We estimate a dynamic learning model of college dropout, taking advantage of unique expectations data to greatly reduce our reliance on standard assumptions. Our simulations show that 45% of dropout in the first 2 years of college can be attributed to what students learn about their academic performance, with this type of learning playing a smaller role later in college. Poorly performing students tend to leave because staying is not worthwhile rather than because they are at risk of failing out of school. Poor performance substantially decreases the enjoyability of school and substantially influences beliefs about postcollege earnings.

Research examining the educational attainment of low-income students has often focused on financial factors such as credit constraints. We use unique longitudinal data to provide direct evidence about a prominent alternative explanation—that departures from school arise as students learn about their academic ability or grade performance. Examining college dropout, we find that this explanation plays a very prominent role; our simulations indicate that dropout between the first and second years would be reduced by 40% if no learning occurred about grade performance/academic ability. The article also contributes directly to the understanding of gender differences in educational attainment.

Unique new data from a college with a mandatory work‐study program are used to examine the relationship between working during school and academic performance. Particular attention is paid to the importance of biases that are potentially present because the number of hours that are worked is endogenously chosen by the individual. The results suggest that, even if results appear reasonable, a researcher should be cautious when drawing policy conclusions about the relationship between hours worked and a particular outcome of interest unless he or she is confident that potential problems associated with the endogeneity of hours have been adequately addressed.

A serious difficulty in determining the importance of credit constraints in education arises because standard data sources do not provide a direct way of identifying which students are credit constrained. This paper differentiates itself from previous work by taking a direct approach, made possible by unique longitudinal data from the Berea Panel Study. The results from our study of Berea College students suggest that, while credit constraints likely play an important role in the drop-out decisions of some students, the large majority of attrition of students from low-income families should be primarily attributed to reasons other than credit constraints. (JEL I21, I22)

We use unique longitudinal data to provide direct evidence about interracial friendships at different stages of college and to provide new evidence about some of the reasons for the observed patterns of interaction. We find that, while much sorting exists at all stages of college, black and white students are, in reality, very compatible as friends; randomly assigned roommates of different races are as likely to become friends as randomly assigned roommates of the same race. Further, we find that, in the long run, being (randomly) assigned a black roommate significantly increases the number of other black friends a white student has.

This paper studies wage determination using the first longitudinal data set containing job-level task information for individual workers. Novel quantitative task measures detail the amount of time spent performing people, information, and objects tasks at different skill levels. These measures suggest natural proxies for on-the-job human capital accumulation and provide new insights about wage determination. Current job tasks are quantitatively important, with high-skilled tasks being paid substantially more than low-skilled tasks. There is no evidence of learning by doing for low-skilled tasks but strong evidence for high-skilled tasks. Current and past high-skilled information tasks are particularly valuable.

Abstract Using novel data from the Berea Panel Study, we show that the beauty wage premium for college graduates exists only in jobs where attractiveness is plausibly a productive characteristic. A large premium exists in jobs with substantial amounts of interpersonal interaction but not in jobs that require working with information. This finding is inconsistent with employer taste-based discrimination, which would favor attractive workers in all jobs. Unique task data address concerns that measurement error in the importance of interpersonal tasks may bias empirical work toward finding employer discrimination. Our conclusions are in stark contrast to the findings of existing research.

We develop and estimate a model of student study time choices on a social network. The model is designed to exploit unique data in the Berea Panel Study. Study time data allow us to quantify an intuitive mechanism for academic social interactions: own study time may depend on friend study time in a heterogeneous manner. Social network data allow us to embed study time and resulting academic achievement in an estimable equilibrium framework. We develop a specification test that exploits the equilibrium nature of social interactions and use it to show that novel study propensity measures mitigate econometric endogeneity concerns.