National Education Longitudinal Study of 1988 (NELS:88)
|
National Education Longitudinal Study of 1988 (NELS:88)
|
NELS 101: An Introduction to NELS |
|
by Shelley Correll, Statistical Software Support (September 8, 1998) |
OverviewThe National Educational Longitudinal Study of 1988 (NELS:88) is a two-stage stratified probability sample of approximately 25,000 8th grade students who were followed and surveyed at two year intervals for the period of 1988 to 1994. While the NELS:88 data set is primarily focused on students, survey respondents also included students’ teachers, school principals and parents. The longitudinal design of NELS:88 allows researchers to examine how changes in the lives of students and the characteristics of schools, teachers and families impact various educational and social outcomes. Structure of the NELS:88 data setNELS:88 is currently composed of data from four longitudinal waves divided into seven components. The base year study design comprised four components: surveys and tests of students, and surveys of school administrators, parents and teachers. The following descriptions are intended as a very brief introduction to the data set. Users should consult the users manuals provided by the National Center for Educational Statistics (NCES). (See "Must Read Publications" below). For examples of variables associated with different waves and components see "Examples of Variables" below. NELS:88 Waves: Currently the NELS:88 data set is comprised of data from four waves:
NELS:88 Components: The dataset consists of seven different components. Each wave contains a slightly different set of these components. Below is a description of each component and the waves in which it was collected. Student component: This is the main component which contains the responses of students to a large number of survey questions. The questions are on diverse topics such as: school life, family background, attitudes towards smoking and plans for the future. ====> Waves collected: ALL Parent component: The parent component contains parents’ responses to questions about parental behaviors and circumstances about which parents may be more knowledgeable than the students, such as parental education and occupation. The parent questionnaire also contains more sensitive items relating to income and family history. The parent component is intended to be the definitive data source for examining the relationship between student measures and family background and characteristics. ====> Waves collected: Base Year and Second Follow-up School component: The school component contains the responses of the principal or head master at the schools selected for the NELS:88 sample to questions about the overall academic climate of their schools. In particular, information was collected about characteristics of the school, student body, and teaching staff and about school policies, programs and governance. ====> Waves collected: Base Year, First Follow-up, Second Follow-up Teacher component: The teacher component was designed to provide teacher information that could be used to analyze the behaviors and outcomes of the student sample. Selected teachers in two of the four subject areas (math, science, English and social studies) completed the questionnaire which elicited teacher evaluations of student characteristics and performance and curricular information about classes taught. ====> Waves collected: Base Year, First Follow-up, Second Follow-up Assessments: In addition to the student questionnaire, students completed a series of cognitive tests which covered four subject areas: reading comprehension, mathematics, science, and history/citizenship/geography. The Educational Testing Service (ETS) developed the cognititve test battery, which included 116 items to be completed in 85 minutes. ====> Waves collected: Base Year, First Follow-up, Second Follow-up Dropouts: Beginning with the first follow-up, a high school dropout component was collected. Students from the base year sample who dropped out of high school were asked questions about their reasons for leaving school and their future plans, including any plans for returning to high school. They were also asked questions about their school experiences, family background, and leisure activities. ====> Waves collected: First Follow-up, Second Follow-up, Third Follow-up Transcripts: This component contains transcript data, such as coursework and grades. collected in the spring of 1992. For budgetary reasons, high school transcripts were not collected for all students. Thus, researchers using this component need to use a separate transcript weight. See the NCES publication "User’s Manual: NELS:88 Follow-up: Transcript Component Data File. The transcript data is only available as part of the "Restricted Use" NELS:88 data set. (See " A Note About the Restricted Release Data File" below). ====> Wave collected: Second Follow-up
Examples of VariablesThe variables associated with the base year and first and second follow-ups were given names that identify the wave and component from which they were collected. Third wave variables utilize names that are more descriptive of the variable itself but which do not link them to a wave/component. Examples of variables are given below. Identifying Variables by Wave Variable names for the base year and first and second follow-ups begin with a 2-digit code to identify the wave. "BY" is used to identify base year variables and "F1" and "F2" are used to represent first and second follow-up variables, respectively. Identifying Variables by Component The 2-digit wave code is followed by a one or two letter code to identify the component or source of the variable (student, school, teacher, parent, dropout). The examples below illustrate this naming scheme. Sequence of Variables The variables in the NELS:88 data set are arranged in the following order:
Examples of Variables: Names and Survey Questions Example 1: This variable is a base year variable asked of a student. Note that the variable’s name begins with "BY," to identify it as a base year variable, followed by an "S" to identify it as a student component variable. Variable Name: BYS42A Question: During the school year, how many hours a day do you watch TV on weekdays? Example 2: This variable is a first follow-up variable asked of a high school dropout student. Note that the variable’s name begins with "F1," to identify it as a first follow-up variable, followed by a "D" to identify it as a dropout component variable. Variable Name: F1D12G Question: How much do you agree with the following statement (about the last school you attended): The teaching was good. Example 3: This variables is a school component variable (asked of a school administrator) from the first follow-up wave. "F1" indicates the wave, and "C" identifies in as a school component variable. Variable Name: F1C19 Question: Approximately how many colleges sent a representative to your school to talk with college bound students during the 1989-90 school year? Example 4: This variable is from the parent component from the second follow-up. "F2 identifies the second follow-up, "P" indicates that the question was ask of one of the student’s parents. Variable Name: F2P13 Question: Which of the following categories comes closest to describing your present or most recent job or occupation? Example 5: The third follow-up variables do not utilize the same naming scheme. Instead, their names tend to be more descriptive of the wording of the variable itself. Below is an example: Variable Name: MARSTAT
How the NELS:88 Sample was ChosenThe NELS:88 is a two-stage stratified random sample of 8th graders from 1988. Schools constitute the primary sampling unit. In 1988, approximately 1000 public and private eighth grade schools were selected from a universe file of approximately 40,000 eighth grade schools. In each of the approximately 1000 sampled eighth grade schools, 24 students were randomly selected for inclusion into NELS:88. An additional 2-3 Asian and Hispanic students were then selected from each school. Although the goal of NELS:88 was to follow a group of 8th graders, there were both additions and deletions to the sample in the subsequent waves. The deletions were primarily the result of sub-sampling at two points in the study. The first sub-sample occurred during the first follow-up when students had transitioned from the original 1000 junior high or middle school settings to around 5000 high school settings. The second sub-sample occurred during the third follow-up when the sample was reduced to approximately 14,000. The additions to the original sample included the augmentations of freshened and Base Year Ineligible students. The NELS:88 sample was freshened with 10th graders in 1990 and seniors in 1992 so that each wave would be representative of the universe of high school sophomores and seniors. Additionally, the approximately 5% of the base year sample who were removed from the sample by the school principal due to some condition or handicap were re-evaluated during the first and second follow-ups. If their condition was determined to no longer be severe enough to keep them from participating, they were added back into the sample.
Types of Analyses that Can be Carried OutThe NELS:88 design enables researchers to conduct analyses on three principal levels: longitudinal, cross-sectional and cross-cohort via comparison with two other data sets, High School and Beyond of the 1980s and The National Longitudinal Study of the 1970s. The longitudinal data gathered from students, and augmented with parent, teacher, school principal and school record accounts of students’ progression and development, facilitate examination of various facets of students’ lives and how these factors impact social, behavioral and educational development. By beginning with a cross-section of eighth graders in 1988 and following a sub-sample of these students at two-year intervals, and freshening the 1990 and 1992 samples, the study provides a statistical cross-section of American eighth graders, high school sophomores and high school seniors. Finally, NELS:88 was designed to allow for cross-cohort comparisons by comparing the NELS:88 sample with those generated by the two previous NCES studies listed above. For more information on the similarities and differences between these data sets see Appendix D of the NELS:88 Second Follow-Up: Student Component Data File User’s Manual.
Getting the Desired SampleDue to the complex nature of the NELS:88 design, researchers must first select the appropriate longitudinal or cross-sectional sample for their analyses. Flags are included in the data set to aid in selecting the appropriate sample. Flag variables give the status of each individual at a given point in time (e.g., dropout status) or identify a condition (e.g., handicapped). Below is a table of some of the possible samples that can be selected for analysis, the flags necessary to select these samples, and the weights that should be used in working with the selected sample. (For more on the use of weights see "An Introduction to Weights" below). Table 1. Flags and weights necessary for working with various NELS:88 sub-samples
*** "2 years beyond" refers to the 1994 wave of NELS:88 when most students were 2 years beyond high school. At this time, most were either in college or the labor force. Other flags that researchers might find useful are those that identify high school graduates, high school dropouts and those sample members who have transcript data available. More detail on flags and weights are provided in the User’s Manuals listed in the "Must Read Publications" section below. An Example of How to Use Flags: The following example is intended to illustrate how to use the weights and flags to select an appropriate sub-sample. Suppose you are interested in examining the improvement in math test scores from 8th to 12th grade. Since you are following a group of students over time, you would like to select a longitudinal sample of 8th to 12th grade students. Using Table 1 above, note that you should select those cases whose flag variable, "F3PNLFLG," equals 1. In analyses done with this sample, you should use the weight, F3PNLWT. (For information about how and why to use weights see "An Introduction to Weights" below). The sample you have selected contains all sample members for whom data is available from the base year, first follow-up and second follow-up. You might wish to further restrict you analysis to those sample members who graduated from high school, omitting those who dropped out. If so, you will need to use a flag to remove dropouts from the sample. Likewise, you might wish to restrict your analysis to those cases for which transcript data is available by using one of the transcript flags. (See the User’s Manuals listed in the "Must Read Publications" section below for a complete description of all weights).
An Introduction to WeightsThis section gives a very general description of what weights are and why they should be used. Weights are variables that are included in a data set to compensate for unequal probabilities of sample selection and to adjust for the effects of non-response. By using weights, the researcher can make generalizations to the national populations represented by NELS:88. Since there are many possible sub-samples that can be analyzed with NELS:88, many different weights are included in the dataset. Depending on the group to whom the data are designed to generalize, the individual weights all have positive values for members of the group and have zero values for non-members. Thus, weights can also be used to select appropriate sub-samples. While it is beyond the scope of this document to fully discuss the use of weights, it should be mentioned that if weights are not used, the estimates produced from NELS:88 will not be representative of the population about which the researcher is attempting to estimate. The User’s Manuals listed under "Must Read Publications" provide extensive guidelines for working with weights when using NELS:88. More information on the use of weights and more general guidelines for working with complex surveys, such as NELS:88, can be found in: Lee, Eun Sul; Forthofer, Ronald N. and Ronald J. Lorimor. 1989 Analyzing Complex Survey Data , Quantitative Applications in the Social Sciences Series No. 71, Michael Lewis-Beck, Series Editor. Thousand Oaks, CA: Sage Publications, Inc.
Must Read PublicationsAccording to the "Guide to Using NELS:88 Data" by NCES, reading the publications from the following list will provide researchers with 99% of the information they need to understand the complexities of the NELS:88 data files.
A Note About the Restricted Release Data FileThis document is intended to support the extraction utility on this web page, which assists Stanford users in extracting variables from the NELS:88 Public Release Files. If you need data such as individual transcript course data or detailed characteristics of students’ neighborhoods you could consider obtaining an NCES license agreement to use the Restricted Release Data File (also called the "Privileged Use File"). Contact Cynthia Barton at Cynthia_Barton@ed.gov for more information on obtaining a license. |