Archived Information

Answers in the Tool Box: Academic Intensity, Attendance Patterns, and Bachelor's degree Attainment — June 1999

Appendix B: Inclusions and Exclusions in the Academic Resources Variable:
a Demographic Portrait

It is natural to ask whether some population groups are more likely to be excluded from analyses based on the "academic resources" variable, and what one does to rectify the situation. Some analysts, afraid to lose cases from relatively small samples, will impute values on the basis of sub-group means. But I find no persuasive evidence to impute a test score or a high school class rank or the details of a high school curriculum. Instead, one must assume that missing evidence is randomly distributed, and the key to adjustment lies in the various panel weights provided in the dataset. Because exclusions require reweighting, the table below uses unweighted proportions to provide some hints as to how much adjustment for missing evidence a modified weight will account, particularly if the universe is limited to on-time high school graduates.

  Who is Missing One or More of the Three Components (Test Score,
Class Rank/GPA, Curriculum Intensity & Quality) of "Academic Resources"
 
 

Proportion of Students Whose Records Show
  All Three
Components		 1 or 2
Components		 No
Components		 Event
Drop-Outs		 H.S. Graduates
On-Time		 Other
Men 85.4% 13.7% 0.9% 16.2% 82.1% 3.9%
Women 85.5 13.8 0.7 14.7 85.6 4.0
 
White 88.1 11.5 0.4 14.8 86.3 3.9
Black 82.3 17.1 0.6 18.2 81.6 3.5
Latino 83.1 16.1 0.8 18.3 79.5 4.2
Asian 89.0 11.0 0.0 2.9 90.0 3.1
AmerInd 81.5 17.9 0.6 18.0 74.6 6.8
 
SES Quintile
   High 88.3 11.5 0.2 5.8 94.8 1.9
   2nd 89.5 10.2 0.3 9.1 90.2 2.7
   3rd 87.5 12.3 0.2 14.0 87.3 3.3
   4th 87.3 12.3 0.4 18.1 82.4 4.6
   Low 83.7 15.6 0.7 26.4 73.2 6.7
 
No HS
Diploma		 69.3 27.0 3.7 60.3 N.A. N.A.

It is obvious that the first filter on analysis must be that of high school graduation. Some 60 percent of the permanent ("status") drop-outs (at age 30) were also event drop-outs, and these students tend to be overrepresented in the lowest SES quintiles (not surprising), where black, Latino, and American Indian students are also overrepresented. A relatively high proportion of status drop-outs will not have senior test scores or full high school records. Once one filters the population, it is necessary to create a new senior year weight to account for non-response by reason of missing records.
-###-

Appendix A: Technical Notes and Guidance [Table of Contents] Appendix C: Gradations of Academic Intensity & Quality of H.S. Curriculum