Co-Development of Education Aspirations and Postsecondary Enrollment Especially Among Students Who Are Low Income and Minority

Abstract

Most adolescents, regardless of their occupational aspirations, are educationally ambitious. However, these high education ambitions are often unaligned with their occupational ones. This misalignment between students’ educational expectations and their desired careers can be personally and institutionally costly. In contrast to earlier studies of educational attainment and occupational status, which relied independently on educational expectations and occupational aspirations, this study views alignment as related and co-developing. Results show that educational expectations and occupational aspirations co-develop during high school. Misalignment between educational and occupational expectations is a significant determinant of eventual educational attainment and has implications for future labor market outcomes.

Notes

1. Data are available upon request.

2. Among 15,857 total respondents, the educational expectation variable at Grade 11 is not available for 1,527 students and the occupational aspiration variable at Grade 11 is missing for 4,548 students. If we drop those missing respondents, we lose 5,321 individuals (34%). Results of our analysis are consistent whether we drop missing respondents or not.

3. Required education level for each expected occupation category was determined using O*NET 98 (U.S. Department of Labor’s Occupational Information Network) released in 1990. Required education levels are (1) high school diploma or lower for health care support, protective service, food preparation, personal care, sales, office and administrative support, construction and extraction, installation and maintenance, production, transportation, and military specific occupations; (2) associates degree; (3) bachelors degree for management, business and financial operations, architecture and engineering, life physical and social science occupations, community and social services, education training occupations, and arts design entertainment occupations; and (4) graduate degree for legal occupations and healthcare practitioners.

4. Student self-reports were used as the transcript data has not yet been released.

5. Stop trying and born ability are both reverse coded in our analysis, so that a 1 indicates disagreement with the statement.

6. One disadvantage of including social and emotional perception measures is that we drop observations with missing data and the sample becomes smaller, with students who are prepared for college being over-represented because the students with lower test scores are less likely to answer those questions. Thus, we included indicators for missing variables. When we exclude dummies for missing values for each question and reestimate the model, the majority of estimates are not statistically different.

7. Highest level of science is in five levels: no science (0), general science (1), special science (2), advanced science (3), and Advanced Placement/International Baccalaureate science (4). Highest level of math and Grade Point Average are also available but have high correlation with standardized math score.

8. We use highest level of science taken and social and emotional measures related to science courses because science is a field where the job market has changed considerably since earlier work on alignment was conducted. Furthermore, math requirements are often more restrictive than science requirements, which should give us more useful information about social and emotional measures. However, no science achievement score is available in HSLS:09, so we use the math score instead because it correlates highly with science achievement.

9. When we include all demographic, family background, school characteristics, and perception measures that we included in the main estimation, those who answered yes to the science purposefulness question are around 13 percentage points more likely to take advanced science courses. The other two perception variables (born, stop trying) are not significant.

10. Estimates for binary dependent variables that are generated by linear probability model turn out to be extremely similar to those generated by the logit model.