AHAA & Add Health: The AHAA Design in Relation to the Add Health Waves

Using the AHAA data requires attention to the temporal order of survey administration relative to the academic year a course that appeared on the transcript was actually taken by a the student. Figure 1, below, illustrates these temporally ordered relationships for each Add Health grade level cohort. Add Health sampled approximately equal numbers of students in each of six grade levels (7-12). Each of these grade level cohorts are identified in the far left column, and the academic years in which each cohort is likely to be enrolled in high school are identified in each row with the designation of Year 1 - Year 4 (with Year 1 referring to the first year of high school, and so forth). The numbering system of Years 1-4 refers to the organization of the AHAA transcript-based constructed variables. The shaded cells in Figure 1 represent academic years in which a survey was administered. Notice that Add Health sample members who entered the study as 7th graders in the 1994-1995 academic year generally began high school coursework after the administration of the Wave II survey. In contrast, students who entered the study as a high school senior in the 1994-1995 academic year generally completed most, but not all of their high school coursework before they responded to the In-School survey.

The varying intersections between survey data and transcript data for the different Add Health cohorts results in several issues of which analysts should be aware. First, not all cohorts of students should necessarily be included in every analysis. Because the Add Health design was grade stratified, students are nationally representative of adolescents in their grade level in 1994-95, allowing for a narrowing of the sample to accurately reflect the research question of interest. If an analyst wants to determine how students’ attitudes early in high school influence subsequent academic achievement for example, only the 9th and 10th grade cohorts would be chosen for the analysis. Second, to use the AHAA academic indicators, analysts will often need to refer to different years of course-taking data for students from different cohorts. For example, if an analyst wanted to examine students’ grades (as reported on their transcript from AHAA) in the year immediately preceding an outcome measured at Wave II, this would necessitate referring to Year 1 data for the 9th grade cohort, Year 2 data for the 10th grade cohort, and Year 3 data for the 11th grade cohort. (Note that most 12th graders were not included in Wave II of Add Health). Third, using the AHAA academic indicators that measure the level of students’ courses (see the section on Math and Science Course Sequence indicators) also requires attention to the fact that the meaning of taking a lower level course changes according to when students take it.

Consider the following basic example to illustrate these three issues that must be considered when using Add Health survey data and AHAA high school transcript data. Suppose a researcher is interested in estimating the effects of parent’s education and student’s grades and math course level on student’s educational aspirations. The dependent variable, educational aspirations, is measured at Wave II. Parent’s education is available from either the In-School or the Wave I survey. To measure students’ grades and course placement prior to the timing of the dependent measure at Wave II, several steps must be taken. First, the analyst should select only grade level cohorts that were enrolled in high school prior to Wave II. This excludes the 7th and 8th grade cohorts. The 12th grade cohort, which was generally not interviewed at Wave II, would also be excluded because the dependent variable is not measured for this group. Thus, the analyst would select only the 9th, 10th and 11th grade cohorts and use as independent variables parents’ education, reported in 1994-95 (from either the In-School or Wave I survey), and grades and math course level from that academic year. The organization of the transcript data would require that the analyst use students’ grades in Year 1 for 9th grade cohort, grades in Year 2 for 10th grade cohort, and grades in Year 3 for 11th grade cohort to predict the Wave II survey item, educational aspirations. Accurately capturing the math course level requires an additional step, since the meaning of taking Algebra I as a 9th grader is different than taking Algebra I as a 11th grader. For this example, the analyst might choose to calculate the modal level of the math course sequence separately for each cohort of 9th graders, 10th graders, and 11th graders, and then create a new variable for whether each student’s math course is above, below, or at the mode for his or her cohort.

Because students from all cohorts have completed their high school course-taking by Wave III (with the exception of a small number of students who were in high school for longer than four years), predicting Wave III outcomes with AHAA data is less complicated. For such analyses, the AHAA data provides common benchmarks for all students, such as cumulative indicators of high school achievement or an educational indicator from students’ last year of high school.

For analysts primarily interesting in examining issues of adolescents’ educational experiences, the AHAA data provides a wealth of information on the complete high school careers of six nationally representative cohorts of students in the 1990’s. The Add Health survey provides related information on students’ family background and history which precede the high school careers of students from all cohorts.

For more general examples and analytic advice visit Analytic Advice