进行平均，而是分别估计教师性别对三个科目中每个科目的考试成绩和科目测量的影响，并确定科目间的明显差异。

Third, our results also contribute to the emerging literature on the operating channels of student-teacher gender matching. Although a number of studies have estimated the impacts of student-teacher gender matching on academic performance, there is less evidence on the mechanisms driving the positive findings on test scores. In particular, researchers argued between student-based channels where students change their behaviors and beliefs as a result of student-teacher gender congruence, versus teacher-based channels where teachers may provide differential attention and responses to students based on a student’s gender. Our dataset sheds some light on this issue by allowing us to directly examine students’ perceived interactions with the subject teacher. Specifically, each student was asked to indicate self-perceived frequency of praise from each subject teacher, as well as frequency of being asked to answer a question in class. Since K-12 classes in China are typically tea-cher-centered lectures with limited opportunity for student group activities, being called out by a teacher to answer a question is not only a key way of student-teacher interaction but also a main venue for students to be intellectually engaged in class. Direct measures of student-teacher interactions could therefore provide insight on how gender behavioral differences may intertwine with social forces to either perpetuate or alleviate gender gaps in a particular subject.
第三，我们的结果也有助于关于师生性别匹配作渠道的新兴文献。尽管许多研究已经估计了师生性别匹配对学业成绩的影响，但关于推动考试成绩积极结果的机制的证据较少。特别是，研究人员在以学生为基础的渠道之间争论，学生由于学生与教师的性别一致性而改变他们的行为和信念，而以教师为基础的渠道则教师可能会根据学生的性别对学生提供不同的关注和回应。我们的数据集允许我们直接检查学生与学科教师的感知互动，从而对这个问题有所了解。具体来说，每个学生都被要求表明每个学科老师自我感知的表扬频率，以及在课堂上被要求回答问题的频率。由于中国的 K-12 课程通常是以茶会为中心的讲座，学生小组活动的机会有限，因此被老师叫来回答问题不仅是师生互动的重要方式，也是学生在课堂上进行智力参与的主要场所。因此，对学生与教师互动的直接测量可以提供关于性别行为差异如何与社会力量交织在一起以延续或缓解特定学科的性别差距的见解。

The remainder of the paper unfolds as follows: Section 2 describes the CEPS data, and Section 3 discusses the randomization checks and
本文的其余部分展开如下：第 2 节描述了 CEPS 数据，第 3 节讨论了随机化检查和
statistical methods. In Section 4 we present our main results, and Section 5 concludes.
统计方法。在第 4 节中，我们介绍了我们的主要结果，第 5 节总结了这一结果。

The China Education Panel Survey (CEPS) is the first nationally representative survey for middle school students in China. The primary analyses of this study are based on the baseline wave of this survey, which was collected during the 2013-2014 academic year. ${}^1$${}^1$^(1){ }^{1} The CEPS employs a stratified, multistage sampling scheme. In the first stage, 28 counties/districts were chosen from 2870 counties/districts. Four schools from each county/district were selected and two classrooms were then randomly selected from 7th grade and another two from 9th grade in each sample school. ${}^2$${}^2$^(2){ }^{2} Finally, all students from these selected classrooms were surveyed, resulting in a nationally representative sample of approximately 20,000 students in 438 classrooms of 112 schools.
中国教育跟踪调查 （CEPS） 是针对中国第一个具有全国代表性的中学生调查。本研究的主要分析基于本调查的基线波次，该调查是在 2013-2014 学年收集的。 ${}^1$${}^1$^(1){ }^{1} CEPS 采用分层、多阶段抽样方案。第一阶段，从 2870 个县/区中选出 28 个县/区。从每个县/区选择四所学校，然后从每个样本学校的 7 年级随机选择两间教室，从 9 年级随机选择另外两间教室。 ${}^2$${}^2$^(2){ }^{2} 最后，对这些选定教室的所有学生进行了调查，得出了具有全国代表性的样本，即 20,000 所学校的 438 个教室中的约 112 名学生。

The CEPS administered five separate questionnaires to the (i) sample students, (ii) homeroom teachers, (iii) subject teachers for the three main subjects (Chinese, English, and math), (iv) their parents, and (v) their school administrators. Hao and Yu (2015) provide a comprehensive summary of CEPS in their recent report; here we briefly review the features of the baseline survey most relevant for our analysis. The school administrator questionnaire solicited information about school recourse, school management, and other school-level statistics about teachers and students. One important question asked in school administrator questionnaire is whether the school randomly assigns students to different classrooms. Among all the 112 schools sampled, 83% ( $N=93$$N=93$N=93N=93 ) reported that students were randomly assigned to classrooms upon entry into the middle school; the homeroom teacher and main subject teachers were then randomly assigned to each class. Both students and teachers remain in the same class throughout the three middle school years. In the methodology section, we present statistical evidence to show that students in these schools indeed seem to be randomly assigned to teachers in each subject area, as indicated by preexisting student demographic and family characteristics.
CEPS 向 （i） 样本学生、（ii） 班主任、（iii） 三个主要科目（中文、英文和数学）的学科教师、（iv） 他们的家长和 （v） 他们的学校管理人员进行了五份单独的问卷调查。Hao 和 Yu （2015） 在他们最近的报告中提供了对 CEPS 的全面总结;在这里，我们简要回顾了与我们的分析最相关的基线调查的特点。学校管理人员问卷征集了有关学校追索权、学校管理以及有关教师和学生的其他学校级统计数据的信息。学校管理者问卷中提出的一个重要问题是学校是否将学生随机分配到不同的教室。在所有抽样的 112 所学校中，83% （ $N=93$$N=93$N=93N=93 ） 报告说，学生在进入中学时被随机分配到教室;然后将班主任和主要科目教师随机分配到每个班级。在三个中学阶段，学生和老师都住在同一个班级。在方法部分，我们提供了统计证据，表明这些学校的学生似乎确实被随机分配到每个学科领域的教师，正如先前存在的学生人口统计和家庭特征所表明的那样。

The student questionnaire collected information about student demographic and family background characteristics, and general experience at school. In addition, a set of five survey questions explicitly asked about students’ experience, attitude, and performance related to a specific main subject, for Chinese, English, and math respectively. These five questions ( 15 items in total) can be divided into four categories: academic performance in a subject, student perceived ability of learning a subject, student perception of how important a subject is for their future, and student perception of their engagement in class. The teacher questionnaire collected information from each homeroom and main subject teacher on teachers’ demographic characteristics, subject taught, teaching experience, and other information with regard to their teaching approach and interactions with parents. In the following section, we explain in more detail the outcome and control variables used in this study and how they are constructed.
学生问卷收集了有关学生人口统计和家庭背景特征以及在学校的一般体验的信息。此外，一组五个调查问题明确询问了学生与特定主要科目相关的经验、态度和表现，分别针对中文、英文和数学。这五个问题（共 15 题）可以分为四类：一门学科的学习成绩、学生对学习一门学科的感知能力、学生对一门学科对他们未来有多重要的看法，以及学生对他们在课堂上的参与度的看法。教师问卷收集了每个班主任和主要科目教师关于教师的人口统计特征、所教科目、教学经验以及有关他们的教学方法和与家长互动的其他信息。在下一节中，我们将更详细地解释本研究中使用的结果和控制变量以及它们是如何构建的。

Academic performance: Students are required to report their most recent mid-term test score for each of the three main subjects. Since mid-term and final exams are administered by each school, student test scores are comparable within a particular grade at each school. The raw score is on a 150 -point scale and we standardize all of our outcome measures including both test and subjective outcomes at the school-by-grade-by-subject level for the regression analysis.
学习成绩：学生需要报告三个主要科目中每个科目的最新期中考试成绩。由于期中考试和期末考试由每所学校管理，因此学生的考试成绩在每所学校的特定年级内具有可比性。原始分数为 150 分制，我们将所有结果测量标准化，包括逐个学校、逐学科级别的测试和主观结果，以进行回归分析。

Self-perceived ability of learning a subject: students were asked how difficult a particular subject is for them on a 4-point Likert scale ( $1=$$1=$1=1= very difficult; $4=$$4=$4=4= not difficult at all). Therefore higher values indicate higher self-perceived ability in learning a subject.
学习一门学科的自我感知能力：学生被问及特定科目对他们来说有多难，采用 4 点李克特量表（ $1=$$1=$1=1= 非常困难; $4=$$4=$4=4= 一点也不难）。因此，较高的值表示学习学科的自我感知能力较高。

Perceived importance of a subject for their future: For each subject, students were asked to respond to the statement “Chinese/English/ Mathematics is important for my future”. Students were asked to indicate their response on a 4-point Likert scale, with 1 indicating strongly disagree, and 4 strongly agree. Higher values therefore indicate higher perceived importance of a subject.
感知到某门学科对他们未来的重要性：对于每个科目，学生被要求回答“中文/英文/数学对我的未来很重要”这句话。学生被要求在 4 点李克特量表上表明他们的回答，其中 1 表示非常不同意，4 表示非常同意。因此，较高的值表示主题的感知重要性较高。

Perceived interactions with subject teachers: Students were asked to respond to the following two statements for each of the three main subjects: “my Chinese/English/Mathematics teacher always praises me” and “my Chinese/English/Mathematics teacher always asks me to answer questions in class”. Students responded to both questions on a 4point Likert scale ranging from strongly disagree to strongly agree. Higher values therefore indicate a higher level of perceived interactions with the teacher in class.
与学科教师的感知互动：学生被要求针对三个主要科目中的每一个回答以下两个陈述：“我的中文/英文/数学老师总是表扬我”和“我的中文/英文/数学老师总是让我在课堂上回答问题”。学生们以 4 点李克特量表回答这两个问题，范围从非常不同意到非常同意。因此，较高的值表示课堂上与教师的感知互动水平较高。

Student and teacher control variables: Control variables include information collected at the student-level, teacher-level, and classroomlevel. Specifically, student background characteristics include gender, age, parents’ education level, family wealth, ${}^3$${}^3$^(3){ }^{3} and potential family issues (such as parental absence, whether the father drinks regularly and the frequency of fight between parents). Summary statistics of student background characteristics is presented in Appendix Table A.
学生和教师控制变量：控制变量包括在学生级别、教师级别和课堂级别收集的信息。具体来说，学生背景特征包括性别、年龄、父母的教育水平、家庭财富和 ${}^3$${}^3$^(3){ }^{3} 潜在的家庭问题（例如父母缺席、父亲是否经常喝酒以及父母之间争吵的频率）。学生背景特征的汇总统计数据见附录表 A。

Teacher background information is collected through the teacher survey, which includes information on their demographic characteristics (such as gender and age), whether graduated from a normal institution versus a comprehensive university, whether having a teaching certificate, highest level of education, prior teaching award, ${}^4$${}^4$^(4){ }^{4} career rank, and years of teaching experience. Summary statistics of teacher background characteristics are presented in Appendix Table B. Finally, since all the students in a sample class are surveyed, we are also able to calculate the classroom size based on headcount and include it as a control variable in the analysis.
教师背景信息是通过教师调查收集的，其中包括有关他们的人口统计特征（例如性别和年龄）、是否毕业于师范机构还是综合大学、是否拥有教学证书、最高教育水平、先前的教学奖项、 ${}^4$${}^4$^(4){ }^{4} 职业排名和教学经验年限。教师背景特征的汇总统计数据见附录表 B。最后，由于样本班级中的所有学生都接受了调查，我们还能够根据人数计算教室规模，并将其作为控制变量包含在分析中。

As noted above, our analytic sample only includes students from schools that randomly assigned students into classrooms. Since students reported their academic performance and perceptions for each of the three subjects, our analysis is conducted for each subject separately. For the purpose of this analysis, we then exclude observations that have missing values on student gender, teacher gender or the outcome variable examined. A small percentage of students are also missing information on other covariates and we include indicators for missing
如上所述，我们的分析样本仅包括将学生随机分配到教室的学校的学生。由于学生报告了他们对这三个科目中每个科目的学习成绩和看法，因此我们对每个科目的分析是分别进行的。出于此分析的目的，我们然后排除了在学生性别、教师性别或检查的结果变量上具有缺失值的观察值。一小部分学生也缺少有关其他协变量的信息，并且我们包括了缺失的指标