Teacher Attrition and Retention - A Meta-Analytic and Narrative Review of The Research

Teacher Attrition and Retention: A Meta-Analytic and Narrative Review of the Research Author(s): Geoffrey D. Borman and N.
Maritza Dowling Source: Review of Educational Research, Vol. 78, No. 3 (Sep., 2008), pp. 367-409 Published by: American Educational Research Association Stable URL: http://www.jstor.org/stable/40071133 . Accessed: 27/05/2013 16:32
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Educational Research Association is collaborating with JSTOR to digitize, preserve and extend access to Review of Educational Research.
http://www.jstor.org
This content downloaded from 200.62.146.210 on Mon, 27 May 2013 16:32:05 PM All use subject to JSTOR Terms and Conditions
Reviewof Educational Research Vol. No. 2008, 78, 3, September pp. 367-409 DOI: 10.3102/0034654308321455 2008 AERA.http://rer.aera.net
Teacher Attrition and Retention: A Meta-Analytic and Narrative Review of the Research
Geoffrey D. Borman and N. Maritza Dowling University of Wisconsin-Madison
This comprehensive meta-analysis on teacher career trajectories, consisting of 34 studies of 63 attrition moderators, seeks to understand why teaching attrition occurs, or whatfactors moderate attrition outcomes. Personal characteristics of teachers are importantpredictors of turnover. Attributes ' of teachers schools, including organizational characteristics, student body composition, and resources (instructional spending and teacher salaries), are also key moderators. The evidence suggests that attrition from teaching is (a) not necessarily "healthy" turnover, (b) influenced by various personal and professional factors that change across teachers' career paths, ' (c) more strongly moderated by characteristics of teachers work conditions than previously noted in the literature, and (d) a problem that can be addressed through policies and initiatives. Though researchers have utilized a number of national and state databases and have applied economic labor theory to questions related to teacher attrition, the authors argue that better longitudinal data on teacher career paths and more nuanced theories are needed.
Keywords:
policy. facultycareers,teacherresearch,educational meta-analysis,
This study involved a comprehensivemeta-analysisof all quantitativestudies relatedto teachers'careertrajectoriesand attritionfrom or retentionin the field. Our analyses focused on those primarystudies that measuredtwo teachercareer persistence outcomes: attrition or retention. Rather than quantifying typical national attritionrates, which has been the subject of prior work by Ingersoll why attritionoccurs or, more for(2001a) and others,our goal was to understand of teachPersonalcharacteristics attrition outcomes. moderate what factors mally, ers, includingtheirbackgroundsand qualificationsto performtheirjobs, and the with which they of theirworkplaces,includingthe "rawproducts" characteristics of their workingenvironcharacteristics work (i.e., students),the organizational ments,and the resourcesavailableto performtheirjobs served as our key moderattrition atorsof persistenceoutcomes. Studies that we reviewed operationalized in varying ways. For instance, some reportedvoluntaryquits, retirements,and outcomes.In these cases, we synthesizedonly to otherschools as separate transfers and those outcomesfor voluntaryquits andexcluded the resultsfor the retirement 367
Borman& Dowling
transfergroups.Specifically, the key outcomethatwe consideredin this review is attritionfrom teachingin the form of voluntaryquits. The Importance of Teachers Accordingto the workof Hanushek(1992), the differencebetweenbeing taught by a highly capable and a less thancapableteachercan translateinto a full grade level of achievementin a single school year. Beyond these potential short-term benefits, the researchof Sandersand Rivers (1996) indicatedthat teachereffects can be enduringand cumulative, whetherthey advance studentachievementor leave childrenbehind.As Sandersand Riversdemonstrated, after2 years,the performance of fifth-gradestudents was still affected by the quality of their thirdgrade teacher. Furthermore,students whose initial achievement levels were can have vastlydifferentacademicoutcomesas a resultof the sequence comparable of teachersto whom they are assigned. Indeed,evidence of the strongeffects of teacherson studentachievementcan be tracedback to the classic Colemanreport, whichconcludedthatteachercharacteristics tendedto explainmorevariancein student achievementthanany otherschool resource(Colemanet al., 1966). Not only areteacherscentralto promotingstudentlearning,theircompensation representsa substantialportionof the nationalinvestmentin public education.In the typical school district,teachersalariesaccountfor at least half of the expenditures (Guthrie& Rothstein, 1999; Speakmanet al., 1996). The resourcedemands associated with employing teachers and the strong empirical evidence linking teacherqualityto improvedachievementare compellingreasonsto focus on policies relatedto teachersas key levers to improveefficiency, equity, andproductivity in public education. and policy landDuringthe past two decades,the educationresearchliterature of for reformingthe preparation scape have been replete with recommendations in-service recruitteacher teachers, enhancing professional development, improving ment and retention,and improvingteacherquality.Fromreportsin the 1980s by such as the NationalCommissionon Excellencein Education (1983), organizations the Holmes Group(1986), andthe CarnegieForumon Education andthe Economy focused on the reformof teacherpreparation to (1986), which primarily programs, the recentfederaleducationlegislation,No Child Left Behind,prominent national effortsto reformeducationin the UnitedStateshave emphasizedthe importance of teacherin every classroom. placinga high-quality Perspectives on Teacher Labor Markets and the Role of Attrition A significantpartof this discussion has revolved aroundthe issues of teacher labor marketsand the supply and demand of elementaryand secondary school teachers.Along with the growing concernsregardingteacherquality,this discussion took root duringthe 1980s and was initiatedin responseto two clear demographic trends:projectedincreases in student enrollmentswith the echo of the post-World War II baby boom and anticipatedincreasesin the numberof retirements among an aging teaching force (Darling-Hammond,1984; Grissmer & Kirby, 1987; Murnane,Singer, & Willett, 1989; NationalAcademy of Sciences, 1987). This researchsuggestedimpendingshortfallsof teachersthatwould,in turn, force manyof the nation's school systems to lower standards to fill the increasing numbersof teaching openings, inevitably resulting in a less qualified teaching 368
TeacherAttritionand Retention
This literature, force andlower school performance. along with an increasingnumber of nationaland local policy initiatives,focused considerableattentionon the problemof staffingclassroomswith qualifiedteachers(NationalCommissionon Teachingand America'sFuture,1997). the 1980s and much of the 1990s, the prevailingpolicy responseto Throughout Effortsweredirected toward was fairlypredictable. thesestaffingproblems primarily innovativemethodsof increasingthe supplyof qualifiedteachers.Initiatives,such namedWendy as Teach for America,which was developedby an undergraduate to attract talentedstudentsfrom room, attempted Kopp in her Princetondormitory disciplinesandfields otherthaneducationto enterthe teachingforce (Kopp,2001). Otherefforts have attemptedto recruitnew teachersfrom among early retirees, to the workforce, womenwho have raisedchildrenandwho wish to enteror return and othersinterestedin midcareer job changes. For instance,the federallyfunded was designedto assistretiringand separating military program Troops-to-Teachers veteransto become teachersin theirnext careers.Finally,in concertwith these iniof alternative tiativesandothers,a widerarray emergedto allow licensingprograms to obtain education without formal emergencyalternative training college graduates thatwould allow themto immediately credentials begin teaching. In additionto efforts to improvethe supply of teachers,an increasingamount fromthe of researchandpolicy rhetorichas addressedthe issue of teacherattrition of the a retain that factors has and greaterproportion may help explored profession Followthe Teacher from on teachers national data force. Using existing teaching rates that,in comparisonto turnover up Survey,Ingersoll(2001a, 2001b) reported - which, like of approximately12.0% in the mid-1990s in the field of nursing female occupationthathas experiencedstaffingprobteaching,is a predominantly rate: 15.0%in 1988-1989, 13.2% lems- teachinghas a somewhathigherattrition in 1991-1992, and 14.3%in 1994-1995. In all of these cases, Ingersollnoted that roughly half of the attritionamong teacherswas because of actually leaving the workforceandthe otherhalf was associatedwith teachersmoving fromone school to another. Also, thoughIngersoll(2001a, 2001b) noted thatthe numberof teachersretiring fromthe professionsteadilyincreasedacrossthe 1990s, he foundthatthe number of retireesin any given year was smallerthanthe numberof teachersleaving the profession for other reasons. This finding has importantpolicy implications of the teacherwork force because it does not highlightthe grayingand retirement was a prominenttheme which teacher to related central as the shortages, problem in earlierresearch.Instead,Ingersoll's work suggests that the largerproblem is related to nonretirementturnoverand that policy efforts may be productively directedtowardretainingthe substantialnumberof teacherswho are leaving the professionfor otherreasons. These findings have been disputedby Harrisand Adams (2007), who recently PopulationSurvey.The authorsnotedthat analyzednationaldatafromthe Current it is surely true that turnoveris an importantproblemfacing certainschools and subjectareas,but theirresultssuggestedthataggregateteacherturnoverwas similar to the ratesof turnoverin threecomparisonprofessions:nursing,accounting, and social work. Their analyses also revealed that teacherturnoveris relatively high among older teachers,reflectingthe fact thatthey retireconsiderablyearlier thando otherprofessionals.Harrisand Adamshypothesizedthatthe high number 369
Borman& Dowling
of early retirementsis in partbecause of the relatively high ratio of pensions to a more salient salaries in teaching, which thereforemakes pension participation factor in labor marketdecisions. Similarly, earlier analyses of the Schools and Staffing Survey databy Grissmerand Kirby(1997) showed that the incidence of occurwith the highestattrition teacherturnover followed a {/-shapeddistribution, ring early and laterin teachers'careers.As a result,these authors,like Harrisand thandid of teacherretirements Adams, also gave greaterweight to the importance Ingersoll(2001a, 2001b). Even if early retirements are a large partof the problem,othercases of voluntary attritionare prevalentand seem to representa significant expense that the nation's school systems must bear. Accordingto a recentreportpublishedby the Alliance for Excellent Education(2005), the Department of Laborestimatedthat attrition costs an employer30%of the departingemployee's salary.Based on this projection and an estimated 1999-2000 average teacher salary of $41,820, the costs Alliance for ExcellentEducationindicatedthateach case of teacherattrition a school system $12,546. With 173,439 nonretireesleaving the professionduring 1999-2000, the total cost of replacingpublic school teacherswho droppedout of the professionwas nearly$2.2 billion for the year.Attritionandits associatedcosts to the system have rarely been addressedby formal policies and interventions. who typically leaves the profession and why they chose to do so Understanding could help policy makersinvest in initiativesthat targetthe teachersmost at risk for quitting and that help amelioratethe conditions that appearmost salient in teachers'quit decisions. Beyond the stage of one's career, over the last two decades researchershave examined numerousother factors associated with attrition,including individual characteristics of teachers,compensationpolicies, and attributes of the places in which teachersworkschools. Though no formal quantitativemeta-analysisof this work has been performed,there are some general concepts and beliefs that have emerged from prior narrativereviews of this literature.A recent narrative review of this literature andDaley (2006) was particularly by Guarino,Santibanez, helpful in summarizingsome of the prominentthemes. These authorsexamined literaturerelatedto teacherentry, mobility, and attritionpatternsand found that teachersexhibitedpreferencesfor higher salaries,betterworkingconditions,and greaterintrinsicrewards.Theirwork suggestedthatteacherswill tend to move to other teaching positions or jobs or activities outside of teaching that offer these characteristics. Descriptiveanalysesof the type of teacherswho leave the professionhave suggested that math and science teacherstend to leave the professionat higherrates than do teachersof other subjects (Arnold, Choy, & Bobbitt, 1993; Grissmer& Kirby, 1992). Because teaching is a decidedly female occupation,one's gender, one's maritalstatus,and the compositionof one's family have been examinedby variousresearchers 1998, 2002). Drivenby (Marso& Pigge, 1997; Stinebrickner, concernsthatthe most qualifiedand talentedteachersmay be leaving the profession to go into other occupations, other researchon teacher characteristicshas focused on variousqualificationsof teachers,includingthe level of educationthey attained,the test scores they achieved, and the years of experiencethey had accumulated(Adams, 1996; Hanushek,Kain, & Rivkin, 2004; Johnson& Birkeland, 2003; Stinebrickner, 1998, 2002).
370
of teachers,a growing numberof studieshave In additionto the characteristics in which schools the may play a role in teacherattrition.Guarino ways explored et al. (2006) pointedout thaturbanschools and schools with high percentagesof minority studentswere difficult to staff and that teachers tended to leave these themselves.It is also evident, schools whenmoreattractive opportunities presented that are amenable to that factors through change policy can have an impact though, on teachers' decisions to migrateto other schools or quit teaching. Specifically, the notion thatthe implementaGuarinoet al. noted some findings thatsupported - particularly those related andinductionprograms tion of school-basedmentoring - may help lower ratesof turnoveramongbeginningteachers. to collegial support Also, schools thatprovidedteacherswith more autonomyand administrative supto have lower levels of teacherattritionand migration. portappeared Conceptualizing the Review on teacher withinthe literature Thereareseveralkey findingsthathave currency of teacher the issue the effort. frame that retention and attrition First, qualhelped ity continues to be of central importancein discussions of school reform and Therefore,questionsregardingthe qualifications,and the potential improvement. were of considerableimportance. Second, quality,of those who arelost to attrition characteristics, includinggender,age, andrace/ethnicity, thoughkey demographic of attrihave often been used as covariatesin models thatfocus on otherpredictors these geninterestandpolicy relevancein understanding tion, thereis a substantive of teacherswho leave the profession.Third,variousresearchers eralcharacteristics of teacherqualityacross schools of varyhave found inequitiesin the distribution concentrations and (Borman & Kimball, 2005; Ferguson, poverty ing minority 1998;Kain& Singleton, 1996). These findingsraisequestionsconcerninghow the school context and the demographicsof studentsmay shape teachers' decisions about where they work and how efforts can be developed to attracthigh-quality teachersto high-needschools. Fourth,researchrelatedto improvingthe retentionof teachershas suggested that one powerful incentive is higher teacherpay, which reduces the probability earnonce differencesin alternative thatteachersleave the profession,particularly der van & account 1999; into 1995, taken Klaauw, are (Dolton ings opportunities Murnane& Olsen, 1989, 1990). Therefore,we examinedhow teachercompensation andotherformsof school andclassroomresourcesmay be relatedto teachers' careerpersistence.Finally,thoughIngersoll(2001a) pointedout thatless is known characteristics andthe organizational betweenattrition the relationships regarding thatdocumentedthe roles thatvarious to locate literature of schools, we attempted of school organizationand culturecan play in teacherretention. attributes Method LiteratureSearch Method and correlatesof on teachercareertrajectories Broadsearchesof the literature andretentionwereconductedusing severalsearchstrategies.First,the folattrition lowing computerizedreferencedatabaseswere searched,focusing on articles or researchreportspublishedfrom 1980 to 2005: EducationResourcesInformation Clearinghouse; PsychlNFO; JSTOR; The Scholarly Journal Archive; Social 371
Borman& Dowling
Sciences Full Text;AcademicSearch;andEducationFull Text. We searchedthese databasesfor all publicationsthatcontainedat least one of severaltermsrelatedto teachersand theircareerpaths(i.e., teachercareer change, teachercareer development, teacher career paths, teacher retention, teacher attrition, teacher turnover,and teacher attitudes). A second method applied general World Wide Web searches,using searchengines such as Google, for referencesto researchor any other publishedor unpublishedstudies or compilationsof data. Throughthe method of citation chasing, we also collected all relatedmaterialcited in recent narrativereviews of this literature,including those published by Guarinoet al. (2006), Johnson,Berg, and Donaldson (2005), and Wilson, Floden, and FerriniMundi(2001). Finally, we reviewed the referencesections of all retrievedarticles in an attemptto identify any additionalreports.After completing this series of search methods, we found no other available quantitativestudies of teachers' careerpaths. Inclusion Criteria The searchstrategiesand exhaustivecitationchasing producedmore than 150 unique references to studies. We examined abstractsof the studies to ascertain whetherthey includedany reportof teacherretention or attrition data.If an abstract or studydid not suggestthese datawere reported, the studywas excludedfromfurtherconsideration. Morethan90 of the studieswere selectedandretrievedandfurtherorganizedinto one or more of the following seven majortopic areas: 1. characteristics of individuals who enter and remain in the teaching profession; 2. characteristics of individualswho leave the teachingprofession; 3. characteristics of schools and districts related to teacher attrition and retention; 4. compensationpolicies, teacherattrition,and retention; 5. preservicepolicies thataffect teacherrecruitment and retention; 6. in-servicepolicies thataffect teacherretention;and 7. othergeneraldescriptivearticlesaddressingissues relatedto teachercareer trajectories. All bibliographicreferenceswere enteredinto an EndNotefile. We developed an Access databasewith descriptiveinformationfor each reference.The descriptive informationincluded(a) the relevanttopic or topics addressedby the study; recruit(b) the study design, data, and sample description,including participant ment; (c) independentvariables; (d) outcome variables;and (e) main research questions. In the second stage, we focused on the subsetof studiesthatprovidedquantitative information on teacherretentionandattrition. This subsetincludedan arrayof statisticalapproachesto dataanalysis, rangingfrom retentionand attrition results presentedas proportionsand means to more sophisticatedapproachesincluding, for example, discriminant functionanalyses and logistic, probabilistic,or predictive models of teacher retentionand/orattrition.From these studies, we further chose those thatallowed us to estimatethe odds of attrition fromteachingandhow
372
the odds varied according to a number of teacher-relatedcharacteristicsand characteristics school- andstudent-related commonlyassociatedwith teacherattriTeacher-related or tion and/orretentionin the researchliterature. characteristics, includedgender,age, maritalstatus, based on teachercharacteristics, moderators school level (elementary,middle, high) taught,educationlevel, teaching experience, certificationtype, ability or achievement,subjecttaught,and salary.School includedlocation, sector based on school characteristics, variables,or moderators (public or private), enrollment, various types of teacher supportprovided, and school studentpopulationcharacteristics, includingpovertylevel, racialcomposition, socioeconomic status,and school averageachievementlevels. More specifically, we deemed studies eligible for furtherconsiderationbased on the following criteria: the study provided sufficient data on teacher attritionor retentionfrom which effect sizes could be computed; charteacherattrition the studyreported by individualand/orschool-related acteristicsfrom which an effect could be determined; in the form the statisticalanalysisemployedprovidedattrition probabilities or logged odds; of counts, proportions, the sampleor dataprovidedwere not duplicatedin anotherstudy accepted for inclusion;and the sample used in the evaluation was composed of teachers from a elementary,middle, or high school in the United States.
This was in Many studiesrevieweddid not meet these eligibility requirements. The most sizes. effect for information insufficient of because calculating largepart common reasons for excluding studies were a data analysis design that did not allow the computationof attritionprobabilitiesand failure to provide a standard variablesof interaboutthe samplesizes by the moderator deviationor information and school povertylevel). A subest (e.g., gender,race, educationalbackground, in other of studieswerebasedon samplesor datathatwere reported stantialnumber studiesacceptedfor inclusion, so they were eliminated.In the end, 34 studies met and were selected for analysis. all requirements of the 34 studies includedin the meta-analysisare summaThe characteristics in the table rized in Table 1. Each of these studies, which are listed alphabetically by the first author'ssurnameand withinauthorby yearof publication,contributed at least one independent sampleof teachersfor analysisof attritionby the various variables.As the table reveals, most studiesexaminedmultiplemodermoderator ators of attritionoutcomes across the five categories of teacher demographics, school resources,and characteristics, teacherqualifications,school organizational in Table 19 of the 34 studies as noted characteristics. 1, student school Also, body from multivariderived ratios odds as outcomes attrition teacher logged reported and means. ate models, and the remaining15 studiesreportedproportions ModeratorVariables datanecessaryfor calculatingthe effect In additionto collectingthe quantitative sizes (e.g., proportions,odds ratios, means and standarddeviations, and sample 373
ti
! g-s ? 8 o
cL O
2
c O
.2
O
5 ?
.2
OO
c-g
.2
c
O
I tj c
"I
c c
OOOO
^8
- f""
|
-| I I
| 8 8
05
i 1 I I * I
1 2 8 S 8 |
<
R S 8
a:
^1
|3
&
il
i -r
SS
^HWW
s
-S
H<
-c -c S
2?
>
<
.5
HooU
-c
^ |
c 13 -a |
**
<n
ts
<n
m*
J^
J3, Ts 3
\ o "6 ai, r,
<n
rn
.52
.=
-8
Wo
Sr^
Site
0QO3
Site
CO
S S -g M
OQUUQ
<^
H VJ
1 I i IS
<<<<
! *3* *3?
HUM
374
i "8
S1
S1j 1I |
-si I I i%
<
-II
g I
II I I
i 1 I g|
| II I
1 1 11 1 8. 1 "S 8. i I "s I1
CO
{lilt
CO
^ i i I
13 -
" 2
I 3I I gg
ri
Sc I -1 I I 1
II N
II 8 * I
.a fe -g.
5 2
I I
I *
I I
S ^ -a
O
<
&
R8
i-
il s I 2 i k p -g
u.
e S > 3 * S g 3 S5f !>
?S
1
i-H u: ^ O
l ?.j
l!
11
Is? * |
S5
375
| |
1_ 1_
oo
1 I
oo
11 I
o "P
2j
" 1 1
i Ii -g a:
i. i i i a:
Si i I < <
Jl
1 |
" B
it lltil g g-g.3
|
|
1 ! It S It 2
|
.s |>
^ .S -c
-Ox <UUQ<
-| | t 11
I ill | I -II o
S H=
.
j
cL co
cL to
.S
2xx HUU
-c
-c
c
x-o
Is
<
-is.^
<oo
< o<<o
"I
"i
c * '7-
&
^ ~
-
K ^
\o
ro
v~i
n^-
oo
P; R.
wi
in
in<ri
rs
î
vo
Ttv
P^^J
2^
2 k*\
'_J
^^
O>
v^
t_i
^^
iMi
^^
C^
on
^^
O^
o*>
*-S
I
376
ii
111
I i
.2
1 C'1
-5 -= 2 8 <8 "g 8 -g 8 < CO CO 8 -g CO 8 -g lyj '8 &
I
_ "5 fl 2
co
|
to
2
m
*l
&
i I
no
fl tli
i
a 5
s 11
Ocj
j|jB iiHjjlliiji
CO "QÔ " J 0^ w **" w '^^ CO CO
si
IB
.
8-a1
is is S
II
8-2 8-
III
S-^ 8-a 8-g

2 2 '* 3
s
S ^ is
*
=S=S=
It
c %'*
>
If
1 1 1 1
o<
!>
oox<
| 1
oo
<*- u
\O
<s
S*S
in
in
9.
'
W g
^ "
^ -CM
O^
CT^
CTv
I
S "
^
^ * %% I Pi
<coco
1
ll
coco
I *
1 1 1 ills |1JI
cococo^
377
Borman& Dowling
in the studies. of the effect sizes reported sizes), we coded a numberof moderators As summarized outcomesfor mulby Table 1, nearlyall studiesreportedattrition and yielded multipleeffect size estimates, which we reportedat tiple moderators the level of the outcomerather thanat the level of the study.We organizedandconthe moderators of attrition withinfive categories:teacherdemographic ceptualized school teacherqualifications; school organizational characteristics; characteristics; resources;and school studentbody characteristics. Teacherdemographiccharacteristics.Teacherdemographicvariableshelped us to examine potentialdifferencesin attritionrelatedto the demographicand backof teachers.As revealedby Table 1, the moderator variables groundcharacteristics includedthe following: gender; teacherrace; age; maritalstatus; whetheror not the individualwas having a child; numberof children.
Across the 34 studies includedin the meta-analysis,19 contrastedthe attrition of male andfemaleteachers,and 12 studiescompared the relativeattrition of White and non-White, minorityteachers.Studies operationalized of age as a moderator attritionin variousways. Fourstudiescomparedattrition of teacherswho were 30 yearsor youngerat entryinto the professionto teacherswho were 3 1 yearsor older. Three studies used age as a continuousvariablepredictingthe likelihood of attrition, three studiescontrastedteacherswho were older thanor youngerthana particular age (e.g., 35, 39, 50), and two contrastedthose whose age fell within particular ranges(i.e., 20 to 24 vs. 25 to 29, and 30 or older vs. 25 to 29). Finally, four studies considered maritalstatus as a possible moderatorof attrition,and a study by Stinebrickner(2002) examined whetherhaving a child was associated with attritionand the extent to which the numberof childrenin the teacher'sfamily moderatedattritionoutcomes. Teacher qualifications. Table 1 also lists the studies that included four general categories of teachers' professional qualifications as moderatorsof attrition. Specifically, the moderatorvariables we identified in our review included the following: teachertraining; experience; teacherability or achievement; teachingspecialtyarea.
A total of 13 studiescomparedattrition ratesof teacherswith a graduate degree to those for teachers with an undergraduatedegree or less. Three studies rates 1998, 1999, 2002) specificallyexaminedthe relativeattrition (Stinebrickner, 378
of teacherswith a bachelor'sdegree in science or mathversusthose with a degree in anotherarea.Threestudiesexaminedteachertrainingandits relationto attrition by comparingthe attritionrates of teachersholding a regularteachingcertificate to those for teacherswithout a certificate.Four studies comparedthe attritionof teacherswith less than5 yearsof experienceto the ratesfor teacherswith 5 or more yearsof experience.Years of experiencewas also measuredas a continuousvariable in five studies and was used as a predictorof the probabilityof attrition. between Table 1 also lists a numberof studiesthatinvestigatedthe relationship and achievement tests and on variousstandardized teachers'performances aptitude attritionoutcomes. Specifically, one study by Marsoand Pigge (1997) compared the attrition outcomes of teachers with a "high" score on the ACT and Test of Basic Skills (CTBS) tests to those with a "low" score on Comprehensive the ACT andCTBStests. A scoreof 2 1 or higherwas considereda high ACT score, and a score of 180 or higherwas considereda high CTBS score. Three studies by Stinebrickner (1998, 1999, 2002) also used scores on the mathand verbalsections of teacherattrition. of the ScholasticAptitudeTest (SAT) as continuouspredictors The specialtyareain which teachersreceivedtheirtrainingwas also the subject of a numberof studies. Six studiescomparedattritionfor teacherstrainedin math or science to teachershaving other subject specialties. Another six studies conratesof trainedspecial educationteachersto those for teachers trastedthe attrition of othersecondary-levelcourses. High school science teacherswere comparedto elementaryteachersin one study (Grissmer& Kirby, 1992), and 14 studies conratesfor high school teachersand elementaryschool teachers. trastedattrition School organizationalcharacteristics. Many studies examined implications for teacherturnoverrates related to school-to-school differences in their organizavaritionalconditions.As notedin Table 1, we identifiedthe following moderator reviewed: ables amongthe 34 studies school location; school sector; school size; administrative support; school mentoringprogramfor beginningteachers; and teachernetwork; collaboration with administrators; regularsupportivecommunication for advancement; opportunities bureaucracy.
Six studiescontrastedthe attritionratesof teachersworkingin schools located areato those of teachersworkingin schools locatedin rural or suburban in an urban areas. Six additionalstudies examined the attritionrates of teachers working in public schools relativeto those for teachersworkingin privateschools. Also, four of teacherturnover,with two studies, studies includedschool size as a moderator and and Smith Shin (2004), contrastinglarge schools of more (1995) Ingersoll by than 1,000 students to schools with smaller enrollments and two studies of the
379
Borman& Dowling
relationshipbetween attritionand continuousmeasuresof school enrollmentby Ingersoll(2001a) and Mont and Rees (1996). Researchersstudied attritionoutcomes associated with several other school characteristics thathad to do with the generalclimate and level of organizational in the Three studies used a measureof level of administrative support workplace. support,which was a 5-point Likert-typescale, for predictingthe probabilityof attrition.Administrativesupportwas generallydefined as the school's effectivemethness in assistingteacherswith issues such as studentdiscipline,instructional ods, curriculum,and adjustingto the school environment.Two studies, by Shen betweenthe per(1997) andSmithandIngersoll(2004), examinedthe relationship from centageof beginningteacherswho were providedwith a mentoring program, either the same field or from a different field, and teacher attrition.Smith and in school-basedinducIngersollalso used the percentageof teachersparticipating tion activities, includingcollaborativeand networkingopportunities, as a predictor of the likelihood of attritionfrom teachingas a profession.More specifically, these collaboration andteachernetworkactivitiesincludedseminarsor classes for commonplanningtime with otherteachersin the subjectarea, teachers, beginning and participation in a networkof teachers.The moderating effects of regularsupwere studied as the percentage of portive communicationwith administrators teachersreporting with the school's princiregularor supportivecommunications other or chairs. In one other administrators, case, a studyby Hall, pal, department Pearson, and Carroll (1992) considered the relationshipbetween attritionand teachers' perceptions of the "opportunitiesfor advancement"provided by the teaching profession. Finally, one study, by Eller, Doerfler, and Meier (2000), examinedthe associationbetweenthe numberof bureaucrats per 100 studentsand the likelihood of attritionfrom teaching. School resources.Also documentedin Table 1 are the studiesof teacherturnover thatcompareddifferencesin teachers'salariesandschool-to-schooldifferencesin the resourcesprovidedto facilitateinstruction. The moderator variablesof primary interestincludedthe following: averageclass size; student-teacher ratio; school expenditurefor supportper teacher; school expenditurefor teachingmaterials; teacheraide or classroomassistants; instructional spending; per-pupilspending; teachersalary.
Threestudiesexaminedhow averageclass size was associatedwith the probability of leaving the teachingprofession,and two others investigatedthe relation between average student-teacherratio and attrition.Two studies, by Smith and Ingersoll(2004) and Shin (1995), comparedthe associationof "large"(1,000 students or more) school enrollment to "smaller"school enrolment with teacher
380
turnover.Two additional studies, by Ingersoll (2001a) and Mont and Rees (1996), used school enrollment as a continuouspredictorof attrition. A study by Gritzand Theobald(1996) looked specifically at the relationships allocatedfor supportper teacher,teachingmaterials, betweenschool expenditures and classroomassistantsand attrition.Anotherstudy,by Imazeki(2005), investigated how instructionalspending and per-pupil spending were associated with teacherattritionrates. Finally, 14 studies noted in Table 1 incorporatedteacher For example,6 studiesanalyzedthe relaof teacherturnover. salaryas a predictor for teacherswith 0 to 5 years annual between teachers' salaryandattrition tionship of experiencein the teachingprofession.Another5 studieslooked at the same relationshipfor a sample of teacherswith tenureor with 6 to 30 years of experience. In addition,a study by Shin (1995) comparedthe attritionrates of teacherswith salaries(morethan$10,000 in 1972) to those of teacherswith "low"salaries "high" ratesof teach(less than$8,500 in 1972). The same study also comparedattrition ers receivinga "medium" salary($8,501 to $9,999 in 1972) to those of teachersin the low-salarycategory.Finally,using a 5-pointLikert-type scale, one studyexambetweenteachers'reportedsatisfactionwith theirsalariesand inedthe relationship outcomes. attrition School student body characteristics. The final category of moderatorslisted in Table 1 includes three general categories: (a) the schools' socioeconomic composition;(b) studentachievementlevels; and (c) the racial/ethniccomposition of the schools. Three studies, by Stinebrickner (1998, 1999) and Shin (1995), comparedattritionratesfor teachersworkingin schools in which the majorityof students (more than50%) were classified as "low"socioeconomic statusto those of teachersworking in schools with studentpopulationsof a higher socioeconomic class. Stinebrickner (1998, 1999) also contrastedattritionratesof teachersworkschools to others.Nine otherstudiesoperationalized ing in largely"middle-class" school socioeconomic compositionusing a continuousmeasureof the percentage of students in the school eligible for the federal free or reduced-price lunch program. theirattriThreestudiesidentifiedschools withhighertest scoresandcontrasted tion ratesto those for schools with lower test scores. A continuousvariableindicating the percentageof studentswith poor achievementperformanceswas used to predictteacherturnoverin two studiesby Boyd, Lankford,Loeb, and Wyckoff (2005a, 2005b). Also, two studies, by Hanushek et al. (2004) and Eller betweendistrictaverageTexas Assessment et al. (2000), examinedthe relationship of AcademicSkills (TAAS) mathand readingoutcomes and teacherattrition. As noted in Table 1, the relationshipbetween the percentageof minoritystudents (i.e., AfricanAmerican,Asian, Native American,or Latino)enrolledat the outcomeswas the topic of six studies.Threeadditional school andteacherattrition studiescomparedattritionoutcomesof teachersworkingin schools with predominantly (i.e., more than 50%) minority student bodies composed of African Americanstudents,Latino students,and a combinationof AfricanAmericanand Latinostudentsto those of teachersworkingin schools with lesser minorityenrollments. Finally, one study, by Shin (1995), measuredthe relationbetween school racial/ethnic diversityand teacherattrition. 381
Results Because the outcome of interest was dichotomous, that is, attritionfrom the teachingprofessionor quittingteachingversus remainingin the teachingprofession, logged odds ratioeffect sizes were calculated.The odds ratiois an effect size statisticthattypicallycomparestwo groupsin termsof the relativeodds of a given status or event (Fleiss, 1994; Lipsey & Wilson, 2001). In the case of attrition, nationaldatasuggest thatapproximately 7% of teachersleave the professioneach year (Ingersoll, 2003). Therefore, for a group of 100 teachers, the odds that a teacherwill leave the profession in a given year are about 7 to 93. This is often expressedby theratioof the numberof times thatthe event of interestoccursto the numberof times that it does not. This ratio from the previous example is 7 -593, or .08. Alternatively,the odds of an event can be defined as, Odds=/?- (1-/?), wherep is the probabilityof an event, such as teacherattrition. Thus, if the prob(1 - .07) also gives the odds as .08. ability of attritionis .07, then .07 -sThe odds ratio is an appropriate summarystatisticfor dichotomousoutcomes thatarereportedin a varietyof ways but thatmost frequentlytake the formof relative frequencies or proportions.An odds ratio generally compares the relative odds of two groups.Forinstance,if the odds of attrition for female teachersare.22, or 2 to 9, and the odds for male teachersare .1 1, or 1 to 9, the ratio of these two odds is .22 -.1 1, or 2. The odds ratio of 2 for this hypotheticalcomparisonsuggests thatthe odds of teacherattritionamong women are 2 times greaterthanthe odds for men. for the odds ratiohas some properties thatmakeit inappropriate Unfortunately, statisticalanalysis. It is centeredaround1 ratherthan0, with 1 indicatingthat an event or outcome is equally likely in both groups. An odds ratio greaterthan 1 implies thatthe event is more likely in the first group,and values between0 and 1 imply thatthe event is less likely in the first group.Because of this asymmetrical distributional form, an odds ratio such as 0.5 is actuallyof the same magnitudeas an odds ratioof 2- which is the inverse of 0.5- but in the opposite direction.To addressthis problem,all statisticalanalysesareperformed on the natural log of the odds ratio. One of the main advantagesof the log odds ratio is that it is easier to derive inferencesfor than the odds ratio directly. In large samples, the sampling distributionof the log odds ratio is approximatelynormal,with a mean of 0 and standard deviationof 1.83. Computation of Logged OddsRatio EffectSizes Differences in the natureof the outcome data reportedacross the 34 studies necessitatedthe use of two primarymethodsfor computingeffect sizes. Despite differencesin the formulasfor computingthe effect sizes, the methodswere algebraicallyequivalentandyieldedconsistentestimatesof the logged odds ratioeffect size. First, many studies reportedprobabilisticmodels in which attritionwas the outcome reportedas a logged odds ratio and predictedby various teacher and school characteristics. The modeledpredictors acrossthese studieswere relatively consistent, and all models included teacher's gender, race/ethnicity, and some 382
indicatorof age or experiencelevel. In these cases, we used the coefficients for the modeled teacherand school characteristics,which are interpretedas the logged odds ratioof the likelihoodof attrition, directlyas the logged odds ratioeffect size. or frequencycountsof persons Second,manyotherstudiesreported proportions in contrastinggroups with the outcome of attrition.When studies providedproportions,the effect size was computedbased on the following formula,
ESuor= log, (p, (1 -p2) +p2 (1 -Pi)),
of individualsin Group 1 (e.g., men) who experienced wherep} is the proportion of people in Group2 (e.g., women) who experiattritionandp2 is the proportion When frequencycounts were reported,the following formulawas enced attrition. employed,
= \o%e(ad + bc), ESLOR
where a is the frequencyof individualsin Group 1 (e.g., men) who experienced b is the frequencyof individualsin Group1 who did not experienceattriattrition, tion, c is the count of individualsin Group2 (e.g., women) who left teaching,and d is the count of individualsin Group2 who did not leave teaching. As Lipsey and Wilson (2001) pointedout, the logged odds ratioeffect size can also be expressedin a formthatis moresimilarto thatof the commonstandardized mean differenceeffect size, known as Cohen's d or Hedges's g. Specifically, the betweenthe logit foreachgroup, as thedifference loggedoddsratiocanbe calculated = lOg,(pi H(1 -/?;)) - lOg,(p2+ (1 -P2, ESL0R of individualsin Group 1 (e.g., men) who experienced wherepi is the proportion of people in Group2 (e.g., women) who experiattritionandp2 is the proportion enced attrition.As this formulasuggests, the logged odds ratio effect size can be odds of attrition directlyas the differencebetweenthe log transformed interpreted in each group. Though the logged odds ratio is the preferredmeasure for meta-analysisof is dichotomousoutcomes(Fleiss, 1994;Lipsey & Wilson, 2001), its interpretation not intuitive.Positive values greaterthanzero imply a positive association,values equal to zero suggest no relationship,and logged odds ratios less than zero indiof the meancate a negativerelationship. However,a moreexplicit understanding ratio to a odds the obtained be size the effect of logged converting by may ing all statisticalanalysesusing the logged odds ratio simpleodds ratio.We performed to convert the mean effect sizes for each of the but then took the antilogarithms studiedto an odds ratio.The formulafor this conversionis, moderators
ire - peslor
is a wheree is the base of the naturallogarithm,or approximately 2.72, and ESLOr logged odds ratio. 383
Computation of StandardErrorsand Weights The 34 studies included in this meta-analysishad varying sample sizes, with some studiesusing large, nationallyrepresentative samplesof teachersandothers using small samplesof teachersfroma specific locale. The logged odds ratioeffect sizes derivedfrom studies thatemployed largersamples have less samplingerror thaneffect sizes from studiesthatused smallersamples.Therefore,the formerare morereliableandpreciseestimatesof attrition thanthe latter,andeffect sizes generatedfrom largersamples shouldbe weighted more heavily in our analysesthan those from smallersamples. We conductedall analyses of the logged odds ratioeffect sizes using weights, which were equal to the inverse of the samplingerrorvarianceof the effect sizes, suchthatthe contribution of each effect size was proportionate to its reliability.For those studies that applied probabilisticmodels that reportedcoefficients for the modeled teacher and school characteristicspredictingattrition,we constructed errorsfor the appropriate coefficient fromthe weights using the reportedstandard models and the formula, w= - 7 sez where se is the standard errorfor the coefficient. For studies reporting proportionsof individuals experiencing attrition, the standard errorwas calculatedas the squareroot of,
p(l-p)(l+nl+\+n2),
wherep is the weighted mean of the proportion experiencingattritionin Group 1 and Group2, specifically, p = (n} x p/ + n2x p2) + (n, + n2), and ft/is the numberof individualsin Group 1 and n2is the numberof individuals in Group 2. Standarderrorsfor the studies that reportedfrequencycounts were computedas,
\ a
wherea, b, c, andd arethe cell frequenciesfroma 2 x 2 contingencytable.Finally, afterusing the above formulasto computestandard errorsfor the logged odds ratio effect sizes from studies reportingproportions and frequencycounts, we used the formulareportedabove for the weight, which was the inverseof the standard error squared.
384
TeacherDemographicCharacteristicsand Attrition The first set of tabulatedresults is presentedin Table 2. This table identifies each of the teacher demographicvariables that we identified as a moderatorof effect size, the numberof studies in which the moderator'srelation to attrition was observed, the effect size expressed as an odds ratio, the value for the logged odds ratio used in the statistical analyses, a 95% confidence interval for the logged odds ratio, and the z test statistic, which tests the null hypothesis that the logged odds ratio effect size is equal to zero. Gender,the first moderator presentedin Table 2, was identifiedin more studThe 19 ies of teacherattrition(19) thanany otherteacheror school characteristic. of attritionsuggest that the differencesbetween studies of genderas a moderator men andwomen arestatisticallysignificant(z = -2369967.00, /?< .01) andthatthe odds of men leaving teaching are approximatelythree fourthsthose for women. Alternatively,by taking the inverse of the odds ratio of 0.77, the result suggests thatthe odds of women leaving the professionare 1.30 times those for men. The effect size for teacher race/ethnicity was also statistically significant (z = 917152.76, p < .01), indicatingthatWhite teachersare 1.36 times more likely to leave teachingthannon-White minorityteachers. variablesthat measureda teacher'sage took on a varietyof forms. Moderator First,a simplecontinuousmeasureof the teacher'sage in yearsrevealedthatolder teacherswere less likely to leave teachingthanwere youngerteachers(z = -8.06, p < .01). The effect of a 1-yeardifferencewas slight,but if one assumeda linearrelafor a teacher5 yearsyounger the oddsof attrition tionshipbetweenage andattrition, would be 5.32 times greaterthanthose for the older teacher.Even when considerbetweenage at entryintoteachthe relationship ing thefourstudiesthatinvestigated who begantheircareersat 31 that teachers the evidence and attrition, suggested ing or older were less likely to leave the professionthanteacherswho began teaching at 30 or younger(z = -7.97, p < .01). Single studiescomparedthe relativeodds of attritionof several otherage groups.Two studies, which comparedteacherswho were 26 or olderto teachers25 or youngerandthose who were40 or olderto teach- older teacherswere less likely than ers 39 or younger,revealedthe same trend to leave teaching.However,when comparing20- to 24-year-old youngerteachers teachersto those 25 to 29 years of age, one study suggestedthatyoungerteachers were slightly less likely to experienceattrition. Finally, anotherstudy suggesteda potentialthresholdat which older teachersare more likely to leave teachingthan The odds of teachers5 1 yearsor olderleavingteaching theiryoungercounterparts. were nearly2.5 times those for teachers50 yearsold or younger. of teachers' families have been the topic of study. First, Three characteristics fourstudiesof teachers'maritalstatusandattrition suggestedthatthe odds of married teachersleaving the profession were 1.40 times greaterthan those for nonmarried teachers. Though this difference was not of considerable practical significance,it was statisticallysignificant(z = 2.40, p < .05). Second, the event of 6.69 times greaterrelative havinga new child was associatedwith odds of attrition to the odds for teachersnot havinga new child (z = 7.92, p < .0 1). Finally,the number of childrenin the teacher'sfamily was not associatedwith any differencefor outcome. the attrition
385
N 8
On ren On CN I
? I I I I I T "^
<5"2 a, < o U
& ^ <n en o d
mm P I I I I I ^ <= I I I I I <zp
't^cn ^9 ^ ^ d<NO
Jj^d
^MIMcp
d - Cj>
^dd
dllllld
odd
"S
Sao S
vo -"
so^nvo^HinON'-^
S o O 2 r-vo d -3 -<t o o oo in m m dddeNodd o O\ ^ - <d d
!
i
"Q
cl
< oôsm
"rt
tj. ^ ^ ^ m ^ Tt ^-^^
si g
is
d
o
CN 00
I i;
o
r^ 00
1 S
^ *2 ?j J3
ooo II II II
)< s*. k. ^^>^> GoGcOcO-r*^ C C C |, ||
||
-q '>$T/^v^_;oovO S II cd^^t >
I I <N m
On vo
iiiî
ij
.Ms
j
386
iiu
iiiiirilii
mm
TeacherQualificationsand Attrition The results for those moderatorvariablesrelatedto teacherqualificationsare summarizedin Table 3. In 13 studies, researchers comparedthe relative attrition of teacherswith a graduate degreeandteacherswithouta graduate degree.The evidence fromthese studiessuggestedthatthe odds of teacherswith a graduate degree leaving teachingwere somewhatgreaterthan those for teacherswithouta graduate degree (z = -672.12, p < .01). As the next tabulatedresult suggests, a science or mathundergraduate degreewas associatedwith odds of attritionapproximately twice those for teacherswith otherundergraduate degrees (z = 3.93, p < .01). On or regularteachingcertificathe otherhand,though,teachersholdinga traditional tion had a lower odds of leaving teaching relative to those with no certification (z = -532.34, p < .01). By taking the inverse of the odds ratio of 0.38, the result suggestedthatteacherswithouta certificatehadodds of leaving the professionthat were 2.63 greaterthanthose for teacherswith a certificate. With regardto teacherexperience level, the odds of attritionamong teachers with 5 or 6 years of teachingexperiencewere 1.57 greaterthanthose for teachers duringthe first 5 years of their careers(z = 213062.19, p < .01). Also, with each additionalyearof experience,the odds of attritionincreasedslightly (z = 3.83, p < .01). For instance,a differenceof 5 years of experiencewas associatedwith odds for the moreexperiencedteacherthatwere 5.10 times greaterthanthose of attrition for the less experiencedteacher. Table 3 also lists attritionoutcomes for four measures of teacher ability or achievement.First, teacherswith high scores on the CTBS or the ACT national college admissionand placementexaminationwere more likely to leave teaching thanwere teacherswith lower scores (z = 12.10,/?< .01), but the magnitudeof this differencewas slight. In threeothercases, differencesin teachers'outcomeson the SAT math, verbal, and combined math and verbal sections were not associated with statisticallysignificant attritioneffect sizes. Similarly, teachers who failed certificationtests on the first attemptor who scored in the bottom 25% had attrition rates that were statisticallyequivalent to those who achieved betterperformanceson the tests. basedon teachers'specialtyareaarelisted at the bottom Differencesin attrition relativeto special eduof Table 3. Secondaryteachershad higherodds of attrition cationteachers(z = 129706.81, p < .01), but the effect size of 1.02 was very slight. Comparedto science and math teachers,the odds of attritionfor teachersof any other secondary-levelsubject were 1.12 greater(z = -72952.62, p < .01). When teachers,though, secondaryscience teacherswere comparedto elementary-school theirodds of attritionwere nearly 1.5 times greater(z = 5.44, p < .01). Comparing for all elementaryteachersto those for all secondaryteachers, the odds of attrition the final outcome in Table 3, which was gleaned from 14 separatestudies, suggested thatelementaryteacherswere only 1.02 times more likely to leave the profession, but this differencewas of statisticalsignificance(z = -79049.05, p < .01). School OrganizationalCharacteristicsand Attrition Table 4 documentsthe role thatdifferencesin school organizationcan play in the firstsugto the nine tabulated outcomes.Referring teacherattrition moderators, schools were for teachersfromurbanand suburban gested thatthe odds of attrition 387
* 2
n c4 rvo
* 8
cn
* S
* 2
cn m n 1
-< o <^ %
* S3
co
* 2
<N *-*
S
d
^ 1 "S
5 8 ^
a
I- I ^
cugo^
*
oo
oo
|gc5d
^5
do
d<i5
c/5c5d
do
do
gi^
do
do
do
S3
-5*^-hon
OS
co
mo
-;
op
^^:
c|
^>
CA
o't3[!2cn
co
Tfo
-^co
1 2
1
I
i
-a
iiiil
j.
I.!
si
388
BiUHfUR lliill
|M!,l|^
X-
in ~
N
^ ^ ^ *~* O
so
. sd
Ov (N
-^
R -
5 00 <O T
^ cS
Tf
-X-
^ iri
tt
P c"5
ro O
.
U
a-good
?i = 8
S2^g dodo
Wgp-
'O
||
???
vO 00 - * - p ^p ooood
? d
'& sO
- '(NO
?
Q
pppp cd d>
j7(U'ca<^rt'-'
<N(N>O<N
êJ)2ppp og>^dod
O dodd
rop
a '3
O2^d-
o on o
oooTtoN
^d-d
^
on
y
'5
.>
g S
&^
so
o 2
S
Tt
rS
oo
5 <n
g Ls S
WiOOOCNÔfs,
O^r^f^^
cS
SOSO-2
U 11 mo
q
^-^J^.
9 d
8 o
2 o
2 So
g 2
o ^
? o
R
2 S
SI
cdoo
mt^^jOfsi
Is
fliiiil ^liSlis Jlilillx. ill!} |sl?t?1i JslUô ilil|illi!li|l

389
Borman& Dowling
only slightly greaterthan those for teachersfrom rurallocations (1.13), but this effect size was statisticallysignificant(z = 176316.27,/?< .01). Second, two studies contrastedlarge schools of 1,000 or more students to smaller schools and revealedthatthe odds of attrition fromsmallerschools were 1.08 times higherthan those for attritionfrom largerschools (z = -52.06, p < .01). Anothertwo studies assessed the relationshipbetween a continuousmeasureof school enrollmentand attritionrates,and these studies also suggestedhigherratesof attritionof smaller schools (z = -2.51, p < .05). Third,six studies suggested that teachersin private schools experiencedodds of attritionthat were 2.27 times those of teachersfrom public schools (z = -33541 13.00, p < .01). Fourth,three studies that operationalized administrative supportusing a 5-point Likert-typescale revealeda reduction in the odds of attritionassociatedwith morepositive ratingsof support(z = -2.09, in the percentageof beginningteachersparticipating p < .05). The fifth moderator, a school mentoringprogram,was also a statisticallysignificantpredictorof attrition (z = -2917.82, p < .01), with greaterparticipation in the programsassociated with a reduced likelihood of attrition.Sixth, a greater reportedprevalence of school-basedteachernetworksand opportunities for collaborationwas relatedto lower attrition rates(z = -3.33, p < .01). The outcomesfor the final threemoderators indicatedthatregularand supportivecommunication and with administrators betteropportunities for advancement ratesand were associatedwith lesser attrition thathigherlevels of bureaucracy were relatedto greaterattrition rates,but each of these outcomes was based on only one study. School Resourcesand Attrition Table 5 provides information regarding education expenditures, teacher of attritionodds ratios. School salaries,and otherschool resourcesas moderators for teachersupportandexpenditures for teachingmaterialsexhibited expenditures no statisticallyreliable relationto attritionoutcomes. In each case, though,these resultscome fromonly one studyandfourindependent estimatesof the effect size. Additionalresourcesin the formof teacheraidesor classroomassistantswereassociated with considerablyhigher odds of attrition(z = 2.33, p < .05). One study examinedthe relationbetween instructional spendingand attritionand per-pupil needs, spendingand the odds of attrition.In the case of spendingon instructional additionalfunds reducedthe likelihood of attrition(z = -3.87, p < .01). A difference of $500 per pupil in instructional spendingwas associatedwith odds of attrition for teachersin the lesser fundedschools thatare5.38 times greaterthanthose for teachersfrom the betterfundedschools. Overallincreasesin per-pupilspend(z = 4.78, p < .0 1). ing, though,were associatedwith higherodds of teacherattrition The outcomesfor schools' averageclass sizes (z = 1.82) andstudent-teacher ratios (z = 1.22) revealedno statisticallysignificantdifferencesfor eithermoderator. At the bottom of Table 5 are the results for 14 studies that included teacher salaryas a moderatorof teacherturnover.In both cases, for teacherswith 0 to 5 years of experiencein the teachingprofessionand those teacherswith tenureor 6 to 30 yearsof experience,highersalariescorrelated with reducedodds of attrition. For those teacherslaterin theircareers,the relationshipbetween salaryand attrition was strongest,with an odds ratioof 0.66 (z = -3.44,/? < .01), butteacherswith 0 to 5 yearsexperiencealso experiencedreducedodds of turnover associatedwith 390
# ** *
#
en
N (N
(N
vq r-
2 ~^
^
IT) I
Ji'?^ |
p n p oo
On I
<n
II
u
IT) v-, n-t
32 w
????
ji s
i i
8S;q8
| |
w
^5Orn w -
q 2 ^>
Tt ^;
m h o\ m on o\ h to
dodo
m vq
roq
cs o ^t <^
d(N
!
s -^
co
cl
2 S
m
.i
g g 5 s
i - r
I
S
$
Q O
1 s
g
22 3 g
fl) cS
On ' *~
c^
<n c a3
s!
J Sifis
all
i ifl-s
Hit
plillSii
391
* *
* *
* *
* *
on
on
? ^--
oo
rôotN
r^ ^
^7
!lS3
-_o
83qg82
d^ddod^;
III
1
"S
u
#
I
S85
o
jl
II
?' ?
^
?
c^
cJ??
^h&^
S3
??
^CN
| ' |' |'

,,,
|
'*3 cd
C O
(N On gq^HfOppOppp-H
'1
a -a.2 I ||
PQ
?
O
7
On CN
?
oo On
?
m On
odd
??
oovo <^^
^7
^ m t ^^^
IS. 5|
^ |j or 1) H Cd
fl o ig'5
\o O\
O2d
no-H OOO
^~*~;
do
odd
'-5 3 *> W)
"g g
8
i_ Tt Tt vo cn csmcN von ^-h^^h
I
S
%
g ^ .8
&
g wo73
2 o &*
^
.
s8
^ a
Jtttg
.
. It
Ifi
|4 j 11 1-
HI
-I
i
392
ilPPI-PlM
till,
JMttHiiiiili
higher salaries (z = -2.76, p < .01). Three other outcomes were associated with teachersalaries, but in each case the moderatorwas the subject of only a single study. One study comparingthe attritionrates of teachers with "high"salaries (more than $10,000 in 1972) to those of teachers with "low" salaries (less than $8,500 in 1972) revealed that lower paid teachers had an odds of attrition 1.85 times greaterthantheirhigherpaid counterparts. The same study comparedattrition rates of teachersreceiving a "medium"salary($8,501 to $9,999 in 1972) to those of teachersin the low-salarycategory and suggested that lower paid teachers had odds of attrition 1.37 times greaterthan those for higher paid teachers. Finally, one study used a 5-point Likert-typescale of teachers'reportedsatisfaction with theirsalariesandshowedthata 1-unitincreasein satisfactionwas related to a decreaseof 2.94 for the odds ratioof attrition School StudentBody Characteristicsand Attrition characteristics of schools' studentpopulationshave been the The demographic variables subjectof researchin a numberof studies.In Table 6, the key moderator are organized according to three general categories: indicators of the schools' the averagestudentachievement socioeconomiccomposition,variablesmeasuring level, and moderatorsrelating to the schools' racial/ethniccomposition. Three studies indicatedodds of attrition1.05 times greaterin predominantlylow-SES schools relative to schools in which low-SES studentswere not in the majority (z = 2131,p< .01). Similarly,the resultsfromeight studiessuggestedthatschools meals had with higherpercentagesof studentsqualifyingfor free or reduced-price = found < researchers In one 3. teacher turnover .01 of odds 19,p ). (z study, higher thatthe odds of attritionin schools composed of 20% or more free-lunch-eligible studentswere 1.73 times greaterthanthe odds for schools with less than20%freeluncheligibility. The evidencerelatedto school-averageachievementlevels andteacherattrition was consistent. Schools that had high or above-averageachievementscores had lower odds of attritionamong theirteachersthandid schools with lower achievement scores (z = 3.72, p < .01). Also, higher percentagesof students with poor achievementperformanceswere associated with increasedodds of attrition{z = outcomes on 8.88, p < .01). As suggested by two studies, higherdistrict-average the TAAS were relatedto a reducedlikelihoodof attrition(z = -80.05, p < .01). Finally, at the bottom of Table 6 are the outcomes for various indicators of schools' racial/ethniccomposition. In five cases, studies examined differences in attritionrates between predominantlynonminorityschools and those schools with a majorityAfricanAmerican,Latino,or African Americanand Latinopopulation. In these cases, the odds of attrition among teachers in predominantly White minority schools were up to 3 times greater than those for majorityschools. One study also suggested that the odds of teachers leaving the profession from schools with diverse student populations were more than 3 times greater than the odds for teachers from schools with more homogeneous racial/ethnic compositions. In seven studies, a continuous measure of the percentage of minority students indicated that schools with higher percentages of minorities suffered higher odds of teacher attritionthan did schools with fewer minoritystudents(z = 5.53, p < .01). 393
* cnoo^l tôcn
* d
rôop co 06
* n î
C s; 8
^c U
qt^q
Hw
oo
dc^d1
cjJdcj)
PQ
^b
ô
^5
oo ^"
qcsql
odd1
qqq
^?
'~H
odd
j) pq
oSoS
o-d o
PiPP
dodo
^ ^
odd
P
d
2-^3
O2 8
ocNoq
f":^1^: ~ d>
d>
1
1
8
cl
1 S
5 .a J S
Ks "
>.^nS<
j=
-S
-a -S 3t
I
H co ^ >
IjiIIIiJj
oo co
Hiili?
co
394
* pi en
00 en r^
* So
I I I I
IJ 5
8 O
^ ^
Mill
o
<t3 W
0
i- 1 "5
& 2
n ^ ""tP
tênvoÔ
en
-!
|.g o2-^
=
^
S2
(NCNrn^H
2 end)
I
c G
o| N -
I
IS u
^C oo o
ft
3H
I
i**-
ta?
ii
ii
in
i n
sislfig?
395
Discussion The goal of this synthesis was to understand why attritionoccurs or, moreformally, what factors moderate attrition outcomes. Personal characteristics of teachers, including their backgroundsand qualifications, are importantpredictors of teacher turnover. Specifically, the odds of attrition are higher among teachers who are female, White, young, and married and who have a child. Regardingteachers' qualifications,the odds of attritionare greateramong those who have no graduatedegree, have specialized degrees in mathor science, have regularcertifications,have more years of experience, and score relatively lower on some standardized tests. The characteristicsof the schools in which teachers work are also important moderators of attrition,with higherattritionin urbanand suburbanschools, private schools, elementaryschools, and schools with a lack of collaboration,teachernetworking,and administrativesupport.Those schools with high enrollmentsof poor, minority,and low-achieving studentsalso suffer high attritionrates. Finally, with respect to resources, higher attritionis associated with lower levels of instructionalspending and lower teacher salaries but higher overall per-pupilspending levels. This formal meta-analysisof 34 quantitative studies related to teacher attrition and retention revealed four major themes that help frame one's interpretation of the overall body of evidence and that might help inform futureresearchin this area. First, the attritionfrom the profession of teaching is not necessarily "healthy" attrition.Past researchhas tendedto framethe issue of teacherattritionas a problematicandhighly negativeoutcome.However,to a greatextent,whetherattrition is bador not dependson the relativecompetenceof those who areleavingandstaying. If those who are leaving the schools arethe worstteachers,then attritionmay be deemed as a healthy and potentially beneficial outcome. Of course, this is because healthy and productiveorganizationstypically retainthe most effective employees and turnover the least effective ones. The evidence from this review is somewhatmixed regardingwhetheror not teacher is healthy. attrition Teachers withpresumably who haveearned better training, a graduate to leave degree,aresomewhatmorelikely thanthose with lessertraining tend to stay in teaching teaching.However, those who have earnedcertifications somewhat moreoftenthanthosewithout a teaching certificate. Thereis evidencefrom one studysuggestingthathigherachievingteachers arethe ones who aremorelikely to leave,butthisfindingis notone thathasbeenconsistent acrossothereffortsto study whichhave typicallyfoundno statistically between attrition, significant relationship teacher test scoresandattrition. The schoolsdo seem to lose moreexperienced teachers andteachers withhigh-demand scienceandmathdegrees.Onbalance,then,there is somewhatmoreevidencesuggestingthatit is the moretalented rather thanthe less - those who arebettertrained, talentedteachers moreexperienced, andmorehighly - who tendto be lost to turnover skilled with greater However,this findfrequency. ing is equivocalbecausethe evidenceis somewhatmixedandbecausethese various teacherqualifications areimperfect indicators of teacherquality. Second, attritionappears to be influencedby a numberof personal and professionalfactors thatare prone to change across the life span and careerpath. Kirby 396
andGrissmer(1991) notedthattheoryrelatedto humancapitaloffers some insight into this problem.Applyingthe humancapitalmodel to attrition,a teacher'sdecision to leave the professionis based on a carefulweighing of the costs and benetendsto be higherduringthe earlystagesof a teacher'scareerbecause fits. Attrition the teacherhas accumulatedless specific capital,or knowledge that is specific to Teacherattrition tendsto diminishlater the occupationandthatis nontransferable. in the careeras a teacherbuilds a greateramountof specific capital. As a result, attrition in the early stages of a teacher's careers is especially problematic. Specifically,Ingersoll(2003) estimatedthatduringthe first5 yearsof theircareers, 45% of teachersleave the profession. Kirbyand Grissmer(1991) also suggestedthatthe decision to accept and keep a teachingjob dependson life cycle factorsrelatedto one's existing family status thatindiandchangesin one's family status.Similarly,Wayne (2000) maintained viduals are more likely to leave teaching for family and personal reasons than decibecausethey aredissatisfiedwith theirjob. Indeed,important family-related sions, such as whetherto have a child, tend to be faced by many teachersearly in theircareers.Teacherswho have childrenare far more likely than those teachers who do not have childrento leave the profession.Notwithstanding,there is also some evidence to suggest thatteacherswho leave the professionto have a child or to pursueotherventuresoften returnto teaching.These temporaryleavers represent significantnumbersnationally,with one recentestimate suggesting that one fourthof the teachershiredeach year arepeople who, thoughnot currentlyteaching, have some priorteachingexperience(Wayne,2000). Beyond personaland family factors thatcan affect new teachers'careerdecisions, anotherreasonnew teachersleave is thatteaching,as a profession,has been slow to develop a systematicway to inductbeginnersgraduallyinto a highly complexjob. Unlike otherdevelopedcountries,beginningteachersin the UnitedStates aretypicallygiven the most difficult assignments,areprovidedlimitedclassroom resourcescomparedto veteranteachers,receive little or no support,and are generally isolated behind classroom doors with little feedback or help (Gordon & Maxey, 2000; Moskowitz & Stephens, 1996). However, this meta-analysissuggests that when more formal organizationalmechanismsare put in place to provide novice teachers with supportnetworks and mentoringopportunities,these rates. efforts are associatedwith decreasedattrition Ourresults and previous summariesof the literature by authorssuch as Kirby and Grissmer(1991) also suggest that attritionrates are high early and later in conformsto a (/-shapeddistribution. teachers'careersandthatthis attrition pattern Ourreview indicatesthat researcherswho attemptto model a simple linear relaarevery likely tionshipbetweenyearsof experiencesor teachers'age and attrition the results the data. the truepatternthatis represented Indeed, to misrepresent by from our meta-analysisthat summarizecorrelationsbetween simple continuous with considermeasuresof experience or age and attritionshould be interpreted able caution. Decisions to leave the profession during the later stages of a career may be duringthe earexplainedby factorsthatare distinctfrom those thatare prominent lier years of teaching.For instance,Harrisand Adams (2007) found that teachers thatthis andhypothesized earlierthanotherprofessionals tendto retireconsiderably is in partbecause of the relatively high ratio of pensions to salaries in teaching, 397
Borman& Dowling
which thereforemakespensionparticipation a moresignificantfactorin labormarket decisions. Obviously, family-relateddecisions, including whetherto have a child, areless relevantto teachersduringthe latteryearsof theircareers.However, care of an aging parentor desire to spend time with grandchildren may be salient factorsto those who leave teachingduringthe twilight of theircareers.It is interesting that althoughhigher salariesare associated with higher retentionrates for teachersat all stages of their careers,the evidence reviewed suggests that higher salariestendto be moreimportant for retainingteacherswho have been in the profession for 6 to 30 years thanfor teachersin the first 5 years of theircareers. Third,the characteristicsof teachers' work conditionsare more salientfor preof the dicting attritionthanpreviouslynoted in the literature.The early literature 1970s and 1980s relatedto the careerpathsof teachersstressedthe rapidlyincreasing numberretireesfrom the baby boom generation,personaland family characteristics of teachers as predictorsof voluntaryattrition,and the draw of higher salaries for retaining more teachers. More recently, Ingersoll (2001a, 2001b) factorswithin a school, includinglack of supportfrom arguedthatorganizational studentdiscipline issues, and lack of input and decision-making administrators, power,cause teachersto leave the profession.However,muchdebatecontinuesto characterize the field. Otherresearchers, includingHarrisandAdams(2007), have are the key probcontinuedto provideevidence suggestingthatearly retirements lem and have speculatedthatthis problemis exacerbatedby a pension and salary structure thatrewardsearly retirements. (2001) blames specific shortHarrington ages of mathand science teacherson a dysfunctionallabormarketheld hostageby and,ironically,inequity. poorallocationof resources,disincentivesto productivity, Equalpay for all teachers,he argues,distortsthe marketfor teachersin these technical subjectareas. The evidencereviewedin this meta-analysis suggeststhattherearea largenumber of characteristics of the environments in which teachersworkthatpredictattrition. These characteristics include those noted prominentlyby Ingersoll (2001a, featuresof schools. However,ourresultsindicate 2001b), includingorganizational that they also include salaries and instructionalresources provided to teachers, other organizational attributes of schools, and, also important, the characteristics of schools' studentbodies. Indeed,one of the more troublingcontemporary problems relatedto equalityof educationalopportunity is staffing every classroomin high-poverty and high-minority schools with a highly qualified and talented teacher.The researchevidencehas continuedto suggestthatpoorandminoritystudents have less access to qualifiedteachersthando more affluentand nonminority children(Borman& Kimball,2005; Ferguson,1998; Kain& Singleton, 1996). A significantreasonfor these disparitiesis, in part,attributable to the fact thatthe teacher attrition rates are found in those schools greatest serving low-achieving, poor, and minoritystudents. Fourth,there are various conditions,such as higher salaries, teacher collaboration and networking, and administrative support,thatare relatedto retentionand that are quite amenableto change. Throughthe early and laterstages of research on teacherattrition, the salaryteachersreceive has continuedto be a highly salient of attrition outcomes.This maybe a factorthatexplainsotherrelationships predictor 398
as well, includingthe large disparitiesbetween attritionratesfor privateand public schools. For instance,accordingto the 1987-1988 TeacherFollow-up Survey, 4.5% of public school teachersstatedsalaryas a main reasonfor leaving the profession. In the privateschools, 9.1% of privateschool teachersstated salary as a main reasonfor leaving the profession(Bobbit, Faupel,& Burns, 1991). In addition, Theobald(1990) foundthatsalariesarepositively relatedto decisions to continueteachingin the samedistrict.Evenpreviousresearchhas suggestedthatsalary providesa reasonfor teachersto changecareers.Accordingto BlolandandSelby's factorassociated (1980) review of the literature, salaryappearsto be an important with the careerchanges of male educators,but not female educators.Obviously, increasedsalaries,and even incentives to teach in high-needs schools, are policy options that requireadditionaleducationalexpenditures.However, each of these options could curtailthe high costs to school systems of recruitingand replacing teachers. Also, with respect to staffing high-needs schools, such policies could have very important consequencesfor advancingeducationalequality. Duringthe early years of teachers'careers,otherproactivepolicy options cenbenefits.Given andmentoringappearto have particular collaboration teredaround the high rate of attritionduringthe first 5 years of teachers'careers,more proactive policies to ease this transitionfrom teacherpreparation programsto the real worldof schools andclassroomsare likely to hold promise.Discoveringthe ways and professionaldevelopmentexperiencescan be in which preservice,internship, to help teachersduringthe crucialfirst5 yearsis a criticalresearch,polstructured icy, and programdevelopmentinitiative for the future.In addition,defining the and more seniorteacherscan play in improvingretention roles thatadministrators and more research development.Fromthe evidence reviewed, it appears requires of schools and school systhatinitiativesthatlessen the bureaucratic organization tems andstrategiesthatpromotemoregenuineadministrative supportfrom school leadersandcollegiality amongteachersare strategiesthatmay improveretention. Limitations of the Existing Research Base and Future Directions on the labormarketfor teachersis vital to monitoring High-qualityinformation trendsand addressingpotentialshortagesin a productiveand preemptivemanner. A numberof nationaldata sets have providedgood periodic evidence regarding to exploremanyof the modthese trends.These datahave also allowedresearchers erators of attrition and retention reviewed here. Examples of widely utilized nationaldatasets on teachersandtheircareersarenotedin the appendix.The inforfor informingresearch mationprovidedby these data sources has been important on careertrajectories. However,four significantlimitationsexist. First,few data sources have providedlong-termlongitudinaldataon teachers. Instead,typical nationaldata sets measureattritionratesfrom one academicyear to the next. Althoughinformative,these datado not capturethe dynamictrajectories of teachers'careers.Two waves of data collection through,for instance,the TeacherFollow-up Survey cannottell us aboutthe shape of teachers'growthtraoutcomes, such as commitmentto teaching,and cannottell jectories on important us the extentto which teachersleave andreenterthe teachingforce. We need truly datawith morethantwo time pointsto capturemorenuancedpictures longitudinal As Singer and Willett (1996) argued,such a stratof teachers'careertrajectories. career teacher for pathsallows investigationsof both the durationof studying egy 399
Borman& Dowling
teacheremploymentitself and the growthand change in the attributes, behaviors, and attitudesof teachers. on teacherattritionand retentionhas developed through Second, the literature a relatively uncoordinated arrayof data collection and analyticalefforts that has focused on many elements of the problemand has, as such, not produceda very compelling body of cumulativeevidence. The 34 studies reviewed in this metaa totalof 63 differentcharacteristics analysishave collected information regarding relatedto teachersandthe schools in which they teach.Together,these studiesprovide a rich and variegated literature.Though this breadthof research may be of viewed as an asset, its limiteddepthwith respectto a numberof key moderators - most important, attrition those thatconcernhow policies andotherinitiativescan - is a significant limitation.In this way, the researchhas help improve retention scratched the surface in many respectsand has not produceda strongcumuonly lative body of evidence and theoryfor informingfuturework. We hope thatthis synthesisof the researchevidence will help sharpenresearch questions and theory applied to teacherattritionand retention.For instance, our conceptualizationof five importantconstellationsof variablesaffecting attrition and retention (i.e., teacher demographiccharacteristics,teacher qualifications, school organizationalcharacteristics,school resources,and school studentbody should provide a more coherentframeworkfor organizingfuture characteristics) data-collection efforts and analytical models. Furthermore, the relatively large numberof moderatorsthat showed only weak associations to attritionprovides researchers with empiricalinformation thatcan help guide effortsto winnowdown the numberof predictors in future research.Ourconclusionthatteacher employed attrition ratesand the reasonsfor attrition vary acrossthe life spanandcareerpath pointsto the need for longitudinalanalysesthataresensitiveto differencesemerging over time in teachers'careerchoices. We also hope thatour observationof the of school organizational of teachcharacteristics andthe characteristics importance ers' work conditions as predictorsof attritionand retentionwill inspire greater attentionto these variables,which areboth amenableto interventionand change. As Guarinoet al. (2006) noted,the conceptualframework most often appliedto the studyof teacherlabormarketsis directlyderivedfromthe economic labormarket theoryof supply and demand.With this conceptualization, demandfor teachers is defined as the numberof teaching positions offered at a particular level of compensation,and the supply of teachersis the numberof qualified individuals level of compensation.In this case, compensation willing to teach at a particular includes not only salary,bonuses, and otherforms of currentand futureearnings but also nonpecuniary rewardssuch as workingconditionsand personalsatisfaction. Though we agree that this general theory provides a logical frameworkfor our review suggests thatresearchers must studyingteacherretentionand attrition, apply it in a nuancedway that,most notably,takes into accountthatthe perceived rewardsof teaching and those of competingoccupationsand activities are likely - although monetaryand to change across the career path of teachers and that - many aspects of teachers'workingconditions materialresourcesare important are of equal or greaterimportancewithin the educationlabormarket. on nationalattrition ratesis sporadicandhas been subject Third,the information to some inconsistencies over time becauseof differencesin datacollectionandsamTheinvestment in a yearlynational datacollectioneffortto collectthis plingmethods.
400
fromschoolsanddistricts couldbe minimal becauseit does notrepbasicinformation burdenand could even be addedto existing surveys, resenta significantreporting including the National Center for EducationStatistics Common Core of Data. would provide Consistentyearlymeasuresof teacheroutcomes,includingattrition, indicatorsof the healthof the educationsystem, which disproporotherimportant activitieson students andschools. tionatelyfocuses its data-collection of the problemof teacherattrition, thereis little Finally,despitesome recognition evidence to guide potentialinitiativesto help ameliorateit. Rigorousexperimental are nonexistent. or policies to addressattrition studiesof programs Thoughthereis with some promisesuggestingthatmentoring research some correlational programs for beginningteachersand teachernetworkingprogramsmay be associatedwith reducedattrition rates,thereis a strongneed for randomized experimentsor wellevaluations of initiativessuch more to crafted convincing quasi-experiments provide characteristics of teachers' as these.The weightof evidencesuggestingthatalterable the criticalneedfor underlines role in attrition workenvironments play an important to helpretain of initiatives andevaluations interventions teachers, espewell-designed in need of are most that schools in those improvement. cially APPENDIX National data sources on teachers and teacher careers
Source Nation-wideSchools and Staffing Survey (SASS), NationalCenterfor Education Statistics(NCES) Description Surveysconcerningschools and school personnelconductedby the NCES. The SASS survey system emphasizesteacher demandand shortage,teacherand school administrator characteristics, in conditions and general programs, schools. SASS also collects dataon many and othertopics, includingprincipals' teachers'perceptionsof school climate and problemsin their schools, teacher compensation,districthiringpractices,and of the student basic characteristics population.SASS has four core the components:the School Questionnaire, the Principal TeacherQuestionnaire, and the School District Questionnaire, which was known as the Questionnaire, TeacherDemandand Shortage until the 1999-2000 SASS Questionnaire are These questionnaires administration. sent to respondentsin public, private,and Bureauof IndianAffairs or tribalschools. In 1999-2000, public charterschools were (continued) 401
APPENDIX (continued)
Source Description
also includedin the sample.For the 2003-2004 SASS, a sample of public charterschools was includedin the sample as partof the public school questionnaire. of Teacher Demand and Part of the general SASS system now Survey Shortage, NCES known as the School District Questionnaire Partof the generalSASS system School Questionnaire from the SASS system, NCES TeacherQuestionnaire from the SASS Partof the generalSASS system system, NCES fro the SASS Partof the generalSASS system PrincipalQuestionnaire system, NCES The TeacherFollow-up Survey (TFS), The TFS is designed to measureattrition NCES from the teachingprofessionand teacher mobility.The TFS seeks to determinehow many teachersremainedat the same school, moved to anotherschool, or left the professionin the year following the The TFS uses two SASS administration. one for teacherswho left questionnaires, teachingsince the previousSASS and anotherfor those who are still currently teachingeitherin the same school as last year or in a differentschool. The topics for the Current TeacherQuestionnaire include teachingstatusand assignments,ratingsof variousaspects of teaching,the time teachersspend on differentaspects of the job, professionaldevelopmentover the past 2 years, and ratingsof various strategiesfor retainingmore teachers.The topics for the FormerTeacher include employmentstatus, Questionnaire ratingsof variousaspects of teachingand theircurrent on jobs, information decisions to leave teaching,and ratingsof variousstrategiesfor retainingmore teachers. The SRCG describesrecentscience and Surveys of Recent College Graduates (SRCG) engineeringgraduatesin 1995, 1997, and 1999. The SRCG surveys providemore (continued)
402
Source Description than 12,000 observations on 20- to 30-year-old full-time workers who are recent science and engineering bachelor's degree graduates. The survey contains information about earnings, hours worked per week, geographic location, educational attainment, age, race, sex, date of college graduation, college major, fields of higher degrees earned, and parents' educational attainment. It also includes information about each individual's 1993 Survey of College Graduates (SCG), National Science Foundation (NSF) college grade point average. Data were collected by the NSF as part of the Scientists and Engineers Statistical Data System. The 1993 National SCG was a once-a-decade baseline survey in that it also covered the non-scientist and engineer population with a bachelor's or higher degree - about 29 million people. Second major longitudinal study after the National Longitudinal Study of 1972 (NLS-72) conducted by NCES. HS&B studied the high school students of 1980, attempting to collect the same types of data gathered in NLS-72. However, the second study differed from the first in two significant ways: It addressed many newer issues of the educational process, and it included a sophomore cohort as well as a senior cohort. Adding the sophomore cohort made it possible to study high school dropouts and analyze changes and processes during high school. NELS:88 was the third major longitudinal study sponsored by NCES. The purpose of the NELS:88 was to expand on the base of knowledge from the first two longitudinal surveys by following young adolescents from an earlier age (eighth grade) and updating information throughout the 1990s. In the base year, four cognitive tests (reading, science, history and government, and mathematics) (continued)
High School and Beyond Study (HS&B), NCES
National Education Longitudinal Study of 1988 (NELS:88), NCES
403
Source Description were administered in addition to the student questionnaire, a parent questionnaire, a teacher questionnaire, and a school administrator questionnaire. ELS:2002 is the fourthin a series of NCESsponsoredsecondaryschool-based longitudinalstudies. ELS:2002 is designed to monitorthe transitionof a national sample of young people as they progress from 10thgradethroughhigh school and educationand/orthe on to postsecondary world of work. Undercurrentplans, cohort memberswill be followed through2014 so - includingaccess to, thatlateroutcomes and persistencein, highereducationor - can be success in the labormarket examinedin termsof students'earlier aspirations,achievement,and high school experiences. CPS, conductedby the U.S. Bureauof LaborStatistics,is the least detaileddata set but is useful because it covers the longest span of time. It containsmonthly samplesof the surveys of representative the past two U.S. populationthroughout decades. CPS mergedoutgoing rotation files assembledby the NationalBureauof Economic Researchand providesmore than400,000 observationson full-time workers,aged 20 to 39, between 1980 and 2001. on public NCES database CCD is the primary in the andsecondary education elementary UnitedStates.Produced it is annually, a nationalstatistical database comprehensive of all publicelementary andsecondary andit contains schools andschool districts, data across all states. The CCD comparable containsthreecategoriesof information: information on schools generaldescriptive andschool districts,dataon studentsand staff, and fiscal data. (continued)
EducationLongitudinalStudy of 2002 (ELS:2002), NCES
CurrentPopulationSurvey (CPS)
CommonCore of Data (CCD), NCES
404
Source Description
Alternative Certification: A National Study The Center for Education Policy at the Stanford Research Institute (SRI) International was awarded a grant from the Carnegie Corporation of New York to conduct a comprehensive study of alternative teacher certification programs (2001 to 2005). The study was designed to identify normative standards for effective alternative certification programs; the SRI study explored the components of various alternative routes to teacher certification and their relative effectiveness in preparing teachers for the classroom. The goal of the study was to determine the key characteristics that render alternative certification programs more or less effective in terms of preparing teachers for the classrooms in which they will be working and to understand the contributions of the programs to both the quantity and quality of new teachers.
References
References marked with an asterisk indicate studies included in the meta-analysis. *Adams, G. J. (1996). Using a Cox regression model to examine voluntary teacher turnover.Journal of Experimental Education, 64, 267-285. Alliance for Excellent Education.(2005, August). Tappingthe potential: Retaining and developing high-quality new teachers. Washington, DC: Author. Available from the Alliance for Excellent Education Web site, http://www.all4ed.org/ *Allred, W. E., & Smith, R. B. (1984). Profile of Utah teachers leaving the teaching profession. Rural Educator, 5(3), 2-5. *Arnold, C. L., Choy, S. P., & Bobbin, S. A. (1993). Modeling teacher supply and demand, with commentary (NCES 93-461). Washington, DC: National Center for Education Statistics. *Beaudin, B. Q. (1993). Teachers who interrupttheir careers: Characteristicsof those who returnto the classroom. Educational Evaluation and Policy Analysis, 15, 5 1-64. Bloland, P., & Selby, T. (1980). Factors associated with career change among secondary school teachers: A review of the literature.Educational Research Quarterly, 5(3), 13-24. Bobbit, S. A., Faupel, E., & Burns, S. (1991). Characteristics of stayers, movers, and leavers: Results from the Teacher Follow-up Survey, 1988-1989 (NCES 91-128).
405
Borman& Dowling Washington, DC: U.S. Department of Education, Office of Educational Research and Improvement. Borman, G. D., & Kimball, S. (2005). Teacher quality and educational equality: Do teachers with higher standards-basedevaluation ratings close student achievement gaps? Elementary School Journal, 106, 3-20. *Boyd, D., Lankford, H., Loeb, S., & Wyckoff, J. (2005a). The impact of assessment and accountability on teacher recruitmentand retention: Are there unintendedconsequences? Washington, DC: U.S. Departmentof Education. *Boyd, D., Lankford,H., Loeb, S., & Wyckoff, J. (2005b). Explaining the shortcareers of high-achieving teachers in schools with low-performing students. American Economic Review, 95(2), 166-171. *Brewer, D. J. (1996). Career paths and quit decisions: Evidence from teaching. Journal of Labor Economics, 14(2), 313-339. Carnegie Forum on Education and the Economy. (1986). A nation prepared: Teachers for the 2 1st century. New York: Carnegie Corporation. *Chapman,D. W., & Hutcheson, S. M. (1982). Attrition from teaching careers: A discriminant analysis. American Educational Research Journal, 19, 93-105. *Clewell, B. C, & Villegas, A. M. (2001). Evaluation of the DeWitt Wallace-Reader's Digest Fund's Pathways to Teaching Careers Program. Washington, DC: Urban Institute. Coleman, J. S., Campbell, E. Q., Hobson, C. J., McPartland,J., Mood, A. M., Weinfeld, F. D., et al. (1966) Equality of educational opportunity. Washington, DC: Government Printing Office. Darling-Hammond, L. (1984). Beyond the commission reports: The coming crisis in teaching. Santa Monica, CA: RAND. Dolton, P. J., & van der Klaauw, W. (1995). Leaving teaching in the UK: A duration analysis. Economic Journal, 105, 431-444. Dolton, P. J., & van der Klaauw, W. (1999). The turnover of teachers: A competing risks explanation. Review of Economics and Statistics, 81, 543-552. *Dworkin, A. G. (1980). The changing demography of public school teachers: Some implications for faculty turnover in urban areas. Sociology of Education, 53(2), 65-73. *Eller, W. S., Doerfler, C. B., & Meier, K. J. (2000). Teacher turnover in Texas: Problems and prospects. A report of the Texas Educational Excellence Project. College Station: Texas A&M University. Ferguson, R. F. (1998). Can schools narrow the Black-White test score gap? In C. Jencks & M. Phillips (Eds.), The Black-White test score gap (pp. 318-374). Washington, DC: Brookings Institution. Fleiss, J. L. (1994). Measures of effect size for categorical data. In H. Cooper & L. V. Hedges, The handbook of research synthesis (pp. 245-260). New York: Russell Sage. Gordon, S. P., & Maxey, S. (2000). How to help beginning teachers succeed (2nd ed.). Alexandria, VA: Association for Supervision and CurriculumDevelopment. Grissmer, D. W., & Kirby, S. N. (1987). Teacher attrition: The uphill climb to staff the nation's schools. Santa Monica, CA: RAND. *Grissmer, D. W., & Kirby, S. N. (1992). Patterns of attrition among Indiana teachers: 1965-1987 (R-4076-LE). Santa Monica, CA: RAND. Grissmer,D. W., & Kirby, S. N. (1997). Teacher turnoverand teacherquality. Teachers College Record, 99, 45-56.
406
TeacherAttritionand Retention *Gritz, R. M, & Theobald, N. D. (1996). The effects of school district spending priorities on length of stay in teaching. Journal of Human Resources, 57(3), 477-512. Guarino,C. M., Santibanez, L., & Daley, G. A. (2006). Teacher recruitmentand retention: A review of the recent empirical literature.Review of Educational Research, 76, 173-208. Guthrie,J., & Rothstein, R. (1999). Enabling "adequacy"to achieve reality:Translating adequacy into state school finance distribution arrangements. In J. Hansen & R. Chalk (Eds.), Equity and adequacy in educationfinance: Issues and perspectives (pp. 209-259). Washington, DC: National Academy Press. *Hall, B. W., Pearson, L. C, & Carroll, D. (1992). Teachers' long-range teaching plans: A discriminant analysis. Journal of Educational Research, 85(4), 221-225. Hanushek, E. A. (1992). The trade-off between child quantity and quality. Journal of Political Economy, 100, 84-117. *Hanushek, E. A., Kain, J. F., & Rivkin, S. G. (2004). Why public schools lose teachers. Journal of Human Resources, 59(2), 326-354. Harrington,P. E. (2001). Attracting new teachers requires changing old rules. College Board Review, 192, 6-11. Harris, D. N., & Adams, S. (2007). Understanding the level and causes of teacher turnover: A comparison with other professions. Economics of Education Review, 26, 325-337. Henke, R. R., Chen, X., Geis, S., & Knepper, P. (2000). Progress through the teacher pipeline: 1992-93 college graduates and elementary/secondary teaching as of 1997 (NCES 2000-152). Washington, DC: National Center for Education Statistics. Holmes Group. (1986). Tomorrow's teachers: A report of the Holmes Group. East Lansing, MI: Author. *Imazeki, J. (2005). Teacher salaries and teacher attrition. Economics of Education Review, 24, 431^49. *Ingersoll, R. M. (2001a). Teacher turnover, teacher shortages, and the organization of schools (No. R-01-1). Seattle: University of Washington, Center for the Study of Teaching and Policy. Ingersoll, R. (2001b). Teacher turnoverand teacher shortages:An organizationalanalysis. American Education Research Journal, 38, 499-534. Ingersoll, R. M. (2003). Is there really a teacher shortage? Seattle: University ot Washington, Center for the Study of Teaching and Policy. *Ingersoll, R. M., & Alsalam, N. (1997). Teacherprofessionalization and teacher commitment:A multilevel analysis (NCES 97-069). Washington, DC: National Center for Education Statistics. Johnson, S. M., Berg, J. H., & Donaldson, M. L. (2005). Who stays in teaching and why: A review of the literature on teacher retention. Boston: Harvard Graduate School of Education, Project on the Next Generation of Teachers. *Johnson, S. M., & Birkeland, S. E. (2003). Pursuing a "sense of success": New teachers explain their career decisions. American Educational Research Journal, Kain, J. F., & Singleton, K. (1996, May-June). Equality of educational opportunity *Kirby, S. (1999). Supply and demand of minority teachers in Texas: Problems and prospects. Educational Evaluation and Policy Analysis, 21, 47-66. Kirby, S. N., & Grissmer,D. W. (1991). Sources of teacher supply: Some new evidence from Indiana.Educational Evaluation and Policy Analysis, 13, 256-268.
revisited. New England Economic Review, pp. 87-1 11.
40,581-617.
407
Borman& Dowling Kopp, W. (2001). One day, all children: The unlikely triumph of Teachfor America and what I learned along the way. New York: Public Affairs. Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks, CA: Sage. *Loeb, S., Darling-Hammond, L., & Luczak, J. (2005). How teaching conditions predict teacher turnover in California schools. Peabody Journal of Education, 80(3), 44-70. *Marso, R. N., & Pigge, F. L. (1997). A longitudinal study of persisting and nonpersisting teachers' academic and personal characteristics. Journal of Experimental Education, 65, 243-254. *Mont, D., & Rees, D. I. (1996). The influence of classroom characteristics on high school teacher turnover.Economic Inquiry, 34(\), 152-167. Moskowitz, J., & Stephens, M. (1996). From students of teaching to teachers of students: Teacher induction around the Pacific Rim. Washington, DC: Pelavin Associates. Murnane, R. J., & Olsen, R. J. (1989). The effects of salaries and opportunitycosts on durationin teaching: Evidence from Michigan. Review of Economics and Statistics, 71, 347-352. Murnane, R. J., & Olsen, R. J. (1990). The effect of salaries and opportunity costs on length of stay in teaching: Evidence from North Carolina. Journal of Human Resources, 25, 106-124. National Academy of Sciences. (1987). Toward understanding teacher supply and demand. Washington, DC: National Academy Press. National Commission on Excellence in Education. (1983). A nation at risk: The imperative for educational reform. Washington, DC: Government Printing Office. National Commission on Teaching and America's Future. (1997). Doing what matters most: Investing in quality teaching. New York: Author. Murnane, R. J., Singer, J. D., & Willett, J. B. (1989). The influences of salaries and "opportunity costs" on teachers' career choices: Evidence from North Carolina. Harvard Educational Review, 59, 325-346. *Rees, D. I. (1991). Grievance procedure strength and teacher quits. Industrial and Labor Relations Review, 45(1), 31-43. *Rickman, B. D., & Parker, C. D. (1990). Alternative wages and teacher mobility: A human capital approach.Economics of Education Review, 9(1), 73-79. Sanders,W., & Rivers, J. C. ( 1996, November). Cumulativeand residual effects of teachers on future student academic achievement. Knoxville: University of Tennessee, Value-Added Researchand Assessment Center. *Schlechty, P. C, & Vance, V. S. (1981). Do academically able teachers leave education? The North Carolina case. Phi Delta Kappan, 63(2), 106-1 12. *Shen, J. (1997). Teacher retention and attrition in public schools: Evidence from SASS91. Journal of Educational Research, 97(2), 81-88. *Shin, H. (1995). Estimating future teacher supply: Any policy implications for educational reform? International Journal of Educational Reform, 4(4), 422-433. Singer, J. D., & Willett, J. B. (1996). Methodological issues in the design of longitudinal research: Principles and recommendations for a quantitative study of teachers' careers. Educational Evaluation and Policy Analysis, 18(4), 265-283. *Smith, T. M., & Ingersoll, R. M. (2004). What are the effects of induction and mentoring on beginning teacher turnover?American Educational Research Journal, 41,
681-714.
408
S. T., Cooper,B. S., Sampieri, R., May,J., Holsomback, H., & Glass,B. Speakman, A systemicresource allocationmodel (1996). Bringingmoneyinto the classroom: (Eds.), appliedto the New YorkCitypublicschools.In L. Picus& J. Wattenberger Where does themoneygo? (pp. 106-131). Thousand Oaks,CA: Sage. of teacher attrition. T. R. (1998).Anempirical Economics *Stinebrickner, investigation Review,77(2), 127-136. of Education of a duration modelin the presenceof missT. R. (1999). Estimation *Stinebrickner, and Statistics, 81(3), 529-542. ing data.Reviewof Economics
*Stinebrickner,T. R. (2002). An analysis of occupational change and departurefrom the labor force: Evidence of the reasons that teachers leave. Journal of Human Resources, 57(1), 192-216. Theobald, N. D. (1990). An examination of the influence of personal, professional, and school district characteristics on public school teacher retention. Economics of Education Review, 9(3), 241-250. Wayne, A. J. (2000). Teacher supply and demand: Surprises from primary research. Education Policy Analysis Archives, 8(47). Available from the Education Policy Analysis Archives Web site, http://epaa.asu.edu/epaa/v8n47.html *Whitener, S. D., Gruber, K. J., Lynch, H., Tingos, K., Perona, M., & Fondelier, S. (1997). Characteristics of stayers, movers, and leavers: Results from the Teacher Followup Survey: 1994-95 (NCES 97-450). Washington, DC: National Center for Education Statistics. Wilson, S. M., Floden, R. E., & Ferrini-Mundi,J. (2001, February). Teacher preparation research: Currentknowledge, gaps, and recommendations. Seattle: University of Washington, Center for the Study of Teaching and Policy.
Authors
GEOFFREYD. BORMAN is a professorof educationalleadershipand policy analysis, educationalpolicy studies, andeducationalpsychology at the Universityof WisconsinWI 53706; SciencesBuilding,1025W. JohnsonSt., Madison, Madison,1161DEducational wise.edu. @education, e-mail: gborman of BiostatisticsandMedical N. MARITZADOWLINGis an assistantscientist,Department Informatics,University of Wisconsin-Madison, D4211 VAH, Madison, WI 53726; e-mail:nmdowlin@biostat.wisc.edu.
409

Teacher Attrition and Retention - A Meta-Analytic and Narrative Review of The Research

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Teacher Attrition and Retention - A Meta-Analytic and Narrative Review of The Research

Transféré par

Droits d'auteur :

Formats disponibles

Teacher Attrition and Retention: A Meta-Analytic and Narrative Review of the Research Author(s): Geoffrey D. Borman and N.

policy. facultycareers,teacherresearch,educational meta-analysis,

e S > 3 * S g 3 S5f !>

S-^ 8-a 8-g

S o O 2 r-vo d -3 -<t o o oo in m m dddeNodd o O\ ^ - <d d

fliiiil ^liSlis Jlilillx. ill!} |sl?t?1i JslU^o ilil|illi!li|l

| ' |' |'

High School and Beyond Study (HS&B), NCES

National Education Longitudinal Study of 1988 (NELS:88), NCES

EducationLongitudinalStudy of 2002 (ELS:2002), NCES

CommonCore of Data (CCD), NCES

Vous aimerez peut-être aussi