WIXLIB

address why the alignment between a candidate’s student teaching experiences and early-careerteaching experiences might matter for teacher and student outcomes. We therefore used the TEPand district survey data to investigate the extent to which the literacy instructional practicesemphasized in special education TEPs and K–12 districts were aligned, and whether thisalignment predicted the ELA achievement of students with high-incidence disabilities taught byspecial education teachers.In summary, this study addressed four primary research questions (RQs), representedgraphically in the conceptual framework in Figure 1:1. To what extent do specific measures of preservice preparation (e.g., student teachingplacements, credentials, and licensure test scores) predict ELA achievement for studentswith high-incidence disabilities taught by early-career special education teachers?2. To what extent do the instructional approaches emphasized by districts and TEPspredict ELA achievement for students with high-incidence disabilities taught by earlycareer special education teachers?3. To what extent is there alignment between the literacy instructional practicesemphasized in special education teacher education programs and K–12 special educationsettings?4. To what extent does this alignment predict ELA achievement of students with highincidence disabilities taught by special education teachers?4

2. Data and Methods2.1 DataWe combined data from three sources for this study: data on teacher candidates collectedfrom the TEPs participating in the Teacher Education Learning Collaborative (TELC), 1 datacollected through surveys of special education TEP faculty and district special educationdirectors in Washington State, and data on K–12 students and teachers provided by OSPI.TELC DataThe broader TELC data set includes information from 15 of the state’s 21 college- anduniversity-based TEPs that were licensed to credential teachers during the years we studied, 13of which had special education endorsement programs. The data provided by these programsincluded information about teacher candidates themselves (e.g., race and gender) as well as dataabout when student teaching occurred, the schools and districts in which teacher candidatescompleted their student teaching, and the cooperating teachers who supervised their internships.Though many of the institutions in TELC provided student teaching data going back to the mid2000s (and in one case, to the late 1990s), we focused on student teaching data from 2009–10 to2017–18 in this analysis because—as described in the section on OSPI data—we were able to linkcooperating teacher information from these years to the students they taught, and use these as aproxy for the students candidates taught during their student teaching placement. Moreover,focusing on these more recent years of data made it more plausible that TEP faculty surveyTELC is a consortium of 15 TEPs providing data on student teaching experiences. The institutions that areparticipating in TELC and provided data for this study include Central Washington University, City University,Evergreen State College, Gonzaga University, Northwest University, Pacific Lutheran University, St. Martin’sUniversity, Seattle Pacific University, Seattle University, University of Washington Bothell, University ofWashington Seattle, University of Washington Tacoma, Washington State University, Western GovernorsUniversity, and Western Washington University. The six institutions that are not participating in TELC include onlyone relatively large (for Washington State) public institution in terms of teacher supply (Eastern WashingtonUniversity) and five smaller private institutions (Antioch University, Heritage University, University of PugetSound, Walla Walla University, and Whitworth University).15

responses (described in the next section) would represent the instruction provided to all candidatesin the sample. Following the work of Theobald, Goldhaber, Naito, and Stein (2021), this analysisfocused on graduates of special education endorsement programs at these 13 institutions (definedas graduating from one of these institutions with a special education endorsement).Survey DataWe designed and administered surveys to special education faculty from the 13 TELCinstitutions with special education endorsement programs, and we administered parallel surveysto special education directors of school districts in the state. The surveys were developed incollaboration with a team of special education TEP faculty and former special education districtadministrators. Questions were primarily derived from the Council for Exceptional Children’sInitial Level Special Educator Preparation Standards, but we also consulted an advisory board ofadditional TEP and district personnel to determine instructional practices that are currentlyemphasized for use with students with high-incidence disabilities in TEP coursework and schooldistricts.The surveys were initially piloted with out-of-state contacts during the 2017–18 schoolyear and were then administered during the 2018–19 school year. Faculty surveys wereadministered by e-mail to contacts within special education TEPs. Importantly, respondentscould opt out of any question and could provide contact information for another faculty memberbetter suited to answer a specific question. The final response rate across all questions in thissurvey and the 13 TEPs was 100%. District surveys were administered by project staff in personat meetings of special education district directors in the nine education service districts in thestate. Unlike the faculty survey, directors were required to respond to all questions in the survey(though in many cases, they completed the survey in consultation with other district6

administrators who were attending the meeting). We received complete district responses for82% of the candidates in the analytic sample.Though tangential to the primary analysis, we highlight one question from these surveysthat provides additional motivation for this study (Figure 2). Specifically, the TEP surveys askedfaculty about their perceptions of the extent to which their TEP graduates were prepared, onaverage, to perform a number of tasks required of special education teachers (e.g., superviseparaeducators and address challenging behaviors), while the district surveys asked directors thesame question about their incoming special education teachers. As shown in Figure 2, specialeducation TEP faculty members’ perceptions of the preparedness of their graduates weredramatically higher than directors’ perceptions of the preparedness of incoming specialeducation teachers. While perhaps not surprising, this finding provides some additionalmotivation for our focus on alignment in this analysis, given that one possible explanation forthese divergent perceptions is that early-career special education teachers are being asked to dosomething different from what they were taught in their TEPs.Although the survey contained additional questions addressing necessary knowledge andskills—such as classroom management and preparing individualized education programs—forthe purposes of this analysis, we primarily focused on one question that asked faculty anddirectors to select all literacy instructional practices for students with high-incidence disabilitiesthat were emphasized or used in their TEP or district. Because this question included a relativelylarge number of potential responses, we performed a factor analysis across TEP and districtsurvey responses to identify combinations of practices that tended to “go together,” and to reducethe dimensionality of the data. Table 1 summarizes the results of this analysis, which identifiedthree principal components with an eigenvalue of at least 1.0 (i.e., that explain more variation7

than the average principal component). 2 We labeled the three principal components “Phonics,Fluency, and Comprehension,” “Guided and Close Reading,” and “Balanced Literacy” to reflectthe instructional practices that load most positively onto these factors (all factor loadings with anabsolute value of at least 0.3 are bolded in Table 1). Interestingly, the five practices that loadonto the first factor—text comprehension, phonemic awareness, vocabulary, fluency, andphonics—are the five literacy instructional areas identified as evidence based by the NationalReading Panel (2000) and in follow-up research (e.g., Castles et al., 2018). We used the factorloadings in Table 1 to create measures of the extent to which each individual special educationteacher’s district and TEP emphasized the practices within each factor.OSPI DataWe merged the TELC and survey data with several sources of data on K–12 students andteachers maintained by OSPI. First, the state’s S-275 database provides annual employmentinformation for all public school employees in the state. We used this data set to identifyindividuals in public school teaching positions, teachers who had a master’s degree or higher,and teachers who were hired into the same district in which they student taught. Second, the S275 database can be linked to the state’s Credential and Endorsement database, which contains acomplete history of all teaching credentials (i.e., the credentials necessary for any public schoolteaching position), teaching endorsements (i.e., the subject areas teachers are endorsed to teach),and licensure test scores in the state. We used this database to identify whether candidates andtheir cooperating teachers held an endorsement in special education, another subject, or both; andto measure candidates’ performance on the state’s Washington Educator Skills Test – Basic(WEST-B) in mathematics, reading, and writing.The scree plot in Appendix Figure A1 also suggests that there is a “kink” in these eigenvalues after the thirdeigenvector, which suggests keeping three principal components in each subject.28

Finally, for 2009–10 through 2018–19 (the most recent year of available data), thesedatabases can be connected to the state’s Comprehensive Education Data and Research System(CEDARS), which allowed us to connect candidates both to the students they taught in theirstudent teaching (i.e., in their cooperating teacher’s classrooms) and to the students they taughtonce they entered the workforce. 3 The CEDARS database also allowed us to identify specialeducation teachers, which (following Theobald, Goldhaber, Naito, & Stein, 2021) we defined asteachers in classroom assignments in which at least 50% of students were receiving specialeducation services. The 50% cutoff is relatively arbitrary, but as the density distribution inFigure 3 shows, the classification of special education teachers was not terribly sensitive to thechosen cutoff as students with disabilities account for less than 40% or more than 90% ofstudents in the majority of classrooms in the state.The CEDARS database also allowed us to connect these special education teachers to thetest performance of students they taught. Students in Washington State take standardized testseach year in mathematics and ELA for Grades 3–8; we standardized these scores across allstudents in the state by grade and year and only considered ELA test scores for students whotook a test aligned with their current grade level. Because our analytic approach (described later)required both current and prior student test scores, our sample of teachers included specialeducation teachers who provided ELA instruction to students with high-incidence disabilities inGrades 4–8. The data also included student demographic information such as gender,race/ethnicity, and program participation. Most importantly for this analysis, the data allowed usCEDARS data include fields designed to link students to their individual teachers, based on reported schedules.However, limitations in reporting standards and practices across the state may result in ambiguities or inaccuraciesaround these links.39

to focus on students with high-incidence disabilities, which we defined as students with anemotional/behavioral disorder, health impairment, or specific learning disability. 42.2 Samples, Measures, and Summary StatisticsTable 2 provides summary statistics for all variables of interest across both of thesamples considered in this analysis. We observed 285 special education teachers (accounting for600 teacher-year observations) who appeared in the TELC data and provided ELA instruction tostudents with high-incidence disabilities. Of these, 243 teachers (and 506 teacher-yearobservations) were linked to survey responses from their TEP and district about the literacyinstructional practices emphasized in each. Column 1 of Table 2 summarizes the 10 variables ofinterest that were observed for all of these teachers. About two thirds of teachers in this samplehad a dual endorsement in special education and another subject, while slightly less than half hadat least a master’s degree. Teachers in this sample tended to score slightly higher on the WESTB mathematics test than the reading and writing tests.Turning to the student teaching variables, about 40% of the special education teacherswere teaching in the same district in which they student taught; this is comparable to estimatesfor all teachers in Washington State reported in Krieg et al. (in press). About 60% of specialeducation teachers in each sample did their culminating student teaching placement in specialeducation, while about 70% student taught with a teacher who had a special educationendorsement. Finally, when we considered the characteristics of candidates’ cooperatingteachers, we found that about 75% were supervised by a cooperating teacher with a master’sdegree, and that candidates’ cooperating teachers had 13 years of teaching experience onaverage. We then calculated summary statistics for these same measures and the principal4Summary statistics of these variables are provided in Appendix Table A1.10

component analyses (PCAs) derived from the TEP and district survey data for the subset ofcandidates who were linked to these survey data (column 2 of Table 2).2.3 Analytic ApproachOur research questions connect the variables of interest in Table 2, measured for specialeducation teachers in these samples, to the ELA achievement of students with high-incidencedisabilities taught by these teachers. We performed these analyses in two steps. First, weestimated first-stage “value-added” models (VAMs) across the full sample of special educationteachers who provided ELA instruction to students with high-incidence disabilities in the state.These models attributed student test score gains to specific teachers. We then used the resultingvalue-added estimates as the outcome variable in second-stage models that used the variables ofinterest described above to predict value added only within the samples in which we observedthese variables. One advantage of this approach is that we used the largest sample possible toremove variation due to student characteristics and prior test scores, classroom composition, andteacher experience that could confound our estimates.First-Stage Value-Added ModelsTo investigate the performance of students with high-incidence disabilities in ELA, weestimated VAMs that have been shown to produce unbiased estimates of the contributions ofindividual teachers to student test performance (e.g., Chetty et al., 2014) but are only rarelyapplied to special education teachers (e.g., Feng & Sass, 2013). One challenge unique to thespecial education context is that about 40% of students with high-incidence disabilities whoreceive ELA instruction from special education teachers also receive ELA instruction from ageneral education teacher. We therefore followed Chetty et al. (2014) and calculated the ELAvalue added of all general education teachers in the state (omitting the year each student in the11

sample was in these teachers’ classrooms) and included the general education teacher’s valueadded as an additional predictor of the performance of students with high-incidence disabilitiestaught by special education teachers. 5In addition to the general education teacher value added, the first-stage models controlledfor lagged student achievement, o

A program of research by the American Institutes for Research with Duke University, Northwestern University, Stanford University, University of Missouri-Columbia, University of Texas at Dallas, and University of Washington TRACKING EVERY STUDENT'S LEARNING EVERY YEAR Special Education Teacher Preparation, Literacy Instructional