the han-minority achievement gap, language, and returns to ... · gap between han and minority...
TRANSCRIPT
The Han-Minority Achievement Gap, Language, and Returns to Schools in Rural ChinaAuthor(s): Yunfan Yang /Huan Wang /Linxiu Zhang /Sean Sylvia /Renfu Luo /Yaojiang Shi/Wei Wang /Scott RozelleSource: Economic Development and Cultural Change, Vol. 63, No. 2 (January 2015), pp. 319-359Published by: The University of Chicago PressStable URL: http://www.jstor.org/stable/10.1086/679070 .
Accessed: 07/01/2015 16:28
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp
.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].
.
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access toEconomic Development and Cultural Change.
http://www.jstor.org
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
The Han-Minority Achievement Gap, Language, and
Returns to Schools in Rural Chinayunfan yangInstitute for Geographical Sciences and Natural Resources Research, ChineseAcademy of Sciences; and University of the Chinese Academy of Sciences
huan wangNorthwest University of Xi’an
linxiu zhangInstitute for Geographical Sciences and Natural Resources Research, ChineseAcademy of Sciences
sean sylviaRenmin University of China
renfu luoInstitute for Geographical Sciences and Natural Resources Research, ChineseAcademy of Sciences
yaojiang shiCenter for Experimental Economics for Education, Shaanxi Normal University
wei wangCenter for Chinese Agricultural Policy, Chinese Academy of Sciences
scott rozelleStanford University
I. Introduction
Over the past 3 decades, China’s rural population has experienced rapid in-come growth and a dramatic reduction in poverty ðRavallion and Chen 2007;
The authors acknowledge financial support from the National Institutes of Health ðR01HL106023-03Þ, the Chinese Academy of Sciences ðKZZD-EW-06-02Þ, the Institute of Geographic Sciencesand Natural Resources Research, Chinese Academy of Sciences ð2012ZD008Þ, and the StanfordCenter for International Development. Contact the corresponding author, Renfu Luo, at [email protected].
Electronically published November 18, 2014© 2015 by The University of Chicago. All rights reserved. 0013-0079/2015/6302-0006$10.00
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Chen and Ravallion 2010Þ. Although the well-being of the population as a
320 E C O N O M I C D E V E L O P M E N T A N D C U L T U R A L C H A N G E
whole has risen sharply, the average economic standing of country’s nearly114 million ethnic minorities has improved relatively less than that of theHan majority ðGustafsson and Li 2003; Gustafsson and Sai 2009a, 2009bÞ.Between 1988 and 1995 the average per capita income of Han living in ruralareas increased by more than 52%, while incomes of the rural minority pop-ulation only grew by nearly 22% ðGustafsson and Li 2003Þ. Over this period,the Han-minority income gap nearly doubled from 19.2% to 35.9% ðGus-tafsson and Li 2003Þ. In 2002, rural minorities remained more than one anda half times as likely as the rural Han majority to be in poverty and twice aslikely to have experienced poverty in the past 2 years ðGustafsson and Sai 2009a;Hannum and Wang 2012Þ.Lagging educational attainment among minorities has undoubtedly played
a significant role in the persistence of the Han-minority income gap ðHannumand Wang 2012Þ. Education is an increasingly important determinant of wagesand access to off-farm employment ðZhang et al. 2005; De Brauw and Rozelle2008Þ. At the same time, educational attainment among minorities lags ðHan-num 2002; Hannum et al. 2008; Hannum and Wang 2012Þ. Analyzing mid-census survey data from 2005, Hannum and Wang ð2012Þ find that—among16–21-year-olds—minorities were nearly one-third as likely as Han to haveattained 9 years of compulsory schooling. Minorities are also significantly lesslikely to enroll at the tertiary level. In a 2008 census of entering freshman atfour tier 1 universities in western China, only 4% were non-Han, while mi-norities comprise approximately 11% of the population cohort ðWang et al.2013Þ. The same survey shows female minority students to be at a particulardisadvantage in college admissions: in this entering class, female minority stu-dents were only 25% of their population share.If the Han-minority differences in educational attainment persist, the rela-
tive well-being of minority populations is likely to continue to fall as China’seconomy increasingly demands a higher-quality workforce. Tightening demo-graphics and a nearly complete transition into off-farm labor in China ðmorethan 80% of 16–30-year-olds are now employed off farmÞ are driving up wagesfor unskilled labor at close to 10% per year ðPark, Cai, and Du 2010; Zhanget al. 2013Þ. As unskilled wages rise and low-paying basic manufacturing jobsare replaced with jobs involving more sophisticated tasks, China’s economy willincreasingly demand a high-quality, educated workforce ðZhang et al. 2013Þ.Educationally disadvantaged minorities will find it more difficult to partici-pate in this new labor market and benefit from the higher wages that will comewith it.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
In the context of rural China, poor academic performance in school may
Yang et al. 321
play a significant role in reducing educational attainment or years of school-ing ðYi et al. 2012Þ. In competitive educational systems—such as China’s—lowerexpectations of poorly performing students to thrive in the system may dis-courage continued enrollment ðChuang 1997; Clarke, Haney, and Madaus2000; Reardon and Galindo 2002; Rumberger and Lim 2008Þ. Heavy em-phasis on testing may further lead teachers to direct more attention to higher-performing children and even lead schools to push at-risk students out in aneffort to raise overall test scores ðVickers 1994; Vélez and Saenz 2001; Fortinet al. 2006Þ. These influences are compounded by rising unskilled wages,which drive up the opportunity costs of schooling ðAngrist and Lavy 2009;Fiszbein, Schady, and Ferreira 2009Þ. Indeed, the available evidence highlightsthe correlation between poor performance and dropout among poor stu-dents in western China ðYi et al. 2012Þ. Thus, if minority students performworse than their Han peers, they are likely to attain fewer years of schoolingas they forgo school and opt to enter the labor force in unskilled jobs.Despite the implications of an achievement gap between Han and minority
students, no study that we know of has compared their achievement ðeithergrades or test scoresÞ. Likewise, we find almost no research on the how thedeterminants of achievement may vary between the two groups. Existing em-pirical work on the disparity between Han and minority educational outcomeshas focused on attainment. Hannum ð2002Þ, for example, using a 1992 nationalsurvey of children in China, finds large differences in enrollment between Hanand minority children of primary school age, with enrollment rates lowestamong minorities in western China. She concludes that much of this differ-ence is attributable to geographic composition and family background. Re-search like this, however, is focused on attainment and, presumably due to theabsence of data, has not examined achievement.The overall goal of this article is to document and analyze the achievement
gap between Han and minority students in rural China. To meet this goal wehave two specific objectives. First, we estimate the overall achievement gapðhenceforth, the “Han-minority achievement gap”Þ. We also measure two othersubgaps: the gap between Han and minority students that speak Mandarin asa first language and the gap between Han and minority student that speakMandarin as a second language. Second, we assess what factors contribute mostto these achievement gaps. To do this, we first decompose the achievement gapinto two parts: one part representing the portion of the gap due Han-minoritydifferences in endowments of student, household, peer, teacher, and schoolcharacteristics and a second part due to differences in returns to these char-
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
acteristics. We then asses what effect specific schools have on the Han-minority
322 E C O N O M I C D E V E L O P M E N T A N D C U L T U R A L C H A N G E
achievement gap and what types of schools narrow or widen this gap. That is,we analyze how returns to attending specific schools ðschool fixed effectsÞ dif-fer between Han and minority students and what school characteristics aremost strongly associated with these Han-minority differences in returns to spe-cific schools.To achieve these objectives, we draw on a large-scale survey of schools
sampled from across rural Shaanxi, Gansu, and Qinghai provinces coveringnearly 21,000 students, approximately 13% of whom are minorities. We mea-sure achievement of Han and minority students using curriculum-based stan-dardized exams in math and Chinese given as part of the survey. To assessfactors that contribute to the Han-minority achievement gap, we use detailedinformation on students, households, teachers, and schools and apply decom-position methods pioneered by Oaxaca ð1973Þ and Blinder ð1973Þ. Oaxaca-Blinder type decomposition, originally used to analyze wage differences be-tween groups, has now been applied in a wide variety of contexts. In education,previous research has used this approach to analyze differences in academicachievement across countries ðe.g., McEwan and Marshall 2004; Ammer-mueller 2007Þ, across time ðBarrera-Osorio et al. 2011Þ, and between indig-enous and nonindigenous students ðMcEwan 2004; McEwan and Trowbridge2007; Sakellariou 2008Þ.Our analysis yields three primary findings. First, we find that minority
students in our sample score significantly below Han students on standardizedexams in math and Chinese. The Han-minority achievement gap is nearly0.3 standard deviations ðSDÞ in math and more than 0.2 SD in Chinese.Among minorities in our sample whose primary language is not standardMandarin ðSalar and Tibetan—henceforth, “Non-Mandarin minorities”Þ, theachievement gap is even more striking: these students score 0.62 SD lowerthan Han in math and 0.65 SD lower than Han in Chinese.Second, our decomposition analysis suggests that the Han-minority
achievement gap for Mandarin-speaking minority students ðHui and Tu—henceforth, “Mandarin minorities”Þ is almost fully explained by differences instudent, peer, teacher, and school characteristics. Of these, the largest con-tributor is student and family background. Differences in school quality playa relatively small role. Endowments, however, explain very little of the achieve-ment gap between Han students and non-Mandarin minorities.Third, we find that—in “mixed” schools with both Han and minority
students—the effects of individual schools play a role in widening the Han-minority achievement gap. In these mixed schools, returns to Han students ofðobserved and unobservedÞ specific school attributes are higher than those for
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
similar minority students attending the same school. Teachers appear to play
Yang et al. 323
a central role in affecting the relative returns of Han and minority students.The rest of this article is organized as follows. Section II reviews the back-
ground of minorities in China. Sections III and IV describe the survey and datathat we use for the analysis. Sections V and VI discuss the empirical approachand results. The final section concludes and discusses the policy implicationsof our findings.
II. Background: The Education of Minorities in ChinaIn addition to the Han majority, there are 55 officially recognized minoritynationalities in China. According to the 2010 census, minorities comprised8.5% of the total national population, approximately 114 million peopleðCherng, Hannum, and Lu 2012Þ. Geographically, minorities in China areconcentrated in relatively poor regions of western China: 71.6% of the mi-nority population lives in western provinces, and 91.6% of ethnic autono-mous counties are located in western China. Approximately 40% of theseautonomous counties are nationally designated poverty counties ðHannumand Wang 2012Þ.Beyond geographically targeted antipoverty funds ðfrom which minorities
disproportionately benefit due to concentration in poor areas; Park, Wang,and Wu 2002Þ, a number of policies and programs have aimed to expandaccess to education among minority groups. For example, the 1980 Lawon Regional and Ethnic Autonomy recommended subsidization of educationin minority areas beyond standard educational funding ðCherng et al. 2012Þ.More recently, as part of the Tenth 5-Year Plan ð2001–5Þ, the central govern-ment invested approximately ¥34.2 billion for boarding schools and ethnicuniversities in western China andminority areas ðCherng et al. 2012Þ. A numberof affirmative action policies have also been implemented in higher education,such as university admissions spots reserved for minority students and accep-tance of minority students with lower entrance exam scores ðHannum andWang 2012Þ.Although certain policies have been designed to improve educational at-
tainment among minorities, the structure of education for minority groups islargely similar to the rest of the country ðCherng et al. 2012Þ. Curriculum andassessment are generally the same for minority and Han students ðChernget al. 2012Þ. One exception is the language of instruction. While official pol-icy regarding language of instruction emphasizes the use of Mandarin, schoolswith more than 50% minority students who speak a local language are per-mitted to use the local language ðCherng et al. 2012Þ. In practice, however,there are significant challenges to bilingual instruction. For example, some
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
minority groups with their own language do not have a written language. In
324 E C O N O M I C D E V E L O P M E N T A N D C U L T U R A L C H A N G E
addition, there seldom are financial resources available to develop a local lan-guage curriculum. Schools also are often integrated with students attend-ing class with Han students or students belonging to other minority groupsðHannum and Wang 2012Þ. In our survey of schools in northwest Chinaðdescribed in the next sectionÞ, only 5% have no Han students. No schoolsprovide instruction or teaching material in minority languages.
III. Survey DesignThe data used in this study come from a survey of 300 schools in Shaanxi,Gansu, and Qinghai provinces in western China during the 2011/2012 aca-demic year. Schools were sampled as follows. We first obtained a list of allschools in the following regions: Haidong ðin QinghaiÞ, Longnan, Dingxi,Tianshui ðin GansuÞ, and Ankang ðin ShaanxiÞ. A map of these regions isprovided in figure 1. In total, 26 counties were included in the samplingframe. Within each township located in these five regions, one school wasselected from among all schools with 150–300 students as reported by thelocal education ministry. The survey is thus roughly representative of pri-mary schools in these regions of northwestern China.
Figure 1. Survey regions
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Due to the survey’s geographical coverage, our sample includes both com-
Yang et al. 325
pletely Han ð37%Þ and completely minority schools ð5%Þ. A significant num-ber of schools ð58%Þ are mixed Han and minority schools. We focus most ofthe study on the full sample but, in some parts of the analysis, restrict thesample to mixed schools only.1 In the analysis, we define “mixed” schools asschools with at least two minority and two Han students. By restricting theanalysis to mixed schools, we are better able to pick up differences betweenHan and minority students not confounded by differences in location; how-ever, limiting the sample both limits variation in the data ðleading to less pre-cise parameter estimatesÞ and reduces representativeness.Within each school, we collected information on all fourth and fifth grad-
ers ðmore than 21,000 students in totalÞ. A survey questionnaire admin-istered to students collected detailed information on students and theirfamilies. Table 1 lists all additional variables that we use in our analysis andprovides descriptions of each. All of these variables were asked or measuredat the beginning of the school year.As our measure of academic achievement, we use student scores on stan-
dardized exams in math and Chinese administered by the survey team at theend of the school year. Within each classroom, half of the students wererandomly assigned to take a math exam, and the rest took a Chinese exam. Toensure coherence with the national curriculum, the tests were developed withassistance from local department/bureaus of education. Questions used in themath exam were drawn from the question bank of the Trends in InternationalMathematics and Science Study, an international assessment of mathematicsand science knowledge of primary and lower-secondary school students.Questions used in the Chinese exam were taken from national fourth or fifthgrade textbooks. To minimize cheating, two versions of each exam ðwithreordered questionsÞ were randomly assigned to students. The exam also wasproctored closely by the enumerators. For analysis, scores for both subjecttests are normalized by the distribution of scores in each grade. Exams weregiven in Mandarin, just like year-end tests usually given in the schools in oursample.
IV. Characteristics of Students, Peers, Teachers, and SchoolsA. Minority StatusWe solicited minority status directly from students as part of the survey. Outof the full sample, 12.5% of students identified themselves as belonging to a
1 Similar strategies of dealing with differences in location in decomposition analysis has been used in
previous studies ðsee, e.g., van de Walle and Gunewardena 2001Þ.This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE 1VARIABLE DESCRIPTIONS
Variable Description
Student and household characteristics:Standardized math exam score Normalized score on standardizedmath exam. Exam
designed using grade-appropriate questions fromthe Trends in International Mathematics and ScienceStudy with assistance from the Chinese Ministry ofEducation.
Standardized Chinese exam score Normalized score on standardized Chinese exam.Exam designed using questions from nationalcurriculumwith assistance from theChineseMinistryof Education.
Female ð0/1Þ Student is female.Boarding student ð0/1Þ Student boards at school.Age ðyearsÞ Student age in yearsHousehold size Total number of individuals living the in the stu-
dent’s householdTravel time to school ðminutesÞ Travel time from student’s home to school in minutesMother has lower secondary degreeor above ð0/1Þ
Student’s mother has completed middle school edu-cation or above.
Father has lower secondary degreeor above ð0/1Þ
Student’s father has completed middle school educa-tion or above.
Father at home ð0/1Þ Father currently living at home ðhas not migratedfor workÞ
Mother at home ð0/1Þ Mother currently living at home ðhas not migratedfor workÞ
Household asset index ð0/1Þ Index of household durable assets. Constructedusing first principal component of motorbike,tractor, car, van, refrigerator, air conditioning,computer, laundry machine, and dummy variablesfor type of housing ðcave house, packed earth,brick, apartment building, otherÞ
Class peer characteristic:Proportion of peers’ mothers with lowersecondary degree or above
Class level mean of “mother has lower secondarydegree or above” excluding student i
Proportion of class peers of same ethnicity Proportion of students in class of same ethnicbackground as student i
Peer average household asset index Class level mean of “household asset index”excluding student i
Teacher characteristic:Female teacher ð0/1Þ Teacher is femaleHan teacher ð0/1Þ Teacher is Han majorityTeacher has higher education degree ð0/1Þ Teacher has completed college or aboveTeacher attended normal college ð0/1Þ Teacher attended normal schoolTeacher has received provincial ornational teaching award ð0/1Þ
Teacher has received a provincial or national levelteaching award
Gongban teacher ð0/1Þ Teacher is a regular teacher, not on a short-term contract.Teacher experience ðyearsÞ Teacher years of teaching experience
School characteristic:School size ðstudentsÞ Number of students in schoolStudent-teacher ratio Student-teacher ratioDistance to farthest village served byschool ðminutesÞ
Travel time to the farthest village in school’scatchment area
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
minority group. Table 2 shows the distribution of each of the five main ethnic
TABLE 1 (Continued )
Variable Description
School has provided teacher training in
past year ð0/1ÞSchool has provided training to teachers in pastyear
School infrastructure index Index of school infrastructure constructed using firstprincipal component of number of classrooms,library, garden, school wall, cafeteria, playground,number of computers for student use
Yang et al. 327
groups in the sample by province. Of all minority students, 56.5% are Hui,12.3% are Salar, 13.5% are Tibetan, and 17.0% are Tu.2
For comparison, table 2 also gives the ethnic composition of 10-year-oldsfor the counties in our sample from the 2010 national census. The composi-tion found in the school survey largely mirrors census data. Our survey covers aslightly larger proportion of minorities overall ð0.8% moreÞ and a slightlylarger proportion of Tu and Salar and smaller proportion of Hui and Tibetans.Note that some difference is to be expected given time trends and that theschool survey covers a wider age range.
B. The Achievement GapAccording to our data, there is a significant achievement gap between Han andminority students ðfig. 2Þ. Leftmost bars show the mean standardized examscores in math ðdark grayÞ and Chinese ðlight grayÞ for Han students; the nextpair of bars shows mean scores for all minority students; and the remainingbars show mean scores by minority group. The gap between all minoritystudents ðall ethnic groups pooled togetherÞ and Han students is substantial:0.29 SD in math and 0.25 SD in Chinese.The data also show a striking amount of heterogeneity in exam scores
among individual minority groups. For example, students from the Tu mi-nority perform comparably to Han students. In contrast, the scores of Salarstudents are nearly 0.75 SD below those of the Han students. Importantly,figure 2 suggests that language may be a factor contributing to China’s Han-minority achievement gap. The students from the two minority groups thattypically speak non-Mandarin languages ðSalar and TibetanÞ perform muchworse than Han students. At the same time, the achievement gap betweenstudents from the two minority groups that generally speak Mandarin as theirprimary language ðTu and HuiÞ and Han students is much narrower. Givensubstantial differences between the achievement of Mandarin-speaking and
2 And 0.75% belong to other minority groups. We exclude the other category from the analysis giventheir small number.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
non-Mandarin-speaking minority students, we analyze these two minority
TABLE 2SAMPLE COUNTY ETHNIC COMPOSITION IN SCHOOL SURVEY AND 2010 CENSUS (%)
Han Hui Tibetan Tu Salar
Census data:Gansu 95.74 3.95 .28 .00 .00Shaanxi 99.16 .80 .00 .00 .00Qinghai 40.62 23.44 13.00 7.72 15.00Total 88.33 7.96 1.79 1.43 .45
School survey data:Gansu 92.50 5.65 .65 .24 1.09Shaanxi 99.25 .63 .06 .00 .00Qinghai 46.38 21.47 9.97 16.46 5.72Total 87.53 7.04 1.69 2.12 1.54
Sources. 2010 census data ðChina Statistics Press, 2012Þ and authors’ survey.
328 E C O N O M I C D E V E L O P M E N T A N D C U L T U R A L C H A N G E
groups separately in addition to analyzing the pooled sample of all minoritystudents.
C. Endowments of Background CharacteristicsThe statistics in table 3 highlight some significant differences between Hanand minority students in terms of student and household characteristics. First,minority students from both categories are significantly older than Han stu-
Figure 2. Standardized exam results by ethnic group. Uses all observations in the data set. “Other mi-nority” group excluded from graph due to small sample size. Error bars give 95% confidence intervalsconstructed using 500 bootstrap replications accounting for clustering at the school level.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
dents by around 0.2–0.3 years ðtable 3Þ. In our sample, Han students were
Yang et al. 329
more likely to have repeated a grade compared to all minority groups; thus,this age difference likely reflects longer delays in primary school enrollment onthe part of minorities.3 Available evidence from other countries suggests thatdelayed enrollment may have a positive influence on academic achievementðGlewwe, Jacoby, and King 2001; McEwan and Shapiro 2008Þ; however,delayed enrollment may be due to malnutrition in early childhood ðGlewweand Jacoby 1995; Glewwe et al. 2001Þ. Second, minority students live insignificantly larger households, likely a reflection of differential treatment un-der family planning policies. Given evidence that there is a strong quality-quantity trade-off in rural China, having more siblings may disadvantageminority students ðLi, Zhang, and Zhu 2008Þ. The third significant differencebetween Han and minority students in our sample is that parents of minoritychildren are significantly less educated themselves. Numerous studies from avariety of contexts have shown evidence that parental education—particularlythe mother’s—has a causal influence on the academic achievement of chil-dren. Interestingly the one area in which minority students appear to be un-ambiguously better off is in terms of household asset ownership ðalthough thiscould reflect cheaper prices in regions where minorities are likely to liveÞ.Table 3 also shows differences in class peer characteristics. Minorities at-
tend classes with peers whose mothers are significantly less educated but whosefamilies possess more durable household assets compared to Han students.They also attend classes with a significantly smaller proportion of peers of thesame ethnicity as their own. To examine the distribution of Han and minor-ity students across schools in more detail, figure 3 plots kernel density esti-mates of this variable. These plots clearly show that Han students are muchmore concentrated in ethnically homogenous schools than are minority stu-dents. Nearly 33% of minority students are in the ethnic minority of theirclass, while this figure is only 1% for Han students.A priori it is unclear what affect peer ethnic composition may have on
student achievement for minority and Han students. Minorities ðand theirminority peersÞ are of generally lower socioeconomic status; however, theremay be advantages to attending school with peers of the same ethnicity.Beyond theories related to social identity ðAkerlof and Kranton 2002Þ, non-Mandarin minority students may benefit from classes in which teachers aremore likely to teach ðentirely or partlyÞ in the local language.4
3
In our sample, 39.5% of Han students repeated a grade. This is significantly more likely than forHui ð7 percentage points, p 5 .04Þ and Tu ð11 percentage points, p 5 .01Þ students.4 In our sample, no teachers report doing so.This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE
3SU
MMARYST
ATIST
ICS Minority
Stud
ents
Full
Sample
ð1Þ
Han
Stud
ents
ð2Þ
All ð3Þ
Man
darin
Spea
king
ð4Þ
Non
-Man
darin
Spea
king
ð5Þ
Differen
ceð4Þ2
ð2Þ
½P-Va
lue�
Differen
ceð5Þ2
ð2Þ
½P-Va
lue�
Stud
entan
dho
useh
oldch
aracteristic
s:Stan
dardized
mathex
amscore
.00
.04
2.25
2.14
2.58
2.17
2.62
ð1.00Þ
ð1.00Þ
ð.98Þ
ð.97Þ
ð.93Þ
½.00�
½.00�
Stan
dardized
Chine
seex
amscore
.00
.03
2.22
2.07
2.62
2.10
2.65
ð1.00Þ
ð.99Þ
ð1.06Þ
ð1.00Þ
ð1.11Þ
½.11�
½.00�
Femaleð0/1Þ
.49
.49
.48
.48
.50
2.01
.01
ð.50Þ
ð.50Þ
ð.50Þ
ð.50Þ
ð.50Þ
½.45�
½.62�
Boa
rdingstud
entð0/1Þ
.07
.07
.10
.08
.16
.02
.09
ð.26Þ
ð.25Þ
ð.30Þ
ð.28Þ
ð.36Þ
½.58�
½.19�
Ageðye
arsÞ
10.83
10.80
11.03
11.01
11.08
.21
.28
ð1.13Þ
ð1.14Þ
ð1.04Þ
ð1.08Þ
ð.90Þ
½.00�
½.00�
Hou
seho
ldsize
5.32
5.27
5.64
5.63
5.68
.35
.40
ð1.55Þ
ð1.52Þ
ð1.71Þ
ð1.64Þ
ð1.92Þ
½.00�
½.01�
Travel
timeto
scho
olðm
inutesÞ
25.72
25.91
24.34
23.85
25.63
22.07
2.29
ð26.26
Þð26.55
Þð24.03
Þð22.90
Þð26.71
Þ½.2
9�½.9
4�Mothe
rha
slower
seco
ndarydeg
reeor
abov
eð0/1Þ
.27
.28
.19
.21
.14
2.08
2.15
ð.44Þ
ð.45Þ
ð.39Þ
ð.40Þ
ð.34Þ
½.00�
½.00�
Father
haslower
seco
ndarydeg
reeor
abov
eð0/1Þ
.46
.48
.36
.37
.32
2.11
2.16
ð.50Þ
ð.50Þ
ð.48Þ
ð.48Þ
ð.47Þ
½.00�
½.00�
Father
atho
með0/1Þ
.58
.58
.55
.55
.52
2.03
2.06
ð.49Þ
ð.49Þ
ð.50Þ
ð.50Þ
ð.50Þ
½.27�
½.18�
Mothe
rat
Hom
eð0/1Þ
.68
.69
.66
.67
.63
2.02
2.05
ð.46Þ
ð.46Þ
ð.47Þ
ð.47Þ
ð.48Þ
½.48�
½.14�
Hou
seho
ldassetindex
ð0/1Þ
.00
2.05
.36
.30
.53
.35
.58
ð1.44Þ
ð1.43Þ
ð1.48Þ
ð1.45Þ
ð1.53Þ
½.00�
½.00�
330
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Minority
Stud
ents
Class
pee
rch
aracteristic
:Prop
ortio
nof
peers’mothe
rswith
lower
seco
ndary
degree
orab
ove
.27
.28
.20
.22
.14
2.06
2.14
ð.17Þ
ð.17Þ
ð.16Þ
ð.17Þ
ð.10Þ
½.02�
½.00�
Prop
ortio
nof
classpee
rsof
sameethn
icity
.93
.96
.68
.70
.64
2.26
2.32
ð.18Þ
ð.10Þ
ð.36Þ
ð.36Þ
ð.36Þ
½.00�
½.00�
Peer
averag
eho
useh
oldassetindex
.00
2.05
.35
.28
.56
.33
.61
ð.63Þ
ð.62Þ
ð.61Þ
ð.54Þ
ð.74Þ
½.00�
½.00�
Teache
rch
aracteristic
:Fe
maleteache
rð0/1Þ
.40
.39
.49
.48
.51
.09
.11
ð.49Þ
ð.49Þ
ð.50Þ
ð.50Þ
ð.50Þ
½.10�
½.18�
Han
teache
rð0/1Þ
.90
.96
.51
.51
.48
2.44
2.48
ð.30Þ
ð.20Þ
ð.50Þ
ð.50Þ
ð.50Þ
½.00�
½.00�
Teache
rha
shighe
red
ucationdeg
reeð0/1Þ
.81
.80
.89
.89
.89
.09
.09
ð.39Þ
ð.40Þ
ð.32Þ
ð.31Þ
ð.31Þ
½.00�
½.05�
Teache
rattend
edno
rmal
colle
geð0/1Þ
.79
.78
.88
.88
.89
.10
.11
ð.40Þ
ð.41Þ
ð.32Þ
ð.32Þ
ð.31Þ
½.01�
½.01�
Teache
rha
srece
ived
provinc
ialo
rna
tiona
ltea
ching
awardð0/1Þ
.08
.08
.10
.11
.10
.03
.02
ð.27Þ
ð.26Þ
ð.30Þ
ð.30Þ
ð.30Þ
½.38�
½.72�
Gong
ban
teache
rð0/1Þ
.87
.88
.81
.82
.78
2.06
2.09
ð.34Þ
ð.33Þ
ð.39Þ
ð.39Þ
ð.41Þ
½.32�
½.09�
Teache
rex
perienc
eðyea
rsÞ
13.39
13.53
12.45
12.16
13.11
21.38
2.42
ð10.96
Þð11.11
Þð9.86Þ
ð10.16
Þð8.79Þ
½.29�
½.74�
Scho
olch
aracteristic
:Sc
hool
size
ðstud
entsÞ
221.05
221.47
218.08
224.03
201.71
2.59
219
.74
ð59.25
Þð58.65
Þð63.25
Þð65.51
Þð53.59
Þ½.8
5�½.1
0�Stud
ent-teache
rratio
17.41
17.29
18.28
18.50
17.63
1.21
.34
ð5.11Þ
ð5.17Þ
ð4.55Þ
ð4.57Þ
ð4.36Þ
½.14�
½.72�
Distanc
eto
farthe
stvillageserved
byscho
olðm
inutesÞ
66.83
67.70
60.77
57.56
69.62
210
.14
1.92
ð50.63
Þð50.75
Þð49.39
Þð39.84
Þð68.78
Þ½.1
7�½.9
2�
331
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE
3(C
ontin
ued)
Minority
Stud
ents
Full
Sample
ð1Þ
Han
Stud
ents
ð2Þ
All ð3Þ
Man
darin
Spea
king
ð4Þ
Non
-Man
darin
Spea
king
ð5Þ
Differen
ceð4Þ2
ð2Þ
½P-Va
lue�
Differen
ceð5Þ2
ð2Þ
½P-Va
lue�
Scho
olha
sprovided
teache
rtraining
inpastye
arð0/1Þ
.92
.95
.77
.77
.77
2.17
2.18
ð.26Þ
ð.23Þ
ð.42Þ
ð.42Þ
ð.42Þ
½.04�
½.12�
Scho
olinfrastruc
ture
index
.00
.01
2.07
.13
2.61
.12
2.62
ð1.21Þ
ð1.18Þ
ð1.38Þ
ð1.16Þ
ð1.77Þ
½.55�
½.22�
Sample
size:
Totaln
umber
ofstud
ents
19,129
16,741
2,38
81,75
361
7Num
ber
ofstud
ents—mathsample
9,46
88,28
61,18
287
130
1Num
ber
ofstud
ents—Chine
sesample
9,66
18,45
51,20
688
231
6Num
ber
ofscho
ols
300
285
191
167
64
Note.Variablesareas
described
intable
1.Stan
darderrors
ðinparen
thesesÞa
ccou
ntforclusterin
gat
thescho
olleve
l.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
In terms of teacher and school quality ðtable 3Þ, minorities appear to be,
Figure 3. Distribution of class peer ethnic composition by ethnic group. Kernel density estimated usinga bandwidth of 0.07.
Yang et al. 333
if anything, better off than their Han counterparts. For example, teachers ofminority students are significantly more likely to have a higher education degreeand to have attended a specialized teaching college. Schools attended by Hanand minority students are similar in terms of size, student-teacher ratio, re-moteness, and infrastructure ðalthough minority schools are slightly less likelyto have provided teacher training in the past yearÞ. This may be a result of sig-nificant government educational investment focused on minority areas.Characteristics of students in mixed Han and minority schools ðwith at
least two Han and two minority studentsÞ are given in the appendix ðta-ble A1Þ. We construct mixed school samples for both types of minority stu-dents. Mixed Mandarin minority schools have at least two Mandarin minoritystudents, and mixed non-Mandarin minority schools have at least two non-Mandarin minority students. For the most part, mean differences in char-acteristics between Han and minority students attending the same schools areless significant than the full sample, as would be expected.
V. Returns to Minority StatusWe take a first look at the relationship between the achievement gap andobserved characteristics directly by estimating how the Han-minority achieve-
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
ment gap changes as we adjust for characteristics collected as part of our sur-
334 E C O N O M I C D E V E L O P M E N T A N D C U L T U R A L C H A N G E
vey. That is, we estimate variants of the following regression
Yis 5 a1 b1Mandarin Minority
1 b2Non Mandarin Minority 1 X 0v1 εis;
ð1Þ
where Yis is the normalized test score of student i in schools;MandarinMinorityis a dummy variable equal to 1 if the student is Hui or Tu; Non_MandarinMinority is a dummy variable equal to 1 if the student is Tibetan or Salar; X 0 is avector of student and household, peer, teacher, and school characteristics; and εisis an error term possibly correlated at the school level. The coefficients ofinterest are b1 and b2. How these two coefficients change as we add char-acteristics to the X 0 vector from the error term provides a first look at the abilityof these characteristics to account for differences in achievement between Hanstudents and Mandarin and non-Mandarin minority students.The results of this analysis for standardized math scores are shown in
table 4.5 The raw mean differences are20.17 SD for Mandarin minority stu-dents and 20.62 SD for non-Mandarin minority students ðcol. 1Þ. Controll-ing for student and household characteristics reduces the size of these es-timates to 20.12 and 20.51 SD, respectively ðcol. 2Þ. Sequentially addingpeer, teacher, and school characteristics ðcols. 3–5Þ shows that once studentand peer characteristics are controlled for, the Mandarin minority coefficientdecreases in size and becomes insignificant. The coefficient on non-Mandarinminorities remains large ð20.2 SDÞ and significant even after controlling forschool fixed effects ðcol. 6Þ. In other words, Mandarin minority students scorean average of 0.2 SD less than Han students with similar individual, peer,and teacher characteristics in the same schools. Adding school fixed effectsði.e., controlling for all observed and unobserved school-level characteris-ticsÞ does reduce the estimated gap for this group by more than half, whichsuggests that—despite detailed controls—unobserved school-level heteroge-neity is an important factor.
VI. Decomposing the Han-Minority Achievement GapTo decompose the Han-minority achievement gap, we first estimate educa-tional productions functions, or achievement regressions, that quantify returnsto individual, family, teacher, and school-level characteristics for each of ourstudent classifications ðHan, Mandarin minority, and non-Mandarin minor-ityÞ. We then use the traditional Oaxaca-Blinder decomposition method
5 Results for Chinese scores are similar. These results are in table A2.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
ðBlinder 1973; Oaxaca 1973Þ to decompose the achievement gap—between
TABLE 4MATH ACHIEVEMENT REGRESSIONS (POOLED, FULL SAMPLE)
ð1Þ ð2Þ ð3Þ ð4Þ ð5Þ ð6ÞMandarin-speaking minority 2.17*** 2.12** 2.09 2.08 2.06 2.03
ð.058Þ ð.056Þ ð.063Þ ð.065Þ ð.063Þ ð.066ÞNon-Mandarin-speaking minority 2.62*** 2.51*** 2.47*** 2.46*** 2.46*** 2.19*
ð.097Þ ð.096Þ ð.101Þ ð.093Þ ð.090Þ ð.111ÞStudent and household characteristics Yes Yes Yes Yes YesClass peer characteristics Yes Yes Yes YesTeacher characteristics Yes Yes YesSchool characteristics YesSchool fixed effects YesConstant .04 1.92*** 1.27** 1.18* 1.57** 1.13*
ð.027Þ ð.623Þ ð.631Þ ð.647Þ ð.643Þ ð.628ÞAdjusted R 2 .013 .091 .100 .101 .114 .209
Note. Each column represents a separate regression. Standard errors ðin parenthesesÞ account for clus-tering at the school level. Student and household characteristics, class peer characteristics, teacher char-acteristics, and school characteristics include those in table 1. Estimation sample includes a randomlychosen half of all sample students ðthose who were given a standardized exam in mathÞ. N 5 9,468.* p < .1.** p < .05.*** p < .01.
Yang et al. 335
Han students and both types of minority students. We decompose the gapinto two components. First, there is a component that can be explained bydifferences in student, peer, teacher, and school characteristics. In the rest ofthe analysis, we refer to this component as that due to “differences in char-acteristics.” The second component is due to between-group differences in re-turns to characteristics.The achievement regressions that we use in the decomposition are based on
the following linearized specification of the educational production function:
Yis 5 a1 b1Iis 1 b2Pis 1 b3Tis 1 b4Sis 1 εis; ð2Þ
where, as above, Yis is the observed test score of student i in schools, Iis is avector of individual student and household variables, Pis is a vector of peergroup variables, Tis is a vector of teacher characteristics, Sis is a vector of schoolvariables, and εis is an error term. The error term is allowed to be correlatedat the school level to account for clustering effects. In some specifications, wesubstitute Sis for school fixed effects ðgsÞ to control for unobserved hetero-geneity at the school level.The Han-minority achievement gap ðdifference in test scoresÞ can be ex-
pressed as
Y *H 2 Y *
M
� �5 X *
H 2 X *M
� �bH 1 X *
M bH 2 bMð Þ; ð3Þ
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
where Y *H and Y *
M are the predicted mean standardized test scores of Han and* *
336 E C O N O M I C D E V E L O P M E N T A N D C U L T U R A L C H A N G E
minority students, XH and XM are the mean characteristics of Han and minor-ity students ðIis, Pis, Tis, and SisÞ; and bH and bM are the returns to characteristicsfor Han and minority students estimated using equation ð2Þ above. Note that,because individual school fixed effects cannot be estimated for minority ðHanÞstudents in schools where no minority ðHanÞ students attend, we restrict thesample to only mixed schools with at least two Han and two minority studentsin analysis that includes school fixed effects.The overall difference in exam scores can, therefore, be decomposed into
two components. One is the portion attributable to differences in the quan-tity of characteristics, evaluated using Han returns: bH X *
H 2 X *Mð Þ. The other
portion, X *M bH 2 bMð Þ, is that attributable to differences in returns to the
characteristics of Han and minority students.
A. Returns to Characteristics by Ethnic GroupTable 5 reports the results of separate math achievement regressions for Hanstudents, Mandarin minority students, and non-Mandarin minority stu-dents.6 The odd-numbered columns in the table include all characteristics intable 3; even-numbered columns substitute school characteristics for schoolfixed effects. The coefficients from these regressions ðwhich are the measuredachievement returns to the characteristicsÞ are used in the Oaxaca-Blinder de-compositions below.A few insights emerge from comparing the estimated returns to inputs
across groups. First, the pattern of returns for Han and Mandarin minoritystudents are similar ðcomparing the coefficients in table 5 cols. 1 and 2 for theHan students with the coefficients in cols. 3 and 4 for the Mandarin minor-ity studentsÞ. While some coefficient estimates for Mandarin minorities arenot significant, point estimates largely coincide. One exception is the coeffi-cient on age: after controlling for school-level fixed effects, it appears thatMandarin-speaking minority students benefit from delayed school enrollment.But, some estimated returns for non-Mandarin minority students differ
from the other two groups. For example, non-Mandarin students appear tobe strongly and negatively affected by a larger proportion of classmates ofthe same ethnicity ðcols. 5 and 6Þ. The differences in estimated returns toclass peer ethnic composition are highlighted in figure 4. While both Han
6 Results for Chinese scores are in table A3. Because power is reduced by separating the two minority
groups ðMandarin and non-Mandarin minoritiesÞ, we also conducted all analyses pooling studentswho were given the Chinese exam and students who were given the math exam to estimate returnswith more precision. Qualitative results of the analyses do not change substantially when using thepooled sample.This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
students and Mandarin minority students benefit slightly from being in classes
Yang et al. 337
with more students of their same ethnicity, there is a negative correlation amongnon-Mandarin minority students even after controlling for fixed school-levelfactors. In other words, having more class peers of a students’ same ethnicityhas a large, negative relationship with achievement of non-Mandarin minoritystudents. Given the large degree of underperformance of students from the non-Mandarin minority group, this correlation may be in part due to the effect ofhaving lower-achieving peers. Non-Mandarin-speaking students also appearto be strongly influenced by the quality of teaching. Both the coefficient onhaving a teacher who has received a teaching award ðcols. 3 and 6Þ and thecoefficient on the school having provided teacher training are large and sig-nificant for this group but not in others.Table 6 repeats these regressions for the sample of mixed schools.7 Com-
pared to the full sample, estimated returns are much more similar for Hanand minority students ðof both typesÞ attending the same schools. This sug-gests that the large differences in returns observed in the full sample are largelydue to differences between Han students in Han-only schools and minoritiesin minority-only schools.
B. Returns to Schools by Ethnic GroupWhile estimated returns to observed school characteristics are similar for Hanand minority students, there may still be differences in estimated school fixedeffects for Han and minority students. That is, returns to specific schoolsðaccounting for observed and unobserved characteristicsÞ may differ betweenHan and minority students. To examine this in more detail, we estimate theschool fixed effect version of equation ð2Þ for Han and minority studentsseparately, using the sample of mixed schools with at least two Han studentsand two minority students:8
Y His 5 aH 1 X H 0
bH 1 gHs 1 εHis ; ð4aÞ
Y Mis 5 aM 1 XM 0
bM 1 gMs 1 εMis ; ð4bÞ
where X includes the same student, peer, and teacher characteristics as above,and gs is a vector of school dummy variables. We interpret the estimatedschool fixed effects for Han students ðgH
s Þ and minority students ðgMs Þ as the
return of attending a specific school for Han and minority students, respectively,
7 Results for Chinese scores are in table A4.8 For this part of the analysis, we pool both types of minority students.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE
5MATH
ACHIEVEMENTREGRESS
IONSBYETH
NICITY
Han
Stud
ents
Man
darin-Spea
king
Minority
Non
-Man
darin-Spea
king
Minority
ð1Þ
ð2Þ
ð3Þ
ð4Þ
ð5Þ
ð6Þ
Stud
entan
dho
useh
oldch
aracteristic
s:Fe
maleð0/1Þ
2.22***
2.22***
2.25***
2.24***
2.10
2.13
ð.021
Þð.0
20Þ
ð.061
Þð.0
72Þ
ð.076
Þð.0
85Þ
Boa
rdingstud
entð0/1Þ
2.17***
2.15***
2.34*
2.28
2.24
.05
ð.042
Þð.0
46Þ
ð.173
Þð.1
98Þ
ð.148
Þð.1
01Þ
Ageðye
arsÞ
2.16
2.13
.49
1.09
*.77
2.72
ð.113
Þð.1
07Þ
ð.533
Þð.5
78Þ
ð.855
Þð1.099
ÞAge2
.00
2.00
2.02
2.05**
2.03
.03
ð.005
Þð.0
05Þ
ð.023
Þð.0
25Þ
ð.038
Þð.0
47Þ
Hou
seho
ldsize
2.01**
2.01
2.02
2.03
.02
.01
ð.007
Þð.0
06Þ
ð.020
Þð.0
23Þ
ð.021
Þð.0
25Þ
Travel
timeto
scho
olðm
inutesÞ
.00***
.00***
.00
.00
.00
.00
ð.000
Þð.0
01Þ
ð.001
Þð.0
02Þ
ð.002
Þð.0
02Þ
Mothe
rha
slower
seco
ndarydeg
reeor
abov
eð0/1Þ
.03
.03
.04
.08
2.35**
2.23
ð.025
Þð.0
25Þ
ð.084
Þð.0
93Þ
ð.141
Þð.1
39Þ
Father
haslower
seco
ndarydeg
reeor
abov
eð0/1Þ
.25***
.20***
.19***
.21***
.13
.14
ð.024
Þð.0
24Þ
ð.071
Þð.0
80Þ
ð.145
Þð.1
66Þ
Father
atho
með0/1Þ
2.00
.01
2.13*
2.11
2.11
2.07
ð.022
Þð.0
22Þ
ð.069
Þð.0
83Þ
ð.140
Þð.1
69Þ
Mothe
rat
homeð0/1Þ
2.00
2.03
.12
.07
2.13
2.08
ð.026
Þð.0
24Þ
ð.076
Þð.0
94Þ
ð.104
Þð.1
15Þ
Hou
seho
ldassetindex
ð0/1Þ
.02**
.02**
.02
.02
2.00
.01
ð.008
Þð.0
08Þ
ð.024
Þð.0
29Þ
ð.030
Þð.0
36Þ
Class
pee
rch
aracteristic
:Prop
ortio
nof
pee
rs’mothe
rswith
lower
seco
ndarydeg
reeor
abov
e.17
2.43**
.23
.60
2.46
1.81
***
ð.154
Þð.2
01Þ
ð.246
Þð.5
37Þ
ð.867
Þð.5
33Þ
Prop
ortio
nof
classpee
rsof
sameethn
icity
.20
.09
.10
.44
2.66**
23.28
***
ð.171
Þð.5
20Þ
ð.109
Þð.5
36Þ
ð.272
Þð1.207
ÞPe
eraverag
eho
useh
oldassetindex
.08**
2.06
.16*
2.05
.13
2.23
ð.039
Þð.0
62Þ
ð.084
Þð.1
53Þ
ð.155
Þð.1
44Þ
338
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Han
Stud
ents
Man
darin-Spea
king
Minority
Non
-Man
dar-
in-
Spea
king
Minority
Teache
rch
aracteristic
:Fe
maleteache
rð0/1Þ
.01
2.01
2.05
2.13
2.28*
.08
ð.047
Þð.0
47Þ
ð.083
Þð.1
39Þ
ð.164
Þð.1
33Þ
Han
teache
rð0/1Þ
.04
.02
2.05
2.13
2.26
2.84***
ð.094
Þð.1
46Þ
ð.077
Þð.1
11Þ
ð.168
Þð.2
16Þ
Teache
rha
shighe
red
ucationdeg
reeð0/1Þ
2.01
.03
.19
2.38*
.12
.20
ð.059
Þð.0
76Þ
ð.132
Þð.2
15Þ
ð.202
Þð.1
20Þ
Teache
rattend
edno
rmal
colle
geð0/1Þ
2.03
2.02
.03
2.32
.13
2.28
ð.049
Þð.0
50Þ
ð.165
Þð.1
99Þ
ð.196
Þð.1
74Þ
Teache
rha
srece
ived
provinc
ialo
rna
tiona
ltea
chingaw
ardð0/1Þ
2.05
.03
2.04
2.24
1.03
***
1.14
***
ð.057
Þð.0
65Þ
ð.119
Þð.1
76Þ
ð.243
Þð.2
13Þ
Gong
ban
teache
rð0/1Þ
2.01
.07
.25
.38
2.11
.05
ð.059
Þð.0
53Þ
ð.169
Þð.3
43Þ
ð.138
Þð.0
85Þ
Teache
rex
perienc
eðye
arsÞ
2.00
2.00
2.00
2.02***
.01
.03***
ð.002
Þð.0
03Þ
ð.005
Þð.0
08Þ
ð.010
Þð.0
08Þ
Scho
olch
aracteristic
:Sc
hool
size
ðstud
entsÞ
2.00
.00*
.00
ð.000
Þð.0
01Þ
ð.002
ÞStud
ent-teache
rratio
2.02***
2.00
2.01
ð.004
Þð.0
10Þ
ð.012
ÞDistanc
eto
farthe
stvillageserved
byscho
olðm
inutesÞ
.00
.00*
2.00
ð.000
Þð.0
01Þ
ð.002
ÞSc
hool
hasprovided
teache
rtraining
inpastye
arð0/1Þ
.02
2.01
.61***
ð.083
Þð.1
10Þ
ð.193
ÞSc
hool
infrastruc
ture
index
.01
.12***
.08**
ð.024
Þð.0
31Þ
ð.040
ÞSc
hool
fixed
effects
No
Yes
No
Yes
No
Yes
Con
stan
t1.81
***
1.59
23.27
25.53
25.31
5.78
ð.676
Þð1.044
Þð3.072
Þð3.356
Þð4.817
Þð6.541
ÞObservations
8,28
687
130
1Adjusted
R2
.109
.205
.107
.153
.078
.172
Note.Ea
chco
lumnrepresentsaseparateregression.
Stan
darderrors
ðinparen
thesesÞa
ccou
ntforc
lusteringat
thescho
olleve
l.Estim
ationsample
includ
esarand
omlych
osen
halfof
allsam
ple
stud
ents
ðthosewho
weregiven
astan
dardized
exam
inmathÞ.
*p<.1.
**p<.05.
***p<.01.
339
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Figure4.
Pred
ictedreturnsto
classpee
rethn
icco
mpositio
nbyethn
icgroup
.A,M
ath,
noscho
olfi
xedeffects;
B,m
athwith
scho
olfi
xedeffects;
C,Chine
se,no
scho
olfi
xed
effects;D,C
hine
sewith
scho
olfi
xedeffects.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
relative to a reference school ðthe school whose dummy variable is omitted9
Yang et al. 341
from the regressionsÞ.We estimate that, on average, school fixed effects estimated for Han stu-
dents are 0.3 SD higher in math and 0.39 SD higher in Chinese comparedto those for minority students. Both of these differences are significant at 1%.To compare the effects of a specific school on Han and minority students di-rectly, figure 5 plots the school coefficients for Han ðgH
s Þ against those es-timated for minority students ðgM
s Þ. Figure 5A does this for math scores,and 5B for Chinese scores. In these figures, the majority of schools ð63% ofschools for math and 72% for ChineseÞ lie below the 45° line ðwhere gH
s andgMs are equalÞ. Individual schools tend to generate larger returns for Han com-
pared to similar minority students attending the same school. In other words,the benefits that Han students receive from ðobserved and unobservedÞ attri-butes of individual schools tend to be larger than the benefits received by mi-nority students.What types of schools have larger differences in their effect on Han and
minority students? We examine school-level differences in Han and minor-ity effects by estimating the following regression:
gHs 2 gM
s
� �5 a1 bXs 1 εs; ð5Þ
where Xs is a vector of school-level characteristics and εs is an error term.Here, the Xs vector includes the same teacher characteristics ðaggregated to theschool levelÞ and school characteristics as above, as well as the proportion ofstudents belonging to a Mandarin minority group and the proportion be-longing to a non-Mandarin minority group. We use White-Huber standarderrors to account for heteroskedasticity.The results of this analysis are in table 7. In the full models for math and
Chinese, observed covariates explain more than 25% of the variation of thedifference between the return of school characteristics to Han students andto minority students. Focusing on the results for math, it appears that teach-ers play the most significant role in reducing the Han-minority difference inreturns. Coefficients on variables related to teachers’ education and experi-ence are negative and highly significant. Assuming that these variables ðhaving ahigher education degree, attending a normal college, and teaching experienceÞreflect teaching quality, these results suggest that pedagogical practice in theclassroom highly influences how much Han and minority students benefit fromspecific schools.
9 This analysis is similar to that used in Meng ð2004Þ to examine the effect of firm-level wage policies
on gender wage gaps.This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE
6MATH
ACHIEVEMENTREGRESS
IONSBYETH
NICITY
(MIXED
SCHOOLS
ONLY
)
Mixed
Man
darin
Minority
Scho
ols
Mixed
Non
-Man
darin
Minority
Scho
ols
Han
Stud
ents
Man
darin-Spea
king
Minority
Stud
ents
Han
Stud
ents
Non
-Man
darin-
Spea
king
Minority
Stud
ents
ð1Þ
ð2Þ
ð3Þ
ð4Þ
ð5Þ
ð6Þ
ð7Þ
ð8Þ
Stud
entan
dho
useh
oldch
aracteristic
s:Fe
maleð0/1Þ
2.20***
2.17***
2.40***
2.31***
2.13
2.12
2.25
2.22
ð.054
Þð.0
58Þ
ð.094
Þð.1
03Þ
ð.088
Þð.0
89Þ
ð.162
Þð.1
76Þ
Boa
rdingstud
entð0/1Þ
2.30***
2.26**
2.70***
2.56***
.13
.10
.17
.14
ð.084
Þð.1
12Þ
ð.186
Þð.1
79Þ
ð.163
Þð.2
09Þ
ð.117
Þð.0
90Þ
Ageðyea
rsÞ
.18
2.12
.24
.31
1.10
**1.34
***
.92
.43
ð.262
Þð .2
51Þ
ð.638
Þð.6
75Þ
ð.421
Þð.4
46Þ
ð1.086
Þð1.271
ÞAge2
2.01
2.00
2.01
2.02
2.06***
2.07***
2.03
2.02
ð.011
Þð.0
11Þ
ð.027
Þð.0
28Þ
ð.017
Þð.0
19Þ
ð.048
Þð.0
56Þ
Hou
seho
ldsize
2.02
2.02
.03
.00
2.00
2.01
.04
.05
ð.018
Þð.0
18Þ
ð.032
Þð.0
36Þ
ð.045
Þð.0
47Þ
ð.040
Þð.0
39Þ
Travel
timeto
scho
olðm
inutesÞ
.00
2.00
.00*
.00**
2.00
2.00
.00
.00
ð.002
Þð.0
02Þ
ð.001
Þð.0
01Þ
ð.003
Þð.0
03Þ
ð.002
Þð.0
02Þ
Mothe
rha
slower
seco
ndary
deg
reeor
abov
eð0/1Þ
.06
.05
.04
.04
2.05
2.08
2.38
2.18
ð.063
Þð.0
65Þ
ð.114
Þð.1
19Þ
ð.157
Þð.1
67Þ
ð.228
Þð.2
28Þ
Father
haslower
seco
ndarydeg
ree
orab
oveð0/1Þ
.32***
.30***
.32***
.39***
.26**
.27**
.22
.21
ð.064
Þð.0
65Þ
ð.114
Þð.1
29Þ
ð.109
Þð.1
15Þ
ð.219
Þð.2
28Þ
Father
atho
með0/1Þ
.06
.08
2.16
2.11
2.09
2.10
2.16
2.10
ð.060
Þð.0
66Þ
ð.106
Þð.1
18Þ
ð.115
Þð.1
15Þ
ð.235
Þð.2
16Þ
Mothe
rat
homeð0/1Þ
.05
.06
.20*
.13
2.12
2.11
2.17
2.09
ð.068
Þð.0
70Þ
ð.108
Þð.1
42Þ
ð.109
Þð.1
15Þ
ð.245
Þð.2
87Þ
Hou
seho
ldassetindex
ð0/1Þ
.03
.03
2.01
2.03
.03
.03
.03
.02
ð.021
Þð.0
21Þ
ð.034
Þð.0
36Þ
ð.052
Þð.0
55Þ
ð.056
Þð.0
65Þ
Class
pee
rch
aracteristic
:Prop
ortio
nof
pee
rs’mothe
rswith
lower
seco
ndarydeg
reeor
abov
e.07
2.10
.08
.65
.05
.98
.02
2.59
ð.304
Þð.5
66Þ
ð.370
Þð.8
76Þ
ð.843
Þð1.154
Þð2.162
Þð3.253
ÞProp
ortio
nof
classpee
rsof
sameethn
icity
2.20
.13
.38
.61
.06
.92
.38
.24
ð.263
Þð.5
92Þ
ð.241
Þð.4
54Þ
ð.328
Þð1.026
Þð.8
44Þ
ð1.802
ÞPe
eraverag
eho
useh
oldassetindex
2.06
2.36**
.30**
2.10
2.22
2.06
2.40
2.53
ð.100
Þð.1
62Þ
ð.135
Þð.3
40Þ
ð.181
Þð.2
03Þ
ð.325
Þð.3
95Þ
342
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Mixed
Man
darin
Minority
Scho
ols
Mixed
Non
-Man
darin
Minority
Scho
ols
Teache
rch
aracteristic
:Fe
maleteache
rð0/1Þ
2.18**
2.22**
2.25**
2.27
.12
2.29
.03
.03
ð.078
Þð.1
07Þ
ð.110
Þð.2
24Þ
ð.139
Þð.3
03Þ
ð.267
Þð.4
69Þ
Han
teache
rð0/1Þ
.06
.09
2.03
2.16
2.41**
2.20
2.75***
2.50
ð.158
Þð.1
82Þ
ð.149
Þð.1
86Þ
ð.148
Þð.1
75Þ
ð.249
Þð.3
04Þ
Teache
rha
shighe
red
ucationdeg
reeð0/1Þ
2.04
2.09
2.27
21.06
***
.23
.47**
2.02
2.29
ð.105
Þð.1
25Þ
ð.210
Þð.3
08Þ
ð.157
Þð.2
17Þ
ð.273
Þð.3
41Þ
Teache
rattend
edno
rmal
colle
geð0/1Þ
2.00
.05
.01
2.37
2.29
.03
2.54
21.34
***
ð.083
Þð.1
24Þ
ð.244
Þð.2
62Þ
ð.187
Þð.2
89Þ
ð.342
Þð.2
96Þ
Teache
rha
srece
ived
provinc
ial
orna
tiona
ltea
chingaw
ardð0/1Þ
2.03
.01
2.19
2.20
2.65
2.85***
.40
.47
ð.097
Þð.1
42Þ
ð.126
Þð.1
73Þ
ð.404
Þð.2
31Þ
ð.398
Þð.4
66Þ
Gong
ban
teache
rð0/1Þ
.23***
.18
.77***
.74
.60**
.58
.41
1.70
ð.072
Þð.1
35Þ
ð.239
Þð.4
75Þ
ð.253
Þð.5
26Þ
ð.731
Þð1.357
ÞTe
ache
rex
perienc
eðyea
rsÞ
2.01*
2.01
2.01
2.03***
2.01
.01
2.01
2.04
ð.004
Þð.0
06Þ
ð.006
Þð.0
08Þ
ð.008
Þð.0
16Þ
ð.020
Þð.0
37Þ
Scho
olch
aracteristic
:Sc
hool
size
ðstud
entsÞ
2.00
.00
2.00**
2.00
ð.001
Þð.0
01Þ
ð.001
Þð.0
03Þ
Stud
ent-teache
rratio
2.03***
2.03*
.00
.06*
ð.011
Þð.0
18Þ
ð.019
Þð.0
27Þ
Distanc
eto
farthe
stvillageserved
byscho
olðm
inutesÞ
.00
.00**
.00
2.00
ð.001
Þð.0
02Þ
ð.001
Þð.0
02Þ
Scho
olha
sprovided
teache
rtraining
inpastye
arð0/1Þ
2.23
2.28
21.07
***
2.16
ð.174
Þð.2
55Þ
ð.278
Þð.2
39Þ
Scho
olinfrastruc
ture
index
.09**
.03
2.16**
2.09
ð.041
Þð.0
47Þ
ð.062
Þð.1
28Þ
Scho
olfix
edeffects
No
Yes
No
Yes
No
Yes
No
Yes
Con
stan
t.50
1.26
21.47
2.80
24.05
28.13
***
26.09
22.85
ð1.566
Þð1.518
Þð3.782
Þð4.038
Þð2.515
Þð2.738
Þð6.261
Þð7.364
ÞObservations
1,33
840
139
113
9Adjusted
R2
.149
.187
.163
.234
.092
.094
.082
.193
Note.Ea
chco
lumnrepresentsaseparateregression.
Stan
darderrorsðin
paren
thesesÞa
ccou
ntforclusteringat
thescho
olleve
l.Estim
ationsampleinclud
esarand
omlych
osen
half
ofallsam
ple
stud
ents
ðthosewho
weregiven
astan
dardized
exam
inmathÞ
inmixed
scho
ols.
*p<.1.
**p<.05.
***p<.01.
343
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
C. Decomposition Results
Figure 5. School fixed effects by ethnicity. A, Math; B, Chinese. Estimated using all mixed schools withmore than two minority students and two Han students.
344 E C O N O M I C D E V E L O P M E N T A N D C U L T U R A L C H A N G E
The results of the Oaxaca-Blinder decomposition for math are presented intable 8.10 The first three columns show results for the full sample, the nextthree for the mixed school sample without including school fixed effects, and
10 Results for Chinese are shown in table A5.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE 7CORRELATES OF DIFFERENCES IN RETURNS TO SCHOOL CHARACTERISTICS (SCHOOL FE)
BETWEEN HAN AND MINORITY STUDENTS
Math Chinese
ð1Þ ð2Þ ð3Þ ð4Þ ð5Þ ð6ÞProportion Mandarin minority students 2.11 .12 .12 2.26 .22 .22
ð.328Þ ð.304Þ ð.417Þ ð.230Þ ð.262Þ ð.375ÞProportion non-Mandarin minority students .55* .54 .69 1.11*** 1.36*** .22
ð.328Þ ð.433Þ ð.558Þ ð.310Þ ð.340Þ ð.571ÞProportion of female teachers 2.06 2.08 .16 .19
ð.175Þ ð.205Þ ð.202Þ ð.200ÞProportion of Han teachers 2.31 2.29 .43** .37
ð.314Þ ð.419Þ ð.206Þ ð.292ÞProportion of teachers with higher
education degree 2.81*** 2.92*** 2.19 2.23ð.294Þ ð.308Þ ð.313Þ ð.303Þ
Proportion of teachers who attendednormal college or university 21.48*** 21.45*** 2.36 2.35
ð.242Þ ð.254Þ ð.250Þ ð.230ÞProportion of teachers who have received
provincial or national teaching awards .25 .28 2.51* 2.63*ð.251Þ ð.312Þ ð.303Þ ð.347Þ
Proportion of Gongban teachers .34 .23 2.58 2.33ð.390Þ ð.389Þ ð.505Þ ð.471Þ
Average teacher experience ðyearsÞ 2.03** 2.03*** 2.01 2.01ð.012Þ ð.011Þ ð.013Þ ð.011Þ
School size ðstudentsÞ .00 2.00ð.002Þ ð.001Þ
Student-teacher ratio 2.02 .02ð.016Þ ð.013Þ
Distance to farthest village servedby school ðminutesÞ 2.00 .00***
ð.002Þ ð.001ÞSchool has provided teacher training
in past year ð0/1Þ 2.10 2.44ð.250Þ ð.278Þ
School infrastructure index 2.07 2.01ð.091Þ ð.079Þ
Constant .46*** 2.59*** 3.17*** .52*** 1.19** 1.24*ð.123Þ ð.598Þ ð.859Þ ð.102Þ ð.560Þ ð.713Þ
Observations 75 75 75 73 73 73Adjusted R 2 .080 .297 .260 .147 .193 .254
Note. Dependent variable is the difference between the estimated school fixed effect ðFEÞ for Han studentsand the estimated school fixed effect for minority students. School fixed effects used to construct the de-pendent variablewere estimated using ordinary least squares regressions of student standardized exam scoreson all student, peer, and teacher characteristics in table1 and school dummy variables using the sample ofmixed schools only. All covariates are at the school level.* p < .1.** p < .05.*** p < .01.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE
8OAXACA
DECOMPOSITIONS
(MATH
)
Mixed
Scho
olsOnly
FullSa
mple
NoSc
hool
FESc
hool
FE
All
Minority
ð1Þ
Man
darin
Minority
ð2Þ
Non
-Man
darin
Minority
ð3Þ
All
Minority
ð4Þ
Man
darin
Minority
ð5Þ
Non
-Man
darin
Minority
ð6Þ
All
Minority
ð7Þ
Man
darin
Minority
ð8Þ
Non
-Man
darin
Minority
ð9Þ
Totalg
ap.29***
.17***
.62***
.10
.01
.24*
.10
.01
.24
ð.058
Þð.0
54Þ
ð.084
Þð.0
66Þ
ð.068
Þð.1
29Þ
ð.073
Þð.0
67Þ
ð.150
ÞDifferen
tch
aracteristic
s.15***
.15**
.18**
2.02
2.00
2.09
.10
.15
2.10
ð.059
Þð.0
61Þ
ð.070
Þð.0
78Þ
ð.105
Þð.0
94Þ
ð.146
Þð.1
64Þ
ð.216
ÞStud
entan
dho
useh
old
.06***
.05***
.10***
.05**
.03
.06
.04*
.03
.06
ð.015
Þð.0
17Þ
ð.019
Þð.0
23Þ
ð.027
Þð.0
68Þ
ð.024
Þð.0
29Þ
ð.073
ÞPe
er.04
.03
.04
2.03
2.06
.05
2.03
.14
.35
ð.050
Þð.0
46Þ
ð.064
Þð.0
80Þ
ð.081
Þð.1
26Þ
ð.241
Þð.2
05Þ
ð.340
ÞTe
ache
r.02
.02
.03
2.00
.05
.12
.01
.07
.23
ð.046
Þð.0
47Þ
ð.046
Þð.0
45Þ
ð.064
Þð.1
51Þ
ð.051
Þð.0
64Þ
ð.206
ÞSc
hool
.03
.04
.02
2.04
2.02
2.32*
ð.023
Þð.0
26Þ
ð.034
Þð.0
31Þ
ð.047
Þð.1
89Þ
Scho
olfix
edeffects
.08
2.09
2.75*
ð.272
Þð.2
90Þ
ð.412
ÞDifferen
treturnsto
characteristic
s.14**
.03
.44***
.12
.01
.33***
2.00
2.14
.34
ð.065
Þð.0
63Þ
ð.099
Þð.0
77Þ
ð.097
Þð.1
15Þ
ð.153
Þð.1
51Þ
ð.259
ÞObservations
9,46
89,15
78,58
72,58
21,73
953
02,58
21,73
953
0
Note.Firstthree
columns
useallsch
oolsin
sample;lastsixco
lumns
usesample
ofmixed
scho
olson
ly.E
stim
ationsample
includ
esarand
omlych
osen
halfof
allsam
ple
stud
ents
ðthosewho
weregiven
astan
dardized
exam
inmathÞ.S
tand
arderrors
ðinparen
thesesÞa
ccou
ntforclusterin
gat
thescho
olleve
l.FE
5fixe
deffects.
*p<.1.
**p<.05.
***p<.01.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
the last three for the mixed sample with school fixed effects included. Within
Yang et al. 347
each set of columns we give results for the comparison between Han andð1Þ all minority students, ð2Þ Mandarin minority students, and ð3Þ non-Mandarin minority students. The first row shows the estimated total gap. Thesecond row gives the total portion of the gap estimated to be due to differ-ences in Han and minority characteristics. Estimated subtotals for each cate-gory of included characteristic ðstudent and household characteristics, peercharacteristics, teacher characteristics, and school characteristics or fixed effectsÞadd up to row 2. The penultimate row gives the portion of the gap due todifferences in returns to characteristics.The first key result of the decomposition analysis is that, for both groups,
differences in student and household endowments are the largest explainedcontributor to the Han-minority achievement gap ðfirst rowÞ. For Mandarinminorities, differences in these variables account for 29.4% of the gap inmath. Likewise, for non-Mandarin minority students these variables explain16.1% of the math gap—more than any of the other explained components.The second key finding is that a much larger portion of the gap can be
explained for Mandarin minority students compared to non-Mandarin mi-nority students. We estimate that differences in endowments explain 88% ofthe math gap for Mandarin minority students while only explaining 29% of thegap for non-Mandarin minority students ðcols. 2 and 3, second rowÞ.A third key finding is that the gap between Han and Mandarin minority
students disappears when we restrict the sample to schools with both Han andminority students but remains large for non-Mandarin minority students ðcols.4–9Þ. The achievement gap between Han and Mandarin minority students isthus nearly entirely due to the high performance of Han students in schoolswithout minority students. The math gap for non-Mandarin minority studentsðcols. 6 and 9Þ is reduced by 0.38 SD ð61%Þ yet remains large with non-Mandarin minority students scoring 0.24 SD below their Han counterparts.This gap is fully due to differences in returns to characteristics. Results changelittle when we substitute school characteristics for school fixed effects.Following this set of findings, the decomposition analysis implies that dif-
ferences in characteristics are unable to explain 0.44 SD ð71%Þ of the gap inmath11 between Han students and non-Mandarin minority students in thefull sample and none of the gap after restricting the sample to mixed schoolsonly. This unexplained gap has several possible interpretations. First, it mayindicate that some inputs that are important determinants of learning for
11 Note that 0.64 SD ð98.5%Þ of the gap in Chinese scores is unexplained by differences in char-
acteristics.This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
these students are omitted. However, it is likely that the influence of these is
348 E C O N O M I C D E V E L O P M E N T A N D C U L T U R A L C H A N G E
limited given the large portion of the gap explained between Han and Man-darin minority students.Another explanation is that, even when given similar educational resources
ðor inputsÞ, non-Mandarin minority students benefit less from these inputs.This could be due to these students facing a different schooling environment,even when in the same class as Han students. For example, lower teacher ex-pectations could lead them to focus instruction on Han who they believe maybenefit more from their instruction ðcf. McEwan and Trowbridge 2007Þ. Thiscould also be due to students having difficulty comprehending instructionin Mandarin. Even though instruction in local ethnic languages is permittedin China, this is often difficult in practice ðCherng et al. 2012Þ. For example,instruction in Salar and Tibetan languages is not feasible when these studentsattend school with students of other ethnicities, a common occurrence in oursample.
VII. Summary and ConclusionsThe goal of this article was to document and explain the gap in educationalachievement between Han and minority students in primary schools inwestern China. In our survey of 300 schools in Shaanxi, Gansu, and Qinghaiprovinces ðinvolving nearly 21,000 fourth- and fifth-grade studentsÞ, we findlarge differences in achievement on standardized exams between Han andminority students. On average, minority students perform 0.25 SD lower inmath and 0.22 SD lower in Chinese. Most strikingly, minority students whodo not generally speak Mandarin as their primary language score 0.62 SDlower than Han in math and 0.65 SD lower than Han in Chinese.Using decomposition methods pioneered by Oaxaca ð1973Þ and Blinder
ð1973Þ, we find that most of the achievement gap between Han and minoritystudents with no alternative ethnic language can be explained by differencesin endowments of student, family, and school characteristics. Of these, dif-ferences in students and family characteristics appear to contribute the mostto differences in achievement. Little of the gap between Han students andnon-Mandarin minority students ðSalar and Tibetan in our sampleÞ, however,can be explained by endowment differences. Comparing these students onlyto Han students in the same schools significantly reduces the size of theachievement gap, yet a difference of more than 0.2 SD persists. None of thisremaining gap is explained by differences in endowments. Although severalexplanations are possible, we believe that a likely explanation is that the abilityof students to learn may be hindered by difficulty comprehending instruction
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
in Mandarin ðgiven that no schools in our sample provided instruction or
Yang et al. 349
texts in minority languagesÞ. While we cannot say with certainty why thesestudents may benefit less from a given amount of schooling inputs, ouranalysis suggests that teachers play a significant role.While we believe that the findings of this article are important, admittedly,
the study has a number of limitations. First, although our sample contains suf-ficient numbers of minority students to conduct analyses, studies involving alarger sample of minority students ðparticularly non-Mandarin minority stu-dentsÞ would provide further insight into the achievement gap. Second, our sur-vey did not collect information on the Mandarin ability of individual studentsðalthough we tested students on the Chinese curriculum, this may be distinctfrom pure language abilityÞ. Future studies should employ such information toassess to what degree language is contributing to the underperformance of stu-dents belonging to groups that do not speakMandarin as their primary language.Despite these limitations, however, our results call for the attention of
policy makers to approaches to address the underperformance of minoritystudents in China’s rural areas. Given the large and increasing importance ofeducational attainment to economic well-being, addressing the large achieve-ment gap between Han and minority students may help to mitigate economicdisparities in the future. On the basis of our results, promising approaches toaddress the achievement gap would include those focused on improving thereturns to minority students of given schooling inputs ðe.g., through peda-gogical practiceÞ. Further, if future studies show language to contribute sig-nificantly to the gap, interventions such as remedial tutoring in Mandarin mayalso yield large benefits.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE
A1
SUMMARYST
ATIST
ICSFO
RMIXED
SCHOOLSA
MPLE
Mixed
Man
darin
Minority
Scho
ols
Mixed
Non
-Man
darin
Minority
Scho
ols
Han
Stud
ents
ð1Þ
Man
darin
Minority
Stud
ents
ð2Þ
Differen
ceð2Þ2
ð1Þ
½P-value�
Han
Stud
ents
ð3Þ
Non
-Man
darin
Minority
Stud
ents
ð4Þ
Differen
ceð4Þ2
ð3Þ
½P-value�
Stud
entan
dho
useh
oldch
aracteristic
s:Stan
dardized
mathex
amscore
2.08
2.06
.03
2.16
2.43
2.27
ð1.00Þ
ð.97Þ
½.68�
ð.95Þ
ð.94Þ
½.08�
Stan
dardized
Chine
seex
amscore
2.04
2.05
2.01
2.02
2.39
2.37
ð1.02Þ
ð1.01Þ
½.95�
ð.99Þ
ð1.1Þ
½.03�
Femaleð0/1Þ
.48
.45
2.03
.5.5
0ð.5
Þð.5
Þ½.1
3�ð.5
Þð.5
Þ½.9
4�Boa
rdingstud
entð0/1Þ
.08
.09
.01
.08
.26
.18
ð.27Þ
ð.28Þ
½.77�
ð.27Þ
ð.44Þ
½.13�
Ageðyea
rsÞ
10.85
10.92
.07
10.72
11.19
.47
ð1.08Þ
ð1.1Þ
½.31�
ð1.08Þ
ð1.06Þ
½.00�
Hou
seho
ldsize
5.24
5.32
.08
5.19
5.33
.14
ð1.46Þ
ð1.54Þ
½.38�
ð1.39Þ
ð1.71Þ
½.36�
Travel
timeto
scho
olðm
inutesÞ
23.52
22.43
21.08
25.53
31.4
5.87
ð25.7Þ
ð25.53
Þ½.6
7�ð22.99
Þð32.27
Þ½.2
9�Mothe
rha
slower
seco
ndarydeg
reeor
abov
eð0/1Þ
.27
.23
2.04
.2.15
2.05
ð.44Þ
ð.42Þ
½.25�
ð.4Þ
ð.36Þ
½.22�
Father
haslower
seco
ndarydeg
reeor
abov
eð0/1Þ
.46
.42.06
.38
.34
2.04
ð.5Þ
ð.49Þ
½.16�
ð.49Þ
ð.48Þ
½.42�
Father
atho
með0/1Þ
.59
.57
2.03
.52
.59
.07
ð.49Þ
ð.5Þ
½.51�
ð.5Þ
ð.49Þ
½.30�
Mothe
rat
Hom
eð0/1Þ
.69
.69
0.68
.61
2.07
ð.46Þ
ð.46Þ
½.94�
ð.47Þ
ð.48Þ
½.34�
Hou
seho
ldassetindex
ð0/1Þ
.01
.4.39
.13
.18
.06
ð1.37Þ
ð1.43Þ
½.00�
ð1.38Þ
ð1.5Þ
½.80�
Class
pee
rch
aracteristic
:Prop
ortio
nof
pee
rs’mothe
rswith
lower
seco
ndarydeg
reeor
abov
e.26
.26
2.01
.2.16
2.03
ð.17Þ
ð.17Þ
½.82�
ð.11Þ
ð.11Þ
½.25�
Appendix
350
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Mixed
Man
darin
Minority
Scho
ols
Mixed
Non
-Man
darin
Minority
Scho
ols
Prop
ortio
nof
classpee
rsof
sameethn
icity
.9.52
2.38
.81
.43
2.38
ð.15Þ
ð.33Þ
½.00�
ð.21Þ
ð.25Þ
½.00�
Peer
averag
eho
useh
oldassetindex
.02
.35
.33
.12
.22
.1ð.5
6Þð.5
3Þ½.0
0�ð.6
3Þð.6
7Þ½.5
5�Te
ache
rch
aracteristic
:Fe
maleteache
rð0/1Þ
.36
.49
.12
.4.48
.07
ð.48Þ
ð.5Þ
½.08�
ð.49Þ
ð.5Þ
½.54�
Han
teache
rð0/1Þ
.92
.62
2.3
.86
.61
2.25
ð.27Þ
ð.49Þ
½.00�
ð.35Þ
ð.49Þ
½.05�
Teache
rha
shighe
red
ucationdeg
reeð0/1Þ
.78
.93
.15
.89
.9.01
ð.41Þ
ð.25Þ
½.00�
ð.31Þ
ð.3Þ
½.86�
Teache
rattend
edno
rmal
colle
geð0/1Þ
.79
.91
.12
.77
.83
.05
ð.41Þ
ð.28Þ
½.00�
ð.42Þ
ð.38Þ
½.46�
Teache
rha
srece
ived
provinc
ialo
rna
tiona
ltea
chingaw
ardð0/1Þ
.09
.19
.1.03
.15
.12
ð.29Þ
ð.39Þ
½.05�
ð.18Þ
ð.36Þ
½.26�
Gong
ban
teache
rð0/1Þ
.9.91
.01
.97
.83
2.14
ð.3Þ
ð.28Þ
½.62�
ð.18Þ
ð.38Þ
½.07�
Teache
rex
perienc
eðyea
rsÞ
13.61
13.47
2.15
15.73
15.13
2.61
ð11.28
Þð9.84Þ
½.90�
ð10.08
Þð9.62Þ
½.70�
Scho
olch
aracteristic
:Sc
hool
size
ðstud
entsÞ
225.77
212.43
213
.28
218.74
206.27
212
.47
ð61.88
Þð66.01
Þ½.4
3�ð58.76
Þð57.87
Þ½.4
7�Stud
ent-teache
rratio
18.06
17.15
2.92
17.39
17.36
2.04
ð5.32Þ
ð4.26Þ
½.30�
ð4.39Þ
ð5.66Þ
½.98�
Distanc
eto
farthe
stvillageserved
byscho
olðm
inutesÞ
68.34
60.12
28.21
65.51
106.05
40.55
ð59.43
Þð42.92
Þ½.4
0�ð48.98
Þð82.67
Þ½.1
6�Sc
hool
hasprovided
teache
rtraining
inpastye
arð0/1Þ
.93
.95
.01
.96
.72
2.24
ð.25Þ
ð.23Þ
½.76�
ð.19Þ
ð.45Þ
½.12�
Scho
olinfrastruc
ture
index
.15
.18
.03
.03
.53
.5ð1.2Þ
ð1.17Þ
½.91�
ð1.35Þ
ð1.31Þ
½.21�
Sample
size:
Totaln
umber
ofstud
ents
4,68
293
21,03
629
4Num
ber
ofstud
ents—mathsample
2,33
047
150
814
2Num
ber
ofstud
ents—Chine
sesample
2,35
246
152
815
2Num
ber
ofscho
ols
8686
2323
Note.Mixed
Man
darin
minority
scho
olsarescho
olswith
atleasttwoHan
stud
entsan
dtw
oMan
darin
minority
stud
ents.M
ixed
non-Man
darin
minority
scho
olsarescho
olswith
atleasttwoHan
stud
entsan
dtw
ono
n-Man
darin
minority
stud
ents.V
ariablesareas
described
intable
1.Stan
darderrorsðin
paren
thesesÞa
ccou
ntforc
lusteringat
thescho
olleve
l.
351
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE A2CHINESE ACHIEVEMENT REGRESSIONS (POOLED, FULL SAMPLE)
ð1Þ ð2Þ ð3Þ ð4Þ ð5Þ ð6ÞMandarin-speaking minority 2.10 2.05 2.02 2.05 2.01 2.25***
ato
Student and household
This content downloadedAll use su
ð.064Þ2.65***
ioch
Han Stu
ð1Þ
from 171.66bject to JSTO
ð.065Þ2.59***
rdic
dents
ð2Þ
.209.5 on WeR Terms and
ð.071Þ2.54***
e
Minor
d, 7 Jan 2015 Conditions
ð.071Þ2.57***
er
ity
16:28:44 PM
ð.067Þ2.55***
tc
ð.062Þ2.20*
Non-Mandarin-speaking minorityð.106Þ
ð.104Þ ð.120Þ ð.121Þ ð.118Þ ð.111Þ Student and household characteristics Yes Yes Yes Yes Yes Class peer characteristics Yes Yes Yes Yes Teacher characteristics Yes Yes Yes School characteristics Yes School fixed effects Yes Constant .03 .22 2.39 2.74 2.74 2.43ð.026Þ
ð.678Þ ð.682Þ ð.653Þ ð.660Þ ð.656Þ Adjusted R2 .014 .068 .077 .088 .099 .201Note. Each column represents a separ
e regress n. Standa errors ðin parenthes sÞ accoun for clus- usehold aracterist s, class pe r characte istics, tea her char- tering at the school level. Student and hacteristics, and school characteristics in
lude thos in table . Estimatio sample cludes a andomly c e 1 n in rchosen half of all sample students ðthose who were given a standardized exam in ChineseÞ. N 5 9,661.* p < .1.*** p < .01.TABLE A3CHINESE ACHIEVEMENT REGRESSIONS BY ETHNICITY
Mandarin-SpeakingNon-Mandarin-
Speaking
ð3Þ ð4Þ
Minorityð5Þ ð6Þ
characteristics:
.10*** .08*** .14* .18* .20* .17 Female ð0/1ÞBoarding student ð0/1Þ
ð.021Þ ð.020Þ ð.077Þ ð.090Þ ð.115Þ ð.146Þ 2.14**ð.056Þ2.15***ð.055Þ
2.37**ð.172Þ
2.32*ð.187Þ
2.03ð.285Þ
.06ð.461Þ
Age ðyearsÞ
.16 .15 2.43 2.59* 1.28 .83 ð.120Þ ð.124Þ ð.344Þ ð.332Þ ð.780Þ ð.980ÞAge2
2.01** 2.01** .02 .02 2.06* 2.04 ð.005Þ ð.006Þ ð.015Þ ð.014Þ ð.033Þ ð.042ÞHousehold size
2.02** 2.01 2.01 2.01 .06* .08*** ð.008Þ ð.008Þ ð.018Þ ð.020Þ ð.029Þ ð.022ÞTravel time to school ðminutesÞ
.00 .00 2.00 2.00 2.00 2.00 ð.001Þ ð.001Þ ð.003Þ ð.003Þ ð.003Þ ð.004ÞMother has lower secondary
degree or above ð0/1Þ 2.02 2.02 .01 .05 2.04 2.03ð.025Þ
ð.025Þ ð.105Þ ð.124Þ ð.205Þ ð.222Þ Father has lower secondary degreeor above ð0/1Þ .20*** .17*** .14** .13* 2.15 2.21ð.023Þ
ð.021Þ ð.068Þ ð.071Þ ð.154Þ ð.171Þ Father at home ð0/1Þ 2.04*ð.023Þ
.00ð.021Þ2.15***ð.055Þ
2.03ð.052Þ
2.11ð.154Þ
2.23ð.158Þ
Mother at home ð0/1Þ
.02 2.03 .05 2.06 .13 .17 ð.027Þ ð.025Þ ð.067Þ ð.072Þ ð.167Þ ð.165ÞHousehold asset index ð0/1Þ
.02*** .03*** .09*** .06* .10** .10** ð.008Þ ð.008Þ ð.030Þ ð.032Þ ð.047Þ ð.042Þ52
3TABLE A3 (Continued )
Non-Mandarin-
Mandarin-Speaking SpeakingHan Students Minority
ð1Þ ð2ÞClass peer characteristic:
tp
This content downloadedAll use su
io r
from 171.66.209.5 on Webject to JSTOR Terms and
ð3Þ ð4Þ
n sh
d, 7 Jan 2015 16:28:44 PM Conditions
Minority
ð5Þ ð6Þ
Proportion of peers’ mothers with
.16 2.19 2.30 .03 .50n
3.70*
lower secondary degree or above ð.135Þ ð.209Þ ð.377Þ ð.473Þ ð.995Þ ð1.873ÞProportion of class peers of sameethnicity
2.07 .34 .73*** 2.06 2.60 3.42ð.168Þ
ð.557Þ ð.141Þ ð.530Þ ð.376Þ ð2.356Þ Peer average household asset index .08**ð.035Þ
.09ð.064Þ.08ð.105Þ
2.04ð.152Þ
2.05ð.159Þ
.07ð.292Þ
Teacher characteristic:
Female teacher ð0/1Þ .14*** .08* .14 .29*** 2.11 2.05ð.041Þ
ð.044Þ ð.106Þ ð.105Þ ð.226Þ ð.456Þ Han teacher ð0/1Þ .06 2.11 .12 .16 2.07 2.47**ð.084Þ
ð.095Þ ð.092Þ ð.116Þ ð.117Þ ð.182Þ Teacher has higher education degree ð0/1Þ .08 .01 .27 2.05 2.24 .12ð.060Þ
ð.086Þ ð.203Þ ð.227Þ ð.269Þ ð.417Þ Teacher attended normalcollege ð0/1Þ .07* .01 .17 .28** 2.15 2.52ð.045Þ
ð.050Þ ð.140Þ ð.129Þ ð.250Þ ð.459Þ Teacher has received provincial ornational teaching award ð0/1Þ 2.01 .18** 2.02 2.24 2.21 2.11ð.055Þ
ð.075Þ ð.105Þ ð.147Þ ð.207Þ ð.448Þ Gongban teacher ð0/1Þ .05ð.061Þ
.01ð.070Þ.23ð.145Þ
2.01ð.157Þ
.11ð.175Þ
.36**ð.159Þ
Teacher experience ðyearsÞ
.00 2.00 2.00 2.01 .01 .05** ð.002Þ ð.003Þ ð.005Þ ð.009Þ ð.011Þ ð.017ÞSchool characteristic:
School size ðstudentsÞ 2.00** .00* .00**ð.000Þ
ð.001Þ ð.002Þ Student-teacher ratio 2.01** 2.01 .00ð.005Þ
ð.009Þ ð.017Þ Distance to farthest village served by school ðminutesÞ .00*** .00 2.00ð.000Þ
ð.001Þ ð.002Þ School has provided teachertraining in past year ð0/1Þ .01 .15 2.14ð.076Þ
ð.130Þ ð.445Þ School infrastructure index .04*ð.021Þ
.11**ð.044Þ.05ð.074Þ
School fixed effects
No Yes No Yes No Yes Constant 2.31 2.39 1.22 3.60* 27.42 27.78ð.679Þ
ð.865Þ ð2.015Þ ð1.911Þ ð4.545Þ ð6.560Þ Observations 8,455 8,455 882 882 316 316 Adjusted R 2 .092 .187 .155 .258 .109 .243e regress
n. Standa d errors ði parenthe esÞ accou t for clus- Note. Each column represents a separatering at the school level. Estimation sam le include s a random ly chosen alf of all sa mple stude nts ðthose who were given a standardized exam in hineseÞ. C* p < .1.** p < .05.*** p < .01.353
TABLE
A4
CHIN
ESE
ACHIEVEMENTREGRESS
IONSBYETH
NICITY
(MIXED
SCHOOLS
ONLY
)
Mixed
Man
darin
Minority
Scho
ols
Mixed
Non
-Man
darin
Minority
Scho
ols
Han
Stud
ents
Man
darin-Spea
king
Minority
Stud
ents
Han
Stud
ents
Non
-Man
darin-
Spea
king
Minority
ð1Þ
ð2Þ
ð3Þ
ð4Þ
ð5Þ
ð6Þ
ð7Þ
ð8Þ
Stud
entan
dho
useh
oldch
aracteristic
s:Fe
maleð0/1Þ
.15***
.15***
.14
.19
.10
.09
.25
.09
ð.043
Þð.0
43Þ
ð.113
Þð.1
38Þ
ð.061
Þð.0
65Þ
ð.228
Þð.2
79Þ
Boa
rdingstud
entð0/1Þ
2.50***
2.52***
2.91***
2.76***
.17
.24
.19
.46
ð.140
Þð.1
67Þ
ð.253
Þð.2
55Þ
ð.214
Þð.2
16Þ
ð.503
Þð.4
31Þ
Ageðyea
rsÞ
2.01
.09
2.71*
2.82*
2.89
2.71
1.29
1.57
ð.323
Þð.3
19Þ
ð.406
Þð.4
19Þ
ð.692
Þð.7
55Þ
ð.941
Þð1.216
ÞAge2
2.00
2.01
.03
.03*
.04
.03
2.06
2.08
ð.014
Þð.0
14Þ
ð.017
Þð.0
17Þ
ð.031
Þð.0
34Þ
ð.039
Þð.0
50Þ
Hou
seho
ldsize
.02
.02
2.02
2.04
.04
.04
.15**
.12*
ð.021
Þð.0
20Þ
ð.030
Þð.0
34Þ
ð.027
Þð.0
27Þ
ð.060
Þð.0
56Þ
Travel
timeto
scho
olðm
inutesÞ
.00
2.00
.00
.00
2.00
2.00
2.00
2.00
ð.001
Þð.0
01Þ
ð.003
Þð.0
03Þ
ð.002
Þð.0
02Þ
ð.006
Þð.0
06Þ
Mothe
rha
slower
seco
ndarydeg
reeor
abov
eð0/1Þ
2.02
2.04
.14
.21
.12
.12
2.17
2.37
ð.066
Þð.0
68Þ
ð.133
Þð.1
69Þ
ð.140
Þð.1
49Þ
ð.238
Þð.2
88Þ
Father
haslower
seco
ndarydeg
reeor
abov
eð0/1Þ
.09*
.06
.09
.15
.18**
.19**
2.10
2.06
ð.050
Þð.0
48Þ
ð.108
Þð.1
14Þ
ð.078
Þð.0
74Þ
ð.201
Þð.2
26Þ
Father
atho
með0/1Þ
2.05
2.02
2.12*
2.02
2.08
2.10
.21
.05
ð.053
Þð.0
60Þ
ð.072
Þð.0
88Þ
ð.100
Þð.1
17Þ
ð.242
Þð.2
34Þ
Mothe
rat
homeð0/1Þ
2.03
2.10
2.06
2.20*
.06
.05
2.21
.14
ð.057
Þð.0
60Þ
ð.093
Þð.1
13Þ
ð.069
Þð.0
78Þ
ð.195
Þð.2
08Þ
Hou
seho
ldassetindex
ð0/1Þ
.02
.03
.09**
.07
.03
.03
.04
.02
ð.020
Þð.0
19Þ
ð.046
Þð.0
57Þ
ð.047
Þð.0
49Þ
ð.060
Þð.0
62Þ
Class
pee
rch
aracteristic
:Prop
ortio
nof
pee
rs’mothe
rswith
lower
seco
ndarydeg
reeor
abov
e2.52*
.04
2.83*
.25
21.80
**2.83
21.65
4.82
ð.282
Þð.5
53Þ
ð.419
Þð.7
20Þ
ð.800
Þð2.121
Þð1.910
Þð4.262
ÞProp
ortio
nof
classpee
rsof
sameethn
icity
2.23
2.33
.41*
2.06
2.46
1.29
2.06
3.24
ð.266
Þð.7
98Þ
ð.231
Þð.7
76Þ
ð.350
Þð1.694
Þð1.332
Þð2.992
ÞPe
eraverag
eho
useh
oldassetindex
.06
.28
.24
.01
.47***
2.24
2.59
2.69
ð.103
Þð.2
01Þ
ð.149
Þð.2
43Þ
ð.095
Þð.4
96Þ
ð.516
Þð.4
68Þ
354
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Mixed
Man
darin
Minority
Scho
ols
Mixed
Non
-Man
darin
Minority
Scho
ols
Teache
rch
aracteristic
:Fe
maleteache
rð0/1Þ
.11
.16
.07
.44***
2.18**
2.57
.25
21.79
**ð.0
98Þ
ð.134
Þð.1
22Þ
ð.145
Þð.0
67Þ
ð.351
Þð.1
59Þ
ð.772
ÞHan
teache
rð0/1Þ
2.02
2.15
2.02
.27
.03
2.14
2.38
2.99**
ð.089
Þð.1
11Þ
ð.138
Þð.2
05Þ
ð.120
Þð.2
88Þ
ð.269
Þð.3
62Þ
Teache
rha
shighe
red
ucationdeg
reeð0/1Þ
.10
.12
.40*
.16
.45***
1.23
2.34
2.89
**ð.1
62Þ
ð.177
Þð.2
11Þ
ð.419
Þð.1
27Þ
ð.805
Þð.3
56Þ
ð1.305
ÞTe
ache
rattend
edno
rmal
colle
geð0/1Þ
.09
.23
.19
2.02
2.30
21.21
*2.19
26.60
**ð.1
17Þ
ð.149
Þð.2
50Þ
ð.297
Þð.1
75Þ
ð.641
Þð.3
61Þ
ð3.037
ÞTe
ache
rha
srece
ived
provinc
ialo
rna
tiona
ltea
chingaw
ardð0/1Þ
.08
.00
2.02
2.28**
21.32
***
21.43
**.96
23.41
ð.172
Þð.2
02Þ
ð.118
Þð.1
37Þ
ð.357
Þð.6
08Þ
ð1.193
Þð2.069
ÞGong
ban
teache
rð0/1Þ
.08
2.22
.09
.12
.73***
.04
.17
...
ð.180
Þð.2
42Þ
ð.179
Þð.2
62Þ
ð.126
Þð.2
87Þ
ð.505
Þ...
Teache
rex
perienc
eðyea
rsÞ
.00
.01
2.00
2.00
2.02*
.03
.04**
.29**
ð.004
Þð.0
05Þ
ð.008
Þð.0
15Þ
ð.009
Þð.0
35Þ
ð.016
Þð.1
23Þ
Scho
olch
aracteristic
:Sc
hool
size
ðstud
entsÞ
.00
2.00*
.00
2.00
ð.001
Þð.0
01Þ
ð.001
Þð.0
02Þ
Stud
ent-teache
rratio
2.03***
2.02
.00
.02
ð.011
Þð.0
18Þ
ð.008
Þð.0
29Þ
Distanc
eto
farthe
stvillageserved
byscho
olðm
inutesÞ
.00***
2.00
.00**
2.01***
ð.001
Þð.0
01Þ
ð.001
Þð.0
04Þ
Scho
olha
sprovided
teache
rtraining
inpastye
arð0/1Þ
.31*
.10
.26
21.82
***
ð.160
Þð.2
79Þ
ð.218
Þð.5
32Þ
Scho
olinfrastruc
ture
index
.02
.11
.00
2.49***
ð.051
Þð.0
69Þ
ð.031
Þð.1
62Þ
Scho
olfix
edeffects
No
Yes
No
Yes
No
Yes
No
Yes
Con
stan
t.52
.17
4.35
4.73
*4.28
3.16
25.54
29.97
ð1.891
Þð2.001
Þð2.638
Þð2.633
Þð3.948
Þð4.665
Þð5.803
Þð7.624
ÞObservations
1,42
938
734
412
2Adjusted
R2
.102
.170
.215
.288
.076
.075
.233
.280
Note.Ea
chco
lumnrepresentsaseparateregression.
Stan
darderrors
ðinparen
thesesÞa
ccou
ntforc
lusteringat
thescho
olleve
l.Estim
ationsample
includ
esarand
omlych
osen
halfof
allsam
ple
stud
ents
ðthosewho
weregiven
astan
dardized
exam
inChine
seÞinmixed
scho
ols.
*p<.1.
**p<.05.
***p<.01.
355
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
TABLE
A5
OAXACA
DECOMPOSITIONS
(CHIN
ESE
)
Mixed
Scho
olsOnly
FullSa
mple
NoSc
hool
FESc
hool
FE
All
Minority
ð1Þ
Man
darin
Minority
ð2Þ
Non
-Man
darin
Minority
ð3Þ
AllMinority
ð4Þ
Man
darin
Minority
ð5Þ
Non
-Man
darin
Minority
ð6Þ
AllMinority
ð7Þ
Man
darin
Minority
ð8Þ
Non
-Man
darin
Minority
ð9Þ
Totalg
ap.25***
.10
.65***
.10
.02
.27
.10
.02
.27
ð.070
Þð.0
63Þ
ð.102
Þð.0
89Þ
ð.086
Þð.1
81Þ
ð.091
Þð.0
83Þ
ð.192
ÞDifferen
tch
aracteristic
s.02
.03
.01
2.20**
2.21**
2.14
2.18**
2.19**
2.08
ð.054
Þð.0
53Þ
ð.072
Þð.0
91Þ
ð.091
Þð.1
27Þ
ð.091
Þð.0
92Þ
ð.211
ÞStud
entan
dho
useh
old
.05***
.05***
.06**
.02
2.01
.06
.02
2.02
.08
ð.014
Þð.0
15Þ
ð.022
Þð.0
20Þ
ð.023
Þð.0
43Þ
ð.021
Þð.0
24Þ
ð.053
ÞPe
er2.04
2.03
2.05
2.18*
2.12
2.29*
2.07
2.23
.43
ð.049
Þð.0
46Þ
ð.060
Þð.0
99Þ
ð.103
Þð.1
54Þ
ð.350
Þð.3
25Þ
ð.668
ÞTe
ache
r2.01
2.01
2.01
2.02
2.05
.09
2.04
2.07*
.28
ð.036
Þð.0
35Þ
ð.042
Þð.0
31Þ
ð.044
Þð.1
61Þ
ð.034
Þð.0
45Þ
ð.241
ÞSc
hool
.02
.03
.01
2.02
2.04
2.00
ð.022
Þð.0
21Þ
ð.045
Þð.0
41Þ
ð.060
Þð.1
02Þ
Scho
olfix
edeffects
2.09
.13
2.88
ð.286
Þð.3
12Þ
ð.698
ÞDifferen
treturnsto
characteristic
s.23***
.08
.64***
.30***
.23***
.42***
.28***
.21**
.36
ð.077
Þð.0
71Þ
ð.113
Þð.0
91Þ
ð.083
Þð.1
44Þ
ð.087
Þð.0
93Þ
ð.223
ÞObservations
9,66
19,33
78,77
12,57
11,81
653
22,57
11,81
653
2
Note.Firstthree
columns
useallsch
oolsin
sample;lastsixco
lumns
usesample
ofmixed
scho
olson
ly.E
stim
ationsample
includ
esarand
omlych
osen
halfof
allsam
ple
stud
ents
ðthosewho
weregiven
astan
dardized
exam
inChine
seÞ.Stan
darderrors
ðinparen
thesesÞa
ccou
ntforclusterin
gat
thescho
olleve
l.FE
5fixe
deffects.
*p<.1.
**p<.05.
***p<.01.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
References
Yang et al. 357
Akerlof, George, and Rachel Kranton. 2002. “Identity and Schooling: Some Lessons forthe Economics of Education.” Journal of Economic Literature 40, no. 4:1167–1201.
Ammermueller, Andreas. 2007. “PISA: What Makes the Difference?” Empirical Eco-nomics 33, no. 2:263–87.
Angrist, Joshua, and Victor Lavy. 2009. “The Effects of High Stakes High SchoolAchievement Awards: Evidence from aRandomizedTrial.”American Economic Review99, no. 4:1384–1414.
Barrera-Osorio, Felipe, Vicente Garcia Moreno, Harry A. Patrinos, and Emilio ErnestoPorta. 2011. “Using the Oaxaca-Blinder Decomposition Technique to AnalyzeLearning Outcomes Changes over Time: An Application to Indonesia’s Results inPISA Mathematics.” Unpublished manuscript, World Bank, Washington, DC.
Blinder, Alan. 1973. “Wage Discrimination: Reduced Form and Structural Esti-mates.” Journal of Human Resources 8, no. 4:436–55.
Chen, Shaohua, and Martin Ravallion. 2010. “China Is Poorer than We Thought, butNo Less Successful in the Fight against Poverty.” In Debates on the Measurement ofGlobal Poverty, ed. Sudhir Anand, Paul Segal, and Joseph E. Stiglitz. Oxford: OxfordUniversity Press.
Cherng, Hua-Yu, Emily Hannum, and Chunping Lu. 2012. “Sociological Perspec-tives on Ethnicity and Education in China: Views from Chinese and EnglishLiteratures.” Unpublished manuscript, University of Pennsylvania.
Chuang, Hwei-Lin. 1997. “High School Youths’ Dropout and Re-enrollment Behav-ior.” Economics of Education Review 16, no. 2:171–86.
Clarke, Marguerite, Walter Haney, and George Madaus. 2000. High Stakes Testingand High School Completion. Chestnut Hill, MA: National Board on EducationalTesting and Public Policy.
De Brauw, Alan, and Scott Rozelle. 2008. “Reconciling the Returns to Education inOff-Farm Wage Employment in Rural China.” Review of Development Economics 12,no. 1:57–71.
Fiszbein, Ariel, Norbert Schady, and Francisco Ferreira. 2009. Conditional CashTransfers: Reducing Present and Future Poverty. Washington, DC: World Bank.
Fortin, Laurier, Dianne Marcotte, Pierre Potvin, Egide Royer, and Jacques Joly. 2006.“Typology of Students at Risk of Dropping out of School: Description by Personal,Family and School Factors.” European Journal of Psychology of Education 21, no. 4:363–83.
Glewwe, Paul, and Hanan Jacoby. 1995. “An Economic Analysis of Delayed PrimarySchool Enrollment in a Low Income Country: The Role of Early ChildhoodNutrition.” Review of Economics and Statistics 77, no. 1:156–69.
Glewwe, Paul, Hanan Jacoby, and Elizabeth King. 2001. “Early Childhood Nutritionand Academic Achievement: A Longitudinal Analysis.” Journal of Public Economics81, no. 3:345–68.
Gustafsson, Björn, and Shi Li. 2003. “The Ethnic Minority-Majority Income Gap inRural China during Transition.” Economic Development and Cultural Change 51,no. 4:805–22.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Gustafsson, Björn, and Ding Sai. 2009a. “Temporary and Persistent Poverty amongEthnic Minorities and the Majority in Rural China.” Review of Income and Wealth
—
H
H
H
Li
M
M
M
M
M
O
Pa
Pa
Ra
Re
Ru
Sa
358 E C O N O M I C D E V E L O P M E N T A N D C U L T U R A L C H A N G E
55:588–606.——. 2009b. “Villages Where China’s Ethnic Minorities Live.” China EconomicReview 20, no. 2:193–207.annum, Emily. 2002. “Educational Stratification by Ethnicity in China: Enroll-ment and Attainment in the Early Reform Years.” Demography 39, no. 1:95–117.annum, Emily, Jere Behrman, Meiyan Wang, and Jihong Liu. 2008. “Education inthe Reform Era.” In China’s Great Economic Transformation, ed. Loren Brandt andThomas Rawski. Cambridge: Cambridge University Press.annum, Emily, and Meiyan Wang. 2012. “China: A Case Study in Rapid PovertyReduction.” In Indigenous People, Poverty, and Development, ed. Gillette Hall andHarry Patrinos. Cambridge: Cambridge University Press., Hongbin, Junsen Zhang, and Yi Zhu. 2008. “The Quantity-Quality Trade-Off ofChildren in aDevelopingCountry: IdentificationUsingChineseTwins.”Demography45, no. 1:223–43.cEwan, Patrick. 2004. “The Indigenous Test Score Gap in Bolivia and Chile.”Economic Development and Cultural Change 53, no. 1:157–90.cEwan, Patrick, and Jeffery Marshall. 2004. “Why Does Academic AchievementVary across Countries? Evidence from Cuba and Mexico.” Education Economics 12,no. 3:205–17.cEwan, Patrick, and Joseph Shapiro. 2008. “The Benefits of Delayed PrimarySchool Enrollment Discontinuity Estimates Using Exact Birth Dates.” Journal ofHuman Resources 43, no. 1:1–29.cEwan, Patrick, and Marisol Trowbridge. 2007. “The Achievement of IndigenousStudents in Guatemalan Primary Schools.” International Journal of EducationalDevelopment 27, no. 1:61–76.eng, Xin. 2004. “Gender Earnings Gap: The Role of Firm Specific Effects.” LabourEconomics 11, no. 5:555–73.axaca, R. 1973. “Male-Female Wage Differentials in Urban Labor Markets.” In-ternational Economic Review 14, no. 3:693–709.rk, Albert, Fang Cai, and Yang Du. 2010. “Can China Meet Her EmploymentChallenges?” In Growing Pains: A Study of China’s Growth and Development, ed.Jean Oi, Xueguang Zhou, and Scott Rozelle. Washington, DC: Brookings.rk, Albert, Sangui Wang, and Guobao Wu. 2002. “Regional Poverty Targeting inChina.” Journal of Public Economics 86:123–53.vallion, Martin, and Shaohua Chen. 2007. “China’s ðUnevenÞ Progress againstPoverty.” Journal of Development Economics 82, no. 1:1–42.ardon, Sean, and Claudia Galindo. 2002. “Do High-Stakes Tests Affect Students’Decisions to Drop out of School? Evidence from NELS.” Paper presented at theAnnual Meeting of the American Educational Research Association, New Orleans.mberger, Russel, and Sun Ah Lim. 2008. Why Students Drop out of School: A Reviewof 25 Years of Research. Policy Brief no. 15, California Dropout Research Project.kellariou, Chris. 2008. “Peer Effects and the Indigenous/Non-indigenous EarlyTest-Score Gap in Peru.” Education Economics 16, no. 4:371–90.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions
Van de Walle, Dominique, and Dileni Gunewardena. 2001. “Sources of EthnicInequality in Vietnam.” Journal of Development Economics 65, no. 1:177–207.
Yang et al. 359
Vélez, William, and Rogelio Saenz. 2001. “Toward a Comprehensive Model of theSchool Leaving Process among Latinos.” School Psychology Quarterly 16, no. 4:445–67.
Vickers, Harleen. 1994. “Young Children at Risk: Differences in Family Function-ing.” Journal of Educational Research 87, no. 5:262–70.
Wang, Xiaobing, Chengfang Liu, Linxiu Zhang, Renfu Luo, and Scott Rozelle. 2013.“College Is a Rich, Han, Urban, Male Club: Research Notes from a Census Surveyof Four Tier One Colleges in China.” China Quarterly 214:456–70.
Yi, Hongmei, Linxiu Zhang, Renfu Luo, Yaojiang Shi, Di Mo, Xinxin Chen, CarlBrinton, and Scott Rozelle. 2012. “Dropping Out: Why Are Students LeavingJunior High in China’s Poor Rural Areas?” International Journal of EducationalDevelopment 32, no. 4:555–63.
Zhang, Linxiu, Xiaofei Li, Scott Rozelle, and Jikun Huang. 2013. “China’s LaborTransition and the Future of China’s Rural Wages and Employment.” Unpublishedmanuscript, Stanford University.
Zhang, Junsen, Yaohui Zhao, Albert Park, and Xiaoqing Song. 2005. “Economic Re-turns to Schooling in Urban China, 1988 to 2001.” Journal of Comparative Economics33, no. 4:730–52.
This content downloaded from 171.66.209.5 on Wed, 7 Jan 2015 16:28:44 PMAll use subject to JSTOR Terms and Conditions