Detailed administrative data from a large and diverse community college are used to examine if academic performance depends on whether students are the same race or ethnicity as their instructors. To identify racial interactions and address many threats to internal validity we estimate models that include both student and classroom fixed effects. Given the large sample sizes and computational complexity of the 2-way fixed effects model we rely on numerical algorithms that exploit the particular structure of the model’s normal equations. Although we find no evidence of endogenous sorting, we further limit potential biases from sorting by focusing on students with restricted course enrollment options due to low registration priorities, students not getting first section choices, and on courses with no within-term or within-year racial variation in instructors. We find that the performance gap in terms of class dropout rates, pass rates, and grade performance between white and underrepresented minority students falls by roughly half when taught by an underrepresented minority instructor. We also find these interactions affect longer term outcomes such as subsequent course selection, retention, and degree completion. Potential mechanisms for these positive interactions are examined.

[go to paper]

This paper analyzes the importance of teacher quality at the college level. Instructors are matched to objective and subjective characteristics of teacher quality to estimate the impact of rank, salary, and perceived effectiveness on student performance and subject interest. Student and course fixed effects, time of day and week controls, and students’ lack of knowledge about first-year instructors help minimize selection biases. Subjective teacher evaluations perform well in measuring instructor influences on students while objective characteristics such as rank and salary do not. Overall, the importance of college instructor differences is small, but important outliers exist.

[go to paper ]

Many wonder whether teacher gender plays an important role in higher education by influencing student achievement and subject interest. The data used in this paper helps identify average effects from male and female college students assigned to male or female teachers. In contrast to previous work at the primary and secondary school level, our focus on large first-year undergraduate classes isolates gender interaction effects due to students reacting to instructors rather than instructors reacting to students. In addition, by focussing on college, we examine the extent to which gender interactions may exist at later ages. We find that assignment to a same-sex instructor boosts relative grade performance and the likelihood of completing a course, but the magnitudes of these effects are small. A same-sex instructor increases average grade performance by at most 5 percent of its standard deviation and decreases the likelihood of dropping a course by 1.2 percentage points. The effects are similar when conditioning on initial ability (high school achievement), and ethnic background (mother tongue not English), but smaller when conditioning on mathematics and science courses. The effects of same-sex instructors on upper-year course selection are insignificant.

[go to paper ]

Parameter estimates from earnings processes are key inputs into life-cycle models with heterogeneous agents, but it remains an open question how to model earnings dynamics appropriately. In this study I interpret the wide range of estimates of key parameters in the literature obtained from the same data, such as the variance of individual heterogeneity or persistence of income shocks, as reflecting a fundamental misspecication problem in two commonly used families of models - HIP- and RIP-models. I show that to obtain credible estimates of these parameters it is crucial to control flexibly for age- and time effects in innovation variances, including a rich specication of initial conditions. Starting from a model that is well-specied and that nests HIP- and RIP-models, I investigate the robustness of key parameters across specifications. To isolate the model-specific identifying variation of a parameter, I compare across specifications the results from novel numerical comparative statics that perturb the parameter around its estimated value. Since identication of my preferred model requires covariance structures that are disaggregated to the cohort-level, I rely on administrative social-security data from Germany on quarterly earnings that follow workers from labor market entry until 27 years into their career. I focus my analysis on an education group that displays a covariance structure with qualitatively similar properties like its North American counterpart. I find that (i) estimates of key parameters fluctuate widely across specifications, (ii) permanent and persistent shocks as well as intercept-heterogeneity are always signicant while transitory shocks are not, (iii) a persistent initial condition matches the complex earnings dynamics early in the life-cycle, (iv) slope-heterogeneity is highly signicant in a standard HIP-process but vanishes once one controls for age-effects appropriately and (v) slope-heterogeneity introduces a problem of "over-tting". These results are unchanged when I allow slopes to vary over the life-cycle and when I estimate the model from an education group with a drastically dierent covariance structure.

[go to paper]

New version coming soon.