View in EDS HTML Full Text PDF Full Text

Peers, Study Effort, and Academic Performance in College Education: Evidence from Randomly Assigned Roommates in a Flipped Classroom

Saved in:
Bibliographic Details
Title: Peers, Study Effort, and Academic Performance in College Education: Evidence from Randomly Assigned Roommates in a Flipped Classroom
Language: English
Authors: Pu, Shi, Yan, Yu, Zhang, Liang
Source: Research in Higher Education. Mar 2020 61(2):248-269.
Availability: Springer. Available from: Springer Nature. 233 Spring Street, New York, NY 10013. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-348-4505; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/
Peer Reviewed: Y
Page Count: 22
Publication Date: 2020
Document Type: Journal Articles
Reports - Research
Education Level: Higher Education
Postsecondary Education
Descriptors: Foreign Countries, College Students, Dormitories, Peer Influence, Academic Achievement, Study Habits, Blended Learning, Homework, Video Technology, Educational Technology
Geographic Terms: China
DOI: 10.1007/s11162-019-09571-x
ISSN: 0361-0365
Abstract: This study provides empirical evidence related to two critical assumptions in the student engagement literature in higher education--namely, the malleability of study effort and the causality of the relationship between study effort and student outcomes. We merged student-level administrative data on dormitory assignments, study effort in a hybrid course, and course performance from a regional college in China. Our results indicate that study efforts are more similar for a pair of randomly assigned roommates than for a pair of randomly assigned non-roommates, indicating the malleability of study effort. In addition, we provide evidence on the causal relationship between study effort and academic performance by using roommates' study efforts as an instrumental variable when estimating the effect of a student's study effort on course performance.
Abstractor: As Provided
Entry Date: 2020
Accession Number: EJ1243986
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwEMqMMqeeFotd2BBWcKwltZAAAA4jCB3wYJKoZIhvcNAQcGoIHRMIHOAgEAMIHIBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDM6rRriaXUlX6HMgHwIBEICBmlwHq5CVmM5axWikRcBD7OvvwLL2xnbOwlncdw9SzldQywNNuVWrYDY2WdljTVZqLMS6AEHk2PDpLBJwRfRqSc9mCO7f6rqNrUe0KVdbw8AsCjlSf6kRTd8GuYmtW3I9dCpQUCHPRvaau2Y6cC2O1iWa4bg-y0Fyk5JgJgwimOo-_Xv8eRMOOCRNWCZGUg64qLQ9BqN0XOn1s3M=
Text:
  Availability: 1
  Value: <anid>AN0141772480;rhe01mar.20;2020Feb19.02:32;v2.2.500</anid> <title id="AN0141772480-1">Peers, Study Effort, and Academic Performance in College Education: Evidence from Randomly Assigned Roommates in a Flipped Classroom </title> <p>This study provides empirical evidence related to two critical assumptions in the student engagement literature in higher education—namely, the malleability of study effort and the causality of the relationship between study effort and student outcomes. We merged student-level administrative data on dormitory assignments, study effort in a hybrid course, and course performance from a regional college in China. Our results indicate that study efforts are more similar for a pair of randomly assigned roommates than for a pair of randomly assigned non-roommates, indicating the malleability of study effort. In addition, we provide evidence on the causal relationship between study effort and academic performance by using roommates' study efforts as an instrumental variable when estimating the effect of a student's study effort on course performance.</p> <p>Keywords: Peers; Study Effort; Academic performance</p> <hd id="AN0141772480-2">Introduction</hd> <p>Student learning and development is a defining feature of college education. A substantial amount of research has been Queryconducted on factors that could potentially contribute to improving student learning in higher education. Among all factors, student involvement (Astin [<reflink idref="bib2" id="ref1">2</reflink>], [<reflink idref="bib3" id="ref2">3</reflink>]) and quality of effort (Pace [<reflink idref="bib33" id="ref3">33</reflink>]) have received extensive attention in the higher education literature. At the risk of being too general, the main finding in this important body of literature is that students derive more benefits from college education when they put more effort into their studies (Kuh et al. [<reflink idref="bib21" id="ref4">21</reflink>]; Webber et al. [<reflink idref="bib43" id="ref5">43</reflink>]). While the positive relationship between study effort and college success (i.e., academic performance, college persistence, and graduation) is supported by decades of research, policy implications critically hinge on two potentially crippling questions: Is study effort malleable? Moreover, is the positive relationship between study effort and college success causal?</p> <p>In this study, we attempt to answer these two questions by investigating peer effects in college education. Specifically, we examine the effects of randomly assigned roommates on study effort and further investigate the effects of study effort on academic performance. We integrate student-level information from several data systems of a regional college in China. The Housing Management System provides information on roommate assignments. Students in this college are randomly assigned to two peer groups (i.e., dormitory roommates and social group-mates) based on a few known observable characteristics. The Course Management System captures students' actions in a "flipped course" (an instructional strategy characterized by self-paced online learning with traditional lecture time used to discuss challenges and solve problems), thereby furnishing reliable information on their study efforts (e.g., number of course materials downloaded, number of videos watched, etc.). Finally, the Student Transcript Data System maintains detailed data on student performance in this course that differentiates points earned for content competence from points earned for class participation. In this study, we attempt to provide evidence for the malleability of study effort by comparing similarity in the level of study effort of a pair of randomly assigned roommates with that of non-roommates. Furthermore, we adopt an instrumental variable approach to assess the effect of a student's own study effort on learning outcomes.</p> <p>This study fills a critical void in the literature on both student involvement and peer effects in higher education and makes important contributions on several fronts. First, while there is ample evidence of a positive association between study effort and academic performance, our goal is to demonstrate that study effort is malleable and the relationship between effort and academic performance is likely to be causal. Second, studies of peer effects in higher education have used academic outcomes such as GPAs as the outcome variable; however, study effort could be an important channel through which peer effects on academic performance occur. In what is, to our knowledge, the only other research aimed at examining peer effects on study effort, Stinebrickner and Stinebrickner ([<reflink idref="bib40" id="ref6">40</reflink>]) assessed students' effort through subjective surveys sent via campus mail. In the current study, we use objective data from a college's course management system that provides detailed and concrete information about study effort, including the number of times a student viewed course material and participated in forum discussions. Finally, we account for the fact that final grades in college-level courses could be partially based on class participation, resulting in an inherent, positive association between effort and academic performance. Detailed transcript data enable us to separate between points earned for class participation and points earned for content competence. Using an instrumental variable approach, we show that study effort has positive effects on both class participation and content competence.</p> <hd id="AN0141772480-3">Literature Review</hd> <p>This study is informed by two important lines of research: student involvement and peer effects in higher education. Substantively, this study contributes to the literature on student involvement and effort by examining whether student involvement and effort in a course can be influenced by peers and whether study effort leads to better academic performance in the course. College student involvement (Astin [<reflink idref="bib2" id="ref7">2</reflink>], [<reflink idref="bib3" id="ref8">3</reflink>]) and quality of effort (Pace [<reflink idref="bib33" id="ref9">33</reflink>]) are among the most researched topics in higher education, which is not surprising because student involvement and effort is viewed as an important factor that can directly or indirectly improve student learning (Coates [<reflink idref="bib12" id="ref10">12</reflink>]; Krause and Coates [<reflink idref="bib19" id="ref11">19</reflink>]; Kuh [<reflink idref="bib20" id="ref12">20</reflink>]; Trowler [<reflink idref="bib41" id="ref13">41</reflink>]). In other words, students derive more benefits from college education when they are more involved and put more effort into their studies. The bulk of evidence supporting this proposition has been derived from college student surveys, including the National Survey of Student Engagement (NSSE) in the United States and Canada, and the Australia Survey of Student Engagement (AUSSE) in Australia and New Zealand. For example, results of studies based on NSSE data indicate that student engagement and effort in a variety of educationally purposeful curricular and co-curricular activities (e.g., first-year seminars, learning communities) is positively related to academic outcomes such as GPAs and college persistence (Hu and Kuh [<reflink idref="bib18" id="ref14">18</reflink>]; Kuh et al. [<reflink idref="bib21" id="ref15">21</reflink>]; Webber et al. [<reflink idref="bib43" id="ref16">43</reflink>]).</p> <p>While this body of literature has established robust and positive correlations between student involvement and college success, some lingering issues remain unresolved. First, college student surveys such as the NSSE typically rely on self-reported data, which raise potential concerns about measurement validity (Bowman [<reflink idref="bib8" id="ref17">8</reflink>]; Porter [<reflink idref="bib34" id="ref18">34</reflink>]). Second, although student involvement varies greatly across participants in college student surveys, it is not clear whether variables such as student involvement and study effort are malleable. The malleability of student involvement and effort has been implicitly assumed, but seldom tested in studies that have examined how these factors influence college outcomes. Third, because studies based on college student surveys typically employ cross-sectional designs and examine variations in student involvement and college success at the individual level, it is unclear whether the observed positive correlations are causal due to unobserved individual heterogeneity.</p> <p>Methodologically, this study builds on the recent literature on peer effects that has provided credible empirical evidence of peer effects in higher education (e.g., Booij et al. [<reflink idref="bib6" id="ref19">6</reflink>]; Carrell et al. [<reflink idref="bib10" id="ref20">10</reflink>]; Sacerdote [<reflink idref="bib36" id="ref21">36</reflink>]; Zimmerman [<reflink idref="bib45" id="ref22">45</reflink>]). Multiple mechanisms proposed in the literature suggest that peers could have significant influence on study effort. The "bad apple" model (Hoxby and Weingarth [<reflink idref="bib17" id="ref23">17</reflink>]), for example, suggests that peers' bad habits might distract a student from putting more effort into learning. Stinebrickner and Stinebrickner ([<reflink idref="bib40" id="ref24">40</reflink>]) found that when a roommate brings video games to college, a student is likely to spend less time studying. Conversely, the "shining light" model (Hoxby and Weingarth [<reflink idref="bib17" id="ref25">17</reflink>]) implies that hardworking peers might serve as role models that inspire a student to put more effort into studying. Stinebrickner and Stinebrickner ([<reflink idref="bib39" id="ref26">39</reflink>]) contended that one important way roommates could affect students' first-year achievement was by inspiring them to spend more time studying.</p> <p>This study follows the peer effect literature on peer effects in college education by utilizing random roommate assignment. Researchers have identified two primary confounders that impede the identification of peer effects: selection bias and common shocks (Manski [<reflink idref="bib28" id="ref27">28</reflink>]). Natural peer networks typically are formed based on selection variables which might not be captured by covariates, yet may have a direct impact on student outcomes. In other words, if peer group formation is not random, students with similar characteristics (e.g., hardworking and high-performing) may self-select into the same room (or group), creating a spurious relationship between peer's characteristics and own outcomes. In studies on peer effects in higher education, researchers have used exogenously (vs. naturally) formed peer groups to study peer effects in multiple educational settings, including conditionally randomly assigned college roommates (e.g., Han and Li [<reflink idref="bib16" id="ref28">16</reflink>]; Sacerdote [<reflink idref="bib36" id="ref29">36</reflink>]; Zimmerman [<reflink idref="bib45" id="ref30">45</reflink>]) and randomly assigned social groups in military academies (Lyle [<reflink idref="bib26" id="ref31">26</reflink>], [<reflink idref="bib27" id="ref32">27</reflink>]).</p> <p>Common shocks refer to environmental factors that affect all members of a peer network (Nye et al. [<reflink idref="bib31" id="ref33">31</reflink>]). For example, an effective teacher increases learning for all students in a class. Students who live in the same dormitory building are affected by the environment of the building. In these two examples, students have similar experiences due to common shocks, which may lead to spurious peer effects. Compared to selection bias, common shocks tend to be less problematic. In cases where common shocks are present, most studies use a fixed effects approach (e.g., Bayer et al. [<reflink idref="bib4" id="ref34">4</reflink>]; Bettinger et al. [<reflink idref="bib5" id="ref35">5</reflink>]; Nanda and Sorensen [<reflink idref="bib30" id="ref36">30</reflink>]) or the instrumental variables method (De Giorgi et al. [<reflink idref="bib13" id="ref37">13</reflink>]) to tackle the problem.</p> <p>Although the existing literature offers abundant guidance on the identification of peer effects, evidence of peer effects on study effort is scant. Using subjective survey data to study peer effects could be potentially problematic because the presence of peers would influence individuals' survey answers (Gardner and Steinberg [<reflink idref="bib14" id="ref38">14</reflink>]). To our knowledge, Stinebrickner and Stinebrickner ([<reflink idref="bib40" id="ref39">40</reflink>]) conducted the only study to offer evidence based on objective data. Using housing data from Berea College and a detailed survey of students' time diaries, the authors found that students whose randomly assigned roommates brought video games to the dorm at the beginning of freshman year studied roughly 0.7 h (male roommate) and 0.5 h (female roommate) less every day. In addition, roommates' study time in high school positively impacted students' use of time during the first year of college.</p> <hd id="AN0141772480-4">Data, Sample, Variables, and Methods</hd> <p></p> <hd id="AN0141772480-5">Data and Sample</hd> <p>We obtained student-level administrative data from a 4-year regional college located in the southwest region of China, which offered 23 majors and had a first-year enrollment of 2250 in the Fall 2014 semester. We began with this freshman cohort and excluded students who: (a) did not persist through the first semester; (b) did not take the course that we used in this study; and (c) lived with previously known friends. Only five students in the initial sample dropped out before the end of the first semester, and 136 students did not take the online course in the Spring 2015 semester for unknown reasons. Because study effort data for these students are not available, they are not included in the final sample. The last exclusion filtered out 216 students who probably gamed the random room assignment process and managed to live with previously known friends. These students may have had unobserved similarities with their roommates, thereby imposing an upward bias on the estimated peer effects. The roommate assignment and randomness checks discussed in the next section detail how these students were identified and the consequences of excluding them from the final sample. Our final analytical sample included 1893 students. Table 1 compares the descriptive statistics for students in the initial sample, the sample with effort data, and the final sample.</p> <p>Descriptive statistics across samples</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" rowspan="2"><p>Variable</p></th><th align="left" colspan="2"><p>Initial sample</p></th><th align="left" colspan="2"><p>Effort sample</p></th><th align="left" colspan="2"><p>Final sample</p></th><th align="left" rowspan="2"><p><italic>t</italic> Test statistics</p></th></tr><tr><th align="left"><p>Mean</p></th><th align="left"><p>S.D.</p></th><th align="left"><p>Mean</p></th><th align="left"><p>S.D.</p></th><th align="left"><p>Mean</p></th><th align="left"><p>S.D.</p></th></tr></thead><tbody><tr><td align="left"><p>% Female students</p></td><td align="left"><p>0.706</p></td><td align="left"><p>(0.456)</p></td><td align="left"><p>0.709</p></td><td align="left"><p>(0.454)</p></td><td align="left"><p>0.703</p></td><td align="left"><p>(0.457)</p></td><td char="." align="char"><p>− 0.645</p></td></tr><tr><td align="left"><p>% Fresh high school graduates</p></td><td align="left"><p>0.244</p></td><td align="left"><p>(0.430)</p></td><td align="left"><p>0.241</p></td><td align="left"><p>(0.428)</p></td><td align="left"><p>0.251</p></td><td align="left"><p>(0.433)</p></td><td char="." align="char"><p>0.913</p></td></tr><tr><td align="left"><p>% from home province</p></td><td align="left"><p>0.769</p></td><td align="left"><p>(0.422)</p></td><td align="left"><p>0.755</p></td><td align="left"><p>(0.430)</p></td><td align="left"><p>0.747</p></td><td align="left"><p>(0.435)</p></td><td char="." align="char"><p>− 0.842</p></td></tr><tr><td align="left"><p>% Racial minority</p></td><td align="left"><p>0.031</p></td><td align="left"><p>(0.173)</p></td><td align="left"><p>0.032</p></td><td align="left"><p>(0.175)</p></td><td align="left"><p>0.033</p></td><td align="left"><p>(0.179)</p></td><td char="." align="char"><p>0.363</p></td></tr><tr><td align="left"><p>% Registered online</p></td><td align="left"><p>0.376</p></td><td align="left"><p>(0.484)</p></td><td align="left"><p>0.373</p></td><td align="left"><p>(0.484)</p></td><td align="left"><p>0.343</p></td><td align="left"><p>(0.475)</p></td><td char="." align="char"><p>− 2.729**</p></td></tr><tr><td align="left"><p>% Science track in high school</p></td><td align="left"><p>0.430</p></td><td align="left"><p>(0.495)</p></td><td align="left"><p>0.450</p></td><td align="left"><p>(0.498)</p></td><td align="left"><p>0.460</p></td><td align="left"><p>(0.499)</p></td><td char="." align="char"><p>0.887</p></td></tr><tr><td align="left"><p># Same major roommates</p></td><td align="left"><p>1.157</p></td><td align="left"><p>(1.136)</p></td><td align="left"><p>1.217</p></td><td align="left"><p>(1.141)</p></td><td align="left"><p>1.193</p></td><td align="left"><p>(1.130)</p></td><td char="." align="char"><p>− 0.917</p></td></tr><tr><td align="left"><p>Dormitory room size</p></td><td align="left"><p>3.991</p></td><td align="left"><p>(0.123)</p></td><td align="left"><p>3.991</p></td><td align="left"><p>(0.125)</p></td><td align="left"><p>3.990</p></td><td align="left"><p>(0.132)</p></td><td char="." align="char"><p>− 0.340</p></td></tr><tr><td align="left"><p>CEE percentile</p></td><td align="left"><p>0.800</p></td><td align="left"><p>(0.110)</p></td><td align="left"><p>0.801</p></td><td align="left"><p>(0.110)</p></td><td align="left"><p>0.798</p></td><td align="left"><p>(0.112)</p></td><td char="." align="char"><p>− 0.898</p></td></tr><tr><td align="left"><p># Times course materials accessed</p></td><td align="left" /><td align="left" /><td align="left"><p>26.305</p></td><td align="left"><p>(25.925)</p></td><td align="left"><p>25.968</p></td><td align="left"><p>(25.470)</p></td><td char="." align="char"><p>− 0.577</p></td></tr><tr><td align="left"><p># Times documents downloaded</p></td><td align="left" /><td align="left" /><td align="left"><p>28.440</p></td><td align="left"><p>(28.516)</p></td><td align="left"><p>28.466</p></td><td align="left"><p>(28.650)</p></td><td char="." align="char"><p>0.041</p></td></tr><tr><td align="left"><p># Times completely watched a course video</p></td><td align="left" /><td align="left" /><td align="left"><p>3.360</p></td><td align="left"><p>(4.651)</p></td><td align="left"><p>3.279</p></td><td align="left"><p>(4.571)</p></td><td char="." align="char"><p>− 0.770</p></td></tr><tr><td align="left"><p># Forum posts replied</p></td><td align="left" /><td align="left" /><td align="left"><p>27.192</p></td><td align="left"><p>(32.855)</p></td><td align="left"><p>27.086</p></td><td align="left"><p>(32.403)</p></td><td char="." align="char"><p>− 0.142</p></td></tr><tr><td align="left"><p># Forum posts reviewed</p></td><td align="left" /><td align="left" /><td align="left"><p>66.462</p></td><td align="left"><p>(68.852)</p></td><td align="left"><p>66.144</p></td><td align="left"><p>(67.568)</p></td><td char="." align="char"><p>− 0.205</p></td></tr><tr><td align="left"><p># Forum posts initiated</p></td><td align="left" /><td align="left" /><td align="left"><p>2.310</p></td><td align="left"><p>(4.939)</p></td><td align="left"><p>2.244</p></td><td align="left"><p>(4.746)</p></td><td char="." align="char"><p>− 0.601</p></td></tr><tr><td align="left"><p># Forum posts liked</p></td><td align="left" /><td align="left" /><td align="left"><p>9.565</p></td><td align="left"><p>(42.691)</p></td><td align="left"><p>9.225</p></td><td align="left"><p>(40.717)</p></td><td char="." align="char"><p>− 0.364</p></td></tr><tr><td align="left"><p>Sample size</p></td><td align="left" colspan="2"><p>2250</p></td><td align="left" colspan="2"><p>2019</p></td><td align="left" colspan="2"><p>1893</p></td><td align="left" /></tr></tbody></table> </ephtml> </p> <p>The "Effort Sample" excludes students who have missing data on study effort. The "Final Sample" further removes rooms with roommates from the same high school. <emph>t</emph>-test statistics compare means between the effort sample and final sample; **<emph>p</emph> < 0.01</p> <p>Because the research question concerns the effects of roommates on study effort, the unit of analysis becomes pairs of students instead of individual students. Therefore, we transformed the final sample of 1893 students into 38,548 pairs of students who were in the same social groups. A social group, typically with 40–60 students, is an administrative unit in the college, comprised of some adjacent rooms in male residential halls and some adjacent rooms in female residential halls. Consequently, roommates were always members of the same social group. Figure 1 shows the hierarchical structure of cohorts, social groups, and dormitory rooms. Each social group was managed by an academic advisor and a student affairs officer who were responsible for students' academic affairs and social well-being. Almost all college organized extracurricular activities centered around social groups.</p> <p>Graph: Fig. 1 Relationship between cohort, social group, and dormitory room</p> <hd id="AN0141772480-6">Roommate Assignment and Randomness Checks</hd> <p>Because random room assignment is the key to the identification of peer effects in this study, we conducted extensive interviews with the Director of Residential Life at the college to gain a thorough understanding of the room assignment process. We learned that the room assignment process followed three simple rules with minor twists. First, male and female students were assigned to different residential halls. Second, except for two special programs (accounting and joint degree programs), college major did not play a role in determining room assignment. Students from these two special majors were assigned to a block of reserved rooms. Third, students who were not enrolled in the two special programs had the option to register online (37.60% of students) or onsite (62.40% of students). The online registration was available for a period of 20 days immediately prior to the onsite registration day. Students who registered online were able to select one of eight random rooms provided by the Housing Management System. The onsite registration was open for a period of 3 days right before class started for the academic year. Students who registered onsite did not have the option to choose their rooms. Instead, available beds were queued (not randomly), and each student was assigned to the next available bed when they registered onsite.</p> <p>For the Fall 2014 cohort, the housing assignment assigned 2250 students to 565 quadruple rooms: 559 with four students, three with three students, two with two students, and one with a single student. It is important to note that once students are assigned into rooms, they remain in their rooms during their entire first year. According to the Director of Residence Life, only in "extremely rare" cases, the college would allow room changes. In other words, the room arrangement that we observe at the end of their first year should be very close to their original room assignment. Another potential issue is attrition. The Director of Residential Life informed us that the attrition rate during the first academic year in this college is extremely low. For example, only five students dropped out before the end of the first semester. And most students dropped out due to personal reasons (e.g., health and family issues, starting up companies). More importantly, when students drop out, their beds will not be re-assigned during the academic year.</p> <p>To summarize, the dormitory room assignment process in this college was only conditionally random: students with the same registration type (i.e., online vs. onsite), similar onsite registration times, the same gender, or enrolled in accounting or joint degree programs were more likely to be assigned to the same dormitory rooms by the system. Hence in theory, after controlling for these five factors, the student distribution to dormitory rooms should have been random. However, after extensive communications with the Director of Residence Life, we discovered two loopholes in the system. First, because students who registered online were given eight random rooms to choose from, friends could potentially select the same room when choices provided by the system overlapped. Second, for those who registered onsite, room assignment was highly dependent on the time of registration. Since friends were more likely to register together, they were more likely to end up in the same room. If friends shared similar characteristics, the clustering of similar students in dormitory rooms would bias our estimates.</p> <p>The possibility of some students selecting their roommates by exploiting these loopholes in the system seems to be supported by the data. Table 2 reports correlations of some observable characteristics between roommates after controlling for the five known factors affecting housing assignments. The first column reveals students with similar scores on the College Entrance Exam (CEE) (equivalent to the SAT in the United States) were more likely to become roommates, suggesting that at least a small portion of students likely gamed the system to select their roommates. Since rooms were assigned quickly both online and onsite, students did not have many opportunities to build social connections. It is therefore reasonable to assume that most sorting did not happen among new acquaintances during the registration process, but among students with well-established social connections. Therefore, if pre-college social connections could be identified with the available data, the selection problem could be solved by excluding the self-selected roommates from the sample.</p> <p>Correlations between observable characteristics of students and roommates</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left"><p>Characteristic</p></th><th align="left"><p>Effort sample</p></th><th align="left"><p>Dropped sample</p></th><th align="left"><p>Final sample</p></th></tr></thead><tbody><tr><td align="left"><p>CEE percentile</p></td><td align="left"><p>0.0489* (0.0227)</p></td><td align="left"><p>0.220** (0.0759)</p></td><td align="left"><p>0.0284 (0.0240)</p></td></tr><tr><td align="left"><p>Fresh high school graduate</p></td><td align="left"><p>0.0185 (0.0355)</p></td><td align="left"><p>0.349*** (0.0987)</p></td><td align="left"><p>− 0.0215 (0.0380)</p></td></tr><tr><td align="left"><p>Science track in high school</p></td><td align="left"><p>0.0631 (0.0351)</p></td><td align="left"><p>0.304** (0.105)</p></td><td align="left"><p>0.0252 (0.0374)</p></td></tr><tr><td align="left"><p>Racial minority</p></td><td align="left"><p>− 0.0193 (0.0359)</p></td><td align="left"><p>− 0.0872 (0.123)</p></td><td align="left"><p>− 0.0194 (0.0378)</p></td></tr><tr><td align="left"><p>Attended urban high school</p></td><td align="left"><p>0.0611 (0.0342)</p></td><td align="left"><p>0.186 (0.103)</p></td><td align="left"><p>0.0421 (0.0363)</p></td></tr><tr><td align="left"><p>Control registration sequence</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Gender</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>ACCA class</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>International class</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Observations</p></td><td align="left"><p>2109</p></td><td align="left"><p>216</p></td><td align="left"><p>1893</p></td></tr></tbody></table> </ephtml> </p> <p>Standard errors in parentheses; *<emph>p</emph> < 0.05; **<emph>p</emph> < 0.01; ***<emph>p</emph> < 0.001</p> <p>The most likely pre-college social connections would have been established among students from the same high school. To check this possibility, we tested whether students from the same high school were randomly assigned to rooms. Specifically, we performed a Monte Carlo simulation with 3000 repetitions to simulate a completely random process of assigning same-gender students to rooms and generate a range of the possible number of rooms with multiple students from same high school under random assignment. Simulation results indicate that the average number of rooms is approximately eight, with a maximum of 20. However, in the real data, 55 rooms had multiple students from same high school, which is highly unlikely under random assignment, suggesting that students from the same high school were indeed more likely to be assigned to the same dormitory room.</p> <p>Removing the 216 students living in these 55 rooms effectively eliminated any significant correlation between roommates. As shown in the third column of Table 2, the randomness check shows no significant correlation among roommates' average characteristics after eliminating the 55 rooms. As a double check, we analyzed the randomness of the dropped sample to test the similarities among these self-selected students. As expected, we found multiple significant correlations between roommates' observable characteristics in the dropped sample: roommates were likely to have similar CEE percentiles, to be fresh high school graduates, to have completed the same high school academic track, and to share the same initial major. Consequently, we removed these students from our analytical sample.</p> <hd id="AN0141772480-7">The Online Course</hd> <p>The online course, entitled "Introduction to Computers," was a compulsory general education course for all freshmen at the college. This general education course focused on providing students with hands-on experience with commonly used Microsoft Office software and introducing basic computer knowledge from the functionality of the CPU to the binary coding system. The course lasted for 18 weeks, was taught every spring semester and was taken by nearly all freshmen (e.g., in 2014, 93.73% of freshmen completed this course). The timing of this course, i.e., the second semester of the first academic year, may potentially introduce an extraneous variable of maturation. One type of maturation that all students experience in their first semester does not affect the results of our study; however, maturation could be influenced by peers. For example, roommates may have developed similar study habits during their first semester together. In other words, had the class been offered in the first semester, the magnitude of the peer effect could be different. While this does not diminish the value of the study, it is important to recognize that peer effects may be time dependent.</p> <p>"Introduction to Computers" was a flipped hybrid course, in which knowledge was transferred via self-paced learning on an online platform, and traditional lecture time was used to discuss challenges and solve problems. Students watched online videos to learn course content, downloaded and uploaded course exercises, and participated in an online forum; they also meet with their teachers face-to-face every other week in a classroom. Although it was a hybrid course, the majority of coursework was completed online. According to the course syllabus, students were expected to spend 3 h every week studying course materials and doing homework on the online platform. By contrast, the face-to-face meeting that was held every other week lasted for just 2 h.</p> <p>The face-to-face component was delivered in medium sized sections, each comprised of 40–50 students. More often than not, these meetings focused on the most frequent errors in the uploaded exercises or questions in the online forum, as well as some key concepts in the course materials. Students spent the majority of class time discussing challenges and questions with other students in groups. Because all course materials are provided online, their activities recorded in the online course system, although maybe inadequate to capture all aspects of student learning, can meaningfully reflect their levels of study effort,. For example, students may download (which is recorded in the system) but not read courses materials (which is not observed), or they may play but not watch course videos. We recognize this as a limitation of studies that rely on online indicators to capture student learning activities.</p> <p>Freshmen were assigned to the class sessions by the Office of Academic Affairs at the beginning of the spring semester. Although the school did not provide us with the exact algorithm, the Director of Residence Life stated that social groups play an important role in session assignment. The heatmap in Fig. 2 shows a strong correlation between students' social group and their class sessions. Students from the same social group are more likely in the same class session than students from different social groups. Even for the most heterogeneous class session, 21.6% of students are from the same social group. Eleven class sessions have more than 50% of students come from the same social group. Because social groups are formed by nearby dormitory rooms, the formation of class sessions should also be close to random after including the same set of random controls (buildings, registration method, and registration sequence). We conducted additional random checks within class sessions. Results in Table 3 show no significant correlation between students' observable characteristics after controlling the random controls.</p> <p>Graph: Fig. 2 Cross-tabulate teaching sessions and social groups. Note The ith row and jth column of the heatmap reflect the percentage of students in teaching session i are from social group j</p> <p>Correlations between observable characteristics of students and classmates</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" /><th align="left"><p>CEE percentile</p></th><th align="left"><p>Fresh high school graduate</p></th><th align="left"><p>Science track in high school</p></th><th align="left"><p>Racial minority</p></th><th align="left"><p>Attended urban high school</p></th></tr></thead><tbody><tr><td align="left"><p>Classmates'</p><p>Average CEE Percentile</p></td><td align="left"><p>− 0.0787 (0.117)</p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p>Classmates'</p><p>% Fresh High School Graduate</p></td><td align="left" /><td align="left"><p>0.0285 (0.153)</p></td><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p>Classmates'</p><p>% Science Track in High School</p></td><td align="left" /><td align="left" /><td align="left"><p>0.0141 (0.160)</p></td><td align="left" /><td align="left" /></tr><tr><td align="left"><p>Classmates'</p><p>% Racial Minority</p></td><td align="left" /><td align="left" /><td align="left" /><td align="left"><p>− 0.129 (0.177)</p></td><td align="left" /></tr><tr><td align="left"><p>Classmates'</p><p>% Attended urban high school</p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /><td align="left"><p>0.155 (0.107)</p></td></tr><tr><td align="left"><p>Random controls</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Observations</p></td><td align="left"><p>1791</p></td><td align="left"><p>1832</p></td><td align="left"><p>1832</p></td><td align="left"><p>1832</p></td><td align="left"><p>1832</p></td></tr><tr><td align="left"><p>R-squared</p></td><td align="left"><p>0.012</p></td><td align="left"><p>0.018</p></td><td align="left"><p>0.012</p></td><td align="left"><p>0.003</p></td><td align="left"><p>0.104</p></td></tr></tbody></table> </ephtml> </p> <p>Standard errors in parentheses; *<emph>p</emph> < 0.05; **<emph>p</emph> < 0.01; ***<emph>p</emph> < 0.001</p> <p>The 40 class sections were taught by five lecturers who used a standard syllabus, course content, homework, and exams. To the extent that course experience could vary across these five lecturers, we included instructor fixed effects to account for their potential differences. Unfortunately, for the face-to-face component of the class, there is no information about students' study effort. We recognize this as a potential limitation, although the unavailability of in-class study effort does not affect the data analysis and results of this study.</p> <hd id="AN0141772480-8">Grades</hd> <p>Students' grades in this course were determined by three activities: class engagement (25%), homework (30%), and an in-class final exam (45%). Each student's class engagement grade was based on in-class and online course participation. For example, each complete review of a course video in the course management system counted toward the final grade, and an absence from class decreased a student's engagement score. For homework, students needed to complete and submit six assignments online, roughly one homework assignment every 3 weeks. Students took the final exam during the last class session, when they were given 100 min to independently complete an exam that assessed whether they had mastered basic computer skills. Course grades reflected not only students' learning outcomes but also their effort levels, which could inherently correlate with study effort. Fortunately, the Student Transcript Data System recorded the overall scores and final exam scores separately. The difference between the two (i.e., overall score minus final in-class exam score) reflected each student's class engagement and homework performance. These detailed scores enabled us to differentiate academic performance based on content competence and class participation. We used these two scores as separate dependent variables in our analysis.</p> <hd id="AN0141772480-9">Study Effort</hd> <p>The course management system recorded seven types of activities: (a) the total number of course materials downloaded, (b) the total number of course videos viewed, (c) the total number of slides viewed, (d) the total number of forum posts viewed, (e) the total number of forum posts replied to, (f) the total number of forum posts liked, and (g) the total number of forum posts initiated. All seven items were positively skewed, so we log-transformed values to remedy this problem; transformed items are internally consistent (alpha = 0.817, n = 1893).</p> <p>We then constructed an effort index, <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>y</mi><mi>i</mi></msub></math> </ephtml> , for each student by standardizing and averaging the seven transformed effort items. The predictive validity of this composite is moderate: the correlation between the effort index and overall course grade is 0.247. This correlation coefficient is slightly higher than the predictive validity of the NSSE for GPAs (Gordon et al. [<reflink idref="bib15" id="ref40">15</reflink>]) and comparable to other academic engagement constructs (e.g., Simons-Morton and Chen [<reflink idref="bib37" id="ref41">37</reflink>]). In addition to this overall effort index, we developed two mutually exclusive sub-indices for individual learning (the first three activities) and forum participation (the last four activities).</p> <p>To measure the similarity of study effort between a pair of students, we constructed the following index:</p> <olist> <item> <ephtml> <math display="block" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><msub><mi>S</mi><mrow><mi>i</mi><mtext>j</mtext></mrow></msub><mo>=</mo><mn>1</mn><mo>-</mo><mfrac><mrow><mi>a</mi><mi>b</mi><mi>s</mi><mfenced close=")" open="("><mrow><msub><mi>y</mi><mi>i</mi></msub><mo>-</mo><msub><mi>y</mi><mi>j</mi></msub></mrow></mfenced></mrow><mrow><mtext>max</mtext><mfenced close=")" open="("><mrow><msub><mi>y</mi><mi>m</mi></msub><mo>-</mo><msub><mi>y</mi><mi>n</mi></msub></mrow></mfenced></mrow></mfrac></mrow></math> </ephtml> </item> </olist> <p>Graph</p> <p>For a pair of students, effort similarity, <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>S</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> , equals 1 minus the absolute effort difference between students <emph>i</emph> and <emph>j</emph> divided by the maximum effort difference between any pair of students. This simple linear transformation scales the observed difference between any two students to an index with a range of 0–1 where the value increases as the effort difference decreases. A similarity index value of 1 is assigned to a pair of students with an identical level of effort, while an index value of 0 is assigned to the pair of students with the largest difference in effort. In our final sample, the mean of similarity index is 0.784 with a standard deviation of 0.165.</p> <p>The advantage of the similarity index is twofold. First, this index standardizes and aggregates different aspects and measures of study effort. While this can also be achieved by usual standardization, the similarity index has an important methodological advantage. In regression models where own study effort is predicted by roommates' study effort, there is a well-known reflection problem (Manski [<reflink idref="bib28" id="ref42">28</reflink>]). In other words, roommates affect each other, which causes serious upward bias in the estimated peer effect. The similarity index assesses and compares the final effort level between a pair of roommates, effectively circumventing the reflection problem.</p> <hd id="AN0141772480-10">Other Variables</hd> <p>In addition to course grades and effort data, we extracted a host of demographic, socioeconomic, and academic variables that have been shown in the literature to be associated with study effort and academic performance. Demographic variables are important covariates for the relationship between any given college experience and academic performance (e.g., Arum and Roksa [<reflink idref="bib1" id="ref43">1</reflink>]; Mayhew et al. [<reflink idref="bib29" id="ref44">29</reflink>]; Kuh et al. [<reflink idref="bib21" id="ref45">21</reflink>], [<reflink idref="bib22" id="ref46">22</reflink>]). In this study we include sex, age, and minority status. Socioeconomic status has been shown to be associated with both college experience and academic performance (e.g., Arum and Roksa [<reflink idref="bib1" id="ref47">1</reflink>]; Brint et al. [<reflink idref="bib9" id="ref48">9</reflink>]; Kuh et al. [<reflink idref="bib21" id="ref49">21</reflink>]). While individual information on socioeconomic status is not available in our data, we use home province and high school location (urban or rural) to account for socio-economic differences across provinces and between urban and rural areas. Academic variables including precollege academic achievement are significant predictors for academic performance in college (e.g., Arum and Roksa [<reflink idref="bib1" id="ref50">1</reflink>]; Brint et al. [<reflink idref="bib9" id="ref51">9</reflink>]; Loes et al. [<reflink idref="bib24" id="ref52">24</reflink>]). In this study, academic variables include high school track (science or humanities), students' CEE percentiles, first semester college GPAs, and percentage of shared credits with roommates. Finally, we use two sets of environmental fixed effects—social group and instructor for the "Introduction to Computers" course—to control for common shocks that might affect roommates in the same social group or taught by the same instructor.</p> <hd id="AN0141772480-11">Methods</hd> <p>The first research question focused on whether two students are more likely to exhibit similar effort levels if they live in the same dormitory room. We used the following baseline model to estimate the difference in effort similarity between roommates and social group mates:</p> <p>2 <ephtml> <math display="block" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><msub><mi>S</mi><mrow><mi>i</mi><mtext>j</mtext></mrow></msub><mo>=</mo><msub><mi>α</mi><mi>s</mi></msub><mo>+</mo><msub><mi>β</mi><mn>0</mn></msub><msub><mi>R</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><mi>θ</mi><msub><mi>C</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><mi>δ</mi><msub><mi>X</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><mi>γ</mi><msub><mi>Z</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><msub><mi>ε</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></mrow></math> </ephtml></p> <p>Graph</p> <p>where <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>α</mi><mi>s</mi></msub></math> </ephtml> represents the average effort similarity within each social group, <emph>s</emph>. <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>R</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> is a dummy variable that equals 1 only when students <emph>i</emph> and <emph>j</emph> are peers, <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>C</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> measures common shocks for both students (e.g., the same instructor for the hybrid course), <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>X</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> represents pair characteristics of students <emph>i</emph> and <emph>j</emph> (e.g., both are female), and <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>Z</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> represents conditional random assignment factors for students <emph>i</emph> and <emph>j</emph> (e.g., registration type and registration sequence). The independent variable <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>R</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> is the main interest in this study because its coefficient measures peer effects on study effort. The direction and significance of peer effects are indicated by <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>β</mi><mn>0</mn></msub></math> </ephtml> : if being peers significantly increases effort similarity between students <emph>i</emph> and <emph>j</emph>, then peer effects on effort must be positive and significant. Bayer et al. ([<reflink idref="bib4" id="ref53">4</reflink>]) used a similar model to examine peer effects in the workplace.</p> <p>The first goal of this study is to establish a causal link between peer status, <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>β</mi><mn>0</mn></msub></math> </ephtml> , and students' effort similarity, <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>S</mi><mrow><mi>i</mi><mtext>j</mtext></mrow></msub></math> </ephtml> . This entails disentangling selection effects and common shocks from peer effects. Conceptually, selection effects reflect the phenomenon whereby individuals select peers based on mutual similarities which usually contribute to similar performances. Common shocks represent environmental variables that not only affect all peers, but also explain their similar performances. A classic example of a common shock is a teacher's effect on students' academic performance (Nye et al. [<reflink idref="bib31" id="ref54">31</reflink>]). Finally, peer effects reflect a complex phenomenon stemming from social interactions such as the effects of peer teaching (Loke and Chow [<reflink idref="bib25" id="ref55">25</reflink>]), peer pressure (Borsari and Carey [<reflink idref="bib7" id="ref56">7</reflink>]), social norms (Unger et al. [<reflink idref="bib42" id="ref57">42</reflink>]), and shared information (Ryan [<reflink idref="bib35" id="ref58">35</reflink>]). In economics, peer effects typically are estimated by eliminating selection bias and common shocks.</p> <p>To illustrate how selection effects and common shocks might confound peer effects in a technical manner, we break down <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>ε</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> in Eq. (<reflink idref="bib2" id="ref59">2</reflink>) into three parts:</p> <p>3 <ephtml> <math display="block" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><msub><mi>ε</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>=</mo><msub><mi>u</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><msub><mi>v</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><msub><mi>w</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></mrow></math> </ephtml></p> <p>Graph</p> <p>where <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>u</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> captures the unobserved individual similarities between <emph>i</emph> and <emph>j</emph> that do not change over time, <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>v</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> captures the common shocks experienced by students <emph>i</emph> and <emph>j</emph> after college enrollment, and <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>w</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> is the true noise. In the presence of selection, the peer status of students <emph>i</emph> and <emph>j</emph> may depend on their unobserved individual similarities <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>u</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> . In that case, not controlling <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>u</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> in the model would result in omitted variable bias. Fortunately, the peer group formation in this study is randomly conditioned on <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>Z</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> in Eq. (<reflink idref="bib2" id="ref60">2</reflink>). Therefore, unobserved similarities <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>u</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> are not correlated with peer status <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>R</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> . In addition to selection bias, common shocks, <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>v</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> , may be correlated with the peer group status of students <emph>i</emph> and <emph>j</emph>. For example, Table 4 shows that a student and his or her roommates are more likely to be taught by the same instructor than by different instructors; however, this does not suggest roommates are sorted into the same teachers. As discussed earlier, because social groups are considered when students are sorted into class sessions, roommates are more likely to share the same class session, hence the same teacher. Since an instructor has a significant influence on students' academic engagement level (Skinner and Belmont [<reflink idref="bib38" id="ref61">38</reflink>]), excluding instructors from the model would create a bias on estimated peer effect, <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>β</mi><mn>0</mn></msub></math> </ephtml> . To remedy this problem, we controlled for detailed social and academic activities that affected students <emph>i</emph> and <emph>j</emph>, including social group fixed effects and course instructor fixed effects.</p> <p>Correlations between Students' and roommates' instructors, linear probability model</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" /><th align="left"><p>Taught by instructor 1</p></th><th align="left"><p>Taught by instructor 2</p></th><th align="left"><p>Taught by instructor 3</p></th><th align="left"><p>Taught by instructor 4</p></th><th align="left"><p>Taught by instructor 5</p></th></tr></thead><tbody><tr><td align="left"><p># Peers taught by instructor 1</p></td><td align="left"><p>0.201*** (0.00938)</p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p># Peers taught by instructor 2</p></td><td align="left" /><td align="left"><p>0.0640*** (0.0128)</p></td><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p># Peers taught by instructor 3</p></td><td align="left" /><td align="left" /><td align="left"><p>0.184*** (0.0102)</p></td><td align="left" /><td align="left" /></tr><tr><td align="left"><p># Peers taught by instructor 4</p></td><td align="left" /><td align="left" /><td align="left" /><td align="left"><p>0.144*** (0.0113)</p></td><td align="left" /></tr><tr><td align="left"><p># Peers taught by instructor 5</p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /><td align="left"><p>0.0720*** (0.0128)</p></td></tr><tr><td align="left"><p>Random controls</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Observations</p></td><td align="left"><p>1876</p></td><td align="left"><p>1876</p></td><td align="left"><p>1876</p></td><td align="left"><p>1876</p></td><td align="left"><p>1876</p></td></tr><tr><td align="left"><p>R-squared</p></td><td align="left"><p>0.260</p></td><td align="left"><p>0.087</p></td><td align="left"><p>0.177</p></td><td align="left"><p>0.114</p></td><td align="left"><p>0.085</p></td></tr></tbody></table> </ephtml> </p> <p>Standard errors in parentheses; ***<emph>p</emph> < 0.001</p> <p>An alternate way to test the existence of peer effects on study effort is to use multinomial logistic models to check whether being roommates increases the likelihood that two students exert similar levels of effort. Following this path, we first divided students into three categories—high, medium, and low effort—based on their observed online effort relative to the cohort. We categorized high effort as greater than the 75th percentile of the sample effort distribution, medium effort as between the 25th and 75th percentiles; and low effort as below the 25th percentile. We estimated the following multinomial logistic regression equation to measure the size of peer effects:</p> <p>4 <ephtml> <math display="block" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mo>ln</mo><mfenced close=")" open="("><mfrac><mrow><mi>P</mi><mfenced close=")" open="("><mrow><msub><mi>Y</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>=</mo><mi>r</mi></mrow></mfenced></mrow><mrow><mi>P</mi><mfenced close=")" open="("><mrow><msub><mi>Y</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>=</mo><mn>0</mn></mrow></mfenced></mrow></mfrac></mfenced><mo>=</mo><msub><mi>α</mi><mi>s</mi></msub><mo>+</mo><msub><mi>β</mi><mi>r</mi></msub><msub><mi>R</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><mi>θ</mi><msub><mi>C</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><mi>δ</mi><msub><mi>X</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><mi>γ</mi><msub><mi>Z</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></mrow></math> </ephtml></p> <p>Graph</p> <p>where <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>Y</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> categorizes student pairs into four groups: both high effort, both medium effort, both low effort, or different levels of effort (e.g., one high effort and the other medium effort). Students who exerted different levels of effort are used as the reference group. <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>R</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> is again the variable of interest, and other independent variables have the same meaning as in Eq. (<reflink idref="bib2" id="ref62">2</reflink>). <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>β</mi><mi>r</mi></msub></math> </ephtml> measures the size of peer effects and is hypothetically larger than 0.</p> <p>The second goal of this study is to estimate the causal effect of student effort on academic performance. We hypothesized that a student's academic performance in this class is affected by study effort as well as other individual and course level factors:</p> <p>5 <ephtml> <math display="block" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><msub><mi>Y</mi><mi>i</mi></msub><mo>=</mo><mi>α</mi><mo>+</mo><mi>β</mi><msub><mi>E</mi><mi>i</mi></msub><mo>+</mo><mi>γ</mi><msub><mi>X</mi><mi>i</mi></msub><mo>+</mo><msub><mi>π</mi><mi>t</mi></msub><mo>+</mo><msub><mi>G</mi><mi>i</mi></msub><mo>+</mo><msub><mi>Z</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub><mo>+</mo><msub><mi>ε</mi><mi>i</mi></msub></mrow></math> </ephtml></p> <p>Graph</p> <p>where <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>E</mi><mi>i</mi></msub></math> </ephtml> is the effort level and <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>X</mi><mi>i</mi></msub></math> </ephtml> includes a rich set of individual controls (e.g., age, high school track, home province, CEE percentile, and minority status). In addition, we included fixed effects for instructors ( <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>π</mi><mi>t</mi></msub></math> </ephtml> ), social group ( <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>G</mi><mi>i</mi></msub></math> </ephtml> ) and covariates that determined dormitory room assignments ( <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>Z</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> ).</p> <p>Note that Eq. (<reflink idref="bib5" id="ref63">5</reflink>) does not address the possibility that students with high ability might systematically exert more effort. To resolve this, we used the average of roommates' effort levels <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>E</mi><mrow><mo>-</mo><mi>i</mi></mrow></msub></math> </ephtml> as an instrumental variable for <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>E</mi><mi>i</mi></msub></math> </ephtml> . This instrumental variable is highly credible for two reasons. First, as we show in the first part of our analysis, roommates' effort levels are significantly correlated. Because of the conditional random assignment used in this study, this correlation is likely to be causal, i.e., the average of roommates' effort levels <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>E</mi><mrow><mo>-</mo><mi>i</mi></mrow></msub></math> </ephtml> affects a student' own study effort <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>E</mi><mi>i</mi></msub></math> </ephtml> . Second, there is no reason to believe that roommates' effort levels would <emph>directly</emph> affect a student's score on an in-class exam. As a result, in the two-stage least squares (2SLS) approach, the source of identification used in the second stage is variation in students' own study effort that results from differences in their roommates' study effort in the first stage. This IV strategy is especially credible since teacher fixed effects ( <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>π</mi><mi>t</mi></msub></math> </ephtml> ), social group fixed effects ( <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>G</mi><mi>i</mi></msub></math> </ephtml> ), and room assignment rules ( <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>Z</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> ) are included in the first stage equation. The two fixed effects account for the potential common shocks that might affect both a student's and peers' efforts and the student's course grade. For example, assuming that a good teacher not only elicits students' interest in learning but also effectively conveys knowledge, then the correlation between a student's academic performance and peers' efforts may be the result of sharing a good teacher. The teacher and group fixed effects address issues associated with common shocks. Furthermore, random controls account for the possibility that students with similar abilities are assigned to the same dormitory room. For example, if high ability students sort into peer groups and systematically exert more effort, the correlation between peers' efforts and a student's academic performance may simply reflect unobserved similarities in abilities. The study's setting precludes this possibility with the inclusion of random controls <ephtml> <math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>Z</mi><mrow><mi mathvariant="italic">ij</mi></mrow></msub></math> </ephtml> .</p> <p>We estimated Eq. (<reflink idref="bib2" id="ref64">2</reflink>) using OLS, Eq. (<reflink idref="bib4" id="ref65">4</reflink>) using maximum likelihood values, and the IV values in Eq. (<reflink idref="bib5" id="ref66">5</reflink>) using 2SLS. Since peer group formation was conditionally random and we controlled for post-enrollment common shocks, the OLS estimator of peer effects is likely to be unbiased. However, as Bayer et al. ([<reflink idref="bib4" id="ref67">4</reflink>]) illustrated, errors in Eqs. (<reflink idref="bib2" id="ref68">2</reflink>) and (<reflink idref="bib4" id="ref69">4</reflink>) are clustered, thus OLS and maximum likelihood approaches likely underestimate the standard errors. To remedy this, we used bootstrapping techniques to calculate the proper standard errors for the models.</p> <hd id="AN0141772480-12">Results</hd> <p>Table 5 illustrates the results of the baseline model in Eq. (<reflink idref="bib2" id="ref70">2</reflink>), which explains effort similarity for a pair of students based on peer status and a set of controls. We gradually added in controls from left to right, so the decrease of peer effect size can be observed after controlling for selection bias and common shocks. The first column presents the simple regression which explains effort patterns based on peer status alone. Without other controls, being roommates significantly increases students' effort similarity. This effect, as expected, plunges from 0.0367 to 0.0210 after including control variables that account for the random assignment condition. Furthermore, we added several demographic and academic controls to determine whether the specification in the second column removes all selection bias. As shown in the third column, the estimated peer effect is stable compared to the previous model (0.0210 vs. 0.0205), implying a lack of selection bias. Finally, the fourth column demonstrates the preferred specification in Eq. (<reflink idref="bib2" id="ref71">2</reflink>), which controls for post-enrollment common shocks by adding fixed effects associated with social groups and course teachers, as well as the percentage of shared course sessions. The estimated peer effects decrease slightly from 0.0205 to 0.0180, suggesting that a small part of roommates' effort similarity is explained by their common college experiences. Overall, Table 5 illustrates that selection bias and common shocks explain about half of the observed peer effort similarity (0.0187 out of 0.0367). Nevertheless, after controlling these two main sources of confounders, being peers still significantly increases students' effort similarity.</p> <p>Peer effects on effort similarity</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" rowspan="2"><p>Characteristic</p></th><th align="left" colspan="4"><p>Effort similarity</p></th></tr><tr><th align="left"><p>(1)</p></th><th align="left"><p>(2)</p></th><th align="left"><p>(3)</p></th><th align="left"><p>(4)</p></th></tr></thead><tbody><tr><td align="left"><p>Roommates</p></td><td align="left"><p>0.0367*** (0.00286)</p></td><td align="left"><p>0.0210*** (0.00301)</p></td><td align="left"><p>0.0205*** (0.00300)</p></td><td align="left"><p>0.0180*** (0.00303)</p></td></tr><tr><td align="left"><p>From the same province</p></td><td align="left" /><td align="left" /><td align="left"><p>− 0.00236 (0.00172)</p></td><td align="left"><p>0.00331 (0.00177)</p></td></tr><tr><td align="left"><p>Attended the same high school</p></td><td align="left" /><td align="left" /><td align="left"><p>0.000379 (0.00358)</p></td><td align="left"><p>0.0000409 (0.00358)</p></td></tr><tr><td align="left"><p>Both fresh high school graduates</p></td><td align="left" /><td align="left" /><td align="left"><p>− 0.00473** (0.00179)</p></td><td align="left"><p>− 0.00476** (0.00181)</p></td></tr><tr><td align="left"><p>Both racial minorities</p></td><td align="left" /><td align="left" /><td align="left"><p>− 0.0903*** (0.0179)</p></td><td align="left"><p>− 0.00765 (0.0237)</p></td></tr><tr><td align="left"><p>Both in the racial majority</p></td><td align="left" /><td align="left" /><td align="left"><p>0.00246 (0.00332)</p></td><td align="left"><p>0.00377 (0.00337)</p></td></tr><tr><td align="left"><p>Both science track in high school</p></td><td align="left" /><td align="left" /><td align="left"><p>− 0.00785** (0.00272)</p></td><td align="left"><p>− 0.00641* (0.00268)</p></td></tr><tr><td align="left"><p>Both humanities track in high school</p></td><td align="left" /><td align="left" /><td align="left"><p>− 0.00599* (0.00285)</p></td><td align="left"><p>− 0.00655* (0.00278)</p></td></tr><tr><td align="left"><p>In the same major</p></td><td align="left" /><td align="left" /><td align="left"><p>0.000487 (0.00178)</p></td><td align="left"><p>− 0.0168*** (0.00306)</p></td></tr><tr><td align="left"><p>First semester standard GPA difference < 0.1</p></td><td align="left" /><td align="left" /><td align="left"><p>0.0129*** (0.00163)</p></td><td align="left"><p>0.0128*** (0.00165)</p></td></tr><tr><td align="left"><p>CEE percentile difference < 0.1</p></td><td align="left" /><td align="left" /><td align="left"><p>0.0114*** (0.00254)</p></td><td align="left"><p>0.0109*** (0.00252)</p></td></tr><tr><td align="left"><p>Age difference < 0.5</p></td><td align="left" /><td align="left" /><td align="left"><p>0.00369* (0.00178)</p></td><td align="left"><p>0.00248 (0.00171)</p></td></tr><tr><td align="left"><p>Both attended urban high schools</p></td><td align="left" /><td align="left" /><td align="left"><p>0.00312 (0.00193)</p></td><td align="left"><p>0.00243 (0.00187)</p></td></tr><tr><td align="left"><p>Both attended rural high schools</p></td><td align="left" /><td align="left" /><td align="left"><p>0.00956*** (0.00226)</p></td><td align="left"><p>0.00945*** (0.00228)</p></td></tr><tr><td align="left"><p>Constant</p></td><td align="left"><p>0.781*** (0.000863)</p></td><td align="left"><p>0.758*** (0.00192)</p></td><td align="left"><p>0.751*** (0.00404)</p></td><td align="left"><p>0.745*** (0.0106)</p></td></tr><tr><td align="left"><p>Random controls</p></td><td align="left"><p>No</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Group fixed effects</p></td><td align="left"><p>No</p></td><td align="left"><p>No</p></td><td align="left"><p>No</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Teacher fixed effects</p></td><td align="left"><p>No</p></td><td align="left"><p>No</p></td><td align="left"><p>No</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>% Shared credits</p></td><td align="left"><p>No</p></td><td align="left"><p>No</p></td><td align="left"><p>No</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>N</p></td><td align="left"><p>38,548</p></td><td align="left"><p>38,548</p></td><td align="left"><p>38,548</p></td><td align="left"><p>38,412</p></td></tr><tr><td align="left"><p>R-squared</p></td><td align="left"><p>0.003</p></td><td align="left"><p>0.029</p></td><td align="left"><p>0.034</p></td><td align="left"><p>0.066</p></td></tr></tbody></table> </ephtml> </p> <p>Standard errors in parentheses; bootstrapped standard errors; *<emph>p</emph> < 0.05; **<emph>p</emph> < 0.01; ***<emph>p</emph> < 0.001</p> <p>Table 6 presents the results of the baseline model in Eq. (<reflink idref="bib2" id="ref72">2</reflink>) by gender and effort type. For all students, peer effects on forum participation are considerably stronger than those on individual learning: being peers increases students' forum participation similarity by 0.0204 yet increases individual learning similarity by only 0.0108. This difference is not surprising because forum discussions are much more cooperative in nature than individual learning activities. For example, a forum post initiated by a roommate is likely to elicit comments by other peers, whereas individual learning activities, such as watching course videos, are unlikely to have the same effect. The majority of findings for the pooled sample extend to the two gender subgroups. On average, effort similarity is stronger for roommates than for non-roommates, with no significant gender based differences.</p> <p>Peer effects by gender and type of effort</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" rowspan="2"><p>Type of effort</p></th><th align="left" colspan="3"><p>Gender</p></th></tr><tr><th align="left"><p>Both</p></th><th align="left"><p>Female</p></th><th align="left"><p>Male</p></th></tr></thead><tbody><tr><td align="left"><p>Overall effort</p></td><td align="left"><p>0.0180*** (0.00303)</p></td><td align="left"><p>0.0187*** (0.00340)</p></td><td align="left"><p>0.0159* (0.00619)</p></td></tr><tr><td align="left"><p>Individual learning activities</p></td><td align="left"><p>0.0108*** (0.00279)</p></td><td align="left"><p>0.0107*** (0.00299)</p></td><td align="left"><p>0.00990 (0.00552)</p></td></tr><tr><td align="left"><p>Forum participation</p></td><td align="left"><p>0.0204*** (0.00288)</p></td><td align="left"><p>0.0205*** (0.00322)</p></td><td align="left"><p>0.0194*** (0.00573)</p></td></tr><tr><td align="left"><p>Random controls</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Group fixed effects</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Teacher fixed effects</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>% Shared credits</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>N</p></td><td align="left"><p>38,412</p></td><td align="left"><p>20,630</p></td><td align="left"><p>5118</p></td></tr></tbody></table> </ephtml> </p> <p>Table 7 presents results from a series of multinomial logistic regressions that predict the peer effect on the odds ratio of being in the same effort category. Recall that we classified effort levels of roommates into four categories based on their similarity: both high effort, both medium effort, both low effort, and different effort levels. Table 7 reports the effect of being roommate on the odds ratio of being in a particular category relative to the referent category of different effort levels. For example, when the overall effort is evaluated, being roommates increases the likelihood that a pair of students both fall into the same effort category relative to the different effort category: 31.7% for low effort, 13.7% for medium effort, and 32.6% for high effort. In other words, being roommates increases the likelihood of having similar effort levels. This positive effect of being roommates holds for both individual learning and forum participation. Again, peer effects appear to be stronger on forum participation than on individual learning activities. For example, being roommates increases the likelihood that a pair of students both fall into the same level of forum participation relative to the different effort category: 39.2% for low effort, 23.7% for medium effort, and 47.8% for high effort.</p> <p>Peer effects on student effort, multinomial logistics model (Odds Ratios Included)</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" rowspan="2"><p>Type of effort</p></th><th align="left" colspan="3"><p>Effort level</p></th></tr><tr><th align="left"><p>Both low</p></th><th align="left"><p>Both medium</p></th><th align="left"><p>Both high</p></th></tr></thead><tbody><tr><td align="left"><p>Overall effort</p></td><td align="left"><p>1.317*** (3.329)</p></td><td align="left"><p>1.137* (2.490)</p></td><td align="left"><p>1.326*** (3.725)</p></td></tr><tr><td align="left"><p>Individual learning</p></td><td align="left"><p>1.254** (2.776)</p></td><td align="left"><p>1.0854 (1.586)</p></td><td align="left"><p>1.239** (2.719)</p></td></tr><tr><td align="left"><p>Forum participation</p></td><td align="left"><p>1.392*** (4.002)</p></td><td align="left"><p>1.237*** (4.269)</p></td><td align="left"><p>1.478*** (4.981)</p></td></tr><tr><td align="left"><p>Random controls</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Group fixed effects</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Teacher fixed effects</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>% Shared credits</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>N</p></td><td align="left"><p>38,412</p></td><td align="left"><p>38,412</p></td><td align="left"><p>38,412</p></td></tr></tbody></table> </ephtml> </p> <p> <emph>t-</emph>statistics in parentheses; bootstrapped standard errors; covariates in Table 5, column 4 are included in the models; *<emph>p </emph>< 0.05; **<emph>p</emph> < 0.01; ***<emph>p </emph>< 0.001</p> <p>Having established that a student's study effort in a course is positively influenced by roommates' study efforts, we move to the next research question: What is the effect of study effort on academic performance in the course? As we discussed in the data section, academic performance in this course can be separated into two components: class participation and the in-class final exam. Table 8 reports both OLS and IV estimates for these two measures. To assess the strength of IV, we followed Olea and Pflueger ([<reflink idref="bib32" id="ref73">32</reflink>]) to calculate the effective first-stage F-statistics. This test is robust to heteroscedasticity, autocorrelation, and clustering. We used the same setting in Olea and Pflueger ([<reflink idref="bib32" id="ref74">32</reflink>]) to calculate the thresholds for 2SLS (tau = 10%). The resulting statistics (75.217 and 79.496 respectively) are larger than the suggested threshold (23.109) to reject the weak instrument assumption. It is noteworthy that the F-statistics for the two regression models are different from each other because of slightly different samples used. This test provides additional evidence that roommate effort helps explain a student' study effort after considering other covariates. Results in Table 8 indicate a positive effect of study effort on class participation, which is not surprising because effort is one of the evaluation criteria for class engagement and homework assignments. Our focus is on the second measure of academic performance, the in-class final exam. A similarly positive and significant effect of student effort is observed in both the OLS and IV regressions.</p> <p>Effects of study effort on standardized grades</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" rowspan="2"><p>Variable</p></th><th align="left" colspan="2"><p>Class participation</p></th><th align="left" colspan="2"><p>In-class exam</p></th></tr><tr><th align="left"><p>OLS</p></th><th align="left"><p>2SLS</p></th><th align="left"><p>OLS</p></th><th align="left"><p>2SLS</p></th></tr></thead><tbody><tr><td align="left"><p>Study effort</p></td><td align="left"><p>0.360*** (0.0233)</p></td><td align="left"><p>0.282* (0.113)</p></td><td align="left"><p>0.200*** (0.0236)</p></td><td align="left"><p>0.317** (0.113)</p></td></tr><tr><td align="left"><p>Fresh high school graduate</p></td><td align="left"><p>0.0442 (0.0546)</p></td><td align="left"><p>0.0442 (0.0539)</p></td><td align="left"><p>− 0.0113 (0.0560)</p></td><td align="left"><p>− 0.0132 (0.0555)</p></td></tr><tr><td align="left"><p>Attended urban high school</p></td><td align="left"><p>− 0.0771 (0.0443)</p></td><td align="left"><p>− 0.106 (0.0561)</p></td><td align="left"><p>0.154*** (0.0454)</p></td><td align="left"><p>0.193*** (0.0574)</p></td></tr><tr><td align="left"><p>From home province</p></td><td align="left"><p>0.0519 (0.0470)</p></td><td align="left"><p>0.0618 (0.0475)</p></td><td align="left"><p>0.0664 (0.0481)</p></td><td align="left"><p>0.0534 (0.0491)</p></td></tr><tr><td align="left"><p>Minority</p></td><td align="left"><p>− 0.150 (0.114)</p></td><td align="left"><p>− 0.174 (0.114)</p></td><td align="left"><p>0.180 (0.117)</p></td><td align="left"><p>0.188 (0.117)</p></td></tr><tr><td align="left"><p>Science track in high school</p></td><td align="left"><p>− 0.0463 (0.0397)</p></td><td align="left"><p>− 0.0535 (0.0398)</p></td><td align="left"><p>0.0809* (0.0407)</p></td><td align="left"><p>0.0875* (0.0409)</p></td></tr><tr><td align="left"><p>CEE percentile</p></td><td align="left"><p>− 0.00508 (0.0200)</p></td><td align="left"><p>− 0.00361 (0.0200)</p></td><td align="left"><p>0.0184 (0.0206)</p></td><td align="left"><p>0.0148 (0.0207)</p></td></tr><tr><td align="left"><p>Age enrolled</p></td><td align="left"><p>− 0.00260 (0.0286)</p></td><td align="left"><p>− 0.00110 (0.0282)</p></td><td align="left"><p>− 0.0383 (0.0294)</p></td><td align="left"><p>− 0.0397 (0.0291)</p></td></tr><tr><td align="left"><p>Constant</p></td><td align="left"><p>0.281 (0.624)</p></td><td align="left"><p>0.262 (0.618)</p></td><td align="left"><p>0.878 (0.639)</p></td><td align="left"><p>0.825 (0.636)</p></td></tr><tr><td align="left"><p>Group fixed effects</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Random controls</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>Teacher fixed effects</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td><td align="left"><p>Yes</p></td></tr><tr><td align="left"><p>N</p></td><td align="left"><p>1831</p></td><td align="left"><p>1828</p></td><td align="left"><p>1849</p></td><td align="left"><p>1846</p></td></tr><tr><td align="left"><p>R-squared</p></td><td align="left"><p>0.332</p></td><td align="left"><p>0.328</p></td><td align="left"><p>0.247</p></td><td align="left"><p>0.237</p></td></tr></tbody></table> </ephtml> </p> <p>Standard errors in parentheses; *<emph>p</emph> < 0.05; **<emph>p</emph> < 0.01; ***<emph>p</emph> < 0.001</p> <p>One intriguing result relates to the relative effect size of the OLS and IV estimates for these two academic performance measures. When the class engagement and homework assignments component is used as the dependent variable, OLS seems to overestimate the true effect of study effort (0.360 vs. 0.282); yet when the final exam is used as the dependent variable, OLS underestimates the true effect of study effort (0.200 vs. 0.317). One possible explanation is that students with basic computer knowledge may have been less motivated to exert high levels of effort in what they perceived as an easy class; however, they could still excel on the final exam and obtain high scores. In other words, omitting the motivation variable may have led to a positive bias in the effect of study effort on class engagement and a negative bias on the effect of study effort on in-class exams. This possibility is consistent with the behaviors of several variables in the model that are related to student ability. For example, the variable <emph>attended an urban high school</emph> might indicate higher levels of exposure to computers because computers are more available in urban schools than in rural schools. Similarly, students in the <emph>science track in high school</emph> were usually required to take courses in information technology. Both variables have significant and positive relationships with final exam scores and negative yet insignificant relationships with class engagement and homework performance.</p> <hd id="AN0141772480-13">Discussion and Conclusion</hd> <p>Two main findings from this study provide empirical support to the student engagement literature in higher education. First, student effort appears to be malleable. Our results indicate that effort similarity is higher for a pair of randomly assigned peers than for a pair of randomly assigned non-peers after controlling for selection bias and common shocks. While we did not directly investigate the mechanism for peer effects on study effort, our results support some of the models proposed by previous studies. Whether in the "bad apple" model (Hoxby and Weingarth [<reflink idref="bib17" id="ref75">17</reflink>]) or the "shining light" model (Hoxby and Weingarth [<reflink idref="bib17" id="ref76">17</reflink>]), the essence is that roommates influence each other.</p> <p>Second, our results suggest that the relationship between study effort and academic performance is likely to be causal. We used peers' effort as an instrumental variable to confirm the positive effect of study effort on academic performance in one course. While in this study we went through detailed analysis to show its internal validity, it will require additional studies of other courses and/or in other institutional contexts to assess its generalizability. Nonetheless, results in this study are consistent with many studies conducted at U.S. colleges and universities that have shown significant peer effects on academic performance (e.g., Booij et al. [<reflink idref="bib6" id="ref77">6</reflink>]; Carrell et al. [<reflink idref="bib10" id="ref78">10</reflink>]; Sacerdote [<reflink idref="bib36" id="ref79">36</reflink>]; Zimmerman [<reflink idref="bib45" id="ref80">45</reflink>]).</p> <p>While these results seem to suggest that roommate assignment could be an inexpensive and inconspicuous strategy to influence students' academic performance, caution must be exercised for a couple reasons. First, the peer effects examined in Eq. (<reflink idref="bib2" id="ref81">2</reflink>) are assumed to operate through means. As such, the peer effects exerted by each student to other students is fixed and any rearrangement of roommates may not improve the collective academic performance of all students. In other words, when a high-effort student shares a room with a low-effort student, they influence each other. Additional studies are needed to examine the boundary conditions under which the positive peer effects of high-effort students can be maximized while the negative peer effects of low-effort students can be minimized.</p> <p>Second, peer group manipulation can be precarious at times. Carrell et al. ([<reflink idref="bib11" id="ref82">11</reflink>]) experiment at West Point suggested that peer group manipulation may not lead to expected results. In their study, low-achievers and high-achievers selected into sub-social networks based on academic ability and did not interact much across these sub-structures. As a result, low-achievers assigned to high-achieving peers performed worse than low-achievers assigned to random peer groups. This is not to say peer group manipulation is impossible. Recent research by Li et al. ([<reflink idref="bib23" id="ref83">23</reflink>]) revealed that peer group manipulation accompanied by group incentives can improve low-achievers' academic performance by promoting meaningful academic interaction between low-achieving and high-achieving students.</p> <p>While this study has focused on the causal link between roommate assignment and study effort and between study effort and academic performance, many questions remain unanswered. Future research may want to examine individual and contextual factors that influence the magnitude of peer effects. For example, results in Table 6 suggest peer effects hold for both male and female groups; however, the estimates for female group is slightly larger than that for male group although the difference is not statistically significant. This result provides one explanation why the effect of peer composition is more pronounced for female than for male students (Han and Li [<reflink idref="bib16" id="ref84">16</reflink>]; Zhang and Pu [<reflink idref="bib44" id="ref85">44</reflink>]). Future research may want to examine potentially heterogeneous peer effects on study effort and academic performance across sex, age, race/ethnicity, and academic performance.</p> <p>As another example, results in Table 7 suggest that being roommates increases the likelihood of having similar effort levels; however, results also indicate that peer influence is a two-way street. While roommates are more likely than non-roommates to both have high effort levels, they are more likely to both have low effort levels too. In fact, Zimmerman ([<reflink idref="bib45" id="ref86">45</reflink>]) shows that students with average SAT scores could do worse when they share rooms with students with low SAT scores. Future research may want to examine individual and contextual factors that may maximize the positive peer effects of sharing a room with a hardworking student while minimize the negative peer effects of sharing a room with an indolent roommate.</p> <hd id="AN0141772480-14">Publisher's Note</hd> <p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p> <ref id="AN0141772480-15"> <title> References </title> <blist> <bibl id="bib1" idref="ref43" type="bt">1</bibl> <bibtext> Arum R, Roksa J. Academically adrift: Limited learning on college campuses. 2011: Chicago; University of Chicago Press</bibtext> </blist> <blist> <bibl id="bib2" idref="ref1" type="bt">2</bibl> <bibtext> Astin A. Student involvement: A developmental theory for higher education. Journal of College Student Development. 1984; 40; 5: 518-529</bibtext> </blist> <blist> <bibl id="bib3" idref="ref2" type="bt">3</bibl> <bibtext> Astin AW. What matters in college? Four critical years revisited. 1993: San Francisco; Jossey-Bass</bibtext> </blist> <blist> <bibl id="bib4" idref="ref34" type="bt">4</bibl> <bibtext> Bayer P, Ross SL, Topa G. Place of work and place of residence: Informal hiring networks and labor market outcomes. Journal of Political Economy. 2008; 116; 6: 1150-1196</bibtext> </blist> <blist> <bibl id="bib5" idref="ref35" type="bt">5</bibl> <bibtext> Bettinger, E., Loeb, S., & Taylor, E. (2014). Remote but influential: Peer effects and reflection in online higher education classrooms. Center for Education Policy Analysis Working Paper. Palo Alto: Stanford University.</bibtext> </blist> <blist> <bibl id="bib6" idref="ref19" type="bt">6</bibl> <bibtext> Booij AS, Leuven E, Oosterbeek H. Ability peer effects in University: Evidence from a randomized experiment. The Review of Economic Studies. 2016; 84; 2: 547-578</bibtext> </blist> <blist> <bibl id="bib7" idref="ref56" type="bt">7</bibl> <bibtext> Borsari B, Carey KB. Peer influences on college drinking: A review of the research. Journal of Substance Abuse. 2001; 13; 4: 391-424</bibtext> </blist> <blist> <bibl id="bib8" idref="ref17" type="bt">8</bibl> <bibtext> Bowman NA Herzog S, Bowman NA. The meaning and interpretation of college student self-reporting gains. The validity of college student self-reported data: New directions for institutional research. 2014: San Francisco; Jossey Bass: 59-68</bibtext> </blist> <blist> <bibl id="bib9" idref="ref48" type="bt">9</bibl> <bibtext> Brint S, Cantwell AM, Saxena P. Disciplinary categories, majors, and undergraduate academic experiences: Rethinking Bok's "underachieving colleges" thesis. Research in Higher Education. 2012; 53; 1: 1-25</bibtext> </blist> <blist> <bibtext> Carrell SE, Fullerton RL, West JE. Does your cohort matter? Measuring peer effects in college achievement. Journal of Labor Economics. 2009; 27; 3: 439-464</bibtext> </blist> <blist> <bibtext> Carrell SE, Sacerdote BI, West JE. From natural variation to optimal policy? The importance of endogenous peer group formation. Econometrica. 2013; 81; 3: 855-882</bibtext> </blist> <blist> <bibtext> Coates H. The value of student engagement for higher education quality assurance. Quality in Higher Education. 2005</bibtext> </blist> <blist> <bibtext> De Giorgi G, Pellizzari M, Redaelli S. Identification of social interactions through partially overlapping peer groups. American Economic Journal: Applied Economics. 2010; 2; 2: 241-275</bibtext> </blist> <blist> <bibtext> Gardner M, Steinberg L. Peer influence on risk taking, risk preference, and risky decision making in adolescence and adulthood: An experimental study. Developmental Psychology. 2005; 41; 4: 625-635</bibtext> </blist> <blist> <bibtext> Gordon J, Ludlum J, Hoey JJ. Validating NSSE against student outcomes: Are they related?. Research in Higher Education. 2008; 49; 1: 19-39</bibtext> </blist> <blist> <bibtext> Han L, Li T. The gender difference of peer influence in higher education. Economics of Education Review. 2009; 28; 1: 129-134</bibtext> </blist> <blist> <bibtext> Hoxby, C. M., & Weingarth, G. (2005). Taking race out of the equation: School reassignment and the structure of peer effects. Unpublished.</bibtext> </blist> <blist> <bibtext> Hu S, Kuh GD. Being (dis)engaged in educationally purposeful activities: The influences of student and institutional characteristics. Research in Higher Education. 2002</bibtext> </blist> <blist> <bibtext> Krause K-L, Coates H. Students' engagement in first-year university. Assessment & Evaluation in Higher Education. 2008; 33; 5: 493-505</bibtext> </blist> <blist> <bibtext> Kuh GD. Excerpt from high-impact educational practices: What they are, who has access to them, and why they matter. 2008: Washington, DC; Association of American Colleges and Universities</bibtext> </blist> <blist> <bibtext> Kuh GD, Cruce TM, Shoup R, Kinzie J, Gonyea RM. Unmasking the effects of student engagement on first-year college grades and persistence. Journal of Higher Education. 2008; 79; 5: 540-563</bibtext> </blist> <blist> <bibtext> Kuh GD, Kinzie J, Schuh JH, Whitt EJ. Student success in college: Creating conditions that matter. 2011: Hoboken; Wiley</bibtext> </blist> <blist> <bibtext> Li T, Han L, Zhang L, Rozelle S. Encouraging classroom peer interactions: Evidence from Chinese migrant schools. Journal of Public Economics. 2014; 111: 29-45</bibtext> </blist> <blist> <bibtext> Loes C, Pascarella E, Umbach P. Effects of diversity experiences on critical thinking skills: Who benefits?. The Journal of Higher Education. 2012; 83; 1: 1-25</bibtext> </blist> <blist> <bibtext> Loke AJTY, Chow FLW. Learning partnership—the experience of peer tutoring among nursing students: A qualitative study. International Journal of Nursing Studies. 2007; 44; 2: 237-244</bibtext> </blist> <blist> <bibtext> Lyle DS. Estimating and interpreting peer and role model effects from randomly assigned social groups at West Point. Review of Economics and Statistics. 2007; 89; 2: 289-299</bibtext> </blist> <blist> <bibtext> Lyle DS. The effects of peer group heterogeneity on the production of human capital at West Point. American Economic Journal: Applied Economics. 2009; 1; 4: 69-84</bibtext> </blist> <blist> <bibtext> Manski CF. Identification of endogenous social effects: The reflection problem. Review of Economic Studies. 1993; 60; 3: 531-542</bibtext> </blist> <blist> <bibtext> Mayhew MJ, Rockenbach AN, Bowman NA, Seifert TA, Wolniak GC. How college affects students: 21st century evidence that higher education works. 2016: Hoboken; Wiley</bibtext> </blist> <blist> <bibtext> Nanda, R., & Sorensen, J. (2006). Peer-effects and entrepreneurship (No. 08-051). Harvard Business School Working Paper. Cambridge, MA: Harvard Business School.</bibtext> </blist> <blist> <bibtext> Nye B, Konstantopoulos S, Hedges LV. How large are teacher effects?. Educational Evaluation and Policy Analysis. 2004; 26; 3: 237-257</bibtext> </blist> <blist> <bibtext> Olea JLM, Pflueger C. A robust test for weak instruments. Journal of Business & Economic Statistics. 2013; 31; 3: 358-369</bibtext> </blist> <blist> <bibtext> Pace CR. Achievement and the quality of student effort. 1982: Washington, DC; National Commission on Excellence</bibtext> </blist> <blist> <bibtext> Porter SR. Do college student surveys have any validity?. Review of Higher Education. 2011; 35; 1: 45-76</bibtext> </blist> <blist> <bibtext> Ryan AM. Peer groups as a context for the socialization of adolescents' motivation, engagement, and achievement in school. Educational Psychologist. 2000; 35; 2: 101-111</bibtext> </blist> <blist> <bibtext> Sacerdote B. Peer effects with random assignment: Results for Dartmouth roommates. Quarterly Journal of Economics. 2001; 116; 2: 681-704</bibtext> </blist> <blist> <bibtext> Simons-Morton B, Chen R. Peer and parent influences on school engagement among early adolescents. Youth & Society. 2009; 41; 1: 3-25</bibtext> </blist> <blist> <bibtext> Skinner EA, Belmont MJ. Motivation in the classroom: Reciprocal effects of teacher behavior and student engagement across the school year. Journal of Educational Psychology. 1993; 85; 4: 571-581</bibtext> </blist> <blist> <bibtext> Stinebrickner R, Stinebrickner TR. What can be learned about peer effects using college roommates? Evidence from new survey data and students from disadvantaged backgrounds. Journal of Public Economics. 2006; 90; 8–9: 1435-1454</bibtext> </blist> <blist> <bibtext> Stinebrickner TR, Stinebrickner R. The causal effect of studying on academic performance. BE Journal of Economic Analysis & Policy. 2008; 8; 1: 14</bibtext> </blist> <blist> <bibtext> Trowler V. Student engagement literature review. Higher Education Academy. 2010; 11: 1-15</bibtext> </blist> <blist> <bibtext> Unger JB, Yan L, Shakib S, Rohrbach LA, Chen X, Qian G. Peer influences and access to cigarettes as correlates of adolescent smoking: A cross-cultural comparison of Wuhan, China, and California. Preventive Medicine. 2002; 34; 4: 476-484</bibtext> </blist> <blist> <bibtext> Webber KL, Krylow RB, Zhang Q. Does involvement really matter? Indicators of college student success and satisfaction. Journal of College Student Development. 2013; 54; 6: 591-611</bibtext> </blist> <blist> <bibtext> Zhang L, Pu S. It takes two shining lights to brighten the room: Peer effects with random roommate assignments. Education Economics. 2017; 25; 1: 3-21</bibtext> </blist> <blist> <bibtext> Zimmerman DJ. Peer effects in academic outcomes: Evidence from a natural experiment. Review of Economics and Statistics. 2003; 85; 1: 9-23</bibtext> </blist> </ref> <aug> <p>By Shi Pu; Yu Yan and Liang Zhang</p> <p>Reported by Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib33" firstref="ref3"></nolink> <nolink nlid="nl2" bibid="bib21" firstref="ref4"></nolink> <nolink nlid="nl3" bibid="bib43" firstref="ref5"></nolink> <nolink nlid="nl4" bibid="bib40" firstref="ref6"></nolink> <nolink nlid="nl5" bibid="bib12" firstref="ref10"></nolink> <nolink nlid="nl6" bibid="bib19" firstref="ref11"></nolink> <nolink nlid="nl7" bibid="bib20" firstref="ref12"></nolink> <nolink nlid="nl8" bibid="bib41" firstref="ref13"></nolink> <nolink nlid="nl9" bibid="bib18" firstref="ref14"></nolink> <nolink nlid="nl10" bibid="bib34" firstref="ref18"></nolink> <nolink nlid="nl11" bibid="bib10" firstref="ref20"></nolink> <nolink nlid="nl12" bibid="bib36" firstref="ref21"></nolink> <nolink nlid="nl13" bibid="bib45" firstref="ref22"></nolink> <nolink nlid="nl14" bibid="bib17" firstref="ref23"></nolink> <nolink nlid="nl15" bibid="bib39" firstref="ref26"></nolink> <nolink nlid="nl16" bibid="bib28" firstref="ref27"></nolink> <nolink nlid="nl17" bibid="bib16" firstref="ref28"></nolink> <nolink nlid="nl18" bibid="bib26" firstref="ref31"></nolink> <nolink nlid="nl19" bibid="bib27" firstref="ref32"></nolink> <nolink nlid="nl20" bibid="bib31" firstref="ref33"></nolink> <nolink nlid="nl21" bibid="bib30" firstref="ref36"></nolink> <nolink nlid="nl22" bibid="bib13" firstref="ref37"></nolink> <nolink nlid="nl23" bibid="bib14" firstref="ref38"></nolink> <nolink nlid="nl24" bibid="bib15" firstref="ref40"></nolink> <nolink nlid="nl25" bibid="bib37" firstref="ref41"></nolink> <nolink nlid="nl26" bibid="bib29" firstref="ref44"></nolink> <nolink nlid="nl27" bibid="bib22" firstref="ref46"></nolink> <nolink nlid="nl28" bibid="bib24" firstref="ref52"></nolink> <nolink nlid="nl29" bibid="bib25" firstref="ref55"></nolink> <nolink nlid="nl30" bibid="bib42" firstref="ref57"></nolink> <nolink nlid="nl31" bibid="bib35" firstref="ref58"></nolink> <nolink nlid="nl32" bibid="bib38" firstref="ref61"></nolink> <nolink nlid="nl33" bibid="bib32" firstref="ref73"></nolink> <nolink nlid="nl34" bibid="bib11" firstref="ref82"></nolink> <nolink nlid="nl35" bibid="bib23" firstref="ref83"></nolink> <nolink nlid="nl36" bibid="bib44" firstref="ref85"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1243986
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Peers, Study Effort, and Academic Performance in College Education: Evidence from Randomly Assigned Roommates in a Flipped Classroom
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Pu%2C+Shi%22">Pu, Shi</searchLink><br /><searchLink fieldCode="AR" term="%22Yan%2C+Yu%22">Yan, Yu</searchLink><br /><searchLink fieldCode="AR" term="%22Zhang%2C+Liang%22">Zhang, Liang</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Research+in+Higher+Education%22"><i>Research in Higher Education</i></searchLink>. Mar 2020 61(2):248-269.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Springer. Available from: Springer Nature. 233 Spring Street, New York, NY 10013. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-348-4505; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 22
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2020
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Audience
  Label: Education Level
  Group: Audnce
  Data: <searchLink fieldCode="EL" term="%22Higher+Education%22">Higher Education</searchLink><br /><searchLink fieldCode="EL" term="%22Postsecondary+Education%22">Postsecondary Education</searchLink>
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Foreign+Countries%22">Foreign Countries</searchLink><br /><searchLink fieldCode="DE" term="%22College+Students%22">College Students</searchLink><br /><searchLink fieldCode="DE" term="%22Dormitories%22">Dormitories</searchLink><br /><searchLink fieldCode="DE" term="%22Peer+Influence%22">Peer Influence</searchLink><br /><searchLink fieldCode="DE" term="%22Academic+Achievement%22">Academic Achievement</searchLink><br /><searchLink fieldCode="DE" term="%22Study+Habits%22">Study Habits</searchLink><br /><searchLink fieldCode="DE" term="%22Blended+Learning%22">Blended Learning</searchLink><br /><searchLink fieldCode="DE" term="%22Homework%22">Homework</searchLink><br /><searchLink fieldCode="DE" term="%22Video+Technology%22">Video Technology</searchLink><br /><searchLink fieldCode="DE" term="%22Educational+Technology%22">Educational Technology</searchLink>
– Name: Subject
  Label: Geographic Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22China%22">China</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1007/s11162-019-09571-x
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0361-0365
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: This study provides empirical evidence related to two critical assumptions in the student engagement literature in higher education--namely, the malleability of study effort and the causality of the relationship between study effort and student outcomes. We merged student-level administrative data on dormitory assignments, study effort in a hybrid course, and course performance from a regional college in China. Our results indicate that study efforts are more similar for a pair of randomly assigned roommates than for a pair of randomly assigned non-roommates, indicating the malleability of study effort. In addition, we provide evidence on the causal relationship between study effort and academic performance by using roommates' study efforts as an instrumental variable when estimating the effect of a student's study effort on course performance.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2020
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1243986
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1243986
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1007/s11162-019-09571-x
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 22
        StartPage: 248
    Subjects:
      – SubjectFull: Foreign Countries
        Type: general
      – SubjectFull: College Students
        Type: general
      – SubjectFull: Dormitories
        Type: general
      – SubjectFull: Peer Influence
        Type: general
      – SubjectFull: Academic Achievement
        Type: general
      – SubjectFull: Study Habits
        Type: general
      – SubjectFull: Blended Learning
        Type: general
      – SubjectFull: Homework
        Type: general
      – SubjectFull: Video Technology
        Type: general
      – SubjectFull: Educational Technology
        Type: general
      – SubjectFull: China
        Type: general
    Titles:
      – TitleFull: Peers, Study Effort, and Academic Performance in College Education: Evidence from Randomly Assigned Roommates in a Flipped Classroom
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Pu, Shi
      – PersonEntity:
          Name:
            NameFull: Yan, Yu
      – PersonEntity:
          Name:
            NameFull: Zhang, Liang
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 03
              Type: published
              Y: 2020
          Identifiers:
            – Type: issn-print
              Value: 0361-0365
          Numbering:
            – Type: volume
              Value: 61
            – Type: issue
              Value: 2
          Titles:
            – TitleFull: Research in Higher Education
              Type: main
ResultId 1