Background: The increasing availability of massive administrative data sets linking postsecondary enrollees with postcollege earnings records has stimulated a wealth of new research on the returns to college and has accelerated state and federal efforts to hold institutions accountable for students’ labor market outcomes. Many of these new research and policy efforts rely on state databases limited to postsecondary enrollees who work in the same state postcollege, with limited information regarding family background and precollege ability. Objectives: In this article, we use recent waves of data from the National Longitudinal Survey of Youth 1997 to provide new, nationally representative, nonexperimental estimates of the returns to degrees, as well as to assess the possible limitations of single-state, administrative data–based estimates. Research design: To do this, we explore the sensitivity of estimated returns to college, by testing different sample restrictions, inclusion of different sets of covariates, and alternative ways of treating out-of-state earnings to approximate the real-world limitations of state administrative databases. Results: We find that failure to control for measures of student ability leads to upward bias, while limiting the sample to college enrollees only leads to an understatement of degree returns. On net, these two biases roughly balance out, suggesting that administrative data–based estimates may reasonably approximate true returns. Conclusions: We conclude with a discussion of the relative advantages and disadvantages of survey versus administrative data for estimating returns to college as well as implications for research and policy efforts based upon single-state administrative databases.

中文翻译：

---

估算大学学历的回报：比较调查和州政府基于数据的估算。

背景：将大专生与大学后的收入记录联系起来的海量管理数据集的日益普及，激发了大量关于大学收益的新研究，并加速了州和联邦政府对机构追究学生的劳动力市场成果负责的努力。这些新的研究和政策措施中的许多工作都依赖于州数据库，这些数据库仅限于在同一州大专院校工作的大专生，有关家庭背景和大学前能力的信息有限。目标：在本文中，我们使用来自1997年全国青年纵向调查的最新数据来提供学位回报率的新的，具有全国代表性的非实验性估计，并评估单一州行政数据的可能局限性基于的估计。研究设计：为此，我们通过测试不同的样本限制，纳入不同的协变量集合以及处理州外收入以近似状态限制现实状态的替代方法，来探索估计的大学回报率的敏感性。管理数据库。结果：我们发现，未能控制学生能力的措施会导致向上偏见，而将样本限制为大学入学率只会导致对学位回报率的低估。从网络上看，这两个偏差大致上是平衡的，这表明基于行政数据的估计值可以合理地逼近真实回报。结论：