high school dropout and the role of career and technical education

High School Dropout and the Role ofCareer and Technical Education

A Survival Analysis of Surviving HighSchool

Stephen B PlankJohns Hopkins University

Stefanie DeLucaJohns Hopkins University

Angela EstacionJohns Hopkins University

This article uses data from the National Longitudinal Survey of Youth 1997 to investigate high

school dropout and its association with the high school curriculum In particular it examines

how combinations of career and technical education (CTE) and core academic courses influ-

ence the likelihood of leaving school Hazards models indicate a significant curvilinear associ-

ation between the CTE-to-academic course-taking ratio and the risk of dropping out for youths

who were aged 14 and younger when they entered the ninth grade (not old for grade) This

finding suggests that a middle-range mix of exposure to CTE and an academic curriculum can

strengthen a studentrsquos attachment to or motivation while in school The same association was

not found between course taking and the likelihood of dropping out for youths who were

aged 15 or older when they entered high school thus prompting further consideration of the

situation of being old for grade in school settings that remain highly age graded in their orga-

nization

Sociology of Education 2008 Vol 81 (October) 345ndash370 345

The phenomenon of students droppingout of high school has recently gainedrenewed attention with researchers and

policy makers wanting to know how manystudents are dropping out what causesdropout and what may be done to prevent it(Bridgeland DiIulio and Morison 2006Heckman and LaFontaine 2007 Orfield2004) Leaving high school before receiving adiploma often represents a culminating eventin a long-term process of disengagement fromformal education (Alexander Entwisle andKabbani 2001 Finn 1989) Understandingthis process requires the consideration of stu-

dentsrsquo traits (some invariant some malleable)and the social organization of the schoolhome and neighborhood

One important aspect of the school envi-ronment is the formal curriculummdashthe set ofcourses taken by a student The high schoolcurriculum can be viewed as a socially struc-tured set of opportunities and constraints Anindividual course may represent an opportu-nity for inspired learning establishing socialconnections boredom or discouragementThe combination of all courses takenthroughout a high school career significantlydefines a studentrsquos place within the social

02 Plank 10808 551 PM Page 345

Delivered by Ingenta to Johns Hopkins University

Mon 17 Nov 2008 190028

346 Plank DeLuca and Estacion

organization of the school and determines inpart the paths that a student follows afterhigh school

One domain of the high school curriculumthat has been both lauded and criticized forits potential effects on educational outcomesincluding dropping out is career and techni-cal education (CTE) also known as vocation-al education Some have argued that CTEoffers a much-needed point of attachmentand demonstrates the relevance of schoolingfor many youths (Advisory Committee 2003)Others have argued that CTE represents amarginalized part of the high school land-scape one in which mixed messages are sentabout the value of schooling and a youthrsquossuccess as a student (Cavanagh 2005) Aseries of educational reforms in the 1990semphasized the importance of taking bothacademic and vocational courses so that stu-dents were prepared for the transition to col-lege and work (Castellano Stringfield andStone 2003)

In this article we use data from theNational Longitudinal Survey of Youth 1997(NLYS97) to estimate hazards models of therelationship between CTE course taking andhigh school dropout for a recent cohort ofstudents Our analyses reveal a curvilinearassociation between the CTE-to-academiccourse-taking ratio and the risk of droppingout for youths who were ldquoon timerdquo for gradeprogression (ie younger than age 15 onentering the ninth grade) We interpret thisfinding as revealing that a middle-range mixof exposure to CTE and the academic curricu-lum can strengthen a studentrsquos attachment toor motivation while in school We do not findthe same significant association betweencourse taking and the likelihood of droppingout for youths who were 15 or older on highschool entry prompting further considerationof the implications of being old for grade inschool settings

BACKGROUND

Research on Dropping Out

Research on dropping out of high school hashighlighted a web of sociological psycholog-

ical economic and institutional factors thatcontribute to students leaving high schoolbefore they receive a diploma (Allensworth2004 Fine 1991 Orfield 2004) Furthermorestudies have shown that the consequences ofdropping out are dramatic and costlymdashforboth individuals and the society The negativeoutcomes associated with dropping outinclude a higher likelihood of unemploymenta greater chance of living below the povertyline and relying on public assistance morefrequent and severe health problems andincreased criminal activity (EducationalTesting Service 1995 Mishel and Bernstein1994 National Research Council NRC 1993Rumberger 1987)

Rumberger (2001) reported that despite along-term upward trend in high school com-pletion in the United States approximately 5percent of all high school students drop outof school in any given year1 A substantiallyhigher proportion of students leave school fora significant span of time sometime duringtheir educational careers before they receive ahigh school credential Klerman and Karoly(1994) estimated that 37 percent of a nation-al sample of young men who were aged14ndash21 in 1979 quit high school for at leastthree months sometime before they complet-ed high school Rumberger and Lamb (2003)reported that 21 percent of students from theNational Education Longitudinal Study of1988 (NELS88) cohort dropped out of schoolat some point after the eighth grade butbefore they completed high school In bothstudies many of those with spells of droppingout ultimately returned to complete a highschool credential However as past researchhas shown and the present analysis reiteratesany time away from high school increases therisk of never obtaining a credential

Finn (1989) offered two social-psychologi-cal perspectives to explain dropping outFollowing a frustrationndashself-esteem model heposited that a record of poor performancewould cause students to question their com-petence and weaken their attachment toschool at the extreme this process can leadto the ultimate disengagementmdashdroppingout Following a participation-identificationmodel he posited that positive experiencesencourage a sense of belonging and thus

02 Plank 10808 551 PM Page 346


Mon 17 Nov 2008 190028

Dropout and Career and Technical Education 347

strengthen attachment to school Rumberger(2004) drew on both individual and institu-tional perspectives to understand decisions toquit school The individual perspective focus-es on a studentrsquos values attitudes and behav-iors and considers dropping out of school tobe the final stage in a cumulative process ofacademic and social disengagement fromschool The institutional perspective situatesindividual attitudes and behaviors within thebroader settings or contexts in which stu-dents live most notably families schools andcommunities

As we consider factors that affect droppingout there are also important questions aboutthe timing of dropout and whether thesalience of various risk factors is variable overtime Research has noted the difficulty of theninth-grade transition for many students andits connection to dropping out (Neild andFarley 2004 Roderick and Camburn 1999)Therefore it is important to ask whether theeffects of various independent variables areconstant or nonconstant as the high schoolcareer progresses As one example of such aninquiry Swanson and Schneider (1999) stud-ied the effects of mobility (both change ofresidence and change of schools) on educa-tional achievement and social outcomesThey found that the effects of mobility (par-ticularly a change of schools) were noncon-stant over time Specifically they detectedeffects that were somewhat negative in theshort term but ultimately beneficial in thelong term

As we model the likelihood of dropoutone can imagine several instances in whichthe effects of independent variables may benonconstant over time For example paststudies have noted that there are both acad-emic and social causes of dropping out(Goldschmidt and Wang 1999 Rumbergerand Larson 1998) There is reason to askwhether the relative importance of the acad-emic and social realms shifts over time If stu-dents are especially reactive to academic suc-cesses and failures early in high school butmore attuned to social attachments later wemay expect the effect of grade point average(GPA) on the likelihood of dropout to bestrong and positive in the early months ofhigh school but weaker in later months

As another example past research hasidentified the role incompatibilities that comewith being ldquooff-timerdquo or older than onersquosclassmates as a significant strain that can leadto dropping out (Entwisle Alexander andOlson 2004 Jimerson Anderson andWhipple 2002 Roderick 1994) One canimagine that such strains for those who areold for grade would become especially pro-nounced as they reach age 17 18 or 19when lures toward adult (and nonstudent)roles and identities become especially pro-nounced If this were the case one wouldexpect to find that being old for grade wouldhave a positive association with the risk ofdropout at all points in the high schoolcareer but with increasing intensity as oneproceeds through high school

A related issue is whether being old forgrade may act as a moderator that affects thestrength of association between other inde-pendent variables and the likelihood ofdropout (Baron and Kenny 1986) For exam-ple students who enter high school old forgrade may carry so many risk factors and aca-demic deficiencies that any potential benefitsof a particular curricular program are lost onthem Given these possibilities we examinewhether the relationship between course tak-ing and the likelihood of dropout differsaccording to age at entry into the ninthgrade

CTE

In this article we explore whether CTE incombination with academic courses predictsdropping out CTE has evolved from what hastraditionally been called vocational educa-tion Whether this evolution has been dra-matic is open to debate (DeLuca Plank andEstacion 2006 Stone 2000) Historically mostvocational education programs weredesigned to prepare students for work andhelp them enter the workforce shortly afterhigh school During the past 15 years therehave been efforts to improve vocational edu-cation programs not only to prepare studentsfor jobs but to increase their educationalattainment Recent federal legislation pre-sents a vision of CTE that involves not only

02 Plank 10808 551 PM Page 347


Mon 17 Nov 2008 190028


the development of labor market skills butthe integration of CTE and academic subjectsthe erasure of the stigma often attached tovocational education and the provision ofpathways to both postsecondary educationand employment (Castellano et al 2003Lynch 2000)2

Formal high school courses that are typi-cally categorized as CTE include (1) familyand consumer sciences (2) general labormarket preparation and (3) various courses in10 specific labor market preparation (SLMP)areas (National Center for EducationStatistics NCES 1999) Family and consumersciences courses include home economicschild development foods and nutrition andfamily relations General labor market prepa-ration courses include basic keyboardingexploratory industrial arts college and careerplanning and introduction to technologyThe 10 SLMP areas are agriculture and renew-able resources business marketing and distri-bution health care public and protective ser-vices trade and industry technology andcommunications personal and other servicesfood service and hospitality and child careand education SLMP courses are a combina-tion of classroom-based learning experiencescooperative education and other workplacelearning

Although not every US high school offerscomprehensive CTE programs most offersome CTE and the majority of high schoolstudents take at least one CTE course (NCES2001) In recent years 90 percent to 96 per-cent of US high school graduates have takenat least one CTE course during high school(DeLuca et al 2006 Levesque et al 2000)The average number of Carnegie units earnedin CTE among graduates in 2000 was 38 outof an average of 262 total credits (NCES2004)

CTE is not without its critics and its placewithin the American high school is not espe-cially secure Recent federal budget debatesoccurring at a time when greater emphasis isbeing placed on testing and core academicareas have questioned the effectiveness andmerit of CTE (Cavanagh 2005) Some haveportrayed CTE as a dumping ground in whichunmotivated youths encounter low expecta-tions and outdated training In past decades

research on tracking has often portrayedvocational education as a domain in whichpoor and minority youths are disproportion-ately concentrated with these students expe-riencing courses with little rigor and limitedlearning opportunities (Gamoran and Mare1989 Oakes 1985 Vanfossen Jones andSpade 1987) However proponents viewCTE when implemented properly as animportant part of the high school environ-ment and a valuable source of attachmentmotivation and learning especially for non-college-bound students (Arum 1998Castellano et al 2003 Rosenbaum 2001)

Part of the disjuncture between the pes-simism of the older literature on tracking andthe optimism of more recent commentarieson CTE stems from perceptions that a genuinetransformation has taken place in the way thatoccupationally based themes are presented tohigh school students The ldquonewrdquo vocationaleducation has been touted as having consid-erable potential to enhance studentsrsquo engage-ment with school and integrate academiccontent with occupational applications (NRCand Institute of Medicine 2004) In schoolsthat have embraced the ldquonewrdquo vocationaleducation fairly broad occupational themesare often featured such as health occupationsrather than nursing or industrial productioninstead of welding These theme-based set-tings are intended to offer students consider-able autonomy in selecting tasks and methodsof learning and in constructing meaning(Advisory Committee 2003 Stone 2000) Thisapproach contrasts with the narrow focus ofthe older model of vocational education thatoften attempted to prepare individuals forspecific entry-level jobs We are limited in ourability to discern from the present data howmuch of the sample membersrsquo exposure toCTE can accurately be called ldquonewrdquo vocation-al education as opposed to the older modelbut we consider these issues again in our con-cluding section

CTE and Possible Links toDropping Out

The question of whether vocational educa-tion or CTE has a causal connection to drop-ping out has been posed by researchers peri-

02 Plank 10808 551 PM Page 348


Mon 17 Nov 2008 190028


odically for several decades (eg Agodini andDeke 2004 Bishop 1988 Catterall and Stern1986 Combs and Cooley 1968 Grasso andShea 1979 Mertens Seitz and Cox 1982Perlmutter 1982 Pittman 1991 Rasinski andPedlow 1998) Despite fairly frequentattempts to address the issue a clear andconsistent answer has not emergedReviewing 30 years of studies Kulik (1998)concluded that participation in vocationalprograms increases the likelihood that non-college-bound youths will complete highschool Specifically he estimated that anyparticipation in CTE decreased the dropoutrate of such youths by about 6 percentDespite Kulikrsquos overall conclusion of a positiveeffect of vocational education on completinghigh school some studies have found nosuch effects (eg Agodini and Deke 2004Pittman 1991)

Another perspective on the relationshipbetween CTE and dropout drawn from theliterature on tracking suggests a negativerelationship Classic work in the field hashighlighted the low quality of the contentand teaching in vocational tracks and hasdemonstrated that the vocational track isoften used as a remedial track serving as away to remove poor-performing studentsfrom classes with their college-bound peers(Oakes 1985 Rosenbaum 1978) Kang andBishop (1989) also noted that many studentswho may be on the verge of dropping outenroll in vocational courses so they canacquire some job-related skills before theyleave Both the channeling and self-selectionprocesses create a situation in which studentsbecome increasingly detached from school asa result of unpleasant experiences in thesevocational courses

However many of these studies were con-ducted before the federal legislative initiativesin the 1990s that attempted to reform thenature of vocational education by emphasiz-ing an integrated approach to CTE and acad-emic courses A 2003 report of the AdvisoryCommittee for the National Assessment ofVocational Education suggested that combin-ing academic courses with CTE courses canbe a powerful experience for students keep-ing them attached to school and motivatingthem to complete their diplomas For exam-

ple CTE can appeal to different learningstyles and interests because the programsdirectly connect academic skills in the class-room with real-world activities in the work-place CTE classes can also clarify the value ofacademic classes by specifying the skills thatare needed to succeed in careers of interestthereby leading students to see a greatervalue associated with staying in schoolParticipation in CTE programs may alsoencourage some students to define theircareer goals thus keeping them interested inschool (Advisory Committee 2003)

Explained another way blending CTE withacademic work may increase studentsrsquoengagement a concept that has been shownto be strongly related to academic achieve-ment and attainment (Connell Spencer andAber 1994 Marks 2000 NRC and Council ofMedicine 2004) Past research directs ourattention to the ways in which curriculararrangements can affect behavioral emotion-al and cognitive domains of engagement (cfFredricks Blumenfeld and Paris 2004) Theeffects of a curriculum can be profound espe-cially if emotional and cognitive connectionsare forged for students as they connect theirpresent academic preparation with futuregoals In our work it is important to considerwhether any observed associations betweenstudentsrsquo course taking and the likelihood ofdropping out can be explained by argumentsabout engagement Our data did not allow usto test important mediating mechanismsinvolving engagement but we use the litera-ture on the topic to guide our interpretationof findings

Models of the RelationshipBetween CTE and Dropout

To study the relationship between highschool course taking and dropping out weneeded an appropriate way to measure a stu-dentrsquos curricular experience In an era inwhich rigid track definitions (eg collegepreparatory or purely vocational) do notalways neatly apply (cf Lucas 1999) it is use-ful to examine curricular exposure along acontinuum of CTE and academic course tak-ing For any academic term or for a highschool career cumulatively a studentrsquos ratio of

02 Plank 10808 551 PM Page 349


Mon 17 Nov 2008 190028


CTE credits to core academic credits can becomputed (assuming the number of academ-ic credits is not zero)3 This ratio allowed us tosummarize a studentrsquos location within thecurriculum when students were observed tocombine CTE and academic courses in manyratios and qualitative patterns

We offer four competing hypotheses aboutthe relationship between the CTE-to-academ-ic course-taking ratio and the likelihood ofdropping out once other relevant factors(prior test scores high school GPA genderrace and socioeconomic status) have beencontrolled The first possibility is a positive lin-ear relationship suggesting that a higher pro-portion of CTE courses leads to dropout Thispossibility follows from some previousresearch on tracking which suggested thatincreased levels of CTE course taking mayrepresent a student being channeled intolow-level classes and away from core acade-mic learning opportunities thus raising thelikelihood of the student leaving school Thesecond possibility is a negative linear relation-ship between CTE and the likelihood of drop-ping out consistent with the expectation thathigh levels of CTE are beneficial This possibil-ity suggests that when students have theopportunity to concentrate in CTE they aremore likely to stay in school because moreindividually relevant and satisfying choicesare available to them The third possibility isno effectmdashconsistent with the idea that thesubstance of course taking itself has no inde-pendent effect on the likelihood of droppingout once other relevant factors have beencontrolled

The fourth possibility is a curvilinear rela-tionship by which the likelihood of droppingout initially decreases as the CTE-to-academicratio increases but only up to a pointBeyond this point the likelihood of droppingout increases as the CTE-to-academic ratioincreases Previous work found that a ratio ofapproximately three CTE courses to everyfour academic courses was associated withthe lowest odds of dropping out of highschool (Plank 2001) Ratios either higher orlower than 3 to 4 were associated with anincreased risk of dropping out and this asso-ciation was especially salient for individualswith poor academic performance

While this earlier finding is intriguingthere are certain methodological shortcom-ings to be addressed The models in Plankrsquos(2001) study were standard logit models ashave been most models of dropping out ofhigh school presented in journals in the pastseveral decades (cf Willett and Singer 1991)When one studies the effects of course taking(and other time-varying independent vari-ables of interest) there is a potential problemof reverse causality (Kang and Bishop 1989)Students generally take more CTE courses inthe last two years of high school (Levesque etal 2000) and therefore any observed associ-ation between course taking and CTE may bedriven by the fact that only those who per-sisted to the final years of high school couldattain a relatively high (or at least middle-range) CTE-to-academic ratio Our studyaddressed these issues by using hazards mod-els to account appropriately for the timing ofcourse taking and its relationship to dropout

METHODS

Sample

We used the NLSY97 which tracks a nation-ally representative sample of 8984 youths inthe United States who were aged 12 to 16 asof December 31 1996 Annual interviews col-lect detailed information about educationaldevelopmental and labor market experi-ences In addition to survey data from thesampled participants we also used interviewswith parents and high school transcripts Wefocused on a subsample of the oldest NLSY97participantsmdashthose born in 1980mdashand useddata from Rounds 1ndash3 extending to Round 5to establish the date of a high school diplomaor receipt of a general equivalency diploma(GED) Round 1 interviews were conductedfrom February 1997 to May 1998 when thesample members were aged 16 years 2months to 18 years 3 months The Round 1questions asked of the youths and their par-ents covered past schooling events and sta-tuses as well as current conditions The Round3 interviews occurred when the sample mem-bers were aged 18 years 10 months to 20years 3 months

02 Plank 10808 551 PM Page 350


Mon 17 Nov 2008 190028


We measured whether an individual expe-rienced a dropout event between his or herinitial entry into the ninth grade and the timeof the Round 3 interview Following the 1980cohort members until they were 1883 yearsto 2025 years allowed for a comprehensiveanalysis of their high school careers droppingout and completion of degrees withoutsevere problems of right-censored dropout orgraduation times (events that occurred afterthe observation ended)

The initial descriptive analyses were for1628 cases representing all NLSY97 samplemembers born in 1980 except for 63 whowere removed from the analyses for validreasons (eg had never enrolled in the ninthgrade or a higher grade by the time of theRound 1 interview) Sampling weights wereused in generating descriptive statisticsbased on the 1628 cases to generalize to anational population For the hazards modelswe were constrained to 846 individuals forwhom transcript data were available (seeAppendix B for the availability of data fromtranscripts and our data-preparation steps)For the estimation of hazards models sam-pling weights were not used as is discussednext

Modeling Strategy

This article presents estimates from Coxregression models the most commonly usedvariant of hazards models (Allison 1995)Hazards models are useful for describing thetiming of life-course events and for buildingstatistical models of the risk of an eventrsquosoccurrence over time (Willett and Singer1991) The influence of early experiences onlater decisions or behaviors the time-varyingnature of key predictors the nonconstantlevel of risk that individuals may experienceover time and various lengths of exposure torisk among the participants in a study areaddressed by hazards models in ways that astandard logit model with one record (orobservation) per participant cannot addressWhen time is treated as continuous (as it is inour analyses) what is modeled is the instan-taneous risk that an event will happen at Timet given that it has not occurred before Timet An intuitive notion of risk will be useful to

keep in mind as we assess whether changes incovariates are associated with increases ordecreases in a studentrsquos risk of dropping outof high school

The final models presented here are non-proportional hazards with time-varyingcovariates4 The general model can be writtenas follows

hi(t) = λ0(t)expββXi + ββXi(t) (1)

Equation 1 indicates that the hazard forindividual i dropping out of school at Timet given that he or she has not dropped outbefore t is the product of two factors (1) abaseline hazard function λ0(t) that is leftunspecified except that it cannot be nega-tive and (2) a linear function of a set offixed and time-varying covariates which isthen exponentiated For our analyses theobserved risk period began with the firstmonth in which an individual attended theninth grade and ended when an individualexperienced a dropout event for the firsttime The period could also end with thereceipt of a high school diploma or right-censoring via the occurrence of the Round 3interview (or the Round 1 or Round 2 inter-view if later interviews were not complet-ed) Our unit of analysis was the person-month with varying start months length-of-risk period and methods of leaving therisk set

Sampling weights were not used in theestimation of the hazards models with oursubsample of 846 cases We followed the rec-ommendations of NLSY97 documentationand technical staff at the Center for HumanResource Research at Ohio State University(2006 J Zagorsky personal communicationJune 26 2007) Given our subsample gener-ated by dropping a large number of observa-tions with unavailable transcript data and ourapplication of regression-based techniquesunweighted analyses are recommended Weview the unweighted subsample analyzed viahazards models as a diverse group comprisedof individuals from across the United Statesnot as a sample that directly reflects thedemographic or social composition of USyouths born in 19805

02 Plank 10808 551 PM Page 351


Mon 17 Nov 2008 190028


Variables

Dependent Variable The dependent vari-able in the hazards models indicates theevent of being away from school for 30 daysor more for a reason other than vacation atsome time after initial enrollment in the ninthgrade Information about a spell away fromschoolmdashand the reason for such a spellmdashcame from the youthsrsquo and parentsrsquo reports6The dropout event was linked to the monthof the high school career in which it occurred

Ascriptive Traits Prior studies led us toexpect that females would be less likely todrop out but that this effect would attenuateas achievement and other school experienceswere introduced into the models (Gold-schmidt and Wang 1999 Rumberger 1987)Therefore we included in our models an indi-cator for female with male as the excludedreference category Indicators for blackHispanic Asian and ldquoother raceethnicityrdquowere used with white as the excluded refer-ence category Past research has shown thatdropout rates in the United States are higherfor black and Hispanic youths than for whiteand Asian youths However some studieshave shown that once family and socioeco-nomic background are controlled minoritydisadvantage in persistence in school is fullyaccounted for or even reversed (HauserSimmons and Pager 2004) When presentingsuch a finding Hauser et al suggested thatother things being equal minorities wouldstay in school longer than whites becausethey lack attractive opportunities outsideschool

Family and Residential Context To reflectcultural and financial resources in the homewe included the highest grade completed bya residential parent and the natural log ofhousehold income Household structure wasrepresented by indicators for living with thebiological mother only living with the biolog-ical father only living with one biological par-ent and one stepparent and ldquootherrdquo house-hold arrangement with living with both bio-logical parents the excluded reference cate-gory Prior research led us to expect that chil-dren who live with single parents (Krein and

Beller 1988 McLanahan 1985) or stepparents(Astone and McLanahan 1991) will drop outat higher rates than will children who livewith both biological parents (called ldquointactrdquofamilies in much of the literature) Our mod-els were structured to examine whetheraspects of academic achievement and coursetaking account for differences in dropoutrates that may be observed between intactand nonintact families An indicator of livingin an urban area with nonurban areas as theexcluded reference category was included asan important control variable becausedropout rates have consistently been shownto be the highest in urban locations (Balfanzand Legters 2004 Hauser et al 2004)

Proxies for Academic Preparation and PriorAchievement As a proxy for academicpreparation and performance we used themathematics knowledge subscore from thecomputer-adaptive form of the ArmedServices Vocational Aptitude Battery (ASVAB)This test was administered between summer1997 and spring 1998 and thus is not an aca-demic performance that was measuredbefore our sample members started highschool Nonetheless we interpret it as anindicator of academic preparation heavilydetermined by pre-high school academicdevelopment Early academic performancewas also captured via a self-reported eighth-grade GPA (which was used as the initial valuefor our time-varying GPA covariate and wasreplaced by another value only when the firsthigh school term was completed)

We also included age at ninth-grade enroll-ment measured in months For some aspectsof our modeling this variable was treated ascontinuous For other purposes we split it atage 15 (180 months) to distinguish those whowere old for grade on entering the ninthgrade from those who were not In mostcases of course those who were old for gradeexperienced retention at some point beforehigh school We used age at enrollmentrather than an indicator of retention becauseprior research has shown that being old forgrade explains virtually all the associationbetween retention in grade and the risk ofdropping out that is not already explained byearly academic performance (Roderick 1994)

02 Plank 10808 551 PM Page 352


Mon 17 Nov 2008 190028


Time-varying Covariates Characterizingthe High School Experience Observedvalues for some independent variables couldchange for each month an individual was inthe risk set for dropping out7 Therefore ourmodels included the following time-varyingcovariates GPA achieved during the mostrecently completed term the ratio of CTEcredits to academic credits earned during themost recently completed term and thesquare of the CTE-to-academic ratio for themost recently completed term Both a first-order and a squared term were included toallow us to detect a curvilinear associationbetween the course-taking ratio and the riskof dropping out

Appendix Table A1 presents unweighteddescriptive statistics for the dependent andindependent variables for the analytic sampleof 1628 individuals and the transcript sub-sample of 846 The transcript subsample hada lower proportion of dropouts than did thefull sample largely because dropouts haveinterrupted or irregular high school careersand thus it is sometimes difficult to obtaincomplete transcripts for the terms they werein school Appendix Table A1 also shows thatthe transcript subsample as compared withthe full sample had somewhat higher pro-portions of white students females andhouseholds with two biological parents andslightly higher family incomes levels ofparental education and ASVAB scores

RESULTS

A First Look at Dropping Out

Table 1 shows that by the Round 3 interview208 percent of the individuals in our full ana-lytic sample had at least one dropout spelldefined as 30 days away from high school fora reason other than vacation Table 1 showsthat 804 percent of the sample members hadattained a traditional diploma by the Round 5interview period an estimate that includesrecipients of diplomas who had experiencedone or more spells of dropout before theygraduated Clearly though even one spellaway from school greatly reduces the odds ofeventually receiving a high school diploma or

another credential Additional analyses (notshown) revealed that those with a dropoutevent by Round 3 had only a 278 percentchance of earning a traditional diploma byRound 5 and a 459 percent chance of attain-ing any high school credential In contrastthe figures for those without a dropout eventby Round 3 were 942 percent and 944 per-cent respectively

Table 1 also shows attainment outcomesby race and ethnicity gender eighth-gradeGPA residential parentsrsquo highest grade com-pleted and age at entry into the ninth gradeBlack and Hispanic students were significant-ly more likely than white students to drop outby Round 3 and significantly less likely toreceive a diploma or any high school creden-tial by Round 5 Males were somewhat morelikely than females to drop out and less likelyto receive diplomas or GEDs Both eighth-grade grades and parental education wererelated to dropping out and the completionof high school in the expected directionswith striking differences across levels Finallythose who were aged 15 or older when theyentered the ninth grade dropped out at amuch higher rate than did those who wereyounger than 15 when they entered the ninthgrade

Figure 1 shows the distribution of age atthe time of the first dropout spell for the 379individuals who had such a spell The valuesrange from 13 years 8 months to 19 years 9months and the median is 17 years 2months There is an increase in the incidenceof dropping out as the 16th birthdayapproaches and again as the 17th birthdayoccurs The sparser concentration of dropoutevents after approximately age 18 years 6months is driven by at least three factors (1)Many of those who were at a relatively highrisk of dropping out had already experiencedthe event by that age and thus exited the riskset for first dropout occurrence (2) manypeople had graduated from high school bythat age and thus exited the risk set and (c)some people had their Round 3 interviewssoon after that time and thus had their obser-vations censored

Figure 2 translates age at the dropout eventinto months after initial entry into the ninthgrade and shows that first dropout spells

02 Plank 10808 551 PM Page 353


Mon 17 Nov 2008 190028


occurred with considerable frequency bothearly and late in high school careers The diffi-culties of the ninth-grade transition experi-enced by many high school students are repre-sented among the individuals who droppedout during the first 9 to 12 months after theirinitial entry into the ninth grade For othersdropout spells occurred after 30 40 or even 48months of high school revealing that the prob-lem of dropping out is not concentrated exclu-sively in the early years of high school8

The wide distribution of the timing ofdropout prompts the question of whether dif-ferent factors are more influential or less influ-

ential at different stages of the high schoolcareer The hazards modeling addressed thisquestion

Results of Hazards Models

Table 2 displays exponentiated coefficientsmdashthat is multiplicative effects on the hazardratiomdashfrom a series of four estimated modelsExponentiated coefficients significantlygreater than 1 indicate that a covariate isassociated with an increased risk of droppingout while those significantly less than 1 indi-cate a reduced risk of dropping out

Table 1 Frequency of High School Dropout High School Diploma or Any High SchoolCredential for Members of the NLSY97 Sample Who Were Born in 1980 (weighted N =1628)

Dropout High School Any High SchoolEvent by Diploma by Credential byRound 3 Round 5 Round 5

Variable Interview Interview Interview

Total Sample 208 804 844

RaceEthnicityBlack 271 702 759Hispanic 288 721 758White 181 838 875

GenderMale 227 760 809Female 187 852 880

Eighth-Grade Grade Point Average0 to 10 597 343 467gt 10 to 20 373 693 769gt 20 to 30 228 772 812gt 30 to 40 74 932 944

Parentsrsquo Highest Grade Completed

le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949

Age at Ninth-Grade Enrollmentlt 15 127 888 909ge 15 399 606 689

02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028


Model A includes sex and raceethnicityOnly the coefficient for Hispanic students isstatistically significant showing that aHispanic student has a dropout risk that is 76percent higher than that of a non-Hispanicwhite student (controlling for gender)Coefficients for females and black studentsare not statistically significant although theyare in the directions that previous researchwould predict

Model B adds household characteristicsand urban location As expected higherparental education is associated with areduced risk of dropping out Living withonersquos biological mother only with onersquos bio-logical father only or with a biological parentand a stepparent are all associated with agreater risk of dropping out relative to livingwith both biological parents While these esti-mates for household composition are signifi-cant in this preliminary model we note thatthey lose their significance as various inter-vening variables are introduced in Models Cand D It appears that the lower academicperformance and delayed entry into highschool that are sometimes associated withnonintact family structures account for much

of the elevation of dropout rates that wereinitially estimated for the children living inthese families Living in an urban area is asso-ciated with a 78 percent greater risk of drop-ping out than what is estimated for living in anonurban area

In Model B Hispanic youths no longershow an elevated risk of dropping out afterparental education and other factors havebeen controlled This finding is consistentwith previous findings that much of theobserved difference in dropout rates amongracialethnic groups can be attributed to dif-ferences in family and community character-istics (Fernandez Paulsen and Hirano-Nakanishi 1989 Hauser et al 2004) Also thelack of significance for household income isnoteworthy especially when coupled withthe significance of parental education Thisfinding suggests that human capital and par-entsrsquo familiarity and comfort with navigatingthe educational system are more salient inunderstanding high school persistence than isfinancial capital

Model C adds the ASVAB score and age atentry into the ninth grade Higher ASVABscores are associated with a significantly

Figure 1 Distribution of Ages at the Time of the Dropout Event (first event after entryinto the ninth grade) (n = 379)

02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028


lower risk of dropping out Age at entry intothe ninth grade significantly predictsdropout with the interpretation that startinghigh school a year later increases the risk ofdropping out by a factor of more than 39 Inthis model black students have a lower risk ofdropping out than do white studentsmdasha find-ing that emerges once both socioeconomicstatus and performance on standardized testsare controlled

Model D adds time-varying covariates forGPA and the CTE-to-academic course-takingratio Higher ASVAB scores and a higher GPAin the most recently completed term are asso-ciated with a reduced risk of dropping outUrban residence and each additional monthof age at entry into the ninth grade are asso-ciated with an elevated risk of dropping out

The estimates for the effects of the CTE-to-academic ratio suggest a U-shaped functionas evidenced by the facts that the first-ordertermrsquos estimated effect on the hazard ratio isless than 10 while the squared termrsquos esti-mated effect is greater than 1010 These twoterms generate a curve with a point of inflec-tion where the course-taking ratio equals046 or where roughly one CTE credit isearned for every two core academic credits

However in contrast to previous research thefirst-order term is not significant and thesquared term is only marginally significant (plt 10) An exploration of the data makes clearthat the differences in the estimates for thecourse-taking ratio between previous researchand the present results are largely driven bythe exclusion or inclusion of age at entry intothe ninth grade as we detail later11 While thefirst-order and squared terms for the course-taking ratio are not significant individuallytheir inclusion as a pair significantly improvesthe fit of the model It is for this reason (hav-ing conducted omnibus tests of significance)that we cite a slight U-shaped pattern for theeffect of the CTE-to-academic ratio on the riskof dropping out12

High School Entry Age as aContingency

Prior studies and our empirical findings sug-gest the need to explore further the contin-gencies that are associated with age at entryinto high school As reflected in Tables 1 and2 students who were old for grade on entryinto high school dropped out at higher ratesOlder students are also more heavily concen-

Figure 2 Distribution of the Month in Which the Dropout Occurred (relative to the ini-tial entry into the ninth grade) (n = 379)

02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028


trated than are younger students at the medi-um and high levels of the CTE-to-academicratio at any given point in the high schoolcareer (see Appendix Table A2)

Since age at entry into high school partlydetermines high school course taking and isassociated with the risk of dropping out wetook the initial step of controlling for it beforewe interpreted the effects of course taking(see Table 2 Model D) However a furtherquestion is whether the association betweenthe course-taking ratio and the likelihood ofdropout is fundamentally different for stu-dents who enter high school at different ages

To address this question we reestimatedModel D separately for two subgroupsmdashthose who were younger than 15 years old atentry into high school (see Table 3 Column1) and those who were age 15 or older atentry (see Table 3 Column 2)13

The estimated models for the two sub-groups are different For the younger sub-group age at entry into high school is not asignificant predictor of dropout While therewas variation in the age at entry into highschool for this subgroup this variation wasnot systematically associated with the likeli-hood of dropping out independent of other

Table 2 Effects on Hazard Ratio (based on the Cox nonproportional hazards model ofdropping out) for Members of the NLSY97 Sample Who Were Born in 1980 with AvailableTranscript Data (n = 846)

Model BWith Model C

Household WithModel A Characteristics ASVAB andGender and Age at

and Urban Ninth-grade Model DRace Location Entry Final

Variable Only Added Added Model

Female 078 073+ 072+ 085Black 145 095 059 053Hispanic 176 086 088 088Asian 046 040 052 066Other raceethnicity (nonwhite) 074 064 075 077Parentsrsquo highest grade completed 083 089 089ln(household income) 095 098 100Biological mother only 167 155+ 139Biological father only 257 138 127One biological parentndashone stepparent 219 161+ 153Other household arrangement 166 108 112Urban residence 178 197 193ASVAB Math knowledge 054 065Grade point average (most recent term) 056CTE-to-academic ratio (most recent term) 027CTE-to-academic ratio2 (most recent term) 413+Age (months) at ninth-grade entry 111 111

Reduction in -2 log likelihood from model without covariates 1011 6042 16812 19334

df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028


variables in the model The effect of ASVABscores is significant with higher scores associ-ated with a lower risk of dropping out Allthree time-varying predictorsmdashGPA the first-order term for course-taking ratio and thesquared term for course-taking ratiomdashare sig-nificant A higher GPA is associated with alower risk of dropping out The associationbetween the course-taking ratio and the riskof dropping out is characterized by a pro-nounced U-shaped function (with a point ofinflection at 054) The lower curve of Figure3 illustrates how the coefficients for the CTE-to-academic course-taking ratio translate intochanges in the log hazard of dropping out forthe younger subgroup

To aid in interpreting this U-shaped pat-

tern for the lower curve consider two hypo-thetical students who are alike on all inde-pendent variables except the course-takingratio Hypothetical Student A took one CTEcourse for every two academic courses in themost recently completed term while StudentB took no CTE courses These students withtheir CTE-to-academic ratios of 05 and 00respectively differ in their predicted log haz-ards by a quantity of 0903 (see Figure 3) Byexponentiating the additive effects from themodel of log hazards we can restate this find-ing Hypothetical Student A who struck themiddle-range mix of CTE and academiccourses has a risk of dropping out that is only405 percent as great as Student Brsquos

For the older subgroup (those who were

Table 3 Effects on Hazard Ratio (based on the Cox nonproportional hazards model ofdropping out) for Two Subsamples Defined by Age at Initial Entry to the Ninth Grade (n= 846)

Younger than At leastAge 15 Age 15

at Initial at InitialEntry Entry to the to the Ninth NinthGrade Grade

Variable (n = 653) (n = 193)

Female 089 080Black 077 036Hispanic 069 108Parentsrsquo highest grade completed 088 093ln(household income) 095 103Biological mother only 133 159Biological father only 190 123One biological parentndashone stepparent 170 181Other household arrangement 104 180Urban residence 185 207ASVAB Math knowledge 049 080Grade point average (most recent term) 057 053CTE-to-academic ratio (most recent term) 004 092CTE-to-academic ratio2 (most recent term) 2142 169Age (in months) at ninth-grade entry 101 117

Reduction in -2 log likelihood from the model without covariates 7414 6523df 15 15p lt 0001 lt 0001

p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028


15 or older at entry into the ninth grade) theASVAB score is not a significant predictor ofdropping out nor are the course-taking coef-ficients The effect of the course-taking ratioon the log hazard of dropping out for theolder subgroup is depicted in the upper curveof Figure 3 The plotted curve sweeps upwardas the course-taking ratio increases but this isnot significant and is therefore statisticallyindistinguishable from a flat line For the olderstudents GPA in the most recently completedterm is significant Finally age at entry intothe ninth grade is significant (with each addi-tional month of age bringing an increased riskof dropping out)

It appears that for those who are old forgrade on entry into high school age is such astrong correlate of dropping out that itswamps most other effects It is not just thatthe older students as a group drop out at rel-atively high rates but that each additionalmonth that a student is older at entry intohigh school brings an increased risk of drop-ping out The findings that grades but notASVAB scores are a significant predictor ofdropping out for the older subsample sug-gests that weaknesses in academic perfor-mance may not be the only factor that leads

older students to disengage from schoolother likely causes include various aspects ofthe social realm and social stigma

Time as a Contingency

In addition to examining variation by age atentry into the ninth grade we also exploredwhether the effects of the independent vari-ables were constant or nonconstant overtime We found only a single significant inter-action effect a negative coefficient for theinteraction of age at entry into the ninthgrade with time for the subsample of 846cases Earlier in this article we offered a sce-nario under which increasing effects of age atentry would be expected as the high schoolcareer progressed Instead the estimatedmodel showed decreasing effects of age atentry as the high school career progressedThe most straightforward explanation forsuch a finding has to do with the distributionof age at a particular point in time and statelaws about the age at which dropping out islegal (generally 16 or 17) Early in highschool younger students are constrained bysuch laws while students who are the oldestfor their grade often are not therefore age

Figure 3 Estimated Effects of the Course-Taking Ratio on the Log Hazard of DroppingOut for Two Subsamples Defined by Age at the Initial Entry into the Ninth Grade

02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028


makes a difference during this stage ofschooling in determining who will decide toleave school Later in the high school careeras almost all students reach the legal age forleaving school age plays less of a role indetermining which students are most likely todrop out

DISCUSSION

Our results suggest that for students who areof the modal age or young for grade at thetime of high school entry some CTE com-bined with core academic course taking maydecrease the risk of dropoutmdashbut only up toa point A combination of approximately oneCTE course for every two core academiccourses is associated with the lowest risk ofdropping out after other variables in ourmodels are controlled Being below this point(taking few or no CTE courses) or beingabove this point (concentrating on CTE to theexclusion of adequate academic preparation)implies an increased risk of dropping out Ourmodels also revealed significant negativeeffects of scores on standardized tests andGPA in the most recently completed term onthe likelihood of dropping out The results forcourse taking test scores and grades onreport cards jointly support Finnrsquos (1989)model of the frustration that can come withpoor performance and the sense of belongingand identification that can come with positivecurricular (and extracurricular) experiences

Although our analyses do not providedirect evidence for these processes we spec-ulate that some combination of behavioralemotional and cognitive engagement isoften maximized for students when CTE andcore academics are experienced together Tounderstand why more engagement may beexpected for students who combine CTE andacademic courses it is necessary to under-stand the positive and negative consequencesof CTE On the one hand CTE may help stu-dents more readily see the value of school inpreparing them for careers of interest and canencourage students to define their careergoals By connecting school with the transi-tion to adulthood and a career CTE can clar-ify the application and value of academic sub-

jects as they pertain to jobs or perhaps thepostsecondary education that is needed for acareer of interest thereby keeping youthsengaged with school However if a student ischanneled toward a segregated vocationalpart of the curriculum where there are fewconnections to academically focused teachersand educationally engaged peers it is easy tosee why persistence in school is tenuous

In contrast to the results for students whoentered high school at normal ages oryounger we found that different concentra-tions of CTE and academic course taking donot have detectable effects on the likelihoodthat students who are old for grade will dropout It seems that the constellation of risk fac-tors or challenges that these older studentsencounter such as the strains of incompatibleroles diminishes the extent to which thesestudents are receptive to any potential effectsof course taking Quite likely to be older thanis normal as one proceeds through highschool is to be attracted to certain behaviorsthat are at odds with the life of a typical highschool student Since an older studentrsquosscores on standardized tests and grades donot completely account for the effect of ageon dropping out we suspect that social com-patibility is also relevant for these adolescentsa possibility that was supported by recentresearch (cf Stearns et al 2007)

Roderick (1994) and Entwisle et al (2004)reached similar conclusions In their study ofhigh school dropouts in Baltimore Entwisleand her colleagues found that the lower aca-demic performance of retained students didnot account for the studentsrsquo elevated risk ofdropping out Therefore they suggested thatthe effects of retention on dropping outstemmed from social sources as well becauseof the pressures that students experiencewhen they are off-time in the age-gradedorganization of the secondary school Thisorganization labels students as ldquobehindrdquo orldquofailuresrdquo and adds to the time they mustspend in school before graduation By drop-ping out old-for-grade students can shed apunishing role and instead seek their desiredstatus through paid work or other means

02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028


LIMITATIONS FUTURERESEARCH AND IMPLICATIONS

Although we believe that the study con-tributes to an understanding of the highschool dropout process there are several lim-itations that warrant future research Firstwhile we strongly suspect that our curvilinearfinding is partly due to genuine effects of thebalance that a student strikes between careerpreparation and core academics it is also like-ly that the U-shaped pattern is partially drivenby unmeasured aspects of what is occurringat a deeper pedagogical level in schoolswhere a middle-range mix is available to stu-dents For example it could be that where wesee students pursuing a middle-range combi-nation of academic and CTE courses theyattend schools that are successfully offeringintegrated programs and broad occupationalawareness and where students have consider-able autonomy to design their learning expe-riences In contrast it could be that where wesee students taking higher concentrations ofCTE these students are in schools that offerprograms that still look a lot like the ldquooldrdquovocational education and all that it implies forlow expectations and disengagementAdditional research is needed to understandwhere and how the principles of ldquonewrdquo voca-tional education have genuinely been incor-porated into US high schools

Second while our use of time-varyingcovariates for GPA and course taking bolsterscausal inference we have not definitivelyestablished a causal relationship betweencourse taking and dropping out It is likelythat there are unobserved characteristics ofstudents that both lead them to select cours-es and determine whether they completehigh school Certainly the members of oursample were not randomly assigned to theircourses but rather guided by their or theirfamiliesrsquo preferences that we have not beenable to model Matches are also nonrandombecause schools make organizational deci-sions about what courses to offer and towhom thus constraining studentsrsquo curricularchoices To understand this process of selec-tion and assignment research is needed todescribe the rigor and breadth of schoolsrsquocurricular offerings in both the core academ-

ic areas and CTE Such research should firstmodel a studentrsquos access to courses across awide range of curricular areas includingadvanced mathematics Advanced Placementcourses CTE generally and CTE organizedinto coherent sequences that build in theirdepth and specialization The effects of takingcertain combinations of courses conditionalon opportunity should then be assessed Suchan endeavor requires high-quality data aboutschoolsrsquo curricular offerings individual stu-dentsrsquo course taking and studentsrsquo enroll-ment or dropout status measured with preci-sion

In terms of practical significance our studysignals the importance of two processes Firstthe types of courses that students take maybe related to their choices to stay in schoolOur findings suggest that some students canbe retained in school if they take courses thatnot only have a traditional academic focusbut that give schooling a connection tocareers and technical applications of knowl-edge Second as is consistent with a growingbody of research attachment to school variesby studentsrsquo retention histories and age onentry into high school This insight suggeststhat educators should try to help studentswho are old for grade reclaim or create ahealthy attachment to school and the goal ofhigh school graduation This goal will not beaccomplished simply by directing such stu-dents to a particular set or combination ofcourses Most likely additional efforts will benecessary to help these students to feel inte-grated into the life and activities of the schooland to believe that completion of a degree isvaluable and possible

NOTES

1 This figure (approximately 5 percent)follows from the event dropout-rate calcula-tion used by the NCES (see Kaufman Alt andChapman 2004) The rate for a given year isthe proportion of youths aged 15ndash24 whowere enrolled in high school in one Octoberbut who had left high school without suc-cessfully completing a high school programmdashwhether a traditional diploma a GED oranother credentialmdashby the next October For

02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028


discussions of alternative ways to calculaterates of dropout or high school completionsee Balfanz and Legters (2004) Heckman andLaFontaine (2007) Miao and Haney (2004)and Swanson and Chaplin (2003)

2 We refer to the Carl D PerkinsVocational and Applied TechnologyEducation Amendments of 1990 and 1998(PL 101-392 and PL 105-332 also known asPerkins II and III respectively) and the School-to-Work Opportunities Act of 1994 (PL 103-239)

3 CTE and core academic courses aredefined in the calculation of the curricularratio as follows CTE includes all courses infamily and consumer sciences general labormarket preparation and the SLMP areas coreacademics include all courses in mathematicsscience English or language arts (althoughnot foreign languages) and social studies(including history)

4 For the final models in our tables it isthe presence of time-dependent covariatesthat makes them nonproportional hazardsmodels Proportional hazards models arecharacterized by the ratio of any two individ-ualsrsquo hazards being constant over time Thusif one would graph the log hazards for anytwo individuals against time the hazard func-tions would be parallel In contrast in non-proportional models the time-dependentcovariates change at different rates for differ-ent individuals so the ratios of their hazardscannot remain constant We also exploredinteractions of nonvarying covariates withtime which is another situation in which non-proportional models are estimated All mod-els were estimated using SAS PROC PHREGspecifying the ldquoexactrdquo method for dealingwith tied event timesmdashie instances in whichtwo or more individuals dropped out of highschool after the same number of months hadelapsed after their initial entry into the ninthgrade

5 For the hazards modeling missing datawere treated via multiple imputation usingSAS procedures MI and MIANALYZE (Schafer1997) The Markov Chain Monte Carlomethod was used and 20 imputations weregenerated Inspection of diagnostic statisticssuggested that the relative efficiency of esti-mates was markedly better with 20 imputa-

tions than with 5 or 10 but that any addi-tional improvements were minimal if morethan 20 imputations were used

6 Details of programming and decisionrules used in the construction of variables areavailable on request from the authors For themajority of identified dropout events infor-mation provided by a youth could be usedand a precise month of the event was knownWhen information provided by a parent hadto be used (which was true for some of theevents identified as occurring before theRound 1 interview) only the grade withinwhich the event occurred was known Toallow for analyses that maintained the per-son-month as the unit of analysis we con-ducted a random draw to impute in whichmonth within the school year indicated bythe parent the event occurred

7 Operationally an individualrsquos values onthese covariates changed each time an acad-emic term was completed as indicated bythe school transcript Depending on onersquosschool these terms might have been quarters(approximately nine weeks in duration)trimesters semesters ormdashrarelymdashentire aca-demic years

8 It is also worth noting in Figure 2 thatsome dropout events occur during what aregenerally the summer months For exampleMonths 11 and 23 would represent July formany of the sample members (specifically forthose who began the ninth grade inSeptember) We allowed ldquosummer monthrdquodropout times for conceptual reasons A stu-dent might have attended school in June1997 but then not returned to school untilJanuary 1998 We did not declare July 1997 tobe the studentrsquos dropout month until we con-firmed that the student was not enrolled ineither September or October of 1997 but ifindeed the student had not enrolled in eitherSeptember or October then we thought itwas appropriate to declare July the month inwhich the student stopped attending school(and perhaps the month in which he or shearticulated this decision to himself or herself)

9 The coefficient in the linear model of loghazards is 00999 Multiplying by 12 (months)and exponentiating to calculate the effect onthe hazard ratio yields exp(1200999) = 332

10 The U-shaped function is perhaps eas-

02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028


ier to visualize if we translate Table 2rsquos effectson hazard ratios back into effects on log-haz-ards to get the following The first-ordercourse-taking ratio term has an estimatedcoefficient of -131 (natural log of 027) Thesquared term has an estimated coefficient of142 (natural log of 413)

11 Including age at entry into the ninthgrade is clearly important for a properly speci-fied model We also pursued further analysis tounderstand differences among a hazards modelusing NLSY97 data a logistic regression usingNLSY97 data and a logistic regression usingearlier NELS88 data (cf Plank 2001) We con-cluded that differences among models weredriven by changing data sets much more thanby changing statistical methods However it isunclear whether this finding reveals historicalchanges and altered educational processes thatoccurred across five or six years or differences inthe methods that were used to measure vari-ables in the two studies Another relatively smallpart of the difference between findings fromNELS88 and an NLSY97 hazards model has todo with changing methods (to the more defen-sible hazards model) The change in methodsseems to imply small changes in the strength ofestimated effects thus giving some credence toconcerns about reverse causality in the logisticregression results but overall this does notseem to be a large source of different results

12 The omnibus tests were used to adju-

dicate among our four competing hypothesesabout the relationship between the CTE-to-academic ratio and the likelihood of droppingout First Model D was compared with amodel that included neither a first-order termnor a squared term a specification that isconsistent with the hypothesis of ldquono effectrdquoof course taking This test yielded a differencein -2LL of 586 with 2 degrees of freedom p= 0586 Second Model D was comparedwith a model with a first-order term but nosquared term a specification that is consis-tent with the hypotheses of linear effects(whether positive or negative) of course tak-ing This test yielded a difference in -2LL of381 with 1 degree of freedom p = 0509Each test suggested that the curvilinear spec-ification is superior to the alternative (thoughonly marginally so)

13 The model specification for each of thetwo subgroups and summarized in Table 3 isidentical to the specification for Model D inTable 2 except that dummy variables forAsian and other raceethnicity are not includ-ed Once the sample was split according toage at entry into high school there were toofew students in these categories to estimatemeaningful coefficients Thus the excludedreference category for raceethnicity for Table3 is all adolescents who were neither blacknor Hispanic

02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028


Table A1 Unweighted Descriptive Statistics for Members of the NLSY97 Sample WhoWere Born in 1980 and for the Transcript Subsample

1980 Cohort ndash 1980 Cohort Transcript

Variable (N = 1628) Subsample (n = 846)

Ever dropped out (by Round 3 interview) 233 139

Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552

Other raceethnicity 12 13

ln(Household income) Mean = 101 SD = 20 Mean = 104 SD = 17

Parentsrsquo highest grade completed Mean = 132 SD = 30 Mean = 136 SD = 29

Living with biological mother only 257 215

Living with biological father only 31 28

Living with one biological parent andone stepparent 119 151

Living with two biological parents 488 535

Other household arrangement 105 71

Urban location 742 723

Age (in months) at enrollment inthe ninth grade Mean = 1769 SD = 79 Mean = 1754 SD = 65

ASVAB Math knowledge (divided by 1000) Mean = 024 SD = 101 Mean = 048 SD = 085

APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028


Table A2 Descriptive Statistics for CTE-to-Academic Ratio and Grade Point AverageDuring Months 7 19 31 and 43 of the High School Career for Two Subsamples Definedby Age at Initial Entry into the Ninth Grade (n = 846)

Variable Month 7 Month 19 Month 31 Month 43

Subsample Younger than Age 15 (n = 653)Number still in risk set as of month 650 644 635 589

CTE-to-Academic Ratio (most recently completed term)

25th percentile 000 000 000 000Median 010 005 017 02575th percentile 025 025 033 067

Grade Point Average (most recently completed term)


Subsample Aged 15 or Older on Entry (n = 193)

Number still in risk set as of month 187 171 149 111





APPENDIX B

Use and Preparation of NLSY97 High School Transcript Data

On the basis of communication with staff at the Center for Human Resource Research atOhio State University NORC and the Bureau of Labor Statistics we understand that the tran-script data are most representative of the 1980 cohort All the data presented in this articledescribe the 1980 birth cohort of the NLSY97 sample While the total sample of youths whowere born in 1980 was 1691 transcript data are available for 873 individuals (or 52)

Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028


APPENDIX B (CONTINUED)

Timing of Course Taking

The transcript data are intended to provide information on every term for each high schoolattended by a student In addition to course titles and grades received for each course weknow the start and stop dates for academic terms For some transcripts dates were incom-plete typically a start or stop month or year or some combination of months and years wasincomplete For a student with missing start or stop dates we used available information forthe academic schedule of that same school information about the type of term (fall or spring)and information about a studentrsquos high school completion date to complete the studentrsquos tra-jectory In some cases we followed decision rules based on knowledge of the start and stopmonths of a typical school system (eg the school year generally begins in September)

Credits Earned

For most courses it is clear how many credits were earned In a few cases we corrected obvi-ous data-entry errors (eg one course earned 50 credits while every other course earned 05or 10 credit) In other cases a course earned a passing grade although no credits were record-ed We identified these cases and assigned what appeared to be the appropriate number ofcredits on the basis of other information on the same individualrsquos transcript

Generalizability

Overall the transcript subsample matched the 1980 cohort closely with a few exceptions (seeTable A1) The transcript subsample was slightly more advantaged socioeconomically and aca-demically had more educated parents a higher family income and more students living withboth parents In terms of academic achievement the transcript subsample was more likely tohave mostly As and less likely to have mostly Ds than the fuller 1980 sample (on the basis ofstudentsrsquo self-reported grades) They scored higher on the ASVAB arithmetic-reasoning testand were less likely to have been suspended However the two samples were similar withrespect to gender composition age household size region of residence school type andabsences They were also comparable in terms of participation in various CTE programs (egjob shadowing or apprenticeships) We also compared our results on course-taking patterns tothose found in other nationally representative data sets such as NELS88 and the High SchoolTranscript Study The results of these comparisons are available on request and confirm thatour NLSY97 1980 cohort was similar in its course-taking patterns to other national samplesduring the mid- and late-1990s

Limitations

As we mentioned earlier the transcript subsample was somewhat limited relative to the entireNLSY97 data set First we are missing transcripts for about half the 1980 birth cohort mak-ing a full-scale analysis of the experiences of that cohort problematic Second despite carefulanalysis it is possible that one could misinterpret a schoolrsquos credit or grading system Thirdthere are some incomplete transcripts which is to be expected given studentsrsquo dropout ratesand transfers but it still poses some problems for validity Nonetheless given the comparisonswe have made with the full 1980 cohort and other national data sets we feel confident thatthese data begin to give us an understanding of current curricular patterns and their associa-tion with dropping out

02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES

Advisory Committee for the National Assessmentof Vocational Education 2003 Report of theAdvisory Committee for the National Assessmentof Vocational Education Washington DC USDepartment of Education Office of the UnderSecretary

Agodini Roberto and John Deke 2004 TheRelationship Between High School VocationalEducation and Dropping Out (report submittedto the US Department of EducationPlanning and Evaluation Service) PrincetonNJ Mathematica Policy Research

Alexander Karl L Doris R Entwisle and Nader SKabbani 2001 ldquoThe Dropout Process in LifeCourse Perspective Early Risk Factors at Homeand Schoolrdquo Teachers College Record103760ndash822

Allensworth Elaine 2004 Ending SocialPromotion Dropout Rates in Chicago afterImplementation of the Eighth-GradePromotion Gate Chicago Consortium onChicago School Research

Allison Paul D 1995 Survival Analysis Using theSAS System A Practical Guide Cary NC SASInstitute

Arum Richard 1998 ldquoInvested Dollars or DivertedDreams The Effect of Resources on VocationalStudentsrsquo Educational Outcomesrdquo Sociology ofEducation 71130ndash51

Astone Nan M and Sara S McLanahan 1991ldquoFamily Structure Parental Practices and HighSchool Completion American SociologicalReview 56309ndash20

Balfanz Robert and Nettie Legters 2004 Locatingthe Dropout Crisis Where Are They LocatedWho Attends Them Baltimore MD Center forSocial Organization of Schools Johns HopkinsUniversity

Baron Reuben M and David A Kenny 1986ldquoThe Moderator-Mediator Variable Distinctionin Social Psychological Research ConceptualStrategic and Statistical ConsiderationsrdquoJournal of Personality and Social Psychology511173ndash82

Bishop John 1988 ldquoVocational Education for At-Risk Youth How Can It Be Made MoreEffectiverdquo (Working paper 88-11) Ithaca NYSchool of Industrial and Labor RelationsCornell University (ERIC Document NoED297150)

Bridgeland John M John J DiIulio Jr and Karen BMorison 2006 The Silent Epidemic Perspectivesof High School Dropouts Available onlinehttpwwwcivicenterprisesnetpdfsthesilen-tepidemic3-06pdf

Castellano Marisa Sam Stringfield and James RStone III 2003 ldquoSecondary Career andTechnical Education and ComprehensiveSchool Reform Implications for Research andPracticerdquo Review of Educational Research73231ndash72

Catterall James S and David Stern 1986 ldquoTheEffects of Alternative Programs on High SchoolCompletion and Labor Market OutcomesrdquoEducational Evaluation and Policy Analysis877ndash86

Cavanagh Sean 2005 February 23 ldquoVocationalEducationrsquos New Job Defend ThyselfrdquoEducation Week

Center for Human Resource Research 2006NLSY97 Userrsquos Guide A Guide to Rounds 1ndash8Data Columbus OH Author

Combs Janet and William W Cooley 1968ldquoDropouts In High School and After SchoolrdquoAmerican Educational Research Journal5343ndash63

Connell James P Margaret B Spencer and JLawrence Aber 1994 ldquoEducational Risk andResilience in African-American Youth ContextSelf Action and Outcomes in Schoolrdquo ChildDevelopment 65493ndash506

DeLuca Stefanie Stephen B Plank and AngelaEstacion 2006 Does Career and TechnicalEducation Affect College Enrollment St PaulMN National Research Center for Career andTechnical Education

Educational Testing Service 1995 DreamsDeferred High School Dropouts in the UnitedStates Princeton NJ Author

Entwisle Doris R Karl L Alexander and Linda SOlson 2004 ldquoTemporary as Compared toPermanent High School Dropout Social Forces821181ndash1205

Fernandez Roberto M Ronnelle Paulsen andMarsha Hirano-Nakanishi 1989 ldquoDroppingOut Among Hispanic Youthrdquo Social ScienceResearch 1821ndash52

Fine Michelle 1991 Framing Dropouts Notes onthe Politics of an Urban High School AlbanyState University of New York Press

Finn Jeremy D 1989 ldquoWithdrawing from SchoolrdquoReview of Educational Research 59117ndash42

Fredricks Jennifer A Phyllis C Blumenfeld andAlison H Paris 2004 ldquoSchool EngagementPotential of the Concept State of theEvidencerdquo Review of Educational Research7459ndash109

Gamoran Adam and Robert D Mare 1989ldquoSecondary School Tracking and EducationalInequality Compensation Reinforcement orNeutralityrdquo American Journal of Sociology941146ndash83

02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028


Goldschmidt Pete and Jia Wang 1999 ldquoWhenCan Schools Affect Dropout Behavior ALongitudinal Multilevel Analysisrdquo AmericanEducational Research Journal 36715ndash38

Grasso John T and John R Shea 1979 VocationalEducation and Training Impact on YouthBerkeley CA Carnegie Council on PolicyStudies in Higher Education

Hauser Robert M Solon J Simmons and Devah IPager 2004 ldquoHigh School DropoutRaceEthnicity and Social Background fromthe 1970s to the 1990srdquo Pp 85ndash106 inDropouts in America Confronting theGraduation Rate Crisis edited by Gary OrfieldCambridge MA Harvard EducationPublishing Group

Heckman James J and Paul LaFontaine 2007ldquoThe American High School Graduation RateTrends and Levelsrdquo NBER (Working paperW13670) Available online httpssrncomabstract=1073650

Jimerson Shane R Gabrielle E Anderson andAngela D Whipple 2002 ldquoWinning the Battleand Losing the War Examining the RelationBetween Grade Retention and Dropping Outof High Schoolrdquo Psychology in the Schools39441ndash57

Kang Suk and John H Bishop 1989 ldquoVocationaland Academic Education in High SchoolComplements or Substitutesrdquo Economics ofEducation Review 8133ndash48

Kaufman Phillip Martha N Alt and ChristopherD Chapman 2004 Dropout Rates in theUnited States 2001 Washington DC USDepartment of Education

Klerman Jacob and Lynn A Karoly 1994 ldquoYoungMen and the Transition to Stable EmploymentrdquoMonthly Labor Review 11731ndash48

Krein Sheila F and Andrea H Beller 1988ldquoEducational Attainment of Children fromSingle Parent Families Differences byExposure Gender and Racerdquo Demography25221ndash34

Kulik James A 1998 ldquoCurricular Tracks and HighSchool Vocational Educationrdquo Pp 65ndash131 inThe Quality of Vocational EducationBackground Papers from the 1994 NationalAssessment of Vocational Education edited byAdam Gamoran Washington DC USDepartment of Education

Levesque Karen Doug Lauen Peter TeitelbaumMartha Alt and Sally Librera 2000 VocationalEducation in the United States Toward the Year2000 (NCES 2000-029) Washington DCUS Department of Education

Lucas Samuel R 1999 Tracking Inequality NewYork Teachers College Press

Lynch Richard L 2000 ldquoHigh School Career andTechnical Education for the First Decade of the21st Centuryrdquo Journal of Vocational EducationResearch 25155ndash98

Marks Helen M 2000 ldquoStudent Engagement inInstructional Activity Patterns in theElementary Middle and High School YearsrdquoAmerican Educational Research Journal37153ndash84

McLanahan Sara S 1985 ldquoFamily Structure andthe Intergenerational Transmission ofPovertyrdquo American Journal of Sociology90873ndash901

Mertens Donna M Patricia Seitz and SterlingCox 1982 Vocational Education and the HighSchool Dropout Columbus OH NationalCenter for Research on Vocational Education(ERIC Document No ED228297)

Miao Jing and Walt Haney 2004 ldquoHigh SchoolGraduation Rates Alternative Methods andImplicationsrdquo Education Policy AnalysisArchives 12(55) Available onlinehttpepaaasueduepaav12n55

Mishel Lawrence and Jared Bernstein 1994 TheState of Working America New York EconomicPolicy Institute

National Center for Education Statistics 1999ldquo1998 Revision of the Secondary SchoolTaxonomyrdquo (Working paper 1999-06)Washington DC US Department ofEducation

mdash- 2001 Features of Occupational Programs at theSecondary and Postsecondary Education Levels(NCES 2001-018) Washington DC USDepartment of Education

mdash- 2004 The High School Transcript Study ADecade of Change in Curricula andAchievement 1990-2000 (NCES 2004-455)Washington DC US Department ofEducation

National Research Council 1993 LosingGenerations Adolescents in High-Risk SettingsWashington DC National Academy Press

National Research Council and Institute ofMedicine 2004 Engaging Schools FosteringHigh School Studentsrsquo Motivation to LearnWashington DC National Academy Press

Neild Ruth C and Elizabeth Farley 2004ldquoWhatever Happened to the Class of 2000The Timing of Dropout in PhiladelphiarsquosSchoolsrdquo Pp 207ndash20 in Dropouts in AmericaConfronting the Graduation Rate Crisis editedby Gary Orfield Cambridge MA HarvardEducation Publishing Group

Oakes Jeannie 1985 Keeping Track New HavenCT Yale University Press

Orfield Gary ed 2004 Dropouts in America

02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028


Confronting the Graduation Rate CrisisCambridge MA Harvard EducationPublishing Group

Perlmutter Deborah E 1982 Career TrainingChoice Project CATCH A Follow-up Study ofStudents Denied Admission to Vocational HighSchools Brooklyn NY New York City Board ofEducation Office of Occupational and CareerEducation (ERIC Document No ED219563)

Pittman Robert B 1991 ldquoSocial FactorsEnrollment in VocationalTechnical Coursesand High School Dropout Ratesrdquo Journal ofEducational Research 84288ndash95

Plank Stephen B 2001 ldquoA Question of BalanceCTE Academic Courses High SchoolPersistence and Student AchievementrdquoJournal of Vocational Education Research26279ndash327

Rasinski Kenneth A and Steven Pedlow 1998ldquoThe Effect of High School VocationalEducation on Academic Achievement Gainand High School Persistence Evidence fromNELS88rdquo Pp 177ndash207 in The Quality ofVocational Education Background Papers fromthe 1994 National Assessment of VocationalEducation edited by Adam GamoranWashington DC US Department ofEducation

Roderick Melissa 1994 ldquoGrade Retention andSchool Dropout Investigating theAssociationrdquo American Educational ResearchJournal 31729ndash59

Roderick Melissa and Eric Camburn 1999 ldquoRiskand Recovery from Course Failure in the EarlyYears of High Schoolrdquo American EducationalResearch Journal 36303ndash43

Rosenbaum James E 1978 ldquoStructure ofOpportunity in Schoolsrdquo Social Forces57236ndash56

mdash- 2001 Beyond College for All Career Paths for theForgotten Half New York Russell SageFoundation

Rumberger Russell W 1987 ldquoHigh SchoolDropouts A Review of Issues and EvidencerdquoReview of Educational Research 57101ndash21

mdash- 2001 Why Students Drop Out of School andWhat Can Be Done Revised version of paperprepared for the Conference ldquoDropouts inAmerica How Severe Is the Problem What

Do We Know About Intervention andPreventionrdquo Harvard University January 132001 Available online httpwwwcivilright-sprojectuclaeduresearchdropoutsrum-bergerpdf

mdash- 2004 ldquoWhy Students Drop Out of SchoolrdquoPp 131ndash55 in Dropouts in AmericaConfronting the Graduation Rate Crisis editedby Gary Orfield Cambridge MA HarvardEducation Publishing Group

Rumberger Russell W and Stephen P Lamb2003 ldquoThe Early Employment and FurtherEducation Experiences of High SchoolDropouts A Comparative Study of the UnitedStates and Australiardquo Economics of EducationReview 22353ndash66

Rumberger Russell W and Katherine A Larson1998 ldquoStudent Mobility and the IncreasedRisk of High School Dropoutrdquo AmericanJournal of Education 1071ndash35

Schafer J L 1997 Analysis of IncompleteMultivariate Data London Chapman andHall

Stearns Elizabeth Stephanie Moller Judith Blauand Stephanie Potochnick 2007 ldquoStayingBack and Dropping Out The RelationshipBetween Grade Retention and SchoolDropoutrdquo Sociology of Education 80210ndash40

Stone James R III (2000) ldquoEditorrsquos Notes Journalof Vocational Education Research 2585ndash91

Swanson Christopher B and Duncan Chaplin2003 Counting High School Graduates WhenGraduates Count Measuring Graduation RatesUnder the High Stakes of NCLB WashingtonDC Urban Institute Available onlinehttpwwwurbanorgUploadedPDF410641_NCLBpdf

Swanson Christopher B and Barbara Schneider1999 ldquoStudents on the Move Residential andEducational Mobility in Americarsquos SchoolsrdquoSociology of Education 7254ndash67

Vanfossen Beth E James D Jones and Joan ZSpade 1987 ldquoCurriculum Tracking and StatusMaintenancerdquo Sociology of Education60104ndash22

Willett John B and Judith D Singer 1991 ldquoFromWhether to When New Methods for StudyingStudent Dropout and Teacher AttritionrdquoReview of Educational Research 61407ndash50

02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028


Stephen B Plank PhD is Assistant Professor Department of Sociology Johns Hopkins UniversityHis main fields of interest are sociology of education urban school reform delinquency and strati-fication Dr Plank is the codirector of the recently formed Baltimore Education Research Consortiuma partnership of the Baltimore City Public Schools Johns Hopkins University Morgan StateUniversity and other partners His current projects focus on solutions to the problem of high schooldropouts predictors of successful transitions to college and school climate

Stefanie DeLuca PhD is Assistant Professor Department of Sociology Johns Hopkins UniversityHer main fields of interest are sociology of education urban sociology residential mobility and thetransition to adulthood Dr DeLuca is engaged in research that involves sociological considerationsof education and housing policy Some of her research examines the transition to college and workfor young adults Other projects focus on the longitudinal effects of housing voucher experimentson employment schooling and long-term neighborhood residence for poor families Dr DeLucawas recently awarded a William T Grant Foundation Scholarrsquos Award to study how the changes infamily school and neighborhood context that accompany residential mobility affect schoolinghealth and delinquency for young people especially poor youths in the South

Angela Estacion MA is a graduate student Department of Sociology Johns Hopkins UniversityHer main fields of interest are sociology of the family sociology of education and social stratifica-tion Ms Estacion is currently working on her dissertation which investigates the relationshipamong young womenrsquos employment and educational trajectories marriage and cohabitation deci-sions and the economic contexts of labor and housing markets during the transition to adulthood

The work reported herein was supported under the National Research Center for Career andTechnical Education PRAward NoVO51A070003 administered by the Office of Vocational andAdult Education US Department of Education However the contents do not necessarily representthe positions or policies of the Office of Vocational and Adult Education or the US Department ofEducation and endorsement by the federal government should not be assumed

Sabbatical support from the Center for Research on Educational Opportunity at the University ofNotre Dame and a fellowship from the National Academy of Education and the Spencer Foundationsupported the second authorrsquos time on this article We thank James Stone James RosenbaumRussell Rumberger Charles Bidwell and colleagues at Johns Hopkins Universityrsquos Department ofSociology for comments on drafts of the manuscript An earlier version was presented at the 2004annual meeting of the American Sociological Association Sarah Barnard provided research assis-tance Address correspondence to Stephen B Plank Department of Sociology Johns HopkinsUniversity 3400 North Charles Street Baltimore MD 21218-2685 e-mail splankjhuedu

02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028


organization of the school and determines inpart the paths that a student follows afterhigh school

One domain of the high school curriculumthat has been both lauded and criticized forits potential effects on educational outcomesincluding dropping out is career and techni-cal education (CTE) also known as vocation-al education Some have argued that CTEoffers a much-needed point of attachmentand demonstrates the relevance of schoolingfor many youths (Advisory Committee 2003)Others have argued that CTE represents amarginalized part of the high school land-scape one in which mixed messages are sentabout the value of schooling and a youthrsquossuccess as a student (Cavanagh 2005) Aseries of educational reforms in the 1990semphasized the importance of taking bothacademic and vocational courses so that stu-dents were prepared for the transition to col-lege and work (Castellano Stringfield andStone 2003)

In this article we use data from theNational Longitudinal Survey of Youth 1997(NLYS97) to estimate hazards models of therelationship between CTE course taking andhigh school dropout for a recent cohort ofstudents Our analyses reveal a curvilinearassociation between the CTE-to-academiccourse-taking ratio and the risk of droppingout for youths who were ldquoon timerdquo for gradeprogression (ie younger than age 15 onentering the ninth grade) We interpret thisfinding as revealing that a middle-range mixof exposure to CTE and the academic curricu-lum can strengthen a studentrsquos attachment toor motivation while in school We do not findthe same significant association betweencourse taking and the likelihood of droppingout for youths who were 15 or older on highschool entry prompting further considerationof the implications of being old for grade inschool settings

BACKGROUND

Research on Dropping Out

Research on dropping out of high school hashighlighted a web of sociological psycholog-

ical economic and institutional factors thatcontribute to students leaving high schoolbefore they receive a diploma (Allensworth2004 Fine 1991 Orfield 2004) Furthermorestudies have shown that the consequences ofdropping out are dramatic and costlymdashforboth individuals and the society The negativeoutcomes associated with dropping outinclude a higher likelihood of unemploymenta greater chance of living below the povertyline and relying on public assistance morefrequent and severe health problems andincreased criminal activity (EducationalTesting Service 1995 Mishel and Bernstein1994 National Research Council NRC 1993Rumberger 1987)

Rumberger (2001) reported that despite along-term upward trend in high school com-pletion in the United States approximately 5percent of all high school students drop outof school in any given year1 A substantiallyhigher proportion of students leave school fora significant span of time sometime duringtheir educational careers before they receive ahigh school credential Klerman and Karoly(1994) estimated that 37 percent of a nation-al sample of young men who were aged14ndash21 in 1979 quit high school for at leastthree months sometime before they complet-ed high school Rumberger and Lamb (2003)reported that 21 percent of students from theNational Education Longitudinal Study of1988 (NELS88) cohort dropped out of schoolat some point after the eighth grade butbefore they completed high school In bothstudies many of those with spells of droppingout ultimately returned to complete a highschool credential However as past researchhas shown and the present analysis reiteratesany time away from high school increases therisk of never obtaining a credential

Finn (1989) offered two social-psychologi-cal perspectives to explain dropping outFollowing a frustrationndashself-esteem model heposited that a record of poor performancewould cause students to question their com-petence and weaken their attachment toschool at the extreme this process can leadto the ultimate disengagementmdashdroppingout Following a participation-identificationmodel he posited that positive experiencesencourage a sense of belonging and thus

02 Plank 10808 551 PM Page 346


Mon 17 Nov 2008 190028







CTE


02 Plank 10808 551 PM Page 347


Mon 17 Nov 2008 190028










02 Plank 10808 551 PM Page 348


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 349


Mon 17 Nov 2008 190028






METHODS

Sample


02 Plank 10808 551 PM Page 350


Mon 17 Nov 2008 190028




Modeling Strategy







02 Plank 10808 551 PM Page 351


Mon 17 Nov 2008 190028


Variables







02 Plank 10808 551 PM Page 352


Mon 17 Nov 2008 190028




RESULTS







02 Plank 10808 551 PM Page 353


Mon 17 Nov 2008 190028















le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949


02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028







CTE


02 Plank 10808 551 PM 47


Mon 17 Nov 2008 190028










02 Plank 10808 551 PM Page 348


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 349


Mon 17 Nov 2008 190028






METHODS

Sample


02 Plank 10808 551 PM Page 350


Mon 17 Nov 2008 190028




Modeling Strategy







02 Plank 10808 551 PM Page 351


Mon 17 Nov 2008 190028


Variables







02 Plank 10808 551 PM Page 352


Mon 17 Nov 2008 190028




RESULTS







02 Plank 10808 551 PM Page 353


Mon 17 Nov 2008 190028















le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949


02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028










02 Plank 10808 551 PM Page 348


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 349


Mon 17 Nov 2008 190028






METHODS

Sample


02 Plank 10808 551 PM Page 350


Mon 17 Nov 2008 190028




Modeling Strategy







02 Plank 10808 551 PM Page 351


Mon 17 Nov 2008 190028


Variables







02 Plank 10808 551 PM Page 352


Mon 17 Nov 2008 190028




RESULTS







02 Plank 10808 551 PM Page 353


Mon 17 Nov 2008 190028















le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949


02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 349


Mon 17 Nov 2008 190028






METHODS

Sample


02 Plank 10808 551 PM Page 350


Mon 17 Nov 2008 190028




Modeling Strategy







02 Plank 10808 551 PM Page 351


Mon 17 Nov 2008 190028


Variables







02 Plank 10808 551 PM Page 352


Mon 17 Nov 2008 190028




RESULTS







02 Plank 10808 551 PM Page 353


Mon 17 Nov 2008 190028















le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949


02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028






METHODS

Sample


02 Plank 10808 551 PM Page 350


Mon 17 Nov 2008 190028




Modeling Strategy







02 Plank 10808 551 PM Page 351


Mon 17 Nov 2008 190028


Variables







02 Plank 10808 551 PM Page 352


Mon 17 Nov 2008 190028




RESULTS







02 Plank 10808 551 PM Page 353


Mon 17 Nov 2008 190028















le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949


02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028




Modeling Strategy







02 Plank 10808 551 PM Page 351


Mon 17 Nov 2008 190028


Variables







02 Plank 10808 551 PM Page 352


Mon 17 Nov 2008 190028




RESULTS







02 Plank 10808 551 PM Page 353


Mon 17 Nov 2008 190028















le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949


02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028


Variables







02 Plank 10808 551 PM Page 352


Mon 17 Nov 2008 190028




RESULTS







02 Plank 10808 551 PM Page 353


Mon 17 Nov 2008 190028















le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949


02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028




RESULTS







02 Plank 10808 551 PM Page 353


Mon 17 Nov 2008 190028















le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949


02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028















le 11 473 575 65812 232 797 83713ndash15 173 833 866ge 16 48 930 949


02 Plank 10808 551 PM Page 354


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 355


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028









02 Plank 10808 551 PM Page 356


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028







Model BWith Model C






df 5 12 14 17p 0723 lt 0001 lt 0001 lt 0001

+p lt 10 p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 357


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028









Variable (n = 653) (n = 193)



p lt 05 p lt 01 p lt 001

02 Plank 10808 551 PM Page 358


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028








02 Plank 10808 551 PM Page 359


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028



DISCUSSION






02 Plank 10808 551 PM Page 360


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028







NOTES


02 Plank 10808 551 PM Page 361


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028













02 Plank 10808 551 PM Page 362


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 363


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028






Female 498 511

Black 276 228

Hispanic 204 187

Asian 21 20

White 484 552












APPENDIX A

02 Plank 10808 551 PM Page 364


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028















APPENDIX B



Continued

02 Plank 10808 551 PM Page 365


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028





Credits Earned


Generalizability


Limitations


02 Plank 10808 551 PM Page 366


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028


REFERENCES


























02 Plank 10808 551 PM Page 367


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028




























02 Plank 10808 551 PM Page 368


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028
























02 Plank 10808 551 PM Page 369


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028







02 Plank 10808 551 PM Page 370


Mon 17 Nov 2008 190028

high school dropout and the role of career and technical education

Documents