![Sample our Education journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5932/TOC-Subject-Sample-2021-Education.jpg"})
ABSTRACT
The literature has shown inconsistent support for the equalization thesis, that is, the idea that a college degree erases the effect of social origin on socioeconomic destination, and suggested higher intergenerational persistence among advanced degree holders compared to those with bachelor’s degrees. The present study sheds light on the origin-destination link by investigating the intergenerational association between parents’ education and offspring’s earnings, paying attention to parents’ education relative to their children’s. Drawing on large samples and multiple waves of data from the National Survey of College Graduates, this study also makes an empirical contribution by analyzing intergenerational persistence across degree types. For women, I find highest intergenerational persistence at the bachelor level, but little evidence of intergenerational association for any advanced degrees. For men, results show intergenerational persistence across educational groups. Differences across respondents holding different types of degree support a theory of intergenerational relative education advantage, in which the effect of parents’ education on offspring’s attainment varies depending on offspring’s education relative to their parents. Educational and labor market-related factors do not change the overall picture.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1. Intergenerational association and intergenerational persistence are used as synonyms across the text.
2. More information can be found at https://www.nsf.gov/statistics/srvygrads/.
3. I thank Lynn Milan and Wan-Ying Chang at NSF for their advice on data sampling and modeling.
4. The lower threshold of age 35 limits the possibility that respondents get an advanced degree at older ages. Only a small number of people acquire an advanced degree after age 35 (Kim et al., Citation2015).
5. Missingness is very limited. In the selected age range, less than 0.5% of the cases have missing parental education and only 0.4% report zero income. Accordingly, list-wise deletion of such cases should not raise major concerns; additionally, that is in line with previous research using the same data (Witteveen & Attewell, Citation2020). In line with previous research (e.g., Manzoni & Streib, Citation2019; Torche, Citation2018; Witteveen & Attewell, Citation2020), the analyses only focus on those with a job; about 15% of respondents do not have a job. Sensitivity analyses evaluating selection in the study sample based on two stage Heckman selection models predicting selection into the sample in the first stage and earnings in the second stage confirmed the picture the current analyses reveal.
6. No distinction can be made within these five categories. Bachelor includes BS, BA, AB; Master includes MS and MA; Professional includes JD, LLB, MD, DDS, DVM; Doctorate includes be PhD, DSc, EdD. The NSCG does not distinguish between MA and MBA. However, following the suggestion of the NSCG staff as well as previous work using the NSCG (Hersch, Citation2019; Oh & Kim, Citation2020), I classified MA with major in Business as MBA.
7. Annual salary is top-coded at $200,000.
8. I checked results against using OLS regression on log-earnings and results were confirmed across models. Results are not shown but are available upon request.
9. Separating across recipients of different types of degrees facilitates comparison of effects across types of degrees, although does not speak of the statistical significance of differences. In additional analyses (not shown but available upon request), I estimated models on the full sample of all degree recipients including interaction terms between type of degree and parental education. Results from such analyses confirmed that effects which are significant in the sub-group analyses and appear different across groups are also significantly different statistically.
10. While I recognize the relevance of race, as well as the association between race and class, the data only allow us to introduce race as control; separate analyses by race are not feasible due to small cell sizes.
11. This includes Natural Resources, Construction, Maintenance, Production, Transportation, Material Moving.
12. I made empirical choices to account for potential multicollinearity among some of the included controls. For example, rather than looking at work experience, which would be highly correlated with age at degree, I focused on tenure in a job. Additionally, I ran tests to address multicollinearity concerns. A high correlation among control variables would be a first cause of concern; however, results are reassuring in that they reveal that correlation among predictors is low. While correlation only identifies issues for pairs of variables, alternative strategies to detect multicollinearity, such as VIF, may not work well in non-linear regression. Additionally, satisfactory standard errors/confidence intervals in the models suggest that multicollinearity is not causing a problem.
13. in the Appendix provides full results for the regression models for men. Given the overall findings of weak or no intergenerational persistence among women, I only show full results for men.
14. Sensitivity tests for women confirmed the picture of weak or no intergenerational association. Thus, I do not show those tests.