Documente Academic
Documente Profesional
Documente Cultură
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=asa.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.
American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
Sociology of Education.
http://www.jstor.org
Tracking, Students' Effort, and
Academic Achievement
William Carbonaro
Universityof Notre Dame
the schooling process (see, e.g., Newmann the relationshipsbetween effortand curricu-
1992). Iarguethat effortcan be studiedapart lar tracking will render a more complete
from this affectivecomponent and that it is account of how curriculartracking affects
beneficialto do so because effort can affect academicoutcomes.
outcomes, regardlessof whetherstudentsare
enthusiastic about, interested in, and/or
attachedto theirschool. RESEARCHAGENDA
Students' Effortand Learning The analyses presented here focus on four
main questions regardingeffortand curricu-
Numerousstudies have found that students'
lartracking:(1) Does effortvaryacrosstracks?
effort is related to academic achievement.
Studiesof "engagement"havetypicallyrelied (2) What explains variationin effort across
tracks?(3) Does variationin effortexplaindif-
heavily on measuresof effort, such as the ferences in learning across tracks?and (4)
completionof homework,attentiveness,and Does efforthave the same effect on learning
preparedness.Generally,the findingsof such across tracks,or does effort mattermore for
studies have indicatedthat studentswho are
more engaged learnmore in school (Johnson learningin some tracksthan in others?Each
researchquestionlinkseffort,curricular differ-
et al. 2001; Marks2000; Smerdon1999; Yair and in that will
entiation, learning ways
2000). Farkaset al. (1990) found that stu- enhance researchers' understanding of
dents' "workhabits,"as measuredby teach-
inequalityin students'outcomes.
ers' reportsof homework,class participation, First, does effort vary across curricular
effort,and organization,were positivelyrelat- tracks?Althoughnumerousstudieshave doc-
ed to students'masteryof coursesand grade umenteddifferencesin learningopportunities
point averages (GPAs).Rosenbaum(2001) across tracks (Gamoran and Carbonaro
alsofound that students'"preparedness" and 2002-03; Oakes 1985, 1990; Page 1991),
absenteeismwas relatedto their GPAs.Time
only a handful of studies have examined
spent on homeworkby students is a com- whether effort varies across tracks. Eder
monly used measure of effort (see, e.g., (1981) and Felmleeand Eder(1983) found
Natrielloand McDill1986), and numerous that firstgradersin lower-abilitygroupswere
studies have indicated that it is related to less attentivethan theirpeersin higher-ability
higher achievement (Alexanderand Cook groups. If "attentiveness"is considered a
1982; Carbonaro and Gamoran 2002; measure of effort, this research suggested
Natrielloand McDill1986; but see Bryk,Lee, that ability grouping is related to effort.
and Holland1993 for an exception).Inshort, Natrielloand McDill(1986) found that stu-
althoughthe labelsand measuresused have dents in the college trackspent moretime on
variedacrossstudies,efforthas been found to their homeworkthan did studentswho were
be positivelyrelatedto achievement. not in the college track.While Natrielloand
Althoughresearchon students'effort has McDillused "timespent on homework"as a
yielded interestinginsights,researchershave proxy for effort, this measure is flawed
paid little attention to the possible connec- because other researchon trackinghas indi-
tion between effort and curriculartracking. cated that teachersassignmore homeworkto
This omission is surprising,given the sheer students who are enrolled in higher-track
volume of research on curriculartracking. classes (Oakes 1985). Thus, it is not clear
Most studiesof trackinghave focused on the whether higher-trackstudents are actually
importance of learning opportunities in exertingmoreeffortor whether they are sim-
explaining differences in learning across abili- ply responding to the greater demands that
ty groups and tracks, and only a few have their teachers place on them. Finally,
examined the links between either ability Smerdon (1999) found that engagement, as
grouping or curricular tracking and effort measured by attendance, preparedness, and
(Eder 1981; Felmlee and Eder 1983; Natriello time spent on homework, was positively relat-
and McDill 1986). Hence, an examination of ed to track placement. Smerdon's measures
30 Carbonaro
Beliefs about
Self and
Future
EffortFuue t
SStudent
Track Student
Placement - - - - - -
> Effort
Achievement
Prior
Achieve-
ment Intellectual
Engagement
OTL,
Background
Factors
1985, 1990), students in these classes may data, samples of students in four subjects-
have to expend less effortto learnthe mater- math, English, history, and science-were
ial than may students in lower-trackclasses created. This design allows students to
who are exposed to low-qualityinstructionby appearin multiplesamples,therebyallowing
less-experiencedteachers. Lecturesthat lack theirtrackstatusand effortto varyby subject.
clarityand excitement are likelyto require Consequently,the design of this study is
students to ask more questionsin class, seek superiorto the designs of other studies of
help outside class, and do more work at trackingand effortin which measuresof track
home to learnthe material.Inshort,students status and effort/engagementwere reported
in lower-trackclassesmay need to exert more without referenceto specific academicsub-
effortto do as well as studentsin higher-track jects (e.g., Smerdon 1999). The results
classes.4Thus,effortmay be a criticalpredic- reportedin this articlefocus on math, since
tor of students'success in lower-trackclasses most of the research in the area of high
but may be less importantfor learning in schooltrackinghasfocused on mathachieve-
higher-trackclasses. Examiningthese ques- ment (e.g., Gamoranand Mare1989; Hoffer
tions will illuminate how the interplay 1992). However, the analyses were per-
between structure (learning opportunities formed in all four subjects, and the overall
provided by different track classes) and findingswere consistentacrossacademicsub-
agency (students'effortsto learn)shapesaca- jects (resultsavailableon request).
demic outcomesfor students.
Students' Effort Variables
DATA The main variablesof interest measurestu-
dents' effort.It is importantto recognizethat
The analyses reportedhere used data from measuring effort is problematic in several
the 8th- to 10th-gradecohortof the National respects.First,it may be difficultfor students
Education Longitudinal Survey of 1988 to reportaccuratelythe effort they exert in
school for several reasons. One problem is
(NELS:88).This data set is well suited for
that "exertinghigh effort"may mean some-
studying the relationshipbetween tracking
and effortfor severalreasons.First,the longi- thing differentfor differentstudents. Some
tudinaldesign of the data makesit possibleto studentsmaythinkthey areexertinghigh lev-
controlfor 8th-gradedifferencesin students' els of effortif they do everythingthe teacher
attitudes, behaviors, and achievements, asks; others may consider such effort only
thereby reducing the likelihoodthat differ- adequate. Such differentdefinitionsof effort
ences in effortand learningsimplyreflectpre- decreasethe reliabilityof self-reportedeffort.
existing differencespriorto the 10th-grade Social desirabilitybias is another problem
track placement. Second, the NELS:88data with self-reportsof effort. Brint,Contreras,
providea vast arrayof student-and teacher- and Matthews(2001) found that elementary
reported indicatorsthat serve as important schools transmittedmany messages to stu-
controlsin the models. dents about the valueof hardworkas partof
Finally,and most important,the NELS:88 the hidden curriculum.If studentsinternalize
data providetwo teacherreportsper student such messages,it may be difficultforthem to
in the 8th and 10th grades. In both the 8th admit that they are not working hard.
and 10th grades,two teacherswere surveyed Alternatively,some evidence has suggested
per student;one teachertaught either math that high schoolstudentsmaydownplaytheir
or science, and the other taught either effortin explainingeithertheiracademicsuc-
English or history. Subject-specific reports by cess or failure (Bishop 1999).
teachers have an important advantage over Teachers serve as an alternative source of
data that are not subject specific: Both stu- information about students' effort. Just as
dents' effort and students' track placement with students, there are advantages and dis-
are allowed to vary across classes. To capital- advantages to using teachers' reports of
ize on this important feature of the NELS effort. On the one hand, teachers are limited
Tracking,Effort,and Achievement 33
in that they must indirectlyassess effort. For (math or science and historyor English)who
example,a student maytry hardand spend a reportedthese items on the basisof theirpar-
great deal of time on a homeworkassign- ticularperceptionsof effortin a given subject.
ment, but still do a poor job because he or It would have been preferable to match
she does not havethe skillsor resourcesneed- reportsof 8th- and 10th-gradeeffortby aca-
ed to succeed. A teacher may incorrectly demic subject(i.e., 8th- and 10th-gradeeffort
attributethe student'spoor performanceon as reported by a student's 8th- and 10th-
the assignmentto insufficienteffortand thus grade math teachers). Unfortunately,it was
providean unreliableestimate of effort. On not possible to do so for many students
the otherhand,a teacheris not susceptibleto because their math teachers in the 8th and
social desirabilitybias in reportingstudents' 10th gradeswere not sampled;instead,these
effortand hence may provideless-biasedesti- studentshad a scienceteachersampledin the
mates of effort. 8th gradeand a math teachersampledin the
Despite their potential limitations, the 10th grade. To deal with this problem, I
teachers' reports of students' effort in the included the reports from both 8th-grade
NELS:88data are preferableto the students' teachers for each student in a single scale,
reportsfor two reasons. First,the range of which made it possibleto get an averagelevel
items on effort reported by the teachers is of effortexertedacrossthe two classes,which
moreextensivethan that reportedby the stu- servesas a proxyfor effortin a given subject.
dents. In addition, the teacher measuresof The reliability for this scale was high
students' effort relied on both subjective (Cronbach's alpha = .85).
assessmentsof students'effortand students' A 10th-grademeasureof effortwas creat-
behaviors that are more tangible, easily ed using three items from the 10th-grade
observed, and reported.Again,when differ- reportsfromteachers(see AppendixTableA).
ent types of informationabout effort are Whilethe scale includesan item that is a sub-
used, the limitationsof any given item will be jective reportof effort, it also uses two items
less important.Twoseparatemeasuresof 8th- that are based on students'behaviors:atten-
and 10th-gradeeffort were created for the tiveness and turning in homework.5Turning
analysesthat follow. in homework is an indicatorof procedural
First,the measureof 8th-grade(or "prior") effort,and attentivenessis a measureof intel-
effort was created from teacher-reported lectual effort. Ideally,separate measuresfor
items of effortthat were collected in the 8th each of the three types of effort-rule orient-
grade. Bycontrollingfor 8th-gradeeffort,it is ed, procedural,and intellectual-could be
possibleto determinewhether differencesin includedin the analyses,but the three avail-
10th-grade effort across tracks reflect the able measuresin NELSare best suited to be
placement of students who expended low combinedin a scale, to maximizethe reliabil-
effort into lower-trackclasses, ratherthan a ity and validityof the effortmeasure.
responseby studentsto their placementin a These measures are subject specific for
given track. The seven items used in the each student, and most students had sepa-
"prioreffort" scale (displayed in Appendix rate reportsfrom two of their teachers.The
TableA) cover both subjectiveassessmentsof major advantage of using subject-specific
effort(e.g., "thisstudent performsbelow his samplesis that effortis allowedto varyacross
or her ability") and reports of concrete, students' classes. Students' effort may vary
observablestudent behaviors(e.g., "Thestu- across their classes because of differencesin
dent is frequentlyabsent").Thevariousitems either intrinsicor extrinsicmotivation(e.g.,
in the scale tap the differenttypes of effort students preferringor valuingone academic
mentioned previously: rule oriented (student subject over another). In addition, if (as I
is absent, tardy, and disruptive), procedural argue later) students' track placements vary
(student rarely completes homework), and across subjects and effort is expected to vary
intellectual (student performs below ability, is across tracks, then a subject-specific measure
withdrawn, and is inattentive). of effort is crucial for the analyses.
Each student had teachers in two subjects In her study, Smerdon (1999) used stu-
34 Carbonaro
Table 1. Means and Standard Deviations for the Variables in the Analyses (n = 6,911)a
Background
Female .50 0 1
Race/ethnicity
White,non-Hispanic
(reference) .75 - 0 1
Black .09 - 0 1
Hispanic .10 - 0 1
Asian .06 - 0 1
SES -.013 0.775 -2.97 2.56
8th-GradeAchievement
Math 37.31 12.09 16.18 66.81
Reading 27.60 8.67 10.61 43.83
Science 19.21 4.87 9.46 32.88
History 29.96 4.54 19.23 41.30
10th-Grade
Achievement
Math 37.31 12.08 16.37 72.76
Track(Math)
Honors/advanced .13 - 0 1
Academic .53 - 0 1
General (reference) .29 - 0 1
Vocational/other .05 - 0 1
Students'Effort
Effortscale (G8) 12.45 2.69 0 14
Effortscale (G10) (math) 6.49 2.07 0 9
Effortscale (G10) (English)b 6.47 2.07 0 9
Effortscale (G10) (history)b 6.61 2.05 0 9
Effortscale (G10) (science)b 6.48 2.05 0 9
Students'Self-Efficacy
Locusof control(G8) 0.049 0.595 -2.89 1.52
Self-concept(G8) 0.028 0.649 -3.61 1.89
Educationalexpectations(G8) 4.62 1.27 1 6
Locusof control(G10) 0.0340 .619 -2.66 1.44
Self-concept(G10) -0.0060 .681 -2.95 1.66
Educationalexpectations(G10) 6.40 2.06 1 9
Students'Intellectual
Stimulation
Studentfeels challenged 12.62 7.74 0 20
Askedto show understanding 10.43 7.93 0 20
a The 8-10 panelweight (F1PNLWVVT)
was used to calculatethe meansand standarddeviationsin this
table.
b The mean and standarddeviationfor this variablewere based on a subject-specificsample.
36 Carbonaro
scaledso that 8th- and 10th-gradetest scores opportunitiesto learn (OTL)are present in
are in the same metric. the conceptual model (see Figure1), direct
Fourcategoriesof variableswere used to measuresof learningopportunitieswere not
explaindifferencesin effortand achievement includedin the analysis.Althoughit wouldbe
acrosstracks.First,since backgroundfactors, preferableto include such measuresin the
such as race/ethnicity,sex, and social class, models,validOTLmeasuresareunavailablein
may be relatedto effort,trackplacement,and NELS.12 Consequently,after adjustmentsfor
achievement, controls for these variables the other mediating variables, the track
were includedin the analyses.Race/ethnicity dummy variablesprobablyrepresentdiffer-
was basedon students'8th-gradereportsand ences in exposure to learningopportunities,
was representedby a series of dummy vari- as well as other unmeasuredfactors.
ables(black,Hispanic,and Asian),whichwere
contrastedwith non-Hispanicwhites (the ref-
erence category). SESwas a composite vari- METHODS
able that was composed of five differentfac-
tors takenfromthe base year:familyincome, Ordinaryleast-squares(OLS)regressiontech-
mother'sand father'seducation, and moth- niqueswere used to analyzethe data.Thedis-
er's and father'soccupations.A dummyvari- tributionfor the effort scale is clearlynega-
able ("female")was created to denote each tively skewed. Consequently,when effort is
students'sex. used as a dependentvariable,the estimatesof
Second, measures of prior achievement the coefficientswill be less efficientthan they
consistedof 8th-gradeIRTtest scoresin read- would be if effortwere normallydistributed.
ing, math, science, and history. Following The models presentedherewere rerunwith a
Jencks's(1985) suggestion,allfourtest scores normalizedversionof the effort scale as the
were includedseparatelyas independentvari- dependent variable.Comparisonsof the two
ables in the regressionanalysesto controlfor sets of analysesindicatedthat the magnitude
priorachievement.The use of fourtest scores and levels of statisticalsignificancefor the
minimizedpossiblereliabilityproblemsin the coefficientswere nearlyidentical.The results
analysis.8 for the analyses using the untransformed
Third, multiple measures of students' effortscale are presentedin the tables,since
beliefs about themselves and their future the interpretationof coefficients is more
were used in the analysis:students' (1) 8th- straightforward in these analyses.
and 10th-grade locus of control,9 (2) 8th- As with any quantitativestudy, missing
and 10th-gradeself-concept,10and (3) 8th- data also presented problemsfor the analy-
and 10th-gradeexpectationsregardinghow ses. Althoughthere was no specificvariable
far studentsthinkthey will go in school.11By that had a high degree of item nonresponse
includingthe 8th- and 10th-grademeasures (percentagesof missingcases rangedfrom0
as separatepredictorsin the regressionmod- to 20 percent),when includedtogetherin the
els, I was able to establishwhethera change "full"regressionmodels, 35 percent to 42
in students'locus of control and/or self-con- percent of the cases were lost with listwise
cept affected effort. Finally,two separate deletion. If the data are "missingcompletely
measures of intellectual stimulation were at random,"listwisedeletion providesunbi-
included:whetherstudentsfelt challengedin ased estimates, but the smallersample size
a given subjectand whetherthey were asked decreasesthe statisticalpower of the models
to show understandingin a given subject. (Alison2002).
Althoughitems that measurewhether a stu- To avoidthis limitationof listwisedeletion,
dent was interested in or stimulated by a class multiple imputation (using AMELIA software)
may be preferable, such measures are was used to deal with missing data that were
unavailable in NELS.Regardless,the measures due to item nonresponse. Multiple imputa-
used here are likely to be strongly correlated tion provides larger sample sizes than does
with these more-direct measures. listwise deletion and requires only the weaker
It is important to note that although "missing at random" assumption to produce
Tracking,Effort,and Achievement 37
unbiased estimates. Likeall imputation proce- first-order matrix Taylorseries expansion that
dures, King et al.'s (2001) multiple imputa- generates the correct standard errors
tion technique uses information from other (Statacorp 1999).
variables in the data set to generate predicted
values for cases with missing data. In this
case, all the variables listed in Table 1 were RESULTS
used to impute missing data in each academ-
ic subject. However, multiple imputation is
Does Effort VaryAcross Tracks?
unique in that it generates multiple data
sets-in this case, five-with different imput- Do students in different track classes exert dif-
ed values in each data set. By imputing miss- ferent levels of effort? Table 2 displays the
ing values five different times, it is possible to means for 10th-grade effort by track. The
account for the uncertainty inherent in the higher the track of the class, the more effort
imputation process. The same models are students exerted. The differences in effort
then run on each data set, and the final across tracks are sizable and statistically sig-
results are then averaged across the five nificant in all four subjects. For example, the
analyses. Since the imputed data sets have no difference in effort between students in the
missing data, the sample sizes remain the honors and academic tracks is roughly a third
same in each of the regression models (unlike of a standard deviation. Even more striking,
listwise deletion). The sample sizes in English, the difference in effort between students in
math, science, and historywere 8,518, 6,911, the honors and general track is between 60
5,896, and 4,351, respectively.13 percent and 85 percent of a standard devia-
Since NELSis a multistage cluster sample, tion. The teachers reported that the students
the true standard errors are actually larger in vocational classes exerted the least effort in
than the standard errors reported by most all four subjects.
statistical software packages. The "survey"
command in Stata was used to calculate the
correct standard errors and account for the
Explaining TrackDifferences in
Effort
design effects in NELS.By using information
about the sampling strata and primary sam- The results presented in Table 2 suggest that
pling unit, Stata is able to generate weighted students in different tracks exert different lev-
point estimates that are then used to create a els of effort. What explains these differences
Track
Honors 1.556*** 1.273*** .917*** .496*** .370*** .330***
(.256) (.210) (.151) (.081) (.061) (.054)
Academic .803*** .642*** .394*** .207** .117 .098
(.193) (.155) (.095) (.050) (.028) (.023)
Vocational .168 .207 .242 .293* .368* .388**
(.013) (.017) (.020) (.024) (.030) (.032)
Background
Female .767*** .610*** .662*** .647*** .643***
(.185) (.147) (.160) (.156) (.155)
Black -.353** -.251* -.059 -.215* -.248**
(-.050) (-.035) (-.008) (-.030) (.034)
Hispanic -.129 -.055 .059 .003 .002
(-.019) (-.008) (.008) (.001) (.000)
Asian .532*** .494*** .454*** .412*** .411***
. 6nnn)(.4f6 (.051) .n40) (.046)
SES .300*** .244*** .099** .001 .006
(.1 12) (.091) (.037) (.000) (.002)
PriorEffort
8th-gradeeffort .184*** .184*** .168*** .166***
(.269) (.240) (.219) (.216)
PriorAchievement
Math .027*** .025*** .026***
(.I61) (.148) (.154)
Reading -.001 -.004 -.004
(-.006) (-.018) (-.017)
History .014 .007 .006
(.031) (.016) (.014)
Science .004 .002 .002
(.011) (.003) (.005)
Students' Beliefs
Educational
expectations .147*** .140***
(.147) (.140)
Locusof control(Gi10) .232*** .213***
(.070) (.063)
Self-concept(G10) .051 .045
(.016) (.015)
IntellectualStimulation
Challenge .011**
(.040)
Show understanding .013***
(.050)
AdjustedR2 .060 .114 .179 .201 .221 .226
and numbersin parenthesesare standardizedcoefficients.
Note: Coefficientsare unstandardized,
*
p < .05, ** p < .01, *** p < .001 (two-tailed tests).
trolled. Thus, female and Asian students factors related to students' different experi-
appear to increase their levels of effort in high ences within different track classes explain the
school, regardless of their previous history of remaining track differences in effort. In Model
working hard and academic achievement. 5, 8th- and 10th-grade measures of students'
The next two models examine whether beliefs about themselves and their future
40 Carbonaro
were added as controlsin the regression.The portion of the differences in effort across
inclusionof 8th-grademeasuresaccountsfor tracks:Whenthe coefficientsin Models4 and
any prior differences in students' beliefs 6 are compared,the honors-trackeffect on
before their placement in their 10th-grade effortis reducedby 33 percent,and the aca-
track. Hence, Model 5 examines whether demic coefficient becomes statistically
changesin beliefsthat are due to trackplace- insignificant.Thus, the results suggest that
ment are related to students' effort. Tenth- differencesin effortacrosstracksreflectmore
grade expectationsand locus of control are than simply the types of students who are
both positively related to students' effort, sorted into differenttracks;rather,students'
while students' self-concept is not.15 experiencesin their classesalso partlyexplain
Controllingfor beliefsdecreasesthe trackdif- why higher-track studentstend to exertmore
ferences in effort. The honors-trackeffect is effortthan do lower-trackstudents.
reducedby 30 percentfrom Model4. Forthe It should be noted that the effects of the
academictrack,the coefficientbecomes sta- honorsand vocationaltracks,althoughsmall,
tisticallyinsignificant.Whilethe trackdiffer- remainsignificantin Model6.16The concep-
ences in effort decrease when students' tual modeldisplayedin Figure1 suggeststhat
beliefs about themselvesand the future are after mediatingvariablesare added as con-
controlled,it should be noted that causality trols in the model, no track differencesin
remainsambiguous:It is possiblethat greater effort should remain significant.There are
effort leads to highergrades,which, in turn, several possible explanations for why the
raisestudents'self-concept,locus of control, honors- and vocational-track coefficients
and expectations, so these results must be remainsignificantin the finalmodel. First,the
interpretedwith caution. mediatingvariablesmay be poorlymeasured
Model
Finally, 6 adds controlsfor intellectu- in the modelsand hence do not serveas ade-
al stimulationto the model.One way in which quate controlsfor the concepts in the analy-
track placement may be relatedto effort is ses. Second, there may have been unob-
through greater intellectual stimulation in served variablesthat were omitted from the
responseto the more-demandingcurriculum model that created track "effects"that are
and stimulating instructionin higher-track spurious.Finally,it is possible that teachers
classes. Both the level of challenge and the are susceptibleto a "halo effect," whereby
degree to which students are requiredto students' track placements affect teachers'
"show understanding" in math are positively judgmentsabout theireffort,independentof
relatedto effort,althoughthe magnitudeof the the actual levels of effortthe studentsexert.
associationis fairlymodest.The resultsindicate Since the effort measurerelieson some stu-
that intellectualstimulationexplainsa modest dent behaviorsthat are observableby teach-
portion(11 percent)of the honorseffect on ers (i.e., completionof homeworkand paying
effort, but the coefficientremainsstatistically attention), this effect is probably minor.
significant.If better measuresof intellectual However,it is stillpossiblethat missedhome-
stimulation,such as indicatorsof a student's work assignmentsand incidencesof inatten-
interestin or stimulationby a given class,were tion by lower-trackstudents are perceived
includedin the model,moreof the trackdiffer- more readily and negatively by teachers,
ences in effortmaybe explained. thereby artificiallystrengthening the track-
Thus, the overallconclusion drawn from effortrelationship.
Table3 is that trackdifferencesin effortin the
10th grade are explained mostly by the Effort and TrackEffects on
process by which students are sorted into Achievement
tracks: Students who were predisposed to
exert more effort in the 10th grade (as evi- The first two sets of analyses suggest that
denced by their 8th-grade effort and achieve- there are important differences in effort
ment) were more likely to be placed in high- across tracks and that although these differ-
er tracks. However, when combined, beliefs ences are largely the result of the different
and intellectual stimulation explain a sizable types of students who are placed in different
Tracking,Effort,and Achievement 41
tracks,they are also partlydue to students' How does controllingfor students' effort
experienceswithin these tracks.The next set and track placement affect the relationship
of questionslinkeffortwith trackdifferences between background characteristics and
in achievement.In particular,are trackdiffer- achievement?Controllingfor effortdoes not
ences in learning partly explained by the change the racial/ethniccoefficientsand only
greater effort exerted by students in higher slightly reduces the SESeffect on achieve-
tracks?The resultsof the analysesexamining ment. It is interestingthat althoughthe base-
this questionare displayedin Table4. line model (Model 1) did not revealgender
The first model (Model 1) estimates the differencesin achievement, the addition of
trackdifferencesin gains in math after back- the effortmeasurerevealedan advantagefor
ground characteristics and 8th-grade male students. Hence, female students are
achievementare controlled.These "adjusted closing the achievement gap in math with
trackeffects"indicatethat regardlessof stu- male studentsby exertingmore effortin their
dents' priormathachievement,studentsgain math classes(see Table3).
morewhen they are placed in a highermath While10th-gradeefforthas a strongeffect
track. Controls for 10th-grade effort were on learning,this effect could be due largelyto
added to the next model (Model2) to estab- the fact that students who exert more effort
lish whether differences in effort partly are morelikelyto be sortedinto highertracks.
explainsome of the trackdifferencesin math To eliminatethis possibility,Model 3 adds the
gains. Efforthas a significant,positiveeffect controlfor 8th-gradeeffortas a predictorof
on math gains. These effects are relatively 10th-grade learning gains. The addition of
largecomparedwith the differencesin math this variableto the model slightly reduces
gains across tracks: A standard deviation changes in the trackdifferencesin learningor
increasein effortproducesan averagegain in the effects of effort on students' learning
math achievementthat is two fifthsand one gains, but both remain sizable and signifi-
quarterof the academic and honors effects cant. When 10th-gradeeffortis removedas a
(respectively).17 predictor of achievement in Model 3, the
Does the positive relationshipbetween resultsindicatethat 8th-gradeeffortexplains
effort and math gains partly explain the some of the trackdifferencesin achievement,
effects of trackon learning?When the track but only about half as much as 10th-grade
coefficientsin Model 1 are compared with effortdoes in Model2.18Thus,it appearsthat
those in Model2, a fairlymodest reductionin the effectsof 10th-gradeeffortdo not simply
the trackeffects on learningis observed:The reflectthe fact that studentswho try hardare
effect of honors trackis reduced by roughly more likely to be sorted into higher-track
10 percent,and the effect of academictrack classes. Rather,students who exert greater
is reduced by 7 percent. In the other three effort in the 10th grade learn more, regard-
subjects,effort explainsslightlymore of the less of how much effort they exerted in the
trackeffects(generallyfrom 10 percentto 20 8th grade.
percent),but most of the trackeffects remain
unexplained (results not shown). Hence, Effects of Effort on LearningAcross
althougheffortplaysan importantrolein pre- Tracks
dicting learning,it explainslittle of the track
effects on learning.If effortdoes not explain The last question of interest focuses on
much of the trackdifferencesin achievement, whetherthe effect of efforton learningvaries
what does? As Figure1 (and priorresearch) across tracks. Model 4 in Table4 examines
suggests, differences in learning opportuni- this issue by adding interaction terms
ties across tracks likely account for track dif- between effort and track placement to the
ferences in achievement. Unfortunately, this model. None of the interaction terms is sta-
interpretation of the results cannot be con- tistically significant at the .05 level. Hence,
firmed through further analysis because of the results suggest that the effects of effort on
the absence of adequate direct measures of learning are the same for all students, regard-
learning opportunities in the NELSdata. less of their track.
42 Carbonaro
Track
Honors 4.866*** 4.451 *** 4.275*** 3.960**
(.119) (.109) (.104) (.097)
Academic 3.474*** 3.252*** 3.090*** 3.515"***
(.124) (.117) (.110) (.126)
Vocational -.670 -.836 -.796 .499
(-.008) (-.010) (-.010) (.006)
Effort
10th-grade effort .602*** .534*** .563***
(.089) (.079) (.083)
8th-grade effort .213*** .212***
(.041) (.041)
Trackx Effort
Honorsx Effort -.030
(.006)
Academicx Effort -.051
(-.017)
Vocationalx Effort -.162
(-.016)
Background
Female -.129 -.732*** -.733*** -.730***
(-.005) (-.026) (-.026) (-.026)
Black -1.573*** -1.513*** -1.479*** -1.477***
(-.033) (-.031) (-.031) (-.031)
Hispanic -.577* -.592* -.555* -.558*
(-.012) (-.013) (-.122) (-.012)
Asian .425 .136 .139 .130
(.007) (.002) (.002) (.002)
SES .920*** .858*** .859*** .859***
(.051) (.047) (.047) (.047)
PriorAchievement
Math .727*** .707*** .702*** .702***
(.632) (.615) (.611) (.611)
Reading .114*** .115** .113*** .113***
(.071) (.071) (.070) (.070)
History .209*** .194*** .186*** .185***
(.068) (.063) (.060) (.060)
Science .1 75*** .1 74*** .1 76*** .1 76***
(.061) (.061) (.061) (.061)
AdjustedR2 .788 .794 .796 .796
Note: Coefficientsare unstandardized,and numbersin parenthesesare standardizedcoefficients.
* p < .05, ** p < .01, *** p < .001 (two-tailed tests).
Tracking,Effort,and Achievement 43
Appendix Table A
Items Included in the 8th- and 10th-Grade Effort Scales
SurveyItems ResponseCategories
8th-GradeEffort
Student performsbelow his or her ability[BYTI_2,BYT42] 0 = no, 1 = yes
Student rarelycompletes homework[BYT1_3,BYT4_3]
Student is frequentlyabsent [BYTI_4,BYT4_4]
Student is frequentlytardy [BYT1_5,BYT4_5]
Student is inattentivein class [BYT1_6,BYT4_6]
Student is exceptionallypassive/withdrawn[BYT1_7,BYT4_7]
Student is frequentlydisruptive[BYT1_8,BYT4_8]
10th-GradeEffort
Does this student usuallywork hard in your class? [F1T1_2,F1T5_2] 0 = no, 1 = yes
How often is this student attentive in class? [F1T1_18,F1T5_18] 0 = never, 1 = rarely,
2 = some of the time,
3 = most of the time,
4 = all of the time
How often does this student complete homework assignments in
your class? [F1T1_15,F1T5_15] Same as above
ses in that paperreliedon combinedteachers' respondents were asked, "how they felt
reportsof effortand global measuresof track about"the foregoingitemsand couldchoose
that were derivedfrom students'transcripts. froma scalethat rangedfrom"stronglyagree
Althoughnot the mainfocus of the paper,the to stronglydisagree."
findings regardingthe relationshipsamong 10. The 8th- and 10th-gradescales were
tracking,effort,and achievementreportedin created using students'8th- and 10th-grade
that study were consistentwith those report- reportson the same items. The items are as
ed here. However,since this articlefocuses follows:(1) "Ifeel good about myself";(2) "1
specificallyon tracking,effort, and achieve- am a person of worth, the equal of others";
ment, subject-specificmeasuresof effortand (3) "1am able to do things as well as most
trackare the most appropriate,both concep- other people";(4) "Onthe whole, I am satis-
tuallyand empirically. fied with myself";(5) "Icertainlyfeel useless
7. Points2 and 3 requiresome elaboration. at times";and (6) "Attimes, I thinkI am no
First,teachershave a betterunderstandingof good at all." The respondentswere asked
the local context and the meaning attached "how they felt about" the foregoing items
to course labelsthan the transcriptdata can and could choose from a scale that ranged
convey. Second, the transcriptdata provide from "stronglyagree to stronglydisagree."
course labels, but in subjects other than 11. The responsecategoriesfor this vari-
math, it is not possibleto createcourse-based able were as follows:for 8th-gradeexpecta-
indicatorsof trackbecause (1) the labelsare tions, (1) "won'tfinishhigh school,"(2) "will
not specific enough to distinguishbetween finishhigh school,"(3) "willattendvocation-
tracklevels(high versuslow), and (2) no clear al/trade/businessschool after high school,"
course-takingsequence is apparent. Further (4) "willattend college," (5) "willfinishhigh
explanationand detailsof the analysesusing school,"and (6) "willattend a higherschool
the transcriptdata are availableon request. after college"; for 10th-grade expectations,
8. Jencks(1985) argued that when con- (1) "less than high school graduation,"(2)
trollingfor priorachievement,it is preferable "highschool graduationonly,"(3) "lessthan
to add controlsfortest scoresin differentaca- two years of trade school," (4) "morethan
demic subjects (e.g., including 8th-grade two yearsof tradeschool,"(5) "lessthantwo
math, reading,history,and science scores as years of college," (6) "two or more years of
predictorsof 10th-grademath scores), rather college," (7) "finishcollege," (8) "master's
than simplycontrollingfor a single test score degree,"and (9) "Ph.D.or M.D."
in one academic subject (e.g., 8th-grade 12. Ideally,adequatemeasuresof learning
math scores as a control predicting 10th- opportunitieswould includevariables,suchas
grade math scores). He contended that prior instructional time, curricular coverage,
achievementwill be measuredmore reliably instructional quality, and teacher quality.
when multipletest scores are added to the AlthoughNELSincludessome informationon
regressionbecause the additionalscores will these aspects of students' schooling experi-
help correct any measurementerror in one ences, the measures are generally crude.
particulartest score. However,it should be noted that Carbonaro
9. The 8th- and 10th-grade scales were and Gamoran(2002) had some success in
created using students'8th- and 10th-grade using these measuresin theirexaminationof
reportson the same items. The items are as achievementin English.It is unclearwhether
follows:(1) "Idon't haveenough controlover similarmeasurescould be createdin math or
my life";(2) "Good luck is more important other subjects.
than hardwork";(3) "Everytime I try to get 13. The samplesizes differacrossacadem-
ahead, something or somebody stops me"; ic subjects because some students did not
(4) "My plans hardly ever work out, so plan- take courses in each of the four subjects in
ning only makes me unhappy"; (5) "When I their junior year.
make plans, I can almost always make them 14. The bivariate correlation between 8th-
work"; and (6) "Chance and luck are very and 10th-grade effort is .346. Eighth-grade
important in what happens in my life." The achievement is also positively correlated with
Tracking,Effort,and Achievement 47
Ravitch. Washington, DC: Brookings Lucas, Samuel, and Adam Gamoran. 2002.
InstitutionPress. "Trackingand the AchievementGap." Pp.
Fordham,Signithia.1996. Blacked Out:Dilemmas of 171-98 in Bridging theAchievement Gap,edit-
Race, Identity,and Successat CapitalHigh. ed by John Chubb and Tom Loveless.
Chicago:Universityof ChicagoPress. Washington,DC:BrookingsInstitutionPress.
Fordham,Signithia,and JohnOgbu. 1986. "Black MacLeod,Jay.1995. Ain'tNo Makin'It:Aspirations
Students' School Success: Coping with the andAttainment in a Low-Income Neighborhood
'Burdenof "ActingWhite."'"UrbanReview (2nd ed.). Boulder,CO:WestviewPress.
18:176-206. Marks, Helen. 2000. "Student Engagementin
Gamoran, Adam. 1986. "Institutional and Instructional Activity: Patterns in the
InstructionalEffects of Ability Grouping." Elementary, Middle,and HighSchoolsYears."
Sociologyof Education 59:85-98. American Educational Research Journal
1987. "The Stratificationof High School 37:153-84.
Learning Opportunities." Sociology of Mickelson, Roslyn. 1990. "The Attitude-
Education 60:135-55. Achievement Paradox Among Black
Gamoran, Adam, and William Carbonaro. Adolescents."Sociologyof Education63:44-61.
2002-03. "High School English:A National Natriello, Gary, and Edward McDill. 1986.
Portrait."HighSchoolJournal86:1-13. "PerformanceStandards,Student Efforton
Gamoran, Adam, and Robert Mare. 1989. Homework, and Academic Achievement."
"SecondarySchool Trackingand Educational Sociologyof Education59:18-31.
Inequality: Compensation, Reinforcement,or Newmann, Fred, ed. 1992. StudentEngagement
Neutrality?"Americanjournal of Sociology and Achievementin American Secondary
94:1146-83. Schools.New York:TeachersCollegePress.
Ginsberg, Margery,and RaymondWlodkowski. Oakes,Jeannie.1985. KeepingTrack:HowSchools
2000. CreatingHighlyMotivatingClassrooms StructureInequality.New Haven, CT: Yale
forAllStudents.San Francisco: Jossey-Bass. UniversityPress.
Hallinan,Maureen.1992. "The Organizationof -. 1990. Multiplying SantaMonica,
Inequalities.
Students for Instruction in the Middle CA:RAND.
School."Sociologyof Education 65:114-27. Ogbu, John.1978. Minority, and Caste.
Education,
Hoffer, Thomas. 1992. "Middle School Ability New York:AcademicPress.
Grouping and Student Achievement in -. 2003. BlackAmericanStudentsin an Affluent
Science and Mathematics." Education Suburb:A Studyof AcademicDisengagement.
Evaluation and PolicyAnalysis14:205-27. Mahwah,NJ:LawrenceErlbaum.
Jencks,Christopher.1985. "HowMuch Do High Page, Reba. 1991. Lower TrackClassrooms:A
SchoolStudentsLearn?" Sociologyof Education and CulturalPerspective.
Curricular New York:
58:128-35. TeachersCollegePress.
Johnson,MonicaKirkpatrick, RobertCrosnoe,and Pallas,Aaron,DorisEntwisle,KarlAlexander,and
Glenn Elder,Jr.2001. "Students'Attachment Maria Stluka. 1994. "Ability-Group Effects:
and AcademicEngagement:The Roleof Race Instructional, Social, or Institutional?"
and Ethnicity." Sociology of Education Sociologyof Education 67:27-46.
74:318-40. Rosenbaum,James.2001. BeyondCollegefor All:
Jones, James, Beth Vanfossen, and Margaret CareerPathsforthe ForgottenHalf.New York:
Ensminger. 1995. "Individual and RussellSage Foundation.
Organizational Predictorsof HighSchoolTrack Rowan, Brian, and Andrew Miracle,Jr. 1983.
Placement." Sociologyof Education68:287-300. "Systems of Ability Grouping and the
Kerckhoff,Alan. 1976. "The Status Attainment Stratificationof Achievementin Elementary
Process:Socializationor Allocation?"Social Schools."Sociologyof Education 56:133-44.
Forces55:368-81. Smerdon, Becky. 1999. "Engagement and
King, Gary, James Honaker,Anne Joseph, and Achievement:DifferencesBetween African-
Kenneth Scheve. 2001. "Analyzing Americanand White High School Students."
Incomplete Political Science Data: An Research in Sociology of Educationand
Alternative Algorithm for Multiple Socialization12:103-34.
Imputation."AmericanPoliticalScienceReview Serensen,Aage, and MaureenHallinan.1977. "A
95:49-69. Reconceptualization of School Effects."
Lucas, Samuel. 1999. Tracking Inequality: Sociologyof Education 50:273-89.
and Mobilityin AmericanHigh Statacorp.1999. StataStatisticalSoftware:Release
Stratification
Schools.New York:TeachersCollegePress. 6.0. CollegeStation,TX:StataCorporation.
Tracking,Effort,and Achievement 49
TheauthorthanksAdamGamoran,MichaelOlneck,WarrenKubitschek, MaureenHallinan,Sean
of Wisconsin-Madison
Kelly,and attendeesof the University sociologyof educationbrownbagseries
fortheirvaluablefeedbackon an earlierversionof thisarticleand BridgetNicholsonforher valu-
able researchassistanceon this project.Addressall correspondencesto WilliamCarbonaro,
Departmentof Sociology,University of NotreDame, 1016 FlannerHall,NotreDame,IN, 46556;
e-mail:wcarbona@nd.edu.