In this research, we investigated whether online discussions can significantly benefit students' learning in online courses. We designed an experiment by dividing 129 students who enrolled in four fully online "Introduction to Microeconomics" courses taught by the same instructor into four groups (one control group and three experimental groups). We adopted econometric analysis to investigate the issue. Our empirical evidence demonstrated that online discussions can significantly benefit students' exam performance in online courses only when instructors are effectively engaged in discussions. Instructors who do not effectively engage in online discussions had a less than significant effect on students' exam performance even when peer-responses were required. On the other hand, the effect was positive and significant when instructors effectively engaged but the level of significance was weak (at the 10% level of significance) and began with a later exam (Exam 4, out of a total of five exams) rather than the first exam or earlier. Detailed discussions about these phenomena are offered as are several possible reasons for these findings. We recommend significantly increasing weights for online discussions in the final course grade and including exam questions that directly reflect online discussion questions (including serious participation in discussions and requiring that students provide answers to instructors' follow-up questions). These additions would significantly benefit students' learning in online courses.