510 likes | 594 Views
Improving U.S. Public Schools: Evidence from San Diego. Julian Betts. Symptoms of Problems in U.S. Public Education. Large achievement gaps between racial/ethnic and socioeconomic groups
E N D
Improving U.S. Public Schools: Evidence from San Diego Julian Betts
Symptoms of Problems in U.S. Public Education • Large achievement gaps between racial/ethnic and socioeconomic groups • In international tests, U.S. students hold their own nationally up to grade 4 but by grade 8 U.S. students have fallen significantly behind their peers in many other countries
Have U.S. Public Schools Improved Over Time? Inputs have Soared:
Implications • The way we have increased public school spending has not led to large test-score gains • Similarly, overall increases in spending have left achievement gaps relatively unchanged over three decades • Federal No Child Left Behind Law requires states to boost achievement of lowest scoring groups
Agenda for this Talk • Review two books that use student-level data from San Diego to analyze “what matters” for student achievement • Our first book (2003) is a general analysis of “inputs” and outcomes • Our second book (2005) evaluates San Diego’s Blueprint for Student Success, a controversial and sweeping reform
Determinants of Student Achievement: New Evidence from San Diego Julian Betts, Andrew Zau, Lorien Rice PPIC 2003
Origins of Project • How can we most effectively improve school quality? • State-level database has severe limitations • Can’t follow students over time • Limited information on teachers • Can’t match students to peers or teachers • Urgent need for longitudinal student-level study
Goals • Document variations in school resources and student achievement • Identify effects on student achievement of: • Teacher characteristics • Class size • Curriculum • Student background • Peer groups • Derive policy implications
Outline • Variations across SDUSD schools • Recent patterns in test scores • Determinants of learning in SDUSD • Policy implications • Agenda for future research
Schools Serving Disadvantaged Tend to Have: • Less experienced teachers • Less educated teachers • Fewer teachers with full credential/subject authorization
Outline • Variations across SDUSD schools • Recent patterns in test scores • Determinants of learning in SDUSD • Policy implications
Patterns in Test Scores 1998-2000 • Huge achievement gap in spring 1998 • Steady reduction in gaps related to race, language, poverty, spring 1998 to spring 2000
Large Variations in Test ScoresRelated to Socioeconomic Status Spring 1998 reading scores by SES quintile of school 750 High SES 700 Low SES Score 650 600 550 500 2 3 4 5 6 7 8 9 10 Initial grade level
Large Variations in Test ScoresRelated to Socioeconomic Status Spring 1998 reading scores by SES quintile of school 750 High SES 700 Low SES Score 650 600 550 500 2 3 4 5 6 7 8 9 10 Initial grade level
Outline • Variations across SDUSD schools • Recent patterns in test scores • Determinants of learning in SDUSD • Policy implications
Statistical Approach • Models explain gains in math and reading test scores for each student • Use test scores spring 1998 to spring 2000 • Link students to classroom peers, teachers • Control for unobserved but fixed aspects of each student, school, zip code • Identify effect of school resources such as class size by small changesin class size experienced by a student in different years
Interpretation of Regression Results • Simulate effect of switching from less qualified to more qualified teacher • Comparison group: • Teachers with full credential, ten or more years’ experience, bachelor’s degree,and in middle / high school, full subject authorization • For class size, peer group achievement, other variables, examine effect of changing between 25th and 75th percentile
Elementary Schools: Peer Groups, Class Size More Important Than Teacher Qualifications Predicted change in rate of learning Reading 30 Math 20 10 0 Gradepeerscores Classpeerscores Classsize Emerg'y(0-1) Masters % daysabsent -10 Interns(0-1) -20
High Schools: Peers, Absences, andCourses Taken Matter in Math Predicted change in rate of learning Reading 40 Math 30 20 10 % 0 Grade peer scores Class peer scores # classestaken(>2) -10 % daysabsent -20 # classestaken(0-1) -30
High School: Teacher Qualifications Matter Sporadically but Importantly Predicted change in rate of learning Reading Math 100 80 60 40 % 20 0 Masters PhD -20 Supple-mental -40 Boardresolution -60 -80 Emerg'y
Outline • Variations across SDUSD schools • Recent patterns in test scores • Determinants of learning in SDUSD • Policy implications
What Do Peer Group Results Suggest? • Educational interventions: Not a “zero sum game” • Ability grouping an issue nationally • Suggests public school choice as policy lever
What Do Mixed Results on Teacher Qualifications Mean for Policy? • Paradox: Low-SES schools have least qualified teachers and largest test-score gains • Teacher qualifications matter less thansome claim • Middle and high school teachers without “full” subject authorization and master’s • A problem, but not a crisis • Differential pay in subject areas with shortages • Effectiveness of teacher credentialing system?
How to Handle CurrentBudget Cutbacks Statewide • Some areas of spending may deserve protection • Smaller classes in elementary schools (especially for English Learner students) • Adjust pay to retain highly qualified teachers in upper grades, especially high school
From Blueprint to Reality: San Diego’s Education Reforms Julian Betts Andrew Zau Kevin King PPIC 2005
Outline • Overview of “Blueprint for Student Success” • Effects of interventions on gains in reading achievement • Concluding observations
Large Differences in Student Achievement • Across nation: Large achievement gaps between races, socioeoconomic groups, language groups • Betts, Zau, Rice (2003) demonstrate thisin context of San Diego Unified School District (SDUSD) • Growing recognition that broad reforms that do not focus on low-performers ─ such as increased overall spending ─ do little to narrow gaps
SDUSD in Forefront of Reforms • SDUSD among several of largest districtsin nation undertaking reforms to narrow achievement gap • “Blueprint for Student Success” implemented in summer 2000 • Massive redeployment of resources to students lagging behind, especially in reading
Blueprint: A High-Profile Program • National attention in press • Tens of millions of dollars in funding from foundations around nation (e.g., Hewlett, Gates, Atlantic Philanthropies) • “This really is the most important urban school reform effort in the country… . If the reforms work here, they will have a national impact because ideas travel.” • Marshall Smith, former U.S. Undersecretary of Education; currently Program Directorfor Education, the Hewlett Foundation
Goals of Project • Provide first student-level analysis of effectof Blueprint on reading achievement (levels and gaps) • Analyze which elements of Blueprint are most effective • Impossible with earlier school-level analysis • Describe student participation • Study first two years of Blueprint
Elements of Blueprint • Three main strategies to help students • Prevention • Interventions(for those lagging behind, based on test scores) • Graderetention(in extreme cases) • Initial emphasis on reading, although district later implemented some math interventions • Significant spending on teacher professional development
Prevention Strategies • Peer coaches in schools to train teachers • Elementary level • API 2 schools – additional resources / second peer coach • Focus schools – same as API 2 schools plus extended school year • Enhanced classes – all K and grade 1 • First year of middle school / junior high • Genre studies – two-period English class
Interventions • Literacy block • Double-length English class in middle / high school, with class size of 20 in entry grade • Literacy core • Triple-length English class for those significantly below grade level in grade 9 and those Blueprint-retained in grades 6 / 7 • Extended-Day Reading Program (EDRP) • Summer school / intersession • Retention
Interventions Are Quite Targeted • About 1 in 3 students participated in interventions in 2000-2001 and 2001-2002 • Majority of those participating were involved in only one intervention per year
Are the “Right” Students Participating? • Based on two separate reading tests:those deemed below or significantly below grade level are eligible for intervention • District uses test scores as publicized, but some flexibility • Not all eligible students participate • Some ineligible students participate
Outline • Overview of “Blueprint for Student Success” • Effects of interventions on gains in reading achievement • Concluding observations
Statistical Approach Models explain gains in reading test scores for each student Use test score gains from spring 1999 to spring 2002 As in earlier study, control for characteristics of each student, teacher, class, school, and for unobserved but fixed aspects of each student, school, home zip code
Overview of Results • Some preventive measures worked (API2 and Focus Schools) and some did not (peer coaches) • Interventions generally worked well – except for extended English classes in high school
Elementary Schools: Blueprint Greatly Boosted Average Gain in Reading Scores Elementary School Students 40 35 30 25 Change(%) 20 15 10 5 0 API 2 school Focus school EDRP Summer school Inter-session
Middle Schools: Many Blueprint Elements Raised Scores Middle School Students 70 60 50 Change(%) 40 30 20 10 0 Literacy block Literacy core EDRP Summer school Inter- session
High Schools: Summer School the Only Element That Improved Reading Scores High School Students 80 60 40 20 Change(%) 0 Summer school - 20 Peer coach as % of enrollment - 40 Literacy block - 60 - 80 - 100 - 120 Block / core for EL students - 140
How Big Are Gains for Individual Participants Over Time? • For students in grades 3, 6, and 9 in fall 2000 (participating in at least one intervention): What is predicted two-year cumulative effect of participation? • As a percentage of a standard deviation in their cohorts’ test scores in spring 2002: • Elementary + 23% • Middle + 6% • High School – 12%
Grade 3: Blueprint Shifts Participants’ Location in Distribution of Reading Scores Elementary Schools 30 If no Blueprint 25 20 % of Blueprintparticipants 15 10 5 0 1 2 3 4 5 6 7 8 9 10 District Test-Score Decile
Grade 3: Blueprint Shifts Participants’ Location in Distribution of Reading Scores Elementary Schools Actual 30 If no Blueprint 25 20 % of Blueprintparticipants 15 10 5 0 1 2 3 4 5 6 7 8 9 10 District Test-Score Decile
Grade 6: Similar but Weaker ImprovementGrade 9: Small Negative Effects Actual No Blueprint Middle Schools High Schools 25 20 % 15 10 5 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Decile Decile
Participation Patterns Imply Reductions in Ethnic Test Score Gaps, Except in High School Two-Year Reduction in Ethnic Test Score GapsAttributable to Blueprint Black -White Hispanic-White 10 5 % 0 Highschool - 5 Middleschool - 10 - 15 Elementaryschool - 20
Outline • Overview of “Blueprint for Student Success” • Effects of interventions on gains in reading achievement • Concluding observations
Overall, Blueprint Boosted Reading Achievement: But Large Variations • Blueprint most successful in elementary schools, least successful in high schools • High schools: Better off without Blueprint? • Summer school worked, pullout classes failed • Issues of stigma among high school students? • Lack of experience with these reforms at high school level?
Overall Summary of What Works • 3 key elements: diagnosis, intervention, professional development • Student time on task:Matters tremendously for reading (EDRP, summer school, double and triple length English classes) • All of these activities are supervised by teachers trained in approaches to literacy