The paper uses a novel data set to illustrate a point which is consistent with the literature and my intuition. However, apart from that, I don't see a significant contribution in theory, empirical methodology, or conclusions. The authors stated in the second to last paragraph in the  introduction the use of such empirical research in general, but not how this specific paper contributes to the literature.

Some major comments:
(1) Apart from finding that CEE is significant in the model for both universities, there is no other theory or conjecture as to how and why the results are the same (or differ). For example, R^2s are very different between the two universities (Table 1). Is this difference significant? If so, why?
(2) Unobserved heterogeneity across individuals and universities could be huge and potentially time varying. But this is not controlled for in the model. Some random paraemter models or quantile regressions may be explored for their use in this context.
(3) The papers uses time invariant individual characteristics (i.e., CEE scores) to predict a variable (annual GPA) for three consecutive years. It would permit a much wider choice of empirical tools if the data is formulated as a panel.
(4) It would be interesting to see how CEE subject scores affect the GPA for courses (or course sequences) in the same subject area, e.g., how CEE score for math affects math related course grades in college.
(5) This study suffers from the exact same problems addressed in Rothstein (JOE, 2004), whose estimator could thus be very useful in this study.

Minor comment:
(1) The authors seem to have never memtioend how the models are estimated.
(2) Notes to the tables say "robust standard errors". Which one? Robust to what?

Possible extension: It would be interesting to check the effect of CEE scores on letter grades (rather than GPA), using models of discrete variables.

谢绝喝茶