We use linked K-12 and community college administrative data from North Carolina to provide descriptive evidence on remediation in math and English, and to evaluate the effect of remediation using across-college variation in placement policies.  Our descriptive evidence examines variation in the likelihood of assignment to remedial coursework conditional on 8th grade standardized test scores.  Students in the same statewide test score quartile who reside in areas served by different community colleges exhibit significant variation in the likelihood of enrolling in either remedial math or English.  Moreover, the 8th grade test score-assignment gradient varies significantly across colleges.  We use this variation to identify the impact of assignment to remedial coursework for students at varying points in the 8th grade test score distribution.  Instrumental variable results suggest that negative OLS estimates of the relationship between remedial course enrollment and measures of success in community colleges are actually biased towards zero.  This pattern suggests that students who enroll in remedial courses are positively selected on unobservables, presumably because higher-risk students opt to defer enrollment.