台灣教育長期追蹤資料庫 taiwan education panel survey (teps) ping-yin kuan national...

55
台台台台台台台台台台台 Taiwan Education Panel Survey (TEPS) Ping-Yin Kuan National Chengchi University 11/16/2004

Upload: marylou-porter

Post on 04-Jan-2016

215 views

Category:

Documents


2 download

TRANSCRIPT

  • Taiwan Education Panel Survey (TEPS)

    Ping-Yin KuanNational Chengchi University11/16/2004

  • Main FeaturesA national longitudinal survey project collecting data from students of junior high cohort and senior high cohort from 2001 to 2007. It also collects data from parents, teachers, and school administrators.To gain a systematic understanding of the main factors affecting students learning. To provide an important resource for both academic research and policy formulation.

  • Project BackgroundLack of good data sets to inform educational reform policies.Past educational researches tended to be cross-sectional, limited in sample sizes, based on adult population, limited in model specifications, gave inconsistent findings, and did not address basic but critical questions.Needs international comparisons.

  • Funding AgenciesMinistry of EducationNational Science CouncilAcademia Sinica

  • Research TeamPrincipal Investigators:

    Ly-Yun Chang and Tony Tam(Academia Sinica)Co-investigators and research fellows are recruited from various academic institutions in Taiwan. Yes, there are UWM graduates in the team.

  • Project ConcernsTheoretical Concern Learning Effects, Behavioral and Psychological Consequences of Schooling InstitutionsPolicy Concern Educational Opportunity

    School Quality

  • Theoretical FrameworkY = f (A, O, E)

    Ability (A)Opportunity (O)Effort (E) Effects Analytical Ability (Y1) Behavior (Y2) Health (Y3)

  • Research Possibility (1)

    Y1

    Y3

    Learning Effects

    Learning Effects

    Learning Effects

    Mental Health

    Mental Health

    Mental Health

    Time 2

    Time 1

    Time 3

  • Research Possibility (2)

    Y1

    E

    Learning Effects

    Learning Effects

    Learning Effects

    Effort

    Effort

    Effort

    Time 2

    Time 1

    Time 3

  • Educational System in Taiwan

  • Basic Research DesignSamples: -- Junior high sample (1st year students)-- Senior high/vocational high sample (2nd year students)-- Junior college (2nd year students)Multiple perspectives (student, parent, and teacher) on selected student, parent, teacher, and school attributes.Longitudinal study and inter-cohort comparisons with comparable sampling designs.

  • Project Timetable

    Nov. 1999 to Dec. 2000

    Year

    2001

    2002

    2003

    2004

    2005

    2006

    2007

    1st

    2nd

    1st

    Semester

    2nd

    1st

    2nd

    1st

    2nd

    1st

    2nd

    1st

    2nd

    1st

    2nd

    1st

    2nd

    1st

    Preparatory operations

    Junior high sample

    Data collection

    Junior high

    Fist year

    (7th grade)

    Data cleaning, compilation and follow-up planning

    Data collection

    Junior high

    Third

    year

    (9th grade)

    Data cleaning, compilation and follow-up planning

    Data collection

    Senior high Second year

    (11th grade)

    Data cleaning, compilation and follow-up planning

    Data collection

    Senior high

    Third

    year

    (12th grade)

    Data cleaning and compil-

    ation

    Senior high sample

    Data collection

    Senior high Second year

    (11th grade)

    Data cleaning, compilation and follow-up planning

    Data collection

    Senior high

    Third year

    (12th grade)

    Data cleaning and compilation

    Junior college sample

    Data collection

    Junior college Second year

    Data cleaning, compilation and follow-up planning

    Data collection

    Junior college Third year

    Data cleaning and compilation

  • Sampling DesignConcerns- Causal Analysis- Multi-level Analysis- Attrition in Follow- upsDesign- Stratified by urban/rural, public/private, and school types- Sample school programs first, then classes, and then students. In principle, 4 classes and 15 students in each class were sampled.- Oversample certain populations

  • Sample Size (2001/2003)

    Total

    Program Types

    Junior High

    Senior High

    Senior Vocational

    Junior College

    Total number of schools/programs sampled

    546

    338

    Regular

    Comprehensive

    62

    27

    163

    49

    Actual number of schools/programs for which data were available

    539/

    535

    333/

    333

    159/

    158

    48/

    46

    62/

    64

    26/

    26

    Actual number of classes for which data were available

    2,303/

    3,213

    1,244/

    1,930

    573/

    733

    130/

    156

    260/

    283

    96/

    111

    Actual number of students for whom data were available

    39,336/

    37,133

    20,004/

    18,903

    8,719/

    8,264

    2,062/

    1,947

    4,066/

    3,842

    4,485/

    4,177

    Actual number of teachers for whom data were available

    6,601/

    6,558

    3,580/

    3,845

    2,728/

    2,486

    312/

    245

  • Data CollectionAbility tests: For students only.Questionnaires: Completed by students, parents, teachers and school administrators.

  • Questionnaire Design

    Students

    Parents

    Teachers

    Schools

    I.Daily Schedule at School

    II.Home Environment

    III.School Environment

    IV.Extracurricular Activities and Peers

    V.Personal Backgrounds, Educational Aspiration, Self-Evaluation

    VI. Civic Orientation

    I.Family Characteristics

    II.About the Student

    III. Students Life During the Primary School and Junior High School Years

    IV. Relationship Between the Family and the Current School

    V.Expectations of the Student

    VI. Civic Orientation

    I.Questions to be Answered by All Teachers

    II.Questions to be Answered by the Homeroom Teacher

    III.Questions to be Answered by Class Teachers (Chinese, English, and Math)

    VI. Civic Orientation

    I.Questions to be Answered by the Principal

    II.Questions to be Answered by the Director of Academic Affairs

    III.Student Affairs

    IV.School Funding and Equipment

    V.Personnel Matters

  • Ability TestsMeasurement of overall analytical ability or problem solving ability that would reflect a students learning achievement and growth.Tests emphasize the ability to solve problems through analysis and deduction rather than through rote learning.Test modules include general deductive reasoning, science, mathematics, and languages.Test results are estimated ability scores based on Item Response Theory.

  • Some Preliminary Research FindingsTwo basic issues:What is the role of the family in the making of educational inequality?What is the relationship between academic achievement and adolescents mental health?

  • The First Issue: Empirical Questions (1): Effects of Parental SES?

    How does parental socioeconomic status, measured in terms of income and education, matter for the cognitive achievement of students? (How does financial constraint compare to parental education?)How do parental SES effects vary across grades?How different are parental SES effects in Taiwan and the U.S.?

  • Measurement StrategiesConstraint: Different countries have different classifications to begin with. Compromise: To facilitate cross-national comparisons, adopt the same number of categorical income and parental education for Taiwan and the U.S.

    Categorical measures of family background to allow for potentially nonlinear effects.

  • DataTEPS

    -- 2001 (Fall)7th Graders (Junior High Cohort)

    11th Graders (Senior High Cohort)

    -- 2003 (Spring)12th Graders (Senior High Cohort, 1stFollow-Up)NELS:88

    -- 1988 (Base year)8th Graders

    -- 1990 (1st Follow-Up)10th Graders

    -- 1992 (2nd Follow-Up)12th Graders

  • CROSS-SECTION ANALYSIS: IRT Score

  • Figure 3C. Net Family Income Effects: K7, K11, & K12 in TEPS R-sqK11 vs K7 = 0.95K11 vs K12 = 0.97K12 vs K7 = 0.89

  • Figure 3D. Net Parental Education Effects: K7, K11, & K12 in TEPS R-sqK11 vs K7 = 0.92K11 vs K12 = 0.998K12 vs K7 = 0.91

  • Summary of FindingsWhat appears to be strong family income effect in Taiwan, despite the emphasis by recent critics of Taiwanese education, is largely spurious (of parental education).Parental education effects are remarkably stable across high school grades.The qualitative and quantitative results are surprisingly similar across two strikingly dissimilar societies.

  • The First Issue: Empirical Questions (2): Effects of Family Structure?

    How do types of family structure affect the cognitive achievement of students?-- Strong evidence has emerged that single-parent and stepparent families have adverse effects on childrens educational achievement. -- Some studies in the U.S. also found that children of single-parent families with cohabitating grandparent(s) performed quite similarly to those of intact families.

  • The First Issue: Empirical Questions (2)

    How about co-residing grandparent(s) in an intact family? Will they bring similar positive educational advantage to their grandchildren? -- In Taiwan, not only nuclear intact families are still the dominant family type, but the multigenerational intact families composed by two biological parents, unmarried children, and at least a grandparent still consist about 11% of households in Taiwan (2000 census).

  • Why Does Family Structure Matter to Childrens Achievement?Economic resources: Non-intact families are often trapped in poverty or have greater economic burden.Socialization resources: Non-intact families are less able or less likely to provide a good environment for children in terms of educational involvement and educational aspiration.Network resources: Non-intact families have fewer network ties for obtaining information and other types of support related to childrens learning.

  • What might a grandparent bring to the family? The case in Taiwan Economic resources? Even though the rate of cohabitating with older parents is declining, non-cohabiting adult children still feels obligated to support their parents financially. Socialization resources? Co-residing grandparents may provide more psychological support for the grandchildren, convey parents expectation, give advice to the grandchildren, and constantly monitor the grandchildrens activities at home.

  • What might a grandparent bring to the family? The case in Taiwan (contd) Network resources? The presence of grandparents may give additional linkage to relatives, communities, and schools and, hence, contribute to grandchildrens learning.

  • Data and MethodDATA: Two cohorts of TEPS 7th Graders (N = 12,442) and 11th Graders (N = 12,320)Measures -- Dependent Variables: IRT Ability Score -- Independent Variables: 5 types of family structure (1) Nuclear intact (reference) (65%; 67%)(2) Multigenerational intact (17%; 18% )(3) Multigenerational single-parent (3 %; 2%)(4) Single-parent (8% )(5) All other types of non-intact (15%; 13%)

  • Data and Method (contd)

    -- Indep. Var.: Indicators of 3 types of resources:Economic: Monthly family incomeSocialization: Attend school events; talk about school; talk about inner thoughts; checking homework; educational expectation Network: visit relatives; know other parentsControl variables: Sex, sib size; ethnicity; parents educational level

  • Data and Method (contd)Method: OLS regressionModel 1: Types of family structure (gross effects)Model 2: Model 1 + control variablesModel 3: Model 2 + monthly family incomeModel 4: Model 3 + indicators of socialization resourcesModel 5: Model 4 + indicators of network resources

  • R2 of 5 Regression Models for Two Cohorts

    Chart9

    0.0540.009

    0.2140.13

    0.2220.148

    0.2940.24

    0.2970.242

    The 7th grader

    The 11th grader

    Regression Models

    R2

    Sheet1

    model-10.0540.009

    model-20.2140.13

    model-30.2220.148

    model-40.2940.24

    model-50.2970.242

    Sheet1

    00

    00

    00

    00

    00

    The 7th grader

    The 11th grader

    Regression Models

    R2

    Sheet2

    7th

    model-1model-2model-3model-4model-5

    multigenerational intact0.0650.0680.0670.0530.052

    multigenerational with single-parent-0.244-0.237-0.182-0.137-0.121

    single parent-0.411-0.34-0.29-0.216-0.213

    all other types-0.685-0.53-0.51-0.407-0.395

    11th

    model-1model-2model-3model-4model-5

    multigenerational intact0.0480.0820.0820.0730.073

    multigenerational with single-parent-0.029-0.0760.0060.0290.035

    single parent-0.142-0.156-0.075-0.044-0.044

    all other types-0.353-0.337-0.312-0.231-0.217

    Sheet2

    0000

    0000

    0000

    0000

    0000

    multigenerational intact

    multigenerational with single-parent

    single parent

    all other types

    Model (7th)

    value b

    Sheet3

    0000

    0000

    0000

    0000

    0000

    multigenerational intact

    multigenerational with single-parent

    single parent

    all other types

    Model (11th)

    value b

  • Effects of Types of Family Structure in 5 Regression Models

    Chart4

    0.065-0.244-0.411-0.685

    0.068-0.237-0.34-0.53

    0.067-0.182-0.29-0.51

    0.053-0.137-0.216-0.407

    0.052-0.121-0.213-0.395

    multigenerational intact

    multigenerational with single-parent

    single parent

    all other types

    7th Graders

    b

    Sheet1

    0.0540.009

    0.2140.13

    0.2220.148

    0.2940.24

    0.2970.242

    Sheet1

    00

    00

    00

    00

    00

    The 7th grader

    The 10th grader

    Regression Models

    R square scores

    Sheet2

    7th

    model-1model-2model-3model-4model-5

    multigenerational intact0.0650.0680.0670.0530.052

    multigenerational with single-parent-0.244-0.237-0.182-0.137-0.121

    single parent-0.411-0.34-0.29-0.216-0.213

    all other types-0.685-0.53-0.51-0.407-0.395

    11th

    model-1model-2model-3model-4model-5

    multigenerational intact0.0480.0820.0820.0730.073

    multigenerational with single-parent-0.029-0.0760.0060.0290.035

    single parent-0.142-0.156-0.075-0.044-0.044

    all other types-0.353-0.337-0.312-0.231-0.217

    Sheet2

    0000

    0000

    0000

    0000

    0000

    multigenerational intact

    multigenerational with single-parent

    single parent

    all other types

    Model (7th)

    value b

    Sheet3

    0000

    0000

    0000

    0000

    0000

    multigenerational intact

    multigenerational with single-parent

    single parent

    all other types

    Model (11th)

    value b

    Chart5

    0.048-0.029-0.142-0.353

    0.082-0.076-0.156-0.337

    0.0820.006-0.075-0.312

    0.0730.029-0.044-0.231

    0.0730.035-0.044-0.217

    multigenerational intact

    multigenerational with single-parent

    single parent

    all other types

    11th Graders

    b

    Sheet1

    0.0540.009

    0.2140.13

    0.2220.148

    0.2940.24

    0.2970.242

    Sheet1

    00

    00

    00

    00

    00

    The 7th grader

    The 10th grader

    Regression Models

    R square scores

    Sheet2

    7th

    model-1model-2model-3model-4model-5

    multigenerational intact0.0650.0680.0670.0530.052

    multigenerational with single-parent-0.244-0.237-0.182-0.137-0.121

    single parent-0.411-0.34-0.29-0.216-0.213

    all other types-0.685-0.53-0.51-0.407-0.395

    11th

    model-1model-2model-3model-4model-5

    multigenerational intact0.0480.0820.0820.0730.073

    multigenerational with single-parent-0.029-0.0760.0060.0290.035

    single parent-0.142-0.156-0.075-0.044-0.044

    all other types-0.353-0.337-0.312-0.231-0.217

    Sheet2

    0000

    0000

    0000

    0000

    0000

    multigenerational intact

    multigenerational with single-parent

    single parent

    all other types

    Model (7th)

    value b

    Sheet3

    0000

    0000

    0000

    0000

    0000

    multigenerational intact

    multigenerational with single-parent

    single parent

    all other types

    Model (11th)

    value b

  • Summary of Findings

    The impact of family structure is mediated by 3 types of resources. The mediating variables related to economic resources and parental involvement, however, have larger effects for the older cohorts than for the younger cohorts. The addition of a grandparent is beneficial to childrens educational achievement. This positive contribution, however, depends on the type of family structure. The effects of family structure types are all smaller for the older cohort.

  • The Second Issue: Empirical QuestionsWould high academic achievement and expectation induce poorer mental health?-- Previous research found a weak positive relationship.Does high family SES induce poorer mental health?-- High SES parents tend to have higher academic expectation and be more involved in the childrens education, which in turn make their children perform better academically.-- Previous studies have found positive relationship between SES and mental health. But some studies also found high SES or high achieving students have more distress. This relationship may be due to the higher achievement pressure of the high family SES.

  • Empirical Questions (contd)How do different parenting styles and parental involvement strategies affect adolescents mental health then?-- Authoritative parenting style (responsive but firm control) has been found to be positive to childrens academic achievement and adjustment in general.-- Psychological control, on the other hand, has been found to be related to adolescents poor psychological and behavioral outcomes. But no report is on its effect on academic achievement.-- Specific parental involvement strategies in education include all three parenting dimensions: support, behavioral control, and psychological control (such as high parental expectation).

  • Chart1

    8.147.614.214.4

    7.148.415.719.6

    5.545.317.826

    1.342.220.532.2

    1.928.127.239.3

    1.221.624.949.7

    213.625.256.3

    0.75.918.173.3

    018.317.263.1

    Junior Col

    Tech. Col/Univ

    Reg. Col/Univ

    Grad. Sch

    Parental Edu. Level

    Parent's Edu. Expectation (%)

    Parent's Edu. Expectation and Parental Edu. Level

    Sheet1

    Junior ColTech. Col/UnivReg. Col/UnivGrad. Sch

    18.147.614.214.4

    1.57.148.415.719.6

    25.545.317.826

    2.51.342.220.532.2

    31.928.127.239.3

    3.51.221.624.949.7

    4213.625.256.3

    4.50.75.918.173.3

    5018.317.263.1

    Sheet2

    Sheet2

    8.147.614.214.4

    7.148.415.719.6

    5.545.317.826

    1.342.220.532.2

    1.928.127.239.3

    1.221.624.949.7

    213.625.256.3

    0.75.918.173.3

    018.317.263.1

    Junior Col

    Tech. Col/Univ

    Reg. Col/Univ

    Grad. Sch

    Parents' Ave. Edu. Level

    Parent's Edu. Expectation (%)

    Parent's Edu. Expectation and Parents' Ave. Edu. Level

    Sheet3

  • Data and MethodData: 11th graders (N = 11,515)Measures-- Dependent Variable: 14 items selected from The Symptom Checklist-90-Revised (SCL-90-R) that measured frequencies of depressive, anxious, aggressive, and psychosomatic symptoms and suicidal ideation. A factor score was derived from the 14 items by using the confirmatory factor analysis modeling.

  • Data and Method (contd)-- Independent Variables:1. Academic achievement: IRT ability score.2. Family SES: Parents educational level and monthly income.2. Authoritative parenting: items related to parents acceptance, non-punitive behavior, and listening to inner thoughts.3. Parental involvement in education:a. Involvement related to support/warmth and behavioral control including helping with school work, checking school work, and supervision after school.b. Involvement related to psychological control including talks about future schooling plans and talks focused on academics.

  • Data and Method (contd)-- Control Variables:1. Students sex2. Stressful family events experienced: Parents divorce, separation, or death; parents very ill; parents with psychological illness; alcoholic parents; sudden economic fall of the family.Method: OLS regressionModel 1: Control + Family SESModel 2: Control + SES + IRT score + (IRT score)2Model 3: Control + Parenting behavior and educational involvementModel 4: Full model

  • Summary of FindingsHigher family SES has negative effects on mental health. But the effect mostly vanishes when academic achievement is controlled.Academic achievement has negative effects on mental health. Although the quadratic term is significant, the trend is linear most of the time.

  • Summary of Findings (contd)Parental support and behavioral control generally have positive effects on both mental health and academic achievement.Parental acceptance (dimension of warmth and support) has effects positive on mental health and negative on academic achievement. The latter effect falls into insignificance once SES is controlled. Non-punitive parenting, however, has positive effects on either achievement or mental health.

  • Summary of Findings (contd)Parents talking about schooling and occupation plans (psychological control) is positively related to academic achievement, but negatively related to mental health. The effect sizes of parenting behaviors remained very much the same after academic achievement is controlled.

  • Access to TEPS: A Public AssetData of the 1st wave (2001) has been released for public access (http://www.teps.sinica.edu.tw). The first follow-up data (2003) will be released soon.Three levels of access:-- Public access: Online application; no school and class id; 70% of the original sample.-- Restricted access: Restricted to academic and governmental institutions; needs to sign an agreement of confidentiality; could study class effects, but not school effects.-- On-site access: Further restriction (at least a Ph.D. candidates with advisors endorsement). Nearly full access to the data.

  • Access to TEPS (contd)English translation of the questionnaires and various handbooks is underway.