Education Longitudinal Study of 2002

1
Education Longitudinal Study of 2002
NSF Summer Data Policy Institute
July 22, 2007
2
Objectives of ELS 2002

• Monitor the critical transitions made by a cohort
  of high school sophomores (10th graders) through
  college into their adult careers
• Maintain comparability with NLS-72, HSB, and
  NELS88 to measure trends in transitions and
  outcomes over time
• Link the assessment of achievement outcomes to
  other high school assessments (NAEP and PISA) for
  purposes of analysis
• Focus 2nd follow-up on issues of college access
  and choice

3
ELS 2002 Basic Survey Design

• Stage 1 School selection
• Target population schools with 10th graders
• Probability of selection proportional to school
  size
• Stratification region x urbanicity x school
  control
• Total strata 96
• Over sampled Catholic schools, and other private
• Final sample size 752 responding schools with a
  10th grade
• Similar to NELS88 but starts in spring of 10th
  grade

4
ELS 2002 Basic Survey Design

• Stage 2 Student selection
• Random selection of about 26 students per school
• From student rosters provided by school
• Number of students per school was limited to
  minimize burden
• Rosters organized by four race/ethnicity groups
• Over-sampled Asians/Pacific Islanders
• All ELL and special education students included
  (expanded sample)
• Sample size 17,591 10th grade students selected
  representing 3.6 million students

5
ELS 2002 Basic Survey Design

• Contextual data sources
• One parent per student
• Two teachers per student
• One a mathematics teacher of the student
• One an English/language arts teacher
• Principal of the school
• School characteristics
• School librarian/media center director
• Planned follow-ups
• First follow-up, high school seniors (2004)
  (done)
• Second follow-up, college sophomores (2006)
• Third follow-up, eight years after high school
  (2012)

6
Sample Design and Weights

• Schools and students are assigned weights
  according to their probabilities of selection.
• Use of these weights in analyses is necessary to
  obtain accurate population estimates.
• Values of the weights for school and student
  sample members are inversely proportional to
  their probabilities of selection.

7
Sample Design and Weights

• School Weights
• E.g., Stratum H in the Pacific region contains 80
  public, urban, schools
• If two schools are chosen (H1 and H2), the school
  weight for each one is 80/240
• So each sampled school in stratum H represents 40
  schools with a 10th grade.

8
Sample Design and Weights

• Student weights
• Roster of all 10th graders enrolled in spring
  2004 was provided by the selected school
• If school H1 had 120 sophomores and 24 were
  selected, then the within school weight for
  each student would be 120/24 5.
• The overall weight for these students would be 40
  x 5 200.

9
Sample Design and Weights

• Base Year student weights
• The result is a Base Year student weight for
  every individual in the sample corresponding to
  the number of individuals in the population that
  person represents.
• The distribution of these weights is

Statistics Base Year student weight, BYSTUWT
Mean 224
Median 205
Quartiles 99, 205, 300
Range 5-978
10
Sample Design and Weights

• Base Year school weights
• For purposes of analyzing schools, there is also
  a school weight for the Base Year schools.
• The distribution of these school weights is

Statistics Base Year school weight, BYSCHWT
Mean 32.7
Median 17.0
Quartiles 7.4,17.0,27.4
Range 0-395.7
11
Sample Design and Weights

• Between the Base Year and First Follow-up,
• Some student do not respond, and
• Some students migrate to new schools
• In ELS, the sample weights are adjusted to
  account for for sample member non-response
• Using model-based methods
• The total weight for Base Year cohort does not
  change.
• All students who transferred to another school
  were included in the First Follow-up data
  collection

12
ELS 2002 Sample Distribution
Change in the size of the ELS2002 sample of schools and students, by data collection stage (unweighted) Change in the size of the ELS2002 sample of schools and students, by data collection stage (unweighted) Change in the size of the ELS2002 sample of schools and students, by data collection stage (unweighted)
Base Year First Follow-up
Original school
Total number of schools 752 747
Total number of students 16,202 13,988
Average number students per school 21.5 17.8
Transfer school
Total number of schools --- 1,130
Total number of students --- 1,248
Average number students per school --- 1.2
Source ELS 2002 data file
13
Sample Design and Weights

• Additional measures to correct for non-response
• Non-respondents are included in follow-ups
• Receive special questionnaires to obtain data
  otherwise collected in previous follow-up
• Universe variable, F1UNIV1, shows the F1
  respondents who were non-respondents
• All of the main student background variables
  (sex, race, SES, etc.) and some status variables
  are imputed

14
Sample Design and Weights

• After imputation and non-response adjustment, a
  panel weight, F1PNLWT, is formed from sample
  members for whom there is response data in both
  the Base Year and First Follow-up.
• This panel weight should be used whenever change
  in a variable from BY to F1 is analyzed.
• This panel weight is zero for all students for
  whom BY-F1 information is not available
• The sample of cases still represents the
  distribution of the original Base Year Cohort

15
Sample Design and Weights

• For all students who completed a First Follow-up
  questionnaire, there is also a questionnaire
  weight, F1QWT
• F1QWT includes the freshened sample and all other
  students in Base Year or Transfer schools who
  completed a First Follow-up questionnaire,
  including dropouts, etc.
• The cross-sectional sample of seniors is formed
  by crossing F1QWT with a cohort flag, G12COHRT,
  identifying all in-school seniors in the First
  Follow-up.

16
ELS Analysis Populations (E4P)
Sample Population Weight
Base year students Cross-section of sophomores BYSTUWT
Expanded sample Cross-section of sophomores, including ineligible students BYEXPWT (restricted)
First Follow-up Cross-section of seniors (freshened) F1QWT x G12COHRT
Longitudinal panel Same group of sophomores followed for 2 years F1PNLWT
Longitudinal panel Same group of seniors followed for 4 years F2PNLWT
Base year schools Schools with 10th grade BYSCHWT
17
Design Effects and Standard Errors

• Because of the selection of multiple students per
  school, the ELS sample is clustered
• Because of this clustering, the standard errors
  of population estimates are larger than they
  would be with simple random sampling (SRS)
• The extent of departure is measured by the
  Design Effect or DEFF
• Because of the DEFF, special software must be
  used to correctly estimate standard errors

18
Design Effects and Standard Errors

• In general, DEFF is defined as the ratio of,
• the square of the correct standard error by
• the square of the standard error of the estimate
  assuming simple random sampling (SRS).
• DEFF (correct S.E.2 / (SRS S.E.)2

19
Design Effects and Standard Errors
Mean and root design effects for NELS88 and
ELS2002 sophomore cohort panel estimates
20
Survey Components

• Base Year survey
• Spring 2002 sophomores
• Student questionnaire
• Student cognitive tests
• Mathematics
• Reading
• Parent questionnaire
• Teacher questionnaires
• Principal questionnaire
• (school information)
• Library/media questionnaire
• School facilities checklist
• Geocoded data available

• First Follow-up survey
• Spring 2004 seniors
• Student questionnaire
• Transfer student quest.
• Dropout questionnaire
• Early graduate quest.
• Home school quest.
• New student quest.
• Student cognitive tests
• Mathematics
• Principal questionnaire
• (school update)
• High school transcripts

21
ELS Results Educational Expectations
Percentage of 12th graders who expected to attain
a bachelors degree or a graduate or professional
degree 1981-82, 1991-92, and 2003-04.
86
86
67
64
63
51
44
32
16
High SES
Low SES
Middle SES
Bachelors degree
Graduate or professional degree
22
Survey Components

• Third Follow-up
• Eight years after high school
• Adult questionnaire
• College enrollment
• College completion status
• Employment
• Life course (living situation, volunteer work,
  expectations)
• College transcripts

• Second Follow-up
• Two years after high school
• Young adult questionnaire
• High school completion update
• College choice
• College enrollment
• College major
• Employment (and military service)
• Life course (living situation, volunteer work,
  expectations
• Administrative records
• Federal student aid received
• SAT/ACT scores
• GEDs received

23
ELS High School Transcripts

• Sample
• Collected from Base Year school, and
• If in another school, from last school of
  attendance
• Provides a complete record of high school
  coursework
• Course titles, grades earned, credits earned,
  year taken, grades 9-12
• Coded according to course classification schemes
• Includes updated high school completion status
• Diploma or certificate of attendance
• GED
• Still in school

24
ELS High School Transcripts
Ninth and eventual 12th grade mathematics course
completed 2004

• Among students who complete geometry in the ninth
  grade, 84 take Advanced mathematics
  (trigonometry-pre-calculus or calculus) by the
  12th grade

• Among students who take general or applied
  mathematics in the ninth grade, 10 take Advanced
  mathematics

Percent completing specific course by end of
high school
25
Trend analysis with ELS, NELS, HSB
Survey and follow-up High school sophomores High school seniors College sophomores College Graduates
ELS 2002/12 3rd FU 2012
ELS 2002/06 2nd FU 2006
ELS 2002/04 1st FU 2004
ELS 2002 BY 2002
NELS 88/2000 4th FU 2000
NELS 88/2000 3rd FU 1994
NELS 88/2000 2nd FU 1992
NELS 88/90 1st FU 1990
HSB 1980 Soph. 4th FU 1992
HSB 1980 Soph. 2nd FU 1984
HSB 1980 Soph. 1st FU 1982
HSB 1980 Soph. BY 1980
CT
HT
CT
HT
CT
HT
26
Availability of Restricted Data

• Restricted data files
• Use requires a licensing agreement with NCES
• Restricted files contain information not on the
  Public Use data files
• Linking variables
• Course records from student transcripts
• No top or bottom coding
• Released on CD-Roms with the same ECB and kinds
  of documentation as Public Use files.
• http//nces.ed.gov/pubsearch/licenses.asp

27
Education Longitudinal Study 2002

• For additional information
• Go to ELS 2002 web site on the NCES web site
• nces.ed.gov/surveys/els2002
• To obtain Public Use data files
• click on Order CDs at the URL above
• Sign up for the High school longitudinal
  studies News Flash on the NCES web site