Skip Navigation

National Education Longitudinal Study of 1988 (NELS 88)



4. SURVEY DESIGN

Target Population


Students enrolled in the 8th grade in "regular" public and private schools located in the 50 states and the District of Columbia in the spring 1988 school term. The sample was freshened in both the first and second follow-ups to provide valid probability samples that would be nationally representative of 10th‑graders in spring 1990 and 12th‑graders in spring 1992. The NELS:88 project excludes the following types of schools: Bureau of Indian Education (BIE)1schools, special education schools for the handicapped, area vocational schools that do not enroll students directly, and schools for dependents of U.S. personnel overseas. The following students are also excluded: mentally handicapped students and students not proficient in English, for whom the NELS:88 tests would be unsuitable; and students having physical or emotional problems that would make participation in the survey unwise or unduly difficult. However, the Base‑Year Ineligible Study (in the first follow‑up) and the Followback Study of Excluded Students (in the second follow‑up) sampled excluded students and added those no longer considered ineligible to the freshened sample of the first and second follow‑ups, respectively.  

Sample Design

NELS:88 was designed to follow a nationally representative longitudinal component of students who were in the 8th grade in spring 1988. It also provides a nationally representative sample of schools offering 8th grade in 1988. In addition, by freshening the student sample in the first and second follow‑ups, NELS:88 provides nationally representative populations of 10th-graders in 1990 and 12th‑graders in 1992. To meet the needs for cross‑sectional, longitudinal, and cross-cohort analyses, NELS:88 involved complex research designs, including both longitudinal and cross sectional sample designs.

Base‑Year Survey. In the base year, students were selected using a two‑stage stratified probability design, with schools as the first‑stage units and students within schools as the second‑stage units. From a national frame of about 39,000 schools with 8th grades, a pool of 1,030 schools was selected through stratified sampling with probability of selection proportional to their estimated 8th‑grade enrollment; private schools were oversampled to ensure adequate representation. A pool of 1,030 replacement schools was selected by the same method to be used as substitutions for ineligible or refusal schools in the initial pool. A total of 1,060 schools cooperated in the base year; of these, 1,060 schools (815 public and 237 private) contributed usable student data. The sampling frame for NELS:88 was the school database compiled by Quality Education Data, Inc., of Denver, Colorado, supplemented by racial/ethnic data obtained from the U.S. Office for Civil Rights and school district personnel.

Student sampling produced a random selection of 26,440 8th‑graders in 1988; 24,600 participated in the base‑year survey. Hispanic and Asian/Pacific Islander students were oversampled. Within each school, approximately 26 students were randomly selected (typically, 24 regularly sampled students and 2 oversampled Hispanic or Asian/Pacific Islander students). In schools with fewer than 24 8th‑graders, all eligible students were selected. Potential sample members were considered ineligible and excluded from the survey if disabilities or language barriers were seen as obstacles to successful completion of the survey. The eligibility status of excluded members was reassessed in the first and second follow‑ups. (See below.)

Top

First Follow‑up Survey. There were three basic objectives for the first follow‑up sample design. First, the sample was to include approximately 21,500 students who were in the 8th‑grade sample in 1988 (including base‑year nonrespondents), distributed across 1,500 schools. Second, the sample was to constitute a valid probability sample of all students enrolled in the 10th grade in spring 1990. This entailed “freshening” the sample with students who were 10th‑graders in 1990 but who were not in the 8th grade in spring 1988 or who were out of the country at the time of base‑year sampling. The freshening procedure added 1,230 10th‑graders; 1,040 of the students in this new group were found to be eligible and were retained after final subsampling for the first follow‑up survey. Third, the first follow‑up was to include a sample of students who had been deemed ineligible for base‑year data collection due to physical, mental, or linguistic barriers to participation. The Base‑Year Ineligible Study reassessed the eligibility of these students so that those able to take part in the survey could be added to the first follow‑up student sample. Demographic and school enrollment information was also collected for all students excluded in the base year, regardless of their eligibility status for the first follow‑up.

While schools covered in the NELS:88 base‑year survey were representative of the national population of schools offering the 8th grade, the schools in the first follow‑up were not representative of the national population of high schools offering the 10th grade. By 1990, the 1988 8th‑graders had dispersed to many high schools, which did not constitute a national probability sample of high schools. To compensate for this limitation, the High School Effectiveness Study (HSES), which was designed to sustain analyses of school effectiveness issues, was conducted in conjunction with the first follow‑up. From the pool of participating first follow‑up schools, a probability subsample of 251 urban and suburban schools in the 30 largest Metropolitan Statistical Areas were designated as HSES schools. The NELS:88 core student sample was augmented to obtain a within‑school representative student sample large enough to support school effects research. The student sample was increased in HSES schools by an average of 15 students to obtain within‑school student cluster sizes of approximately 30 students.

Second Follow‑up Survey. The second follow‑up sample included all students and dropouts selected in the first follow‑up. From within the schools attended by the sample members, 1,500 12th‑grade schools were selected as sampled schools. Of these, the full complement of component activities occurred in 1,370 schools. For students attending schools other than these 1,370 schools, only the Student and Parent Questionnaires were administered. As in the first follow‑up, the student sample was augmented through freshening to provide a representative sample of students enrolled in the 12th grade in spring 1992. Freshening added into the sample 243 eligible 12th graders who were not in either the base‑year or first follow‑up sampling frames. Schools and students designated for the HSES in the first follow‑up were followed up again—as part of both the main second follow‑up survey and a separate HSES. The Followback Study of Excluded Students was a continuation of the first follow‑up Base‑Year Ineligible Study. In addition, two new components—the High School Transcript Study and the Course Offerings Component—were added to the second follow‑up.

Top

Third Follow‑up Survey. The third follow‑up student sample was created by dividing the second follow‑up sample into 18 groups based on students’ response history, dropout status, eligibility status, school sector type, race, test scores, SES, and freshened status. Each sampling group was assigned an overall selection probability. Cases within a group were selected such that the overall group probability was met, but the probability of selection within the group was proportional to each sample member’s second follow-up design weight. Assigning selection probabilities in this way reduced the variability of the third follow‑up raw weights and consequently increased the efficiency of the resulting sample from 40.1 to 44.0 percent.

Fourth Follow‑up Survey. The fourth follow‑up student sample was the same as the third follow‑up student sample. Data collection for the NELS:88 fourth follow‑up survey ended in September 2000, providing a final respondent population of approximately 12,100 individuals.

The Postsecondary Education Transcript Study, conducted as part of the fourth follow‑up in 2000, followed those who reported having attended at least one postsecondary institution according to either the third follow‑up survey in 1994 or the fourth follow‑up survey in 2000. A total of approximately 9,600 fourth follow‑up survey respondents (79 percent of the overall respondent population) reported postsecondary experience since high school. Approximately 21 percent of the NELS:88 respondent population did not participate in postsecondary education.

Within this sample of students, the transcript data collection further targeted students who attended only postsecondary institutions identified in the Integrated Postsecondary Education Data System (IPEDS) institutional data file, thus excluding postsecondary information collected from foreign institutions, non-degree‑granting programs, and non‑credit‑granting institutions. Transcripts were requested from a total of 3,200 postsecondary institutions.  

Top

Data Collection and Processing

NELS:88 compiled data from five primary sources: students, parents, school administrators, teachers, and high school administrative records (transcripts, course offerings, and course enrollments). Data collection efforts for the base year through third follow‑up extended from spring 1988 through summer 1994. Self‑administered questionnaires, cognitive tests, and telephone or personal interviews were used to collect the data. The follow‑up surveys involved extensive efforts to locate and collect data from sample members who were school dropouts, school transfers, or otherwise mobile individuals. Coding and editing conventions adhered as closely as possible to the procedures and standards previously established for NLS:72 and HS&B. The contractor National Opinion Research Center (NORC) at the University of Chicago was the prime contractor for the NELS:88 project from the base year through the third follow‑up, but Research Triangle Institute conducted the fourth follow‑up.

Reference dates. In the base‑year survey, most questions referred to the student’s experience up to the time of the survey administration in spring 1988. In the follow‑ups, most questions referred to experiences that occurred between the previous survey and the current survey. For example, the second follow‑up largely covered the period between 1990 (when the first follow‑up was conducted) and 1992 (when the second follow‑up was conducted).

Top

Data collection. Prior to each survey, it was necessary to secure a commitment to participate in the study from the administrator of each sampled school. For public schools, the process began by contacting the Council of Chief State School Officers and the officer in each state. Once approval was gained at the state level, contact was made with district superintendents and then with school principals. For private schools, the National Catholic Educational Association and the National Association of Independent Schools were contacted for endorsement of the project, followed by contact of the school principals. The principal of each cooperating school designated a School Coordinator to serve as a liaison between contractor staff and selected respondents—students, parents, teachers, and the school administrator. The School Coordinator (most often a guidance counselor or senior teacher) handled all requests for data and materials, as well as all logistical arrangements for student‑level data collection on the school premises. Coordinators were asked to identify students whose physical or learning disabilities or linguistic deficiencies would preclude participation in the survey and to classify all eligible students as White, Black, Hispanic, Asian/Pacific Islander, or “other” race.

For the base‑year through second follow‑up surveys, Student Questionnaires and test batteries were primarily administered in group sessions at the schools on a scheduled Survey Day. The sessions were monitored by contractor field staff, who also checked the questionnaires for missing data and attempted data retrieval while the students were in the classroom. Makeup sessions were scheduled for students who were unable to attend the first session. In the first and second follow‑ups, off‑campus sessions were used for dropouts and for sample members who were not enrolled in a first follow‑up school on Survey Day. The School Administrator, Teacher, and Parent Questionnaires were self‑administered. Contractor field staff followed up by telephone with individuals who had not returned their questionnaires by mail within a reasonable amount of time.

The first follow‑up data collection required intensive tracing efforts to locate base‑year sample members who, by 1990, were no longer in their 8th‑grade schools but had dispersed to many high schools. Also, in order to derive a more precise dropout rate for the 1988 8th-grade cohort, a second data collection was undertaken 1 year later, in spring 1991. At this time, an attempt was made to administer questionnaires-by telephone or in person-to sample members who had missed data collection at their school or who were no longer enrolled in school. The first follow‑up also included the Base‑Year Ineligible (BYI) Study, which surveyed a sample of students considered ineligible in the base year due to linguistic, mental, or physical deficiencies. The BYI Study sought to determine if eligibility status had changed for the excluded students so that newly eligible students could be added to the longitudinal sample. If an excluded student was now eligible, an abbreviated Student Questionnaire or a Dropout Questionnaire was administered, as appropriate. For those students who were still ineligible, their school enrollment status was ascertained and basic information about their sociodemographic characteristics was recorded.

Tracing efforts continued in the second and third follow‑ups. In the second follow‑up (conducted in 1992), previously excluded students were surveyed through the Followback Study of Excluded Students. The second follow‑up also collected transcript, course offerings, and course enrollments from the high schools; reminder postcards were sent to principals who did not respond within a reasonable period. Data collection for the High School Effectiveness Study (HSES) was conducted concurrently with the collection for the second follow‑up. Because of the overlap in school and student samples, survey instruments and procedures for the HSES were almost identical to those used in the NELS:88 second follow-up survey.

By 1994, when the third follow‑up was conducted, most sample members had graduated from high school and it was no longer feasible to use group sessions to administer Student Questionnaires. Instead, the dominant form of data collection was one‑on‑one administration through computer‑assisted telephone interviewing (CATI). In‑person interviews were used for sample members who required intensive in‑person locating or refusal conversion. Only the Student Questionnaire was administered in the third follow‑up.

By 2000, when the fourth follow‑up was conducted, most sample members who attended college and technical schools had completed their postsecondary education. Data collection for the fourth follow‑up survey was conducted almost exclusively with computer‑assisted interviewing, primarily by telephone (i.e., using CATI). However, in‑person field interviews were also completed with this technology. Field interviewers used the same computer‑assisted interview and online coding software as the study’s telephone interviewers, but on a laptop computer‑based platform (i.e., computer‑assisted personal interviewing, or CAPI). Thus, all of the entry of interview data was accomplished by the NELS:88 fourth follow‑up CATI‑CAPI system.

High school transcripts were collected as part of the second follow‑up. The groundwork for the collection of high school transcripts was laid in the spring and fall of 1991, during pre‑data collection activities for the second follow‑up. Principals were asked to provide any materials—such as course catalogs, student manuals or handbooks, course lists, and registration forms—that would aid transcript course coding. In August 1992, transcript survey materials were mailed to the principals of the NELS:88 and non‑NELS:88 schools attended or most recently attended by sample members eligible for the survey. Two weeks after survey materials were mailed, nonresponding principals were prompted for the return of transcripts with a postcard reminder. Principals who did not return transcripts within 3 weeks of the postcard prompt were prompted over the telephone. Telephone prompting of nonresponding principals continued from October 1992 through February 1993. Field visits to schools requesting assistance in the preparation of transcripts were conducted in February and March.

The Postsecondary Education Transcript Study was carried out at the conclusion of CATI and CAPI data collection for the fourth follow‑up survey. Data collection began in September 2000, and over the next 5 months project staff requested transcripts from postsecondary institutions that NELS:88 fourth follow‑up respondents reported attending during either the NELS:88 third follow‑up or NELS:88 fourth follow‑up studies. Requests for transcripts were sent to the registrars or other contacts at the schools. Telephone follow‑up with nonresponding institutions took place 2 weeks after transmission of the package. Data collection procedures were designed to follow, where possible, each institution’s typical procedures for producing and distributing student transcripts. Returned transcripts and related school catalogs and bulletins were inventoried, transcript identification numbers affixed to each, and unique identifying information removed.

Top

Processing. Data processing activities were quite similar for the base‑year survey and the first and second follow‑ups. An initial check of student documents for missing data was performed on‑site by contractor staff so that data could be retrieved from the students before they left the classroom. Special attention was paid to a list of “critical items.” Once the questionnaires and tests were received at the contractor, they were again reviewed for completeness, and a final disposition code was assigned to the case indicating which documents had been completed by the sample member. Postsecondary institutions reported by the student were coded using the standard IPEDS codes. Data entry for both Student Questionnaires and cognitive tests was performed through optical scanning. New Student Supplements and Dropout Questionnaires were converted to machine‑readable form using key‑to‑disk methods. All cognitive tests were photographed onto microfilm for archival storage.

In the third follow‑up, a CATI system captured the data at the time of the interview. The system evaluated the responses to completed questions and used the results to route the interviewer to the next appropriate question. The CATI program also applied the customary edits, described below under “Editing.” At the conclusion of an interview, the completed case was deposited in the database ready for analysis. There was minimal post‑data entry cleaning because the interviewing module itself conducted the majority of necessary edit checking and conversion functions.

Verbatim responses were collected in the third follow‑up for a number of items, including occupation and major field of study. When respondents indicated their occupation, the CATI interviewers recorded the verbatim response. The system checked the response using a keyword search to match it to a subset of standard industry and occupation codes, and then presented the interviewer with a set of choices based on the keyword matches. The interviewer chose the option which most closely matched the information provided by the respondent, probing for additional information when necessary. Quality control was ensured by a reading and recoding, if necessary, of the verbatim responses by professional readers.

In the fourth follow‑up, data were collected and edited almost exclusively with computer‑assisted interviewing, primarily by telephone (i.e., using CATI).

For the High School Transcript Study, student‑ and course‑level data were abstracted from transcripts. Transcript courses were coded using the course catalog for the school or district, in accordance with the Classification System of Secondary Courses, updated for the 1990 NAEP High School Transcript Study. When a school or district catalog was unavailable, courses were coded by title alone.

Information from the postsecondary education transcripts, including terms of attendance, fields of study, specific courses taken, and grades and credits earned, was coded and processed using a transcript control system developed specifically for this purpose. Specially trained research personnel then coded and tabulated these academic documents.

Top

Editing. In the base‑year through second follow‑up surveys, detection of out‑of‑range codes was completed during scanning or data entry for all closed‑ended questions. Machine editing was used to (1) resolve inconsistencies between filter and dependent questions; (2) supply appropriate missing data codes for questions left blank (e.g., legitimate skip, refusal); (3) detect illegal codes and convert them to missing data codes; and (4) investigate inconsistencies or contradictions. Frequencies and cross‑tabulations for each variable were inspected before and after these steps to verify the accuracy and appropriateness of the machine editing. Items with unusually high nonresponse or multiple responses were further checked by verifying the responses on the questionnaire. A final editing step involved recoding Student Questionnaire responses for some items to the codes for the same items in earlier NELS:88 waves or in HS&B. Once this was done, codes that differed in the Dropout Questionnaire were recoded to coincide with the codes used for Student Questionnaire responses.

In the third and fourth follow‑ups, machine editing was replaced by the interactive edit capabilities of the CATI system, which tested responses for valid ranges, data field size, data type (numeric or text), and consistency with other answers or data from previous rounds. If the system detected an inconsistency because of an interviewer’s incorrect entry, or if the respondent simply realized that he or she had made a reporting error earlier in the interview, the interviewer could go back and change the earlier response. As the new response was entered, all of the edit checks performed at the first response were again performed. The system then worked its way forward through the questionnaire using the new value in all skip instructions, consistency checks, and the like until it reached the first unanswered question, and control was then returned to the interviewer. When problems were encountered, the system could suggest prompts for the interviewer to use to elicit a better or more complete answer.    

Top

Estimation Methods

Sample weighting is required so that NELS:88 data are representative of the full population. Imputation for missing nonresponses, however, has not yet been systematically provided for data analysis.

Weighting. Weighting is used in NELS:88 data analysis to accomplish a number of objectives, including (1) expanding counts from sample data to full population levels; (2) adjusting for differential selection probabilities (e.g., the oversampling of Asian and Hispanic students); (3) adjusting for differential response rates; and (4) improving representativeness by using auxiliary information. Multiple “final” (or nonresponse‑adjusted) weights have been provided for analyzing the different populations that NELS:88 data represent (i.e., base-year schools; 8th-graders in 1988 and 2, 4, 6, and 12 years later; 1990 sophomores; 1992 seniors; and 2000 college graduates). Weights should be used together with the appropriate flag in order to analyze the sample for a particular targeted population.

Weights have not been constructed for all possible analytic purposes. In cases where no specific weight is available, existing weights may provide reasonable approximations. For instance, base-year parent and cognitive test completion rates were so high relative to Student Questionnaire completion that the student weight can be used for them with minimal bias.

NELS:88 weights were calculated in two steps: (1) unadjusted weights were calculated as the inverse of the probabilities of selection, taking into account all stages of the sample selection process; and (2) these initial weights were adjusted to compensate for nonresponse, typically carried out separately within multiple weighting cells. For detailed discussions of the calculation of weights for each wave, users are referred to the methodology reports for the individual surveys.

Scaling (Item Response Theory). Item Response Theory (IRT) was used to calibrate item parameters for all cognitive test items administered to students in NELS:88 tests. The tests conducted in each NELS:88 survey generated achievement measures in standardized scores.

Imputation. NELS:88 surveys have not involved large-scale imputation of missing data. Only a few variables have been imputed: student’s sex, race/ethnicity, and school enrollment status. For example, when sex was missing in the data file, the information was looked for in earlier school rosters. If it was still unavailable after this review, sex was assumed from the sample member’s name (if unambiguous). As a final resort, sex was randomly assigned.

1 These were referred to as Bureau of Indian Affairs (BIA) funded schools.

Top